IMPROVING HEALTH SERVICE DELIVERY: USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS VOLUME 2: AN INTRODUCTION TO PATIENT PATHWAY ANALYSIS AND PRACTICAL EXAMPLES FROM THE FIELD AND PUBLISHED LITERATURE © 2025 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions: The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. IMPROVING HEALTH SERVICE DELIVERY: USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS VOLUME 2: AN INTRODUCTION TO PATIENT PATHWAY ANALYSIS AND PRACTICAL EXAMPLES FROM THE FIELD AND PUBLISHED LITERATURE Ahmad Hegazi Jessica Helen Watson Paul Ouma Nicole Fraser-Hurt Zara Shubber Table of Contents List of Figures iii List of Boxes iv List of Tables v Acknowledgments vi Acronyms vii Foreword  viii What is in This Guide x PART 1: Introduction to Patient Pathway Analysis 1 Why Patient Pathway Analysis? 1 What is a “Patient Pathway”? 2 What is Patient Pathway Analysis? 3 PART 2: The Building Blocks of PPA Study Design 5 Planning for PPA Outcomes 5 The Necessary Data 6 PART 3: Practical Examples of PPA Methods 13 Introduction 13 Module 1: How to Map the Patient Pathway 14 Module 2: How to Map and Profile Healthcare Providers and Services 21 Module 3: How to Assess Delays in Care Using PPA 30 Module 4: How to Assess Healthcare Costs Incurred Along the Patient Pathway 37 Module 5. How to Assess Patient Journeys Against Specified Clinical Pathways 45 Module 6. How to Integrate Geospatial Mapping into a PPA 54 Methods Spotlight: Carbon Accounting in PPA 62 PART 4: Reflecting on Experience in the Field 65 Conclusion 70 References 71 List of Figures Figure 1. Framework of the different considerations when profiling healthcare providers 6 Figure 2. The building blocks of PPA studies  7 Figure 3. Number of screened households and interviewed participants 18 Figure 4. Key touchpoints across the continuum of care for NCDs and MNH 19 Figure 5. Sankey diagram showing maternal care pathway across all three urban study areas 20 Figure 6. Key components of the methodology used in Case Study 1A and 1B 24 Figure 7. Concept map for analyzing alignment between care-seeking patterns and health services 25 Figure 8. Key components of the methodology used in Case Study 2 27 Figure 9. Key components of the methodology used in Case Study 3 35 Figure 10. Adapted WHO survey time points for longitudinal assessment of multiple treatment tracks for drug-resistant and drug-susceptible TB 40 Figure 11. Key components of methodology used in Case Study 4 41 Figure 12. Key components of methodology used in Case Study 5 42 Figure 13. Total cancer-related and total medical costs (Ksh and US$) 44 Figure 14. Itemized expenses by total medical cost 44 Figure 15. Standard steps in patient testing in Adams et al., 2014 49 Figure 16. Summary of study design for Case Study 1 with an integrated geospatial component 57 Figure 17. Origin-to-destination maps for MNH first visit in Dhaka and Chattogram, Bangladesh 58 Figure 18. Distance decay for different touchpoints across the maternal and newborn care continuum 59 Figure 19. Percentage of the population with spatial access to emergency care 60 Figure 20. Heat map showing travel time to the nearest public health facility for the dry (left) and wet (right) seasons 61 Figure 21. Care Delivery Components comprising a care pathway 63 Figure 22. Ghana: Evaluating the NoP effect on health centre use through PPA 67 iii List of Boxes Box 1. Sampling strategy for a PPA study to assess primary healthcare services in Gujarat, India  15 Box 2. Sankey diagrams as a tool of visualizing flow in a PPA  20 Box 3. Examining alignment between care-seeking behavior and service availability 25 Box 4. Examples of different definitions for delays in care from the literature 32 Box 5. Additional example on delays 36 Box 6. Definitions of cost terms 38 Box 7. Definition of clinical pathways from a systematic review 46 Box 8. Example of advanced analytics: Process mining 52 Box 9. Example of advanced analytics: Process mining and discrete event simulation 53 Box 10. A practical application: PPA as a methodology for evaluation of a national priority initiative in Ghana, the Networks of Practice  66 iv List of Tables Table 1. Brief descriptions of the utility of cascade analytics and patient pathway analysis ix Table 2. Distinctions between the use of the word “mapping” in this guide 3 Table 3. Common variables for demand-side data in PPA 9 Table 4. Examples of common variables of supply-side data 11 Table 5. Definitions for levels of care 12 Table 6. Data sources for Case Study 1 17 Table 7. Content of the provider profiling tool 22 Table 8. Examples of provider classification codes by provider type, facility type, health sector and health system level and how it relates to touchpoints in a PPA 23 Table 9. Summary of indicators calculated in Case Study 2  29 Table 10. Sample from questionnaire to assess delays  33 Table 11. Showing how data variables for delays in care were integrated in demand-side data collection 34 Table 12. Classification of health facilities based on the three-tier hospital system  35 Table 13. Steps involved in the current and proposed testing/treatment pathways 50 Table 14. Summary of completed PPA studies by the authors 65 v Acknowledgments The team would like to express its sincere gratitude to the many contributors to the implementation of the Patient Pathway Analytical (PPA) studies in Bangladesh, Ghana and Gujarat State featured in this guide. Contributors include, among others, healthcare providers and administrators in the public, NGO and private sectors, community leaders in the study sites, and research participants who are clients for critical maternal and neonatal health and NCD services. At the World Bank, these PPA studies were led by Zara Shubber, Nicole Fraser, Kojo Twum Nimako, Sanam Roder-DeWan and Michael Peters with technical support from Ahmad Hegazi, Neena Kapoor, Najmul Hussein, Muhammod Abdus Sabur, Katie L. McWilliams, Paul Ouma, Jessica Watson, Elina Pradhan, Navneet Kaur Manchanda, Guru Rajesh Jammy, Andrew Sunil Rajkumar, Kajali Goswami and Mengxiao Wang. Generous advice and strategic guidance were provided by Patrick Mullen, Iffat Mahmoud, Bushra Binte Alam, Atia Hossain, Rahul Pandey, Gil Shapira and Mickey Chopra. We thank our government partners for entrusting the PPA research to us and providing leadership, technical input and liaison with stakeholders. The collaboration with the Bangladesh Directorate General of Health Services, the Ghana Health Service, and in Gujarat, the Health and Family Welfare Department and the State Health System Resource Center is gratefully acknowledged. We also thank the respective Ethics Review Committees for authorizing data collection and providing research clearance. This practical guide is built on the experiences and lessons gained from these PPA studies conducted with partner institutions: In Bangladesh, we thank the team at icddr,b (Shehrin Shaila Mahmood, Khadija Islam Tisha, Fahim Tazware Himel, Zahid Hasan, Sabrina Rasheed, Rumayan Hasan, Gazi Golam Mehdi, Mohammad Abdus Selim, Orin Akter, Md Golam Rabbani, Mohammad Wahid Ahmed, Zerin Jannat, Nabila Mahmood, Kamrun Nahar and Halima Akter Prova) for implementation of the PPA in Dhaka North City Corporation, Dhaka South City Corporation and Chattogram City Corporation. In Ghana, we thank IQVIA (Almas Shamim, Olaoluwa Akinloluwa, Sekinat Amoo, Franklin Glozah, Eugene Kallson, Sushant Malhotra, Kwasi Torpey, Valentine Adaiwo, Hemant Chaudhry, Naim Hage, Chijioke Kaduru and team), the University of Ghana School of Public Health (Kwasi Torpey, Franklin Glozah and team) and NTT-Data (Ramon Vila, Gonzalo Llende, Fanny Fourestier, Liliana Ramalho Inacio and team) for implementing the PPA in Hohoe, Dormaa Central, Ketu North, Tain, Ayawaso Central and Atwima Nwabiagya districts. In Gujarat, we thank the team at Sambodhi (Shubham Gupta, Rakesh Parashar, Aayushi Rastogi, Piyush Kumar, Shipra Prakash, Debrupa Bhattacharjee, Sumaira Khan, Kultar Singh and team) for implementing the PPA in Banaskantha, Dahod, Rajkot and Vadodara districts. This guide was prepared under the technical guidance and supervision of Nicole Fraser and Zara Shubber. The guide also benefitted from initial work by Sebastian Christ and the technical and editorial reviews by Katherine Ward and Valerie Scott at the World Bank. We also gratefully acknowledge the financial support provided for the development of this guide by the Gates Foundation, as well as for the financial support provided for the conduct of the studies included in the guide by the Gates Foundation and Access Accelerated. vi Acronyms BTHFT The Bradford Teaching Hospital LMICs Low- And Middle Income Foundation Trust Countries CKD Chronic Kidney Disease LwPS Living With Pain Service CoC Continuity Of Care MDR-TB Multi-Drug Resistant Tuberculosis CP Clinical Pathway MDT Multidisciplinary Team CPT Current Procedural Terminology MS Multiple Sclerosis CPW Clinical Pathway NAAT Nucleic Acid Amplification Test DM Diabetes Mellitus NHFPC National Health And Family EF Emission Factor Planning Commission EHR Electronic Health Record NHIF National Health Insurance Fund FGD Focus Group Discussion NHIRD National Health Insurance GFR Glomerular Filtration Rate Research Database GHG Greenhouse gas NHS National Health Service GIS Geographic Information System NITI The National Institution For GPS Global Positioning System Transforming India HbA1c Hemoglobin A1c NoP Network Of Practice HMIS Health Management Information PHC Primary Health Care Systems POC Point Of Care HR Hazard Ratio POCT Point of Care Test HSD Health System Delay PPA Patient Pathway Analytics HTN Hypertension RCT Rotator Cuff Tear IASP International Association For The SEP Service Evaluation Project Study Of Pain SPSS Statistic Package For Social ICCC WHO’s Innovative Care For Sciences Chronic Conditions SSA State Sequence Analysis ICD-9-CM International Classification Of TAT Turnaround Time Diseases, Ninth Revision, Clinical TB Tuberculosis Modification TLH Total Laparoscopic Hysterectomy IDI In-Depth Interview UHC Universal Health Coverage IPPA Individual Patient Pathway UK United Kingdom Analysis UNAIDS United Nations Programme On KII Key Informant Interview HIV/AIDS LDD Long Diagnostic Delay WHO World Health Organization vii Foreword The Sustainable Development Goals (SDGs) set a global commitment to achieve universal health coverage (UHC) by 2030. UHC means that all people receive the promotive, preventive, curative, rehabilitative, and palliative health services they need, with sufficient quality, and without financial hardship (WHO, 2018). Primary health care (PHC) is a necessary foundation for the delivery of sustainable high-quality universal health care (WHO & UNICEF, 2018). Better analytics are needed to understand where and how we can improve obtaining these goals. Achieving UHC becomes especially challenging in the face of an aging population with an increasingly complex burden of chronic diseases. Health systems were primarily designed and deployed to manage acute disease and injury, but the disease burden has changed worldwide. This is especially true in low- and middle-income countries (LMICs), where the epidemiological transition of decreasing fertility, increasing life expectancy, and changing morbidity and mortality patterns has been rapid, and the burden of disease outweighs available resources. Chronic and complex diseases, including infectious diseases requiring continuity of care and multiple touch points, place new demands on health system functions and on the efforts to achieve UHC (Knaul et al., 2015). Addressing the needs of the increasingly complex patient population calls for a multi- sectoral approach that puts the patient first (WHO & UNICEF, 2018). Today’s health systems must be patient-centered, holistic, and fit for shifting demographics and changing disease burdens. Health system redesign requires systematic changes to improve the quality, efficiency, and effectiveness of patient care. Implementation cascade analyses point to significant gaps when it comes to screening and diagnosis, as well as to long-term adherence to chronic care medication, pointing to failures in primary, coordinated and continuous care. To better understand why, where, and how this is happening, we can look to those that know it best – the clients at the receiving end of these healthcare services. In LMICs, constrained resource settings and weak systems are often accompanied by unique healthcare-seeking patterns and preferences for the demographics they serve. Clients of the health system—meaning anyone seeking a service from the health system, whether an ill patient or a healthy individual—must be at the center of the changes and improvements being made to health care provision. This means health systems can greatly benefit from understanding clients’ journey in seeking care, their preferences, and the barriers to effective and efficient service delivery. System redesign will require innovative changes to improve service delivery, access, coverage, and quality across the continuum of care. By recognizing that the patient’s care-seeking journey is determined by multiple factors, we can guide the transition of the health system to be more client-oriented and focused on health rather than disease. Understanding the patient pathway can help produce better results across the continuum: increasing access and quality of care, reducing inefficiencies and costs, and making the healthcare experience more personalized for clients. In 2020, we published a practical guide titled Using Care Cascades Analytics (Volume 1) (Fraser- Hurt et al., 2020) to promote this type of decision-support analytics and provide technical guidance viii Foreword and tools for cascade analyses. The analytical framework of the care cascades is an easy-to-apply method to quantify service coverage and breakpoints in care continuity and investigate their causes. The cascade framework has been used widely in health programs and has helped to identify strategies to improve health outcomes in diverse populations and settings. This volume, Using Patient Pathway Analytics for Person-Centered Health System Assessments, is a complement to Volume 1 and is meant to introduce research methods that can help you gain a detailed understanding of the patient’s healthcare journey. This guide, as a companion to the Cascades Analytics Guide, is equally aimed at program implementers, researchers, decision-makers, and other professionals interested in analyzing and improving health service access. However, to encompass the many ways patient pathway analysis (PPA) can be approached, it is broken down by practical questions users may have to undertake the analysis. Each module focuses on specific aspects of patients’ care-seeking journeys and provides guidance on how to generate knowledge which can inform healthcare redesign. Health systems redesign requires the use of multiple viewpoints of the health system and patient experience. The combined guides (Volume 1 and 2) offer multiple approaches to analysis and can be used in conjunction to assess outcomes in key service delivery programs while also capturing the where, how, when and why of care-seeking, as well as the landscape of healthcare providers. While Care Cascades Analytics provides an overarching assessment of performance across the continuum of service delivery, Patient Pathway Analysis allows for a patient-focused approach (Figure 1). By recognizing the pathway patients take, we can improve efficiency, thereby reducing costs, increasing access and quality of care, and making the healthcare experience more personalized for patients. Using both together allows for a strong analysis of both sides of the health system (supply and demand), with the goal of actionable improvement to health outcomes. TABLE 1. BRIEF DESCRIPTIONS OF THE UTILITY OF CASCADE ANALYTICS AND PATIENT PATHWAY ANALYSIS Cascade analytics Patient pathway analysis ■ Captures the summative advancement of clients ■ Captures how clients move through the stages through different stages of care (screening, of care within the healthcare system (or a part of diagnosis, treatment initiation, treatment it), characterizing individual patient flow through maintenance and long-term management). different touchpoints. ■ Focuses on care outcomes of a program ■ Focuses on the care journey and its determinants ■ Advancement to a specific ‘care status’ is ■ Advancement of the journey is captured as a translated into a simple visual showing effective process of care-seeking, akin to clients’ flow coverage (the bar) and attrition (the gap) across through the health system, with more details on the service chain. how and where care is sought, including access ■ Quantifies service coverage and losses across to services, provider preferences, delays in care, the service chain in different sub-populations and costs incurred, and potentially spatial information provides information on barriers and facilitators. on care providers and their clients. ■ Quantifies and describes multiple dimensions of care-seeking behaviors, the healthcare demand and supply interaction, and client experiences, based on the research objectives. ix What is in This Guide In this guide we propose Patient Pathway Analysis (PPA) as an innovative approach for field research to uncover how patients navigate the system and ultimately help care systems to better meet client needs. The guide provides an overview of how patient pathway studies can be designed and implemented to meet this goal. The studies featured in the guide offer a wealth of insights and lessons from research carried out in different health systems. By examining the design options of different study types, implementers and decision-makers can incorporate valuable knowledge into their own assessments. The selected studies encompass different world regions, target populations, health programs and implementation settings, underscoring the importance of tailoring research approaches to the specific contexts in which they are implemented. Throughout the process of compiling the guide, the writing team was astonished by the wealth of available research studies on care-seeking journeys and people’s decision-making on provider use. Drawing on the available literature and implementation experience, the guide highlights both design issues and practical aspects of conducting patient pathway analytics. While every researcher will be faced with a new set of questions and local health system characteristics, there are some common study features and generalizable lessons. Through this guide, the reader can learn about different options ranging from qualitative behavioural studies by social scientists to large quantitative assessments, potentially using large routine databases which require data scientist skills. The guide aims is to present study methods, tools, resources, and carefully selected examples of PPA methods to assist with study design. It considers that needs vary based on health system, target populations, disease focus, and policy question. This guide provides the ingredients (resources) and some recipes (methods) from our literature search. However, there is room for adaptation, innovation and expansion of the PPA methods to answer the research questions of a particular assessment. Using Patient Pathway Analytics presents a wide range of tools used for PPAs, with varied methodologies and therefore varying skill sets and competencies needed for successful implementation. A researcher can determine more effective and efficient pathways for clients to take within the local, regional, or national health care system. One can answer one research question within this guide or many simultaneously. The guide has four parts as follows: ■ Part 1 introduces the broad spectrum of definitions and uses of PPA. It aims to familiarize readers with PPA methods and theoretical considerations ■ Part 2 introduces the basic building blocks of a PPA study, including the needed data, data collection tools, and some basic concepts that will help you understand the methods presented in this guide, as well as a general framework for PPA study design ■ Part 3 showcases methods and analytical approaches used to answer research questions, including: ☐ Module 1: How to map the patient journey (the basic PPA) ☐ Module 2: How to map and profile healthcare providers and services x What is in This Guide ☐ Module 3: How to assess delays in care-seeking ☐ Module 4: How to assess costs incurred along the patient pathway ☐ Module 5: How to assess patient journeys against specified clinical pathways ☐ Module 6: How to integrate geospatial mapping into a PPA ■ Part 4 presents summary learnings and considerations from the author’s own implementation of PPAs in the field Case Study Title Case Study 1A A patient pathway analysis to understand the experiences and preferences of low-income, urban residents in accessing NCD and MNH-related PHC services in Bangladesh Case Study 1B Provider profiling and mapping as part of a PPA on NCD and MNH care-seeking by urban low-income clients in Bangladesh Case Study 2 Using secondary data to analyze the alignment between TB-related care-seeking and services in Pakistan Case Study 3 A multi-center PPA study to examine delays in diagnosis of tuberculosis, in China Case Study 4 Using surveys to determine patient costs over time for tuberculosis patients in Africa (South Africa, Mozambique, Tanzania, and The Gambia) Case Study 5 A case study approach using focus group discussions, in-depth interviews, and patient-reported cost information to explore the economic and social consequences of cancer in Kenya Case Study 6 Developing a new clinical pathway, comparing costs and provider time Case Study 1C Use of the Euclidean Distance method to enhance a PPA in urban Bangladesh Case Study 7 Using network analysis to define geographic accessibility to health care in Namibia and Haiti Case Study 8 Accounting for the effect of rainfall on travel routes in Uganda Case Study 9 Carbon accounting in PPA – Exploring the climate cost of care pathways in Bangladesh xi PART 1: Introduction to Patient Pathway Analysis WHY PATIENT PATHWAY ANALYSIS? Every patient embarks on their own individual journey when they decide to “Patient flow – ensuring that seek health care for a specific reason. The collective journeys of different patients receive the care they individuals represent the flow of a patient population through a health need, when and where they system. Patient flow is influenced by patient characteristics, preferences, need it – is one of the and health system factors. Patient flow problems may reflect a fundamental misalignment between a health system’s capacity and population needs, greatest challenges facing requiring fundamental system redesign and attention to both the supply healthcare today. Stagnant and demand sides of the health system (Kriendler 2021). However, efforts to flow has myriad destructive improve patient flow often focus on specific, localized interventions rather consequences: for patients, than addressing system-level causes. delayed care, with protracted PPA evaluates patient flow through various touchpoints in the health care suffering, anxiety and risk; for system. Taking a comprehensive health systems approach, PPA studies providers, stress, overload, analyze patient trajectories of care, beginning with the patient’s initial care- and burnout; for the health seeking, on to diagnosis and treatment, and if needed, long-term care for system, reduced quality and chronic illness. Evidence generated by PPA studies can then be used to sustainability.” inform service delivery redesign, deciding where, when, and how healthcare services should be provided. For example, care coordination and referral — S. A. KREINDLER, 2017 between healthcare professionals is a major challenge in the patient pathway. Disease management is often a continuous process with multiple touchpoints, sometimes requiring a multidisciplinary approach and coordinated care among different facilities, providers, levels, and locations of care. Patients may be delayed or lost to follow up along pathways to care. PPA can shed light on where and why this happens, dig deeper into patient experiences of care, and reveal bottlenecks in the health care system that may worsen quality of care and outcomes and drive costs. PPA can contribute to service delivery redesign goals by: ■ Comparing available health services with client demand ■ Assessing variation in patterns of care-seeking and service delivery across different patient groups ■ Identifying major barriers to care-seeking and service delivery ■ Exploring possible determinants of health outcomes ■ Generating evidence to inform interventions to improve quality and outcomes 1   USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Evidence generated through PPA is invaluable to inform policy and decision-making aimed at improving health care systems and may lead to revising existing service delivery models or implementing a new ones (Kriendler 2021). By addressing barriers to care and building on existing strengths, service delivery redesign efforts informed by PPA can optimize the flow of patients through a health system, aiming to improve efficiency, quality of care, and outcomes, while reducing costs. WHAT IS A “PATIENT PATHWAY”? As mentioned above, client in this guide refers to anyone seeking a service from the health system, whether an ill patient or a healthy individual. For example, a client may be a woman seeking antenatal care. The term patient typically carries a connotation of someone who may be ill and is seeking health care. We will use these two terms throughout. In this guide, the two terms are used interchangeably for patient (client) pathway. “Patient pathway” refers to a series of patient contacts with the health system along the continuum of care, from initial care-seeking to diagnosis, treatment, and potentially ongoing maintenance of care. Patient pathways may involve visits to different health care facilities or providers or could involve multiple visits with the same health care provider for evaluation and treatment. Further, patient pathways may involve significant coordination and communication across a range of health care providers, including physicians, clinical officers, nurses, laboratory technicians, and pharmacists, among others, either within or across facilities. Electronic medical records and digital mobile health applications can be useful for capturing patient demographic and clinical information and tracking patient visits across the health care system to support effective follow-up care. “Clinical pathways” tends to refer to clinical processes of care—evidence-based, best practices that health professionals aim to follow (Rotter et al., 2019). Clinical pathways aim to standardize care (for a specific problem, procedure, or health episode) based on evidence to improve health outcomes, costs, and efficiency (Rotter et  al., 2019). Existing recommended clinical pathways within a health system relate to patient pathways, as they will influence how patients are directed through the health system. For example, a patient with symptoms of hypertension will decide to seek care and then be guided by their provider at this initial point of care through the recommended clinical pathway of evaluation, diagnosis, and treatment. In some cases, the patient pathway may align closely with recommended clinical pathways—in others, it may deviate, due to any number of factors, including patient behaviors and preferences and health system bottlenecks and breakdowns. For the purposes of this guide, PPA refers to a set of methods to analyze patient pathways—their journey through the health care system. 2 PART 1: Introduction to Patient Pathway Analysis WHAT IS PATIENT PATHWAY ANALYSIS? PPA should include clear and defined Put simply, PPA is a patient-centered method of field touchpoints: research that describes and analyzes how patients Touchpoints refer to key interactions between the interact with the health system at various client/patient and the healthcare system along the “touchpoints,” or points of contact with the health continuum of care. Common examples of system, at a population level. Various study designs touchpoints in PPA are the initial care visit, and methods are used as part of PPA to describe and diagnosis, treatment initiation, referrals for analyze patient pathways, including qualitative and complications or specialty care, and ongoing quantitative data from primary and secondary data treatment maintenance. Depending on the study collection. Combining quantitative data with qualitative scope and objective, utilization of prevention data reflecting patient perspectives and experiences is services (e.g., mammography, community-based ideal to develop a more nuanced understanding of screening) or emergency visits for acute illness episodes may also be included as touchpoints on patient pathways and how and why they unfold as they the patient pathway. do in practice. DEFINING TERMS Table 2 reviews several “mapping” terms and how they are defined in this guide. PPA always involves mapping of the patient journey and may also involve process mapping and geospatial mapping. Notably, patient pathways involve dynamic and complex processes that may be influenced by individual, social, and health system factors. Undertaking a PPA can shed light on these complexities and reveal how patients navigate health systems in practice. TABLE 2. DISTINCTIONS BETWEEN THE USE OF THE WORD “MAPPING” IN THIS GUIDE Mapping of This refers to identifying the major touchpoints of a patient’s journey throughout the patient the health system and gathering information on individual steps taken to go journey through these touchpoints, including redundancies and delays. This does not involve geographical mapping, but rather patient movement through a health system. Process This refers to the identification of different steps of a process taken to achieve a mapping particular aim. It is applicable to a wide range of fields and is primarily intended to detail a process from start to finish and identify where improvements in efficiency can be made. Geospatial This refers to the use of location data (precise GPS locations or approximate mapping locations based on landmarks or regions) and mapping software tools. This may involve using information on road networks and travel routes to establish distances that patients travel to access health care. Geospatial data can be used to create maps that may shed light on both available care and patient demand for care. 3 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS In the following sections of this guide, we will illustrate the process of conducting a PPA. Part 2 of this guide covers the “building blocks” of data that are needed. Part 3 reviews the various methods that can be used to gather these types of data. Part  4 will present summary learnings and considerations from the authors’ own work implementing PPAs, followed by the final conclusions section. 4 PART 2: The Building Blocks of PPA Study Design PLANNING FOR PPA OUTCOMES PPA study designs can vary widely depending on the outcome of interest and the feasibility of the methods for the researchers. In general, PPA studies share a lot of elements and components (or building blocks) that this chapter illustrates. The chapter brings together material from primary work carried out by the World Bank and partners, as well as a review of the current literature on PPA studies. The guide aims to support the overall design and conceptualization steps. It provides information on key data variables, data collection, analysis, and the insights that can be gained from the chosen approach. The data variables discussed in the guide reflect the study designs and research questions covered in the modules in Part 3 addressing issues around patient pathway mapping, care provider profiles, delays in care, cost, and geospatial studies. At the planning stage of a PPA, much of the focus should be on the study objectives and what types of information and data are required to address them. A researcher will have to consider what the appropriate (and feasible) data sources are to get the desired information. While primary data collection or data acquisition from secondary sources can differ based on the context of each study, the data variables may be the same or similar. It is important to first consider what kind of data one has access to and is available either through public datasets or a research partner. The main advantages of newly collected, primary data are that they are recent and designed to answer specific PPA research questions. Figure 1 illustrates the different elements and considerations that influence the methodology and study design of PPA. Figure 2 illustrates commonly gained insights from different PPA designs. Each module of this guide is focused on answering a specific type of research question. At the core of a PPA is the mapping of patient pathways and care-seeking behavior (Module 1). This type of data can be combined with provider mapping and profiling (Module 2) for further insights on the alignment between available and preferred care. PPA study objectives can also be broadened to assess delays in care (Module 3) or cost of care (Module 4). The components may be important to inform decision-making on intervention design or service delivery improvements. Figure 2 is a conceptual framework that connects the building blocks of PPA studies. It is intended to be a guide for the conceptualization of the methodology. It is not a prescriptive framework to follow exactly, but highlights the main components to consider. This framework stems from World Bank’s experience on PPA study design and a review of relevant literature on how PPA studies have been previously designed and the types of results these studies yielded. It is intended to 5 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS FIGURE 1. FRAMEWORK OF THE DIFFERENT CONSIDERATIONS WHEN PROFILING HEALTHCARE PROVIDERS Public By health system sector Private formal Private informal Level 1 By health care level Level 2 Level 3 Initial contact PPA Framework Diagnosis By care stage Medication Monitoring Maternal By condition/disease NCD Preventative/routine By type of contact Illness Acute emergency Source: Authors’ own. summarize in an easy-to-read way the make-up of diverse PPA studies while acknowledging that this guide is not exhaustive; PPA methods are evolving, especially due to the growing potential for advanced analytics of big data from electronic medical records and insurance claims. THE NECESSARY DATA Mapping of the patient pathway can include quantitative and/or qualitative data collection on the different touchpoints on the patient journey. Quantitative data on patient pathways provides the information needed to construct the flow that represents where people go and how they switch between providers (or don’t). Qualitative data can add important depth and context into why and how patients make decisions across the care-seeking pathway. For example, a survey can help uncover where people enter care for a specific disease, whether that is a PHC facility, a pharmacy, or an informal care provider, and the main consideration underlying provider choice. Qualitative research can add nuance and understanding on the patient’s decision-making process, perceptions of provider choices, and experiences of care. Qualitative data collection methods, such as focus group discussions (FGDs), in-depth interviews (IDIs), and/or key informant interviews (KIIs), 6 PART 2: The Building Blocks of PPA Study Design FIGURE 2. THE BUILDING BLOCKS OF PPA STUDIES Data category Data collection Analysis Insights & sources Do you want to assess Socio-demographic Questionnaires & Descriptive Patient pathway patient care-seeking characteristics surveys and with quantified behavior? statistical flow between Registries & analysis on touchpoints national/regional patient databases Mapping the Basic care-seeking flow patient-pathway data Quantifying Quantitative delays breakdown of delays observed in the time intervals between Do you want to assess Assessing Timing of touchpoints of delays between delays in major interest touchpoints of the care touchpoints If yes, add care-seeking journey? Costing Quantified actual relevant analysis costs incurred by variables to Do you want to assess Assessing Direct/ patients along the the data costs associated with cost of indirect care-seeking collection the care-seeking care- costs pathway tools journey? seeking Euclidean Geographic Do you want to Geospatial Location data distance accessibility- include an analysis of analysis & additional origin-destination the movements and geospatial maps - facility travel for care? determinants bypassing Network analysis Do you want to assess Available providers by Phone calls - site Descriptive Distribution of alignment of care- type, sector and level visits - surveys & analysis of providers and seeking with availability questionnaires provider availability of of care services? data services Registries & national/regional Provider mapping and Type and availability databases profiling of health services Qualitative Context and IDIs, Thematic Contextualization of care-seeking behavior & data experience of KIIs & analysis factors influencing patient choices care-seeking FGDs can add important context and explanatory power to any part of a PPA study, including care- seeking behavior, facility services, and client satisfaction. Regardless of the exact study design used to assess the patient pathway, the data needed to answer the research questions is important to consider. Data collection is a labor-intensive process and requires meticulous attention to detail, so it is critical to define the required data at the outset. This guide aims to support the conceptualization of PPA study designs. 7 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS There are two broad categories of data to collect for a PPA: demand-side data and supply- side data. Demand-side data: Throughout this guide, demand-side data will be used to refer to the information that represents the experience and perspective of a patient or a client of the health system as they seek their needed care and services from the system’s providers and facilities. Demand-side data can be sub-grouped as follows: (Table 3) 1. Socio-demographic characteristics 2. Basic care-seeking behavior 3. Additional details on care-seeking behavior: a. Timing of major touchpoints reached b. Cost of care-seeking c. Locations and movements Socio-demographic characteristics: This category explains who the clients are and their communities. This data can be individualized and/or grouped into particular populations of interest to define and describe those whose care- seeking behavior the PPA is intended to reflect, and consequently, who the resulting policy and programmatic recommendations are intended to benefit. Basic care-seeking behavior: The most basic care-seeking data for a PPA explains how the client interacts with the health system from their own perspective and experience. This sub-group includes the most important component of data to conduct a PPA. This is the foundation of assessing the care-seeking behavior of clients participating in the PPA study, and these data can be analyzed on their own or in combination with the supply side data. Additional details on care-seeking behavior: Basic care-seeking data can also be augmented by the following three sub-categories of data: ■ Timing of major touchpoints reached: This category refers to the timing of main events (touchpoints) that will be examined as part of the pathway. This will include the timing each major touchpoint was reached and calculating the duration between the touchpoints (for example, the duration between initial care-seeking and diagnosis or treatment). This is useful for assessing delays along the pathway and time-to-care. ■ Cost of care-seeking: This category of data reflects the out-of-pocket costs incurred by the clients during the examined journey in the healthcare system. This data can also be used to highlight the total financial burden associated with diagnosis or treatment, complications, delays, and/or incorrect referrals, depending on the research question. As this data reflects the patients’ out-of-pocket costs, it needs to be collected from the patients (clients) themselves, either through primary data collection or from available secondary data (for example, prevalence surveys with patient out-of-pocket spending data). 8 PART 2: The Building Blocks of PPA Study Design TABLE 3. COMMON VARIABLES FOR DEMAND-SIDE DATA IN PPA Data category Data variables Socio-demographic ■ Age characteristics ■ Gender ■ Occupation/employment ■ Marital status ■ Highest educational level achieved ■ Monthly household income per capita ■ Relevant risk behaviors (e.g., smoking history, alcohol use) Basic care-seeking ■ Reason for seeking care (symptoms, condition, or preventative care) behavior ■ Where the first care visit happened ■ When and where selected subsequent care visits happened ■ Where clients were referred to ■ Which providers were consulted and why ■ When diagnostic tests were initiated ■ When final diagnosis was reached ■ When treatment was initiated ■ How and where treatment was maintained (medicines, check-ups) ■ Preferences & drivers of choice Timing of major ■ Date symptoms started Additional categories for care-seeking data: touchpoints reached ■ Date of first care visit ■ Date of diagnosis ■ Date treatment was initiated ■ Date of any referrals Costs incurred ■ Direct medical costs (e.g., cost of service, such as consultant across the journey fees, laboratory tests, medications, imaging) ■ Direct non-medical costs (e.g., cost of transportation, food, accommodation) ■ Indirect costs (e.g., missed school or work and/or productivity losses by patients or caregivers, travel time) Locations and ■ Locations of residences (GPS or approximate locations) movements relevant ■ Locations of reported providers (GPS or approximate locations) to touchpoints ■ Travel duration ■ Mode of transport ■ Locations and movements: This category of data refers to the exact or approximate geographic locations of the client’s place of residence and providers that were consulted and mentioned as part of the assessment. This data can include GPS locations of where people live, the nearest public establishment or landmark, and the location of providers used by the client. It also includes the reported travel duration and mode of transport by clients to access care providers. Depending on the type of intended analysis, the data can contain simple origin and destination locations combined with travel duration, or additional data such as exact routes, 9 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS modes of transport, and GPS coordinates. This can be used for visualizing the pathway using maps and cartograms and will be explored in more detail in the geospatial mapping chapter. Depending on the condition or disease of interest, data on care-seeking behavior can be either general or disease-specific. The respective data sources can differ; for example, in the case of tuberculosis (TB), prevalence surveys have data that is more specific to the condition, while sources like healthcare utilization surveys and Demographic and Health Surveys can reflect more general care-seeking habits. Disease-specific data can also be obtained from disease registries, medical records, and insurance claims. Whenever possible, linking and combining different databases can be considered to enrich the available dataset and triangulate for more robust results. Supply-side data: Throughout this guide, “supply-side data” will be used to refer to information that describes the health care system, providers of care (across both formal and informal sectors and at all levels of the system), and health care services. In other words, this data characterizes the supply of healthcare. Supply-side data can be sub-grouped as follows: 1. Basic provider profiling 2. Health services available (per condition of interest) 3. Locations of providers 4. Cost of delivering care Patient pathways should ideally be analyzed in the context of the provider landscape to better understand the demand-supply dynamics. To examine service coverage and utilization, the PPA must first generate a census of the available providers in the region of interest, identify the health services they provide, and classify providers according to their sector and level. Table 4 illustrates common data variables for each sub-group. Basic provider profiling: This category refers to data that describes who the health care providers are within the health system. This includes provider name, provider type (hospital, clinic, village healthcare worker, etc.), provider sector (public, private, NGO), and level of care provided (preventative care, community level care, primary care, etc.). This data may be collected separately from demand-side data and describes the landscape of healthcare in the study region. In some instances, this data may already be available through existing censuses or databases of providers. In other instances, additional data collection, such as provider mapping surveys on site may be required. Health services available (per condition of interest): This category describes the type of health services that are related to the health condition of interest. For example, a PPA for an infectious disease like TB would require data on availability of diagnostic tests and medications. A PPA on maternal and neonatal health (MNH) care may require data on the availability of emergency obstetric and newborn care services at facilities that manage deliveries. For non-communicable diseases (NCDs), required data might be on preventative care services, such as screening, as well on diagnosis and treatment. 10 PART 2: The Building Blocks of PPA Study Design TABLE 4. EXAMPLES OF COMMON VARIABLES OF SUPPLY-SIDE DATA Data category Data variables Basic provider profiling ■ Name and type of facility/provider ■ Level of the facility or provider (community, primary, secondary, tertiary) ■ Sector of provider or facility (public, private, NGO) ■ Type and number of healthcare providers in a facility Health services ■ Type of patients treated (screening for conditions of interest) (per condition of ■ Type of services provided by the facility/provider interest) ■ Presence of a referral system ■ Availability of basic equipment to assess conditions of interest ■ Presence of specific tests (laboratory and radiology; e.g., rapid diabetes test kit, functional nebulizers etc.) Locations of providers ■ Location of provider/facility (GPS or approximate location) Cost of delivering care ■ Direct costs (e.g., cost of provider time, medications, labs) ■ Indirect costs (e.g., overhead, administrative costs, cost of split staffing time, such as nursing for more than one patient) This data is combined with the provider profile data to examine the availability of specific services in the study area. In the following chapter the guide will discuss how this can be further combined with demand-side data to analyze access to care, utilization, and alignment between care-seeking behavior and service availability. Locations of providers: This category refers to the geographic locations of care providers in the study area. This information is used to match provider location with client location information within the study area. The analytical possibilities—ranging from calculation of origin-to-destination distances to more complex analyses using transportation, road networks, time and distance of travel routes—are explored in more detail in the geospatial mapping chapter. Cost of delivering care: Costs incurred by the health system (the supply side) can also be measured. The examples used in this guide focus on costs in health service delivery facilities (clinics/hospitals) using clinical pathways. Primary and secondary data sources can be used for provider profiling; in some instances, a national database may exist with information on providers by sector with the healthcare services of interest. However, it is also possible to collect primary data on providers, which offers the advantage of bypassing potential gaps or outdated information in national-level databases and 11 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS allowing collection of particular datapoints of interest, such as geographic location and operating hours. This exercise should be done in collaboration with local government, or the relevant health authorities, avoid duplication of efforts and to benefit the relevant stakeholders if there is interest in updating the available information. Grouping providers according to sector is important, especially if the study aims to inform health system strengthening efforts. Learning more about available health services and where and why people seek care can shape investments in the health system to serve people better. Public sector providers belong to the government’s health care system, which usually has a hierarchy of care facilities with assigned roles. Private sector providers can be classified into formal and informal categories. Formal providers include hospitals and clinics where care is provided by trained medical professionals such as doctors, nurses, and midwives, while the informal sector may include drug vendors, pharmacies, and traditional healers, among others. Additionally, the NGO sector providing relevant care should be included in the research. Providers should also be categorized according to the level of care they provide. Below is an example of levels of care used in Pakistan (Hanson et al., 2017). Combining supply-side data on available providers and services in a study area with demand-side data from patients on the providers and services they have used allows for a richer picture of patient pathways to emerge. Part 3 of this module will explore each of the building blocks of PPA studies in more detail and present case studies to illustrate the diversity of PPA designs. TABLE 5. DEFINITIONS FOR LEVELS OF CARE (HANSON ET AL., 2017) Level 0 (L0) Refers to basic, community-based care. L0 services include basic triage, provision of health information, and essential prevention and care services. Examples include community health workers and traditional healers, mobile clinics, and voluntary counseling and testing service facilities. Level 1 (L1) Refers to primary healthcare. L1 services are commonly provided on an outpatient basis by nurses, midwives, or private physicians. Some basic diagnostic services, including basic microscopy and essential medicines may be available. Examples include public health centers and private clinics. Level 2 (L2) Refers to primary healthcare, as well as more-advanced care. L2 facilities commonly have more extensive diagnostic and treatment options and can provide both outpatient and inpatient care. Examples include public district hospitals and rural or nongovernmental organization–affiliated private hospitals. Level 3 (L3) Refers to specialized care with large inpatient capacity. L3 facilities provide access to specialized physicians and have more sophisticated diagnostic and treatment capabilities. Examples include public referral or teaching hospitals and urban private hospitals. 12 PART 3: Practical Examples of PPA Methods INTRODUCTION This part of the guide uses the building blocks discussed in Part  2 to show how PPA study components were integrated in practical examples of studies from the authors’ primary work and other relevant literature. Part 3 is divided into six modules: 1. Module 1: How to Map the Patient Pathway (basic PPA) 2. Module 2: How to Map and Profile Healthcare Providers and Services 3. Module 3: How to Assess Delays in Care-Seeking 4. Module 4: How to Assess Costs Incurred Along the Patient Pathway 5. Module 5: How to Assess Patient Journeys Against Specified Clinical Pathways 6. Module 6: How to Integrate Geospatial Mapping into a PPA Module 1 is considered the most basic approach to PPA. Module 5 covers the specific case of PPA studies in the context of defined clinical pathways in specific care delivery situations. Module 2 is a great compliment to Module 1, and hence, the reader can consider Modules 1 and 2 as the core of the PPA methodology presented in this guide, based on the primary work of the authors. The remaining modules can be considered extensions/additions in order to acquire a richer PPA dataset and produce a higher yield of insights, if there is interest in the research questions they address. However, across all modules, a necessary step for any PPA is the determination of an appropriate sampling approach. SAMPLING FOR PPA STUDIES Sampling of a PPA differs depending on the population of interest and the condition of interest. Collecting information on care-seeking patterns in urban slums will differ from villages or rural areas. There is an argument that can be made for the geographic focus as well: one can design a study examining client journeys within a single state or municipality, which allows for a deeper understanding of care-seeking within a well-circumscribed area and reflects the reality of the health system in that area. However, the result is only locally applicable and cannot be generalized to other locations within the same country or a different country. Another approach is to sample more widely and strive to have a more inclusive sample, which can serve the goal of understanding care-seeking behavior more broadly in a country or across multiple subnational regions. The design can include rural and urban areas, and within each 13 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS category sampling units can be selected based on economic or demographic characteristics, or reasonable proxies. In practical terms, the sample needs to consider the method of data collection; the sample size for a study extracting care-seeking info from a huge database may be much larger than a study that uses household surveys to collect care-seeking information because of the time-consuming nature and higher cost of collecting primary data. Ultimately, choosing a probability vs a non-probability sampling method depends on the research question, as well as the intended generalizability of the findings. However, even if a statistical analysis for probability is not relevant to the study design, the sample should be large enough to show enough differences in pathways if quantitative data collection is a main part of the study. Box 1 is a practical example from the primary work of the authors describing the sampling strategy for a PPA in the state of Gujarat, India. This example illustrates an example of sampling for a PPA study while considering the administrative structure of a district. MODULE 1: HOW TO MAP THE PATIENT PATHWAY In this part of the guide, we showcase how to combine information from patients (demand-side) MODULE 1 – AT A GLANCE with information from the health system (supply- ⊲ The value of primary data side). The result of combining this information and collection for PPA when feasible analyzing it together is a detailed pathway that ⊲ Multiple diseases of interest can explains what happens to people from the moment be combined in the same overall they choose to commence seeking healthcare for a study design certain condition until they reach the last touchpoint ⊲ Practical examples showcasing in the examined pathway. defining and examining Characterizing and contextualizing care-seeking touchpoints of care ⊲ Secondary data can either be the behavior—both where and when, across the only source or a supplementary continuum of care—is central to a PPA. Data source of data variables for a PPA collection strategies should be tailored to the ⊲ Visualization of PPA results is a research question and may draw on both primary valuable tool for communicating and secondary data sources. Literature reviews and results secondary data sources such as rigorous health management information systems (HMIS), population- based surveys, and other programmatic data collection efforts are often a good place to start becoming familiar with what is already known. Benefits of secondary data analysis include low cost, quicker time to analysis, and, depending on the data source, access to rigorous data with large and representative sample sizes that would have been too expensive or time-consuming to collect. Disadvantages to relying on secondary data sources can include delays in permissions or access and less control or insight into quality. Available secondary data sources may also not be sufficient to answer your research questions around patient care-seeking, and in this case, additional primary data collection should be considered. It is also important to consider what data sources could be triangulated or linked to help answer the key research questions. 14 PART 3: Practical Examples of PPA Methods BOX 1. SAMPLING STRATEGY FOR A PPA STUDY TO ASSESS PRIMARY HEALTHCARE SERVICES IN GUJARAT, INDIA (WORLD BANK – SAMBHODI) This study was conducted across four districts in Gujarat Three villages/towns nearest to the facility, three villages/ (Banas Kantha, Dahod, Rajkot and Vadodara Municipal towns farthest away from the facility and four villages/ Corporation), and the design was intended to account for towns at a median distance would be considered. the differences between care-seeking in urban and rural settings. The study utilized a household survey to collect Step 3: Selection of secondary sampling units information on care-seeking journeys for the conditions ■ Rural: In rural areas, a comprehensive list of of interest in the selected districts. Accredited Social Health Activists (ASHAs) could In order to reach the number of households agreed help us identify population clusters of 1,000 persons on between the Gujarati government and the research (since each ASHA serves around 1,000 persons) partners, a multi-stage stratified sampling design was which would serve as secondary sampling units. adopted for the selection of the Primary Sampling ■ Urban: Identification of clusters of 1,000 persons Units (PSU); villages for rural areas and towns for urban in census towns could be done by using areas. Additionally, PSUs were divided into Secondary Urban Frame Survey blocks that account for Sampling Units (SSU) based on population clusters of 80–200 households 1,000 individuals to facilitate the household selection process. A logical sampling approach in this case would Step 4: Selection of respondents follow the administrative hierarchy at the district level: For identification and developing the sample frame of District —> Block —> Village/census town (PSU) target households, an enumeration exercise would be conducted. The enumeration exercise would include Step 1: Selection of blocks: screening questions on pregnancy history, age profile of Three blocks selected from each district based on the the household members, incidence of communicable and weights derived from relevant census indicators (in this non-communicable diseases, etc. case, percent of marginal and total working populations, Nine households (HH) with at least one woman or child percent of tribal population, and female literacy rate). meeting the below MNCH eligibility criteria will be randomly The weights have been decided based on the need to selected from each PSU. The key eligibility criteria are: sample for under-served and tribal regions. All blocks 1. Women with pregnancy outcomes in the past within the district would be ranked in a descending twelve months. This will include women with an order, on their scores, and the three blocks selected active pregnancy, as well as women who had other would be: pregnancy outcomes in last twelve months. ■ Block with the highest score In addition to this, three other eligibility criteria will be ■ Block with median score used to draw additional household sample: ■ Block with the lowest score 2. Households with at least one member with hypertension/ heart disease/ stroke – 5 HH per Step 2: Selection of villages/census towns (PSU) PSU Ten villages selected from each block, based on their 3. Households with at least one member with distance to the comprehensive emergency obstetric diabetes – 3 HH per PSU and neonatal care (CEmONC), or the district/subdistrict 4. Households with at least one adolescent member - hospital if a CEmONC does not exist in the block. 7 HH per PSU 15 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Case Study 1 presents a PPA undertaken by the authors of this report to explore patient pathways to NCD and MNH primary health care services among a poor, urban population in Bangladesh. This mixed-methods, multi-component case has been split into three parts across the guide—Case Study 1A presented in Module 1 provides an example of primary data collection for care-seeking data; Case Study 1B in Module 2 illustrates primary data collection to profile healthcare providers; Case Study 1C in Module 6 integrates a geospatial component into a PPA. Case Study 1 demonstrates how different study components can be considered together to gain insight, while highlighting that not all components are required for PPA. CASE STUDY 1A: A patient pathway analysis to understand the experience and preferences of low-income, urban residents in accessing NCD and MNH-related PHC services in Bangladesh – (World Bank and icddr,b) CASE STUDY 1A HIGHLIGHTS: An example of a basic PPA that involves primary data collection using a household survey to gather information on patient pathways. Background Bangladesh’s rapidly expanding urban population has put an enormous strain on the local healthcare system. A poorly coordinated and fragmented urban health system leads to gaps in healthcare service delivery. This study focused on the pathways of individuals navigating the health system to seek care for chronic non-communicable diseases (NCDs) and/or maternal and newborn health (MNH) services. To improve quality of primary health care and health outcomes through an urban health project in Dhaka and Chattograma city corporations, stakeholders sought to better understand how low-income residents interact with the health system and where they prefer to seek their care. The objectives of the PPA were to: 1. Assess how low-income MNH and NCD patients (and clients) navigate the urban health systems of Dhaka and Chattogram city corporations across different care providers. 2. Explore drivers of care-seeking behaviors, experiences, and preferences for care among low-income individuals for MNH and NCD in these study areas. 3. Identify opportunities for reorganization of care and promising service delivery models that could address bottlenecks in accessing high quality MNH and NCD services for low-income, urban populations in these study areas. Data sources This PPA covered three city corporations in Bangladesh—Dhaka North, Dhaka South, and Chattogram—and targeted several NCDs (hypertension/cardiovascular disease (CVD), chronic respiratory illnesses (CRI), and diabetes) and MNH services. Table 6 presents data sources, both qualitative and quantitative, that were included in the basic PPA. 16 PART 3: Practical Examples of PPA Methods TABLE 6. DATA SOURCES FOR CASE STUDY 1 Demand side Supply side Household survey To understand NCD/MNH patient Provider profiling To understand availability and pathways, experiences, and and mapping geographic access to NCD/MNH preferences. services. This is the core method for this PPA. Discrete choice To quantify the drivers of patient Key informant To understand bottlenecks in experiment (DCE) preferences for NCD/MNH care interviews (KII) healthcare delivery. providers. In-depth To contextualize the NCD/MNH interviews (IDI) care-seeking pathway and better understand factors influencing the choice of NCD/MNH care provider. Focus group To understand community discussions (FGD) experiences with care-seeking for NCD/MNH services. Household survey for patient care-seeking behavior and preferences (demand-side) While several data sources were used for Case Study 1, the household survey is the foundation of the PPA. A household survey alone is the most basic form of PPA. If the goal is a simple analysis of patient navigation through the health system—where do patients go for care and when?—this data collection method alone may be sufficient. Methods This survey aimed to map patient pathways through the health system. The PPA methodology we developed was centered around pathway interviews with carefully selected respondents who were asked to recall their care journey based on the concept of ‘touchpoints.’ Although resource- intensive, this meant community-based enrollment was preferable to facility-based enrollment, and if the latter was used, care was taken to enroll across a mix of facility types. Eligible respondents had either been diagnosed with one of the three included NCDs in the past five years and/or been pregnant in the past two years. An initial screening exercise identified households in the study areas with eligible individuals, and eligible individuals who consented to participate were then interviewed using the household survey questionnaire. Figure 3 shows the study flow of number of households screened and number of NCD and MNH respondents, respectively, who were interviewed. The household survey was completed over a 3 months period and collected data on patient sociodemographic characteristics and care-seeking organized by key touchpoint across the continuum of care (Figure 4). 17 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS FIGURE 3. NUMBER OF SCREENED HOUSEHOLDS AND INTERVIEWED PARTICIPANTS Dhaka North Dhaka South Chattogram Total CC CC CC Numbers of HH screened 5,741 5,772 6,410 17,923 for NCD respondents Number of NCD 1,297 1,251 1,304 3,852 respondents interviewed Numbers of HH screened 3,467 3,094 2,201 8,792 for MNH respondents Number of MNH 664 661 652 1,977 respondents interviewed Findings A total of 3,852 and 1,977 respondents were interviewed regarding NCD and MNH services, respectively, across all three city corporations. Within the hypertension/CVD patient group, approximately three-quarters (77.1%) of patients had hypertension, followed by 15.6% of patients with other heart/circulatory system diseases, 5.8% with stroke, and 1.6% with heart disease. No sub-types were distinguished in the diabetes group. In the CRI group, over half (54.9%) had an asthma diagnosis, followed by 41.7% with chronic cough, and only 3.4% with a COPD diagnosis. The household survey data on hypertension patient pathways reflected that pharmacies and drug sellers were major providers of hypertension care. The majority of hypertension patients first sought care at pharmacies and drug stores (often remaining in care there), second only to private hospitals. Pharmacies were a common point of care for measuring blood pressure and obtaining medication refills (current disease management), regardless of the point of care for initial contact, diagnosis, and treatment initiation. In contrast, most diabetes patients preferred to use private hospitals as the first point of entry to care, with more than half being initially diagnosed there. This was followed by pharmacies for initiating care, and pharmacies surpassed public hospitals as the reported location for diagnosis. Additionally, similarly to hypertension, most respondents maintained their treatment with pharmacies (conducting monitoring tests and providing medication); however, one in five reported using private facilities for current disease maintenance and management which could be related to complex needs. On the MNH side, more than three-quarters (78%) of the survey respondents had a live birth in the last two years, while 21% were currently pregnant. The study team did not collect care-seeking details from 24 (1%) of respondents who had a stillbirth, neonatal death, menstrual regulation, and miscarriages, leaving 1,953 respondents in the pathway interview sample. Overall, most (74%) study respondents were aged between 20–34 years. By city corporation, the average age of 18 PART 3: Practical Examples of PPA Methods FIGURE 4. KEY TOUCHPOINTS ACROSS THE CONTINUUM OF CARE FOR NCDS AND MNH NCD touchpoints Refers to point of initial care-seeking for an NCD from a healthcare TOUCHPOINT 1: provider (HCP) (includes name and type of provider, timing of visit, Initial contact diagnostic and referral advice received at initial visit). Refers to point of care for diagnosis of NCD (includes name and TOUCHPOINT 2: type of provider, timing of care, diagnostic and referral advice Diagnosis received at diagnosis visit). Refers to treatment initiation for the NCD once diagnosed (includes TOUCHPOINT 3: name and type of provider, timing of care, treatment prescribed and Treatment initiation referrals). Refers to point of care for ongoing management of NCD(s) (includes TOUCHPOINT 4: name and type of provider, treatment adherence, follow-up and Current disease management monitoring). TOUCHPOINT 5: Refers to point of care for acute, or emergent, NCD episodes in the Acute episodes past 12 months (includes name and type of provider). MNH touchpoints Refers to the first point of contact for maternal care. This could be for TOUCHPOINT 1: a pregnancy test, antenatal care (ANC), or initial care sought for any Initial contact other pregnancy-related issue (includes name and type of provider, timing of visit, and reasons for seeking maternal care). Refers to point of care for regular check-ups during pregnancy TOUCHPOINT 2: (includes name and type of provider, total number of ANC visits, Regular check-ups adherence to diagnostic or treatment advice). Refers to point of care for management of complications experienced TOUCHPOINT 3: during pregnancy and/or childbirth (includes name and type of provider, Complications timing of care, adherence to diagnostic or treatment advice). TOUCHPOINT 4: Refers to point of care for delivery (includes name and type of Childbirth provider, type of delivery, and reason for choosing a particular facility). Refers to point of care for postnatal care (PNC) check ups for mother TOUCHPOINT 5: and baby (includes name and type of provider, timing of visit, postnatal Postnatal care complication, and services offered). 19 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS BOX 2. SANKEY DIAGRAMS AS A TOOL OF VISUALIZING FLOW IN A PPA Visual presentation is a strong tool for summarizing shows the type of provider by level and the number patient pathways. Sankey diagrams were used in this of participants that reported contact with that provider study to visualize the flow of patients through key at each touchpoint. (GP 19 means 19 people reported touchpoints, while also providing information on the initiating care at Public Community/Primary care classification of providers (by type and sector) at each providers; and while 19 reported the same at postnatal touchpoint as reported by the patients. Figure 5 below care, the graph clarifies how some of the initial 19 were was generated using Sankeymatic.com. Figure 5 replaced by individuals from a different provider) FIGURE 5. SANKEY DIAGRAM SHOWING MATERNAL CARE PATHWAY ACROSS ALL THREE URBAN STUDY AREAS Initial Regular Pregnancy Postnatal Contact Check-up Childbirth Care Public Community/ Primary (GP) GP 19 GP 20 GP 18 GP 19 Public Hospital (GH) GH 29 GH 32 GH 93 GH 94 Private Chamber (PC) PC 41 PC 30 PC 9 Private Hospital (PH) PH 212 PH 249 PH 415 PH 412 NGO (NG) NG 629 NG 305 NG 301 NG 623 Traditional Healer (TH) TH 2 Drug Store/ Pharmacy (DS) DS 68 DS 25 Other (OT) OT 4 OT 9 OT 5 OT 1 20 PART 3: Practical Examples of PPA Methods respondents ranged from 24 to 25 years. The majority (>60%) of the women had one to two children and mean parity was 2.14 children. The household survey data on MNH patients highlighted that non-governmental organizations (NGOs) played a significant role in initial pregnancy care, with 63% of respondents reporting they sought initial care at NGOs, 21% at private hospitals, and 7% at drug stores. Women usually continued their ANC check-ups with the same initial provider, although some women switched to NGO-provided ANC from other types of facilities. Among women who delivered at a facility (56% of total deliveries), 35% and 11% transitioned for delivery from NGO-based ANC to private hospitals and public hospitals, respectively. PNC was usually received from the same provider where women delivered. Out-of-pocket expenses were the most cited barrier to accessing care across all touchpoint. Perceived quality of NCD/MNH services was the main driver of provider choice across all study locations and health conditions. Ideally, the household survey method presented here would be paired with the methods presented in Module 2 (Case Study 1B) on provider mapping. Provider mapping helps to better understand patient preferences given the landscape of available care in a specific region. Without additional profile mapping of the healthcare providers, there is no information available on which facilities were not chosen, or bypassed, by patients. Additionally, patient recall bias is an inherent limitation that can be worsened by changes in facility names and location and similarly named facilities. Combining Modules 1 and 2 together can help clarify any gaps that result from these changes. MODULE 2: HOW TO MAP AND PROFILE HEALTHCARE PROVIDERS AND SERVICES Building on the basic PPA design presented in the first module, Module 2 covers profiling and mapping MODULE 2 – AT A GLANCE healthcare providers in the study areas to under­ stand the availability and accessibility of specific ⊲ The value of a dedicated component healthcare services. for examining providers in the area of interest, as part of the PPA study Provider profiling and mapping involves collecting ⊲ Examples of methods for data to characterize local healthcare providers and examining providers by type and services, such as facility type and level, health level of provided care services provided, commodity stocks and prices, ⊲ A practical example from and/or available infrastructure and human resources, the author’s work combining among others. Provider profiling and mapping household surveys (patient results in a robust dataset of provider services interviews) and provider profiling that can provide valuable information for health to yield greater results system redesign efforts and expand the scope of a PPA (e.g., examining patterns of facility bypassing, contextualizing choice of provider, and adding more detail to a geospatial analysis). Case Study 1B illustrates how provider profiling was a valuable addition to the household survey explained in Case Study 1A. 21 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS CASE STUDY 1B: Provider profiling and mapping as part of a PPA on NCD and MNH care- seeking by urban low-income clients in Bangladesh – (World Bank and icddr,b) CASE STUDY 1B HIGHLIGHTS: • An example of an additional PPA component that involves primary data collection using a questionnaire to gather information on available providers and health care services in a given area. • This case study also covers methods to link care-seeking data (Module 1) and provider mapping/ profiling data (Module 2). Background In this PPA focused on urban health systems in Bangladesh, a provider profiling and mapping component was included. Most provider data were collected before the household survey was implemented; however, final data collection was completed once the household survey had provided additional information on reported NCD and MNH care providers. Results from the initial provider profiling work in each of the three study areas (Dhaka North, Dhaka South, and Chattogram city corporations) were used to inform the design of the household survey. With both household survey results on care-seeking behavior and a database of available providers and health services offered, an integrated analysis of demand-supply dynamics, healthcare utilization choices, and bypassing of specific facilities was possible. Methods The data collection tool used for provider profiling and mapping was a semi-structured questionnaire administered to the managers and owners of healthcare facilities and providers. A total of 905 facilities/providers were identified across the three study areas. The tool collected different types of information as follows: (Table 7) After the providers are characterized according to level and sector, information on their health services can then be analyzed to present service availability by provider type, health system sector TABLE 7. CONTENT OF THE PROVIDER PROFILING TOOL Section 1 Provider identification including address, location, and GPS coordinates. Section 2 Information related to health facility or service provider, including the name of the facility or provider, the manager, the classification by type of provider and level of the health system, whether it is a facility or an individual provider (e.g., private office), and its location relative to the area it serves. Information on whether the provider receives patients with the conditions of interest, the presence of a referral system, healthcare staff and their schedule (working hours and days). Section 3 Information on the availability of services for specific services, in this case specific NCDs, associated operating hours, the presence of essential functional equipment, relevant diagnostic tests, and medications 22 PART 3: Practical Examples of PPA Methods and level, as well as location. This provides a description of the landscape of health services in the study areas. Importantly, this taxonomy can be used in a household survey questionnaire on care-seeking. For example, response options in the survey may be informed by the provider profiling data. TABLE 8. EXAMPLES OF PROVIDER CLASSIFICATION CODES BY PROVIDER TYPE, FACILITY TYPE, HEALTH SECTOR AND HEALTH SYSTEM LEVEL AND HOW IT RELATES TO TOUCHPOINTS IN A PPA Type of provider Type of facility Health system sector Level of facility a. MBBS doctor a. Drug store a. Public a. L0: Community b. SACMO/Paramedic b. NGO clinic b. Private formal b. L0: PHC outreach/ c. Drug seller c. Hospital c. Private informal mobile d. Pharmacist d. UPHCSDP d. NGO c. L1: Primary e. Traditional healer e. Private clinic d. L2: Secondary f. Spiritual healer f. Private doctor’s e. L3: Tertiary chamber g. Pharmacy h. Telehealth service i. Others, specify Application of framework in Bangladesh: How do the urban poor with chronic illness navigate the local healthcare system? Initial care Initial Additional Treatment Medicine Monitoring seeking diagnosis diagnostics start refill L3-Public L3-Public L3-Public L3-Public L3-Public L3-Public sector sector sector sector sector sector CVD [N=1995] hypertension/ L2-Public L2-Public L2-Public L2-Public L2-Public L2-Public stroke/heart sector sector sector sector sector sector disease L1-Public L1-Public L1-Public L1-Public L1-Public L1-Public sector sector sector sector sector sector Diabetes L0-Public L0-Public L0-Public L0-Public L0-Public L0-Public [N=924] sector sector sector sector sector sector L3-Private L3-Private L3-Private L3-Private L3-Private L3-Private Chronic formal sector formal sector formal sector formal sector formal sector formal sector respiratory L2-Private L2-Private L2-Private L2-Private L2-Private L2-Private disease [N=882] formal sector formal sector formal sector formal sector formal sector formal sector COPD/asthma, excluding TB L1-Private L1-Private L1-Private L1-Private L1-Private L1-Private formal sector formal sector formal sector formal sector formal sector formal sector L0-Private L0-Private L0-Private L0-Private L0-Private L0-Private informal informal informal informal informal informal Digital* = Documentation of all digital and mobile tools used when navigating health market, including phone Self/Digital* Self/Digital* Self/Digital* Self/Digital* Self/Digital* Self/Digital* 23 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Findings In this study, the provider profiling component provided important insights into the landscape of providers in each of the three city corporations, such as: 1. Providers in the area: drug sellers and pharmacies made up the majority of healthcare providers, followed by private hospitals and traditional healers. Some variations between the three city corporation areas in the makeup of the provider mix were also described. 2. Availability of health services: 77% of mapped facilities provided services for hypertension and 74% provided services for diabetes. These proportions were similar across the three study areas. The database had information on what services were available and which providers offered these services. Analyzing demand- and supply-side together: Figure 6 illustrates how different data components of the study were analyzed (combines Case Study 1A and Case Study 1B). In addition to the household survey and provider mapping, the study also had a Discrete Choice Experiment (DCE) component, and geospatial analysis was completed based on provider location data and respondent residences where care-seeking journeys originated (discussed in Module 6). Quantitative data for the household survey and provider mapping were analyzed using statistical analysis software to generate descriptive statistics and conduct inference testing and logistic regression. Excel and manual coding was used for thematic analysis FIGURE 6. KEY COMPONENTS OF THE METHODOLOGY USED IN CASE STUDY 1A AND 1B Data category Primary data Analysis Insights collection Mapping HH Sankeymatic.com for Sankey Quantitative: Visualized flow through the questionnaire diagrams explaining the flow Socio-demographic touchpoints by level & patient- Basic care-seeking of patients between sector pathway/ behavior touchpoints journey Self-reported delays Discrete Choice (demand in care Experiment Choice of provider per side Stata 16 for descriptive touchpoint data) analysis and statistical inference tests Qualitative: In-depth Contextualizing the Challenges to accessing interviews care-seeking pathway care and factors Key informant Stata 16 for conditional Identifying factors influencing choice of interviews logit regression model influencing the choice care provider Focus group of healthcare provider discussions To further understand community experiences Additional context on of obtaining NCD care quantified care-seeking Coding in Excel for thematic insights analysis and triangulation Provider Providers by type, Phone calls, site Categorized providers by profiling sector and level visits, surveys, type level and sector (supply Health services and and side their availability questionnaires Stata 16 for descriptive data) Provider location data analysis Availability of relevant health services 24 PART 3: Practical Examples of PPA Methods BOX 3. EXAMINING ALIGNMENT BETWEEN CARE-SEEKING BEHAVIOR AND SERVICE AVAILABILITY Analyzing alignment between care-seeking and available of preventative care, diagnosis, and treatment) and health services—or in other words, determining whether comparing it against care-seeking behavior data. This people seek care where appropriate services are is an essential concept to examine and is tied to how available—involves assessing the supply-side (e.g., place clients navigate the healthcare system (Figure 7). FIGURE 7. CONCEPT MAP FOR ANALYZING ALIGNMENT BETWEEN CARE-SEEKING PATTERNS AND HEALTH SERVICES Provider mapping Alignment between care-seeking and profiling patterns and health services Basic care-seeking behavior Type of provider Level of provider Place of initiation of care Sector of provider Utilization and bypassing of primary care access points Availability of Place of diagnosis services Availability of diagnostics Availability of Access to diagnosis Place of treatment treatment start Access to treatment (continues) 25 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS BOX 3. EXAMINING ALIGNMENT BETWEEN CARE-SEEKING BEHAVIOR AND SERVICE AVAILABILITY (CONTINUED) This figure illustrates the concept of alignment; however, 4. Do clients remain with the same provider after it does not illustrate the depth of the data that you can initiating care? Where does provider change get by examining and comparing demand- and supply- happen? side. This perspective can help answer questions like: 5. What is the role of informal providers? 6. Where are clients going the most and the least? 1. Are clients seeking care at the appropriate level of care? These questions not only provide insights but add 2. Which sector are clients choosing to initiate opportunities to ask additional questions about the care? (public, private formal, private informal, or behavior; namely, the reason clients go where they go, NGO) What are the reasons for choosing certain especially if their care-seeking behavior is misaligned providers? with service availability. Additional qualitative methods 3. Do the most frequented providers have the like FGDs and IDIs can help explore clients’ perspective diagnostic and treatment capabilities needed to about what matters to them, their care experiences and serve this population? What challenges do clients perceptions of quality, and the underlying reasons for experience when using them? care-seeking behaviors. for qualitative data from interviews with patients and providers. Thematic analysis was conducted using inductive coding. Triangulation was achieved by comparing and contrasting findings from the KIIs, FGDs, and IDIs. ArcGIS was used to analyze the location data and to examine distances and movements to add another layer of understanding to the care-seeking data and clients’ choice of provider. CASE STUDY 2: Using secondary data to analyze the alignment between TB-related care-seeking and services in Pakistan – (Fatima et al., 2017) CASE STUDY 2 HIGHLIGHTS • A good example of a PPA that assesses alignment between care-seeking behavior and available health services (integrating Modules 1 and 2). • The authors do so using only secondary data. Background Pakistan has the fifth-greatest burden of TB globally, with the fourth highest prevalence of multidrug- resistant TB (MDR-TB). In 2015, the estimated number of new drug-susceptible TB cases was about 510,000, yet only 65% of that number were notified to the National Tuberculosis Program (NTP). Pakistan had set a goal to reach zero deaths due to TB by 2020, and an essential part of achieving this goal was to treat un-notified or “missing” TB cases. Hence, a PPA study would help in mapping 26 PART 3: Practical Examples of PPA Methods where TB patients go to seek care and help contextualize where the “missing” patients are (Hanson et al., 2017). Analyzing demand- and supply-side data in this study can give insights into the alignment of TB care-seeking behavior with TB service availability by examining different elements of the alignment, as seen in Figure 8: 1. Place of initial care-seeking by facility sector and level 2. Access to TB diagnosis at the initial care-seeking location 3. Access to TB treatment at the initial care-seeking location 4. Coverage of sputum microscopy services among health facilities 5. Coverage of TB treatment services among health facilities 6. Location of notification of TB cases compared to estimated local incidence 7. TB treatment outcomes compared to estimated incidence FIGURE 8. KEY COMPONENTS OF THE METHODOLOGY USED IN CASE STUDY 2 Data category Data collection Analysis Insights Mapping Basic care-seeking 2011–2012 National Initial care-seeking the behavior Health Accounts patient- report Diagnostic coverage pathway/ journey Diagnostic access (demand- Treatment coverage side data) Calculated 7 indicators Treatment access Provider Providers by type, National Health Notification location profiling sector and level Accounts report for (supply- most updated data Treatment outcome side data) on private sector facilities (2009–2010 report) and public sector facilities (2011–2012 report) Nearly 90% of patients with tuberculosis initiated care- Coverage of National Tuberculosis seeking in the private sector microscopy for Program laboratory TB diagnosis database Fewer than half of public sector facilities had diagnostic capacity Treatment availability did not reflect patient care-seeking preferences Treatment availability did not reflect patient care-seeking preferences 27 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Data Sources The study used the following readily available data sources: 1. The 2009–2010 National Health Accounts report provided the most up-to-date data on the number of private sector health facilities. 2. The 2011–2012 National Health Accounts report provided the most up-to-date data on the number of public sector health facilities and basic care-seeking behavior for general illness. 3. The National Tuberculosis Program (NTP) provided a laboratory database on all tuberculosis microscopy facilities throughout Pakistan and a list of general practitioners with TB treatment services. Supply-side data The supply-side data for this study included coverage of tuberculosis microscopy and the number of providers available by type, sector, and level. Tuberculosis microscopy coverage at health facilities was calculated by dividing the number of health facilities offering microscopy services (obtained from the National Tuberculosis Program database) by the total number of public and private sector health facilities (obtained from the two National Health Accounts reports from 2009–2010 and 2011–2012). Facilities were then categorized by type, level and sector, the number of facilities with microscopy services was divided by the total facilities for each province, sector, and level. Demand-side data Demand-side data included location of initiation of care for general illness (obtained from the 2011 – 2012 National Health Accounts report). Data specific to tuberculosis care-seeking was not available. When collecting data on place of initial care, the NHA survey gave respondents different types of public and private facilities (grouped as described above). This provided an estimate for the PPA on where individuals presumed to have tuberculosis initiated care. In a separate paper published on their methods, the authors explain the reasoning behind this choice: TB-related care-seeking data from prevalence surveys is specific, but often offers a small sample size that is not sufficient for robust sub-national statistical analysis (Hanson et al., 2017). Data analysis The analysis of this study is explained in detail in the published methods paper, and the methodology itself was replicated in five countries with similar results. In summary, the analysis revolves around calculating the indicators illustrated in Table 9 (Hanson et al., 2017). Findings Almost 90% of patients initiated care in the private sector, which accounts for only 15% of facilities with tuberculosis diagnosis and treatment capacity. While nearly half of tuberculosis microscopy laboratories were located in public sector basic health units and regional health centers, few patients initiated care in these facilities. 28 PART 3: Practical Examples of PPA Methods TABLE 9. SUMMARY OF INDICATORS CALCULATED IN CASE STUDY 2 (HANSON ET AL., 2017) Indicator Calculation Initial care- Place of initial care-seeking by facility sector and level seeking Diagnostic Coverage of microscopy services was calculated by dividing the number of facilities with TB microscopy coverage (categorized by type, level, and sector) by the total number of facilities for each province, level and sector. Diagnostic Access to diagnosis at initial care-seeking location. This shows how likely an individual seeking TB access care is to access a facility with microscopy on the first visit. This was estimated by multiplying the share of initial visits per sector and level (column 1) by the coverage of microscopy per the relevant sector and level (column 2). Treatment Coverage of treatment services among health facilities. Calculated similarly to column 2, the data coverage represented the share of facility per level and sector with TB treatment services. Treatment Access to treatment at initial care-seeking location. Similarly to column 2, this column represents how access likely an individual seeking TB care to access a facility with TB treatment on the first visit. This was calculated by multiplying the share of initial visits per sector and level (column 1) by the coverage of TB treatment services per the relevant sector and level (column 4). Notification This column indicates the proportion of cases reported to Pakistan’s NTP by both the public and location private sector, as well as the unreported estimates. Treatment This column indicates the proportion of cases that either completed or did not complete treatment, outcome as well as the missing cases that are not notified to NTP. Key takeaways from modules 1 & 2 ■ Lack of alignment between where people seek care and which facilities have tuberculosis diagnosis and treatment capacity results in a low likelihood of patients reaching facilities with capacity for tuberculosis services during their first visit. ■ Patient pathway involves understanding patient touchpoints and flow through the health care system. Some conditions (like TB and HIV) have vertical monitoring systems which can be harnessed for PPA. ■ Secondary data (e.g., from HMIS or DHS) can be used for PPA, depending on the research questions and available diseases-specific data on the demand- and supply-side. This requires good understanding of the analytical possibilities these secondary data offer, the ability to link datasets in a meaningful way, and the careful documentation of caveats, limitations and assumptions. ■ Primary data collection for care-seeking behavior can provide comprehensive detail on patient flow and follows each individual patient across different providers and levels of care. Provider profiling can be done first to characterize the health service landscape (levels of care, distance, types of care) before primary collection of care-seeking data. ■ In select cases like this TB PPA, a PPA with survey or routine data can complement a Cascade Analysis and provide a wider picture of people’s care-seeking within a health system. 29 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Practical tips ⊲ Different data sources may categorize health facilities differently or use different naming conventions, making it difficult to compare or merge data. For example, in Case Study 3, each data source used a different naming convention for the health facilities, and there were differences in the classification system used for level of care. For example, data from the National Tuberculosis Program on health facilities with tuberculosis microscopy services included a Level 3 classification, whereas the National Health Accounts reports with the total number of public and private sector health facilities combined Levels 2 and 3. To address this, the researchers categorized each health facility as public, private informal, or private formal, and then further categorized health facilities according to the level of the health care system (e.g., Level 0, Level 1, Level 2, etc). The researchers merged the Level 2 and 3 categories for their health facility mapping work since the NHA data could not be disaggregated into these levels. ⊲ Consider care-seeking that may not have been captured in available official lists of formal and informal health care providers. For example, in Case Study 3, the list of public and private health facilities provided by the National Health Accounts report did not include providers such as pharmacies, drug vendors, or other traditional care providers. If care- seeking in these locations was highly prevalent among people with potential tuberculosis symptoms and microscopy services are not available, the estimates generated on the percentage of people seeking care in facilities with tuberculosis microscopy may be upwardly biased. ⊲ Disease-specific care-seeking data may not be available in all settings. Care-seeking for general illness can be used as proxy in this case, but there may be differences in where tuberculosis patients seek care vs. other patients. MODULE 3: HOW TO ASSESS DELAYS IN CARE USING PPA Delays in care across each stage of patient pathways for NCDs and MCH are important determinants of MODULE 3 – AT A GLANCE health outcomes. Myriad factors can contribute ⊲ Collecting data to assess delays to  delays in care during pregnancy or for NCDs, in care as part of care-seeking including inability to pay, long distances to facilities, behavior assessment and risk perception, among others. Delays in initial ⊲ Defining delays in care and care-seeking may be related to lack of awareness of common intervals examined by the disease, with patients being asymptomatic or PPAs in the literature experiencing only mild symptoms. For example, ⊲ Example of methods that combine patients may not be aware of a slow-progressing primary data collection and chronic disease—for example, prostate cancer—or available secondary data could experience only mild or moderate symptoms in the days before an acute medical emergency, such as a stroke. Patients who have been screened and diagnosed may also wait to pursue follow-up care, particularly if they don’t have symptoms. Throughout the continuum of care for NCDs and 30 PART 3: Practical Examples of PPA Methods MCH, patients may be delayed in care-seeking at any stage, either for systematic reasons, or due to individual behaviors, with accumulated delays potentially posing more of a threat to patient health. Behavioral economics research reveals that individuals tend to heavily discount the distant consequences of chronic conditions, especially in the early stages of illness, which may be related to delays in initial care-seeking (Ghosh et al., 2021). Delays in MCH care Progress in improving maternal and newborn morbidity and mortality has stalled in recent years, and there are persistent disparities in maternal and newborn quality of care and outcomes in particularly in sub-Saharan Africa (WHO, 2023). The risk of maternal death is more than 400 times higher in LMICs relative to high-income countries (Shah et al., 2020). Timely high quality obstetric and neonatal care at both the primary care level (antenatal care and postnatal care) and higher levels of care (skilled deliveries in facilities equipped to manage emergencies) can reduce preventable deaths and disability. In 1994, Thaddeus and Maine developed the Three-Delays Framework for maternal health to examine (1) delays in the decision to seek care; (2) delays in arrival at a health facility; and (3) delays in receiving adequate care (Shah et al., 2020). Delays in NCD care For conditions with long-term treatment and monitoring needs, any delays in the continuous management need to be considered, such as late check-ups and delayed medication refills. A BRIEF INTRODUCTION TO ASSESSING DELAYED CARE-SEEKING WITHIN A PPA Assessing delays in care-seeking can be an important component of a PPA to inform health system redesign or strengthening efforts. This module will focus specifically on assessing delays in care- seeking by the patient, which may be due to factors at the individual, community, and/or health system level. Defining delayed care-seeking Delays in care are defined in this guide as an excessive time lag between two events in care. This could lead to an unfavorable outcome such as additional costs, complications or suffering. However, not all delays are unreasonable or significant. The definition of a delay is thus context- dependent and often related to standards of established clinical practice and guidelines. Early in the design of a PPA component to assess care-seeking delays, researchers should review available clinical guidelines and relevant literature to define appropriate thresholds or cut-off points for delayed care-seeking for a specific health condition in their study area across each interval of the patient pathway. Data collection instruments and analysis should integrate these pre-defined thresholds for delayed care-seeking. 31 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS BOX 4. EXAMPLES OF DIFFERENT DEFINITIONS FOR DELAYS IN CARE FROM THE LITERATURE A study in Thailand looking at cancer care defined and defined it as the time interval between the first delay as a patient waiting more than 3 months after medical consultation of respiratory-related diseases symptom onset to seek medical care (Bhosai et al., and the initiation of TB treatment (Chen et al., 2015). 2011). However, within studies examining TB care, the The objective was to measure the performance of definition of delays vary; for example, Zhang et al., TB services after their integration into the general focused on the interval between initial care-seeking healthcare system and the phase out of a specialized and diagnosis of TB (referred to as Long Diagnostic TB care system. Rather than defining a cut-off for an Delay, or LDD), and defined the delay as an excess acceptable duration, the study compared the median of of 2 weeks from the initial care visit (Zhang et al., 2021). observed HSD between 2003 and 2008, and found it A different study analyzed Health System Delay (HSD) to be 26 days and 33.5 days, respectively. Data collection methods Researchers should begin by breaking down the patient pathway into sequential time intervals for assessment of delays. Commonly, studies looking at delays in care examine the following time intervals (or combinations of them) between: 1. The appearance of symptoms and the initial care visit, which could be further divided into: a. Time between appearance of symptoms and deciding to seek care b. Time between deciding to seek care and arriving at the facility 2. Initial care visit and the provider requesting an appropriate diagnostic test 3. Provider request of the required diagnostic test and acquiring test results (confirmation of diagnosis) 4. Diagnosis and initiation of treatment 5. Treatment monitoring check-ups 6. Treatment targets and delay in scale-up of treatment intensity Beyond quantifying delays, a PPA study can also examine the factors associated with delayed care-seeking using quantitative and/or qualitative methods. The cases presented below provide examples of studies that used a mixed methods approach to assess diagnostic delays for tuberculosis in China (Case Study 3). CASE STUDY EXAMPLES CASE STUDY 3: A multi-center PPA study to examine delays in diagnosis of tuberculosis, in China – Zhang et al. 2021 CASE STUDY 3 HIGHLIGHTS • Incorporates provider mapping, as well mapping of patient care-seeking behavior (Modules 1 and 2) • Assesses delays in care along the patient pathway (Module 3). • Uses a combination of primary and secondary data sources. 32 PART 3: Practical Examples of PPA Methods Background In China, at the time of the study, once any TB diagnosis is confirmed, patients were hospitalized and isolated while observed for adverse effects of TB drugs. About 42% of patients with pulmonary TB delayed seeking care by a month or more, especially in low- and middle-income areas. Patients are often stressed by multiple care visits before diagnosis is reached and treatment is started— delayed diagnosis increases costs and allows patients to remain an active source of transmission in the community. Therefore, there is a need to optimize the diagnostic pathway by understanding the health-seeking patterns of patients, as well as service provision. The focus of this PPA study was the time interval between the initial care visit and confirmation of a TB diagnosis. The study team examined risk factors that contribute to Long Diagnostic Delay (LDD), in addition to evaluating the availability and utilization of TB equipment in hospitals. LDD was defined as a delay greater than 2  weeks from the initial visit at any health facility to diagnosis of TB. This definition was developed using the Tuberculosis Control and Assessment Protocol established by the Chinese Ministry of Health and relevant literature (Yang et al., 2020) where a patient survey was used to quantify delays using the following variables. The guide will not explore the details of this study; however, the Table 10 provides a sample from the questionnaire used by the authors to collect information on delays from the study participants. The full study can be found here. TABLE 10. SAMPLE FROM QUESTIONNAIRE TO ASSESS DELAYS (YANG ET AL., 2020) 1 What is the time interval between the date of onset of suspicious symptoms of TB (e.g., cough, hemoptysis, night sweat, fever, and chest pain) and your first presentation to a professional health provider? A. <1 week; B. < 2 weeks; C. < 3 weeks; D. 3 weeks–2 months; E. 2–3 months; F. >3 months 2 What is the time interval between the date of your first presentation to a professional health provider and being diagnosed as TB? A. <1 week; B. < 2 weeks; C. < 3 weeks; D. 3 weeks–2 months; E. 2–3 months; F. >3 months 3 What is the time interval between the date of TB diagnosis and initiation of treatment? A. <1 week; B. < 2 weeks; C. < 3 weeks; D. 3 weeks–2 months; E. 2–3 months; F. >3 months Accordingly, TB-related delays were classified into one of three types: 1. Delay in care-seeking: the time interval between onset of suspicious symptoms of TB and first presentation to a professional health care provider is longer than three weeks. 2. Delay in diagnosis: the time interval between the date of first presentation to a professional health care provider and TB diagnosis is longer than two weeks. 3. Delay in treatment: the time interval between the date of TB diagnosis and initiation of treatment is longer than one week. 33 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Data collection This case study was the first multi-center PPA with data on clients’ perspectives in China. Data collection took two months to complete and represented both the demand and supply side of the care-seeking journey. Data on timing of care was collected with data on care-seeking behavior using the same instrument. Demand-side data Data on care-seeking behavior was collected by interviewing TB patients, either face-to-face while they were hospitalized after their diagnosis or after being discharged through a live online interview (happening in real time). The interview questionnaire had two parts: (1) sociodemographic characteristics and medical information, including symptoms and other existing conditions; (2) details of the care-seeking pathway (Table 11). TABLE 11. SHOWING HOW DATA VARIABLES FOR DELAYS IN CARE WERE INTEGRATED IN DEMAND-SIDE DATA COLLECTION1 Data category Data variables Sociodemographic ■ Age, gender, occupation, marital status, the highest educational level characteristics achieved, annual household income per capita, smoking history, and alcohol consumption Basic care-seeking ■ Date of symptom onset data + Delays in ■ TB history or contact care-seeking ■ The date a health facility was visited and its name ■ Length of time and distances to hospital ■ Laboratory and radiological findings ■ Diagnosis (including misdiagnosis, suspected TB, clinical diagnosis of TB, and biologically confirmed TB) ■ treatments prescribed and administered for TB Supply-side data For the provider profiling portion of this study, data was collected on 20 TB hospitals across 17 provinces in China through site visits and telephone inquiries. This included medical records and availability of diagnostic tests. Information on the level of health facilities was obtained from the website of the Chinese National Health Commission and facilities and providers were classified into levels accordingly as shown in Table 12. Information on the availability of diagnostic TB laboratory equipment was obtained from the hospitals frequented by the respondents in the study, either by site visits or telephone calls by the trained research staff. The full data variables used in this study can be found in the accompanying published supplemental 1 material. 34 PART 3: Practical Examples of PPA Methods TABLE 12. CLASSIFICATION OF HEALTH FACILITIES BASED ON THE THREE-TIER HOSPITAL SYSTEM (ZHANG ET AL., 2021) Health facility level Health facility type Beds Equipment Staffs Level 0 (L0) Private clinics or Not available Not available 2–3 clinicians trained for community health basic medical care sectors Level 1 (L1) Hospitals among 20–99 Basic clinical departments, 5–10 clinicians and several communities including laboratories, X-ray several paramedical staffs room, etc. Level 2 (L2) Hospitals among a 100–499 A diversity of clinical and 150–800 clinicians and district or town auxiliary departments, paramedical staffs including dedicated TB departments. Level 3 (L3) Hospitals among a >500 Comprehensive clinical >1000 clinicians and city or province departments, as well as the paramedical staffs most advanced laboratories and radiologic rooms Data analysis FIGURE 9. KEY COMPONENTS OF THE METHODOLOGY USED IN CASE STUDY 3 Data category Data collection Analysis Insights Mapping Client flow (Sankey Socio-demographic Primary data: in-person Statistical analysis the patient- diagram) data and online questionnaire Student's t- test; pathway/ Mann-Whitney U test; journey Chi-squared tests; (demand Fisher’s exact test; Utilization of services side data) Basic care-seeking Secondary data: Medical Univariate logistic behavior records regression analysis; Multivariate logistic regression model Quantified delays (ordinary logistic regression); Multi-level logistic regression Provider models Providers by type, Primary data: Site visits/ Risk factors associated profiling with delays sector and level phone calls (supply Data visualization: side data) (flow figure, heatmap, Availability of TB lab Secondary data: Chinese Kaplan-Meier curves) Alignment between and radiographic National Health care visits and equipment Commission website availability of services 35 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Insights More than half of the participants were found to have experienced LDD, with a mean delay of 20 days. In Key takeaways from module 3 China, clients can choose any health­ care facility for initial care, regardless of its level. The clients’ choice ■ Place of initiating care (the right care level and in where they initiated care was an important facility with relevant equipment) is crucial for determinant of the occurrence of LDD; clients who minimizing delays and essential to examine in PPA initiated care in facilities at level L0 and L1 (about study design. Defining time intervals to examine one-third of clients) experienced more delays and enables researchers to assess what happens more frequent medical visits. TB-specific laboratory between PPA touchpoints in terms of timely equipment was often not available in L0-L1 facilities, progress and immediate improvements could be made by ■ EMR databases are a valuable asset for conduct­ improving diagnostic and treatment resources in these ing PPAs (the same concept can be applied facilities. The remaining two-thirds of clients initiated to other databases like disease registries and their care at L2-L3 facilities, where despite the availability sur­veillance data) of TB resources, utilization of these resources was ■ Surveys for primary data collection of care-seeking very low, and the diagnostic processes were non- behavior are a flexible tool that can incorporate standardized. This could be an effect of the centralization other dimensions such as the timing of care and of TB diagnosis and treatment in TB-designated hospitals potential delays for facilitating patient management and fee-waiver, which might have weakened awareness of TB among clinicians in non-TB-designated hospitals. Some of the clients were misdiagnosed with pneumonia instead of TB and started on antibiotic therapy, delaying correct diagnosis of TB. Considering that clients have the freedom to choose to seek their initial care at any level of facility, strengthening awareness around TB among clinical staff working in non-TB-designated hospitals can help reduce diagnostic delays. BOX 5. ADDITIONAL EXAMPLE ON DELAYS Delays can also happen with overarching changes The study provides a good example of using PPA in the health system. It is worth mentioning a study to examine delays in care (specifically HSD), using published by Chen et al. in 2019 entitled, “Patient secondary data [in this case, from the National TB and health care system characteristics are registry and the National Health Insurance Research associated with delayed treatment of tuberculosis Database (NHIRD)]. It is also a good example of cases in Taiwan.” This study sought to understand how using routine data systems can yield a large the effects of integrating TB services into the dataset for research purposes. Patient records in two general health services, instead of having a vertical databases were linked using a patient identifier, with TB care system. The study examined Health System a sample size of 10,932 patients. Delay (HSD), defined as the interval between the first Complete details on the methods can be found medical consultation for respiratory-related diseases here. and the initiation of TB treatment. 36 PART 3: Practical Examples of PPA Methods MODULE 4: HOW TO ASSESS HEALTHCARE COSTS INCURRED ALONG THE PATIENT PATHWAY The burden of health care costs Health care costs impose a considerable economic burden on patients, health systems, and societies at large. As life expectancy MODULE 4 – AT A GLANCE increases, there is corresponding increase in chronic disease, which ⊲ The costs of care-seeking shape has dramatic implications for health care and national budgets. patient journeys in multiple ways Patients with chronic or complex conditions have multiple touchpoints and are a main determinant with the health care system, with diagnosis and long-term treatment of provider choice, while also often requiring multiple and costly health care visits. NCDs frequently impacting the timing and frequency affect patients who are still working-age adults, leading to lower of care-seeking productivity and loss of skilled workers through absenteeism and ⊲ Information on direct costs early retirement. incurred by clients at each touchpoint with the health system can be added to surveys used for Costs for patients patient care-seeking (Module 1) Costs shouldered by patients are important to estimate as they and provider mapping (Module 2) may be a barrier to care-seeking and contribute to the ⊲ In-depth interviews or focus groups impoverishment of families. In LMICs, where there is significant out- with patients can be used to better of-pocket health care spending, catastrophic health expenditures understand patient perspectives and experiences of the economic are common even for essential services. Catastrophic health and social impacts of health care expenditure has been defined as out-of-pocket payments above the spending share of total household expenditure or non-food expenditure that forces households to sacrifice other basic needs, sell assets, incur debts, or become impoverished (Eze et al., 2022). The burden of out-of-pocket spending is inversely related to the economic status of households and can quickly lead to impoverishment. For example, in Kenya the poorest quintile spends 15% of their household budget on health care on average (Lehmann et al., 2020). This disproportionate burden of out- of-pocket medical costs further perpetuates the cycle of poverty as households can no longer afford necessities, potentially leading to more illness. Costs for health systems and societies Cost-effective health service delivery—both across the health system as a whole and within individual facilities—is important to support the provision of high-quality care to a growing NCD patient population. The first step to improving cost-effectiveness is understanding where and why spending happens along the patient journey, and what are its consequences. PPA can be used to detail costs incurred by patients (demand) and by the health delivery system (supply) along the care pathway. A brief introduction to costing within a PPA Cost data collection can be added to various other components of a PPA outlined in Modules 1–3. 37 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS BOX 6. DEFINITIONS OF COST TERMS For health service users For the health system Direct costs Costs directly incurred for health Costs incurred by the health care, such as medications, co-pays, system that are directly tests, or lab services. attributable to patient care (e.g., nursing services) Indirect costs Non-medical costs incurred such Costs that are not directly related as transportation or childcare, or to patient care (e.g., building rent, losses in productivity, as well as security, overhead costs) opportunity costs of spending on one thing instead of another. (for example, spending on medical care rather than a child’s education) Data collection methods Data related to cost can be either quantitative (e.g., surveys, expenditure tracking) or qualitative (e.g., in-depth interviews or focus group discussions). A literature review can help to identify existing cost data in the published or gray literature. Ideally, mixed-methods can be used to gain insight into research and policy questions related to cost. By combining quantitative and qualitative findings within a PPA, rich cost data is generated to better understand the economic and social impacts of health care costs on patients and how costs may influence care-seeking and other health-related decision-making. Patient self-report and recall bias Recall bias may affect patient’s ability to remember precise cost data, and the recall period for such data should be carefully considered and as short as possible to reliably capture costs incurred in care journeys. Additionally, patients recalling their expenditures may not provide full detail to understand the breakdown of patients’ spending on care. CASE STUDY EXAMPLES The cases presented below provide examples of studies that analyzed patient health expenditures along the care pathway through a longitudinal survey (Case Study 4) and through a case study approach using focus group discussions, in-depth interviews, and patient-reported retrospective cost information (Case Study 5). 38 PART 3: Practical Examples of PPA Methods CASE STUDY 4: Using surveys to determine patient costs over time for tuberculosis patients in Africa (South Africa, Mozambique, Tanzania, and The Gambia) – Evans et al. 2020 CASE STUDY 4 HIGHLIGHTS • Integrates costing questions into PPA surveys. • Collects cost data by patient journey touchpoints. • Adapts an existing tool for data collection for the PPA. Background The WHO Tuberculosis (TB) Patient Cost Surveys: A Handbook is a cost survey instrument that collects patient cost data from a single point in time (diagnosis, treatment, etc.). This data can be extrapolated to estimate the total costs incurred throughout the TB care journey and determine the proportion of patients experiencing catastrophic costs. By adapting this survey for longitudinal use, costs from the different phases of TB care were captured. The authors intended to adapt this tool to provide a longitudinal instrument that can be applied to an ongoing multi-country, multi- center cohort study (The TB Sequel Study) (Rachow et al., 2019). This study was conducted in four countries—South Africa, Mozambique, Tanzania, and The Gambia—to compare costs for TB patients over time and determine the proportion suffering from catastrophic costs. The surveys were adapted to capture appropriate direct and indirect costs based on the phase of disease and to reflect local context While this study illustrates a method used with an infectious disease, the longitudinal study design can be used for any PPA costing with repeat care contacts. Patients with chronic diseases will also have multiple touchpoints with the healthcare system, requiring an understanding of when and where along their health journey costs are incurred. Data collection The original WHO TB Patient Cost Survey was designed as a cross-sectional survey to collect cost data on the current treatment phase (at the time of interview) and later adapted for the longitudinal TB Sequel Study (Figure 10: the adapted WHO survey). The survey instrument collects costs across different spending categories, and the data can be analyzed to generate average costs and ranges of spending for different patient types and spending categories. The survey tool also allows for a cumulative cost estimate of the disease trajectory for each patient. The adapted tools are available to download at this link. The original survey instrument is designed to be administered once per study participant (after checking the TB treatment card for verification) and collects data on: 1. Socio-demographic status (e.g., age, gender, employment status, household composition, etc.) and economic data (e.g., individual and household income, household assets, household food security, etc.). 2. Direct costs: out-of-pocket medical expenses (e.g., consultation fee, laboratory tests, medication, etc.) and non-medical (e.g., food, accommodation, transport) payments. 39 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS FIGURE 10. ADAPTED WHO SURVEY TIME POINTS FOR LONGITUDINAL ASSESSMENT OF MULTIPLE TREATMENT TRACKS FOR DRUG-RESISTANT AND DRUG-SUSCEPTIBLE TB (EVANS ET AL., 2020) Long term follow-up Onset of Diagnosis Treatment Treatment symptoms start completed Intensive Continuation M0 M2 M6 M12/24/36 Drug-susceptible TB Pre-treatment (a) On-treatment (b) Post-treatment (c) Intensive Continuation M0 M2 M6 M12 M24/M36 Drug-resistant TB Pre-treatment (a) (short-course) On-treatment (d) Post-treatment (e) Intensive* Continuation M0 M2 M6 M12 M18 M24 M36 Drug-resistant TB Pre-treatment (a) (long-course) On-treatment (f) Post-treatment (g) *an intensive phase of 6–7 months is suggested for patients on regimens that contain amikacin or streptomycin. 3. Indirect costs (i.e., income loss or time cost as a result of TB), and guardian costs. Patients’ estimates of expenses for nutritional/food supplements, health insurance reimbursements, social assistance/protection (e.g., welfare, disability grants, cash transfers, etc.). 4. Opportunity costs: The patient is also asked about the financial consequences of TB and how they cope with these (e.g., dissaving, borrowing funds, selling assets to cover the costs of health-care expenditure) and the social consequences of TB. The adapted tool can be used for longitudinal assessment by surveying patients during multiple points of the TB treatment phase: at study enrollment (Day 0) and months 2, 6, 12, and 24 after enrollment. Patients were recruited based on their enrollment in the National Treatment Programme (NTP) at the study facilities. Paper-based questionnaires were used, and data was subsequently captured in an electronic database (OpenClinica®). The interview duration averaged 45 minutes, and the interviews were also carried out in the community, in addition to at facilities (the original tool was facility based only). 40 PART 3: Practical Examples of PPA Methods The adapted tool still collects data on direct and indirect medical costs, as well as contextual data on socio-economic status, income and food security. Costs were captured at the following stages of the patient pathway: 1. All patients receive questions on the costs incurred prior to diagnosis at the enrollment/ M0 visit. Pre-treatment costs were examined by asking patients to recall any out-of- pocket expenses or indirect costs incurred from the beginning of TB symptoms until the first day of TB treatment (Day 0 of enrollment of the study). 2. Patients received questions on the total costs incurred at each phase of TB treatment (Figure 10). FIGURE 11. KEY COMPONENTS OF METHODOLOGY USED IN CASE STUDY 4 Data category Data collection Analysis Insights Mapping Patient surveys with: Adapted Online tool for Cost per patient based Sociodemographic data the WHO TB patient cost surveys: costing on treatment outcome patient- a handbook, found at TB Sequel (OpenClinica®) Individual and/or pathway/ Care-seeking behaviour questionnaire resources. aggregate costs journey Statistical (demand- Economic impact of Cost of care-seeking analysis side data) TB costs This adapted tool can therefore estimate the costs for each TB patient by treatment phase, through the specific touchpoints based on the treatment algorithm. Most studies are limited to a single survey with a patient and retrospectively assess all significant costs incurred over time. The interviewer therefore needs to carefully support recall of care costs and expenditures incurred by the patient in the past. CASE STUDY 5: A case study approach using focus group discussions, in-depth interviews, and patient-reported cost information to explore the economic and social consequences of cancer in Kenya – Lehmann et al. 2020 CASE STUDY 5 HIGHLIGHTS • Qualitative analysis using focus group discussions and in-depth interviews with patients, family members, and expert key informants. • Basic analysis of patient-reported cancer-related medical and non-medical expenditures. Background Cancer is the second-leading cause of NCD mortality in Kenya. Most cases are diagnosed at a late stage, with poor prognosis and high out-of-pocket spending. In 2015/16, Kenya’s National Hospital Insurance Fund (NHIF) expanded the outpatient benefits package to include cancer care. In the May 2020 discussion paper Economic and Social Consequences of Cancer in Kenya: Case 41 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Studies of Selected Households, Lehmann and colleagues present a case study using qualitative methods to better understand direct and indirect costs faced by selected patients with cancer and their families, their decision-making, and the socioeconomic and psychological impact of having a family member with cancer. Figure 12 provides a brief snapshot of the key study components. FIGURE 12. KEY COMPONENTS OF METHODOLOGY USED IN CASE STUDY 5 Data category Data collection Analysis Insights Mapping Patient, family, and Thematic qualitative Social and economic impacts on the policymaker experiences data analysis using patients and their families patient- In-depth interviews NVivo Cancer-related medical pathway/ expenditure journey Cost calculation in Total, medical, and non-medical (demand- Cancer-related Microsoft Excel costs side data) non-medical expenditure Data collection Qualitative data Two exploratory focus group discussions—one in Nakuru county with seven participants and one in Kisumu county with nine participants—were conducted with a participants across a range of different backgrounds, geographic settings, and socioeconomic groups. Participants for the FGDs were recruited with the support of the Kenya Hospices and Palliative Care Association and the Kenyan Network of Cancer Organizations. Focus group discussions were used to explore emerging themes and refine selection criteria for in-depth interviews with case study households. Semi-structured in-depth interviews were conducted with eight experts on cancer policy, treatment, and financing and eight households. Eligible households for the study were identified and recruited with the support of health care providers, cancer support groups, and non-government organizations supporting cancer patients. Eligibility criteria for case study households included having a household member who currently has cancer, had survived cancer, or died of cancer. Households were purposively recruited using maximum variation sampling—a sampling strategy in which participants are selected to represent a wide range of variation in backgrounds (Moser, A., & Korstjens, 2017)—to represent a range of cancer types, disease progression, socioeconomic status, and NHIF membership. The study included both households that experienced a cancer episode before and after the expansion in the NHIF benefit package to better understand risks of impoverishment and the benefits of financial protection. Recruiters approached patients and household members to ask about their interest in participating in the study and verbally inform them about study objectives, methods, and potential risks and benefits. All households that agreed to participate signed written informed consent forms. 42 PART 3: Practical Examples of PPA Methods All interviews were conducted by study investigators in quiet and private settings, in either Kiswahili or English, audio-recorded, and transcribed in English. Quantitative data Retrospective cost information was captured on paper questionnaires and transferred to Microsoft Excel. Data analysis For qualitative data analysis, the software Nvivo was used to store and code interview transcripts. As for quantitative data, median, mean, and range of cancer-related total, medical, and non- medical costs—including transportation, accommodations, food, and interests on medical loans—were calculated. The proportion of total medical costs incurred by sub-category (e.g., diagnostic, consultation, medicines etc.) was also calculated. Results Qualitative data Challenges identified included lack of awareness and poor knowledge of cancer, late care-seeking, inadequate health insurance coverage and gaps in the benefit package limiting access to critical diagnosis and treatment, and socio-cultural barriers, including stigma, fear and myths that impede patients from seeking care early. Several key themes emerged on the social and economic consequences of cancer from the qualitative findings. Patients in this study were diagnosed at an advanced stage of disease, and most households experienced economic losses due to lost wages, sold assets, and in some cases, debt. All households described losses in income and significant medical and non-medical spending on cancer care. Patients reported additional spending from inaccurate diagnoses, multiple tests performed at different facilities, and the use of traditional healers. While none of the study participants were enrolled in the NHIF prior to the cancer diagnosis, several subsequently enrolled in coverage for a limited but costly set of interventions such as chemotherapy, radiotherapy, and surgery. Even households with NHIF coverage experienced significant out-of-pocket spending on items with more limited coverage, such as diagnostics and other drugs. Formal employment, asset ownership, and access to the NHIF were individual economic factors that influenced the patient’s cancer journey, among others. Quantitative data The two tables in Figure 13 below show descriptive statistics on cancer-related total costs in (on the left) and cancer-related total medical costs (on the right), respectively (Ksh and US$) (Lehmann et al., 2020). Itemization of expenses showed that medicines, which have limited coverage under the NHIF, accounted for more than one-third (36%) of total medical costs reported by these eight patients/ family members of patients, followed by chemotherapy/radiotherapy (24%) and surgery/ hospitalization (20%). 43 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS FIGURE 13. TOTAL CANCER-RELATED AND TOTAL MEDICAL COSTS (KSH AND US$) Cancer-related total costs Cancer-related total medical costs Insured (5) Uninsured (3) All (8*) Insured (5) Uninsured (3) All (8*) Median, Median, Ksh 260,480 128,950 215,990 Ksh 289,320 146,050 235,110 Median, Median, US$ 2,605 1,290 2,160 US$ 2,893 1,461 2,351 Average, Average, Ksh 569,826 192,917 428,485 Ksh 723,576 262,430 550,646 Average, Average, US$ 5,698 1,929 4,285 US$ 7,236 2,624 5,506 Min, Ksh 151,150 93,200 93,200 Min, Ksh 170,300 94,640 94,640 Min, US$ 1,512 932 932 Min, U$ 1,703 946 946 2,426,260 546,600 2,426,260 Max, Ksh 1,785,300 356,600 1,785,300 Max, Ksh Max, US$ 24,263 5,466 24,263 Max, US$ 17,853 3,566 17,853 FIGURE 14. ITEMIZED EXPENSES BY TOTAL MEDICAL COST (LEHMANN ET AL., 2020) Insights Based on the case study findings, the authors made several recommendations related to reducing financial barriers to care, health-seeking behavior, strengthening early screening and diagnosis in primary care, and generating evidence. Focusing on cost-related recommendations, the authors propose: (1) expanding the NHIF benefit package progressively to cover a broader range of interventions; (2) expanding NHIF coverage for informal sector workers; (3) identifying strategies to address non-medical costs associated with use of specialized services at referral facilities, such as transportation costs and accommodation; (4) enhancing knowledge of the NHIF benefit package and promoting early enrollment; and (5) conducting further research on the cost of care to inform improvements to the NHIF benefit package. This small study provides insight into high out-of- pocket spending on cancer care to impoverish patients and their families and suggests the need to generate more robust evidence on the costs incurred by patients and potential strategies to mitigate those costs 44 PART 3: Practical Examples of PPA Methods Estimating health system costs Costing at the health systems level can be done using Key takeaways from module 4 either top-down or bottom-up costing methods. In top- ■ Patients face considerable direct and indirect down costing, data are collected at the organizational healthcare costs that impact their decisions to budget or expenditure level. Activity-based costing, seek care. Low-income patients are especially a sub-set of top-down costing, estimates detailed costs vulnerable to the impact of medical and related for each specific activity to improve overall accuracy of non-medical costs. the costing exercise (Špacírová et  al., 2020). For example, activity-based costing can be done across a ■ PPA provides an opportunity to collect cost data well-defined clinical pathway (see Module 5 for CPW). from patients and care providers which can inform programmatic and policy changes on In bottom-up costing, data is collected on all resources issues such as user changes, cost exemption, used to treat an individual patient. This data can be and insurance cover collected using questionnaires or hospital records while following individual patients along their pathway ■ Mixed methods data can provide richer insight into (as seen in the example using the WHO TB Patient the social and economic impact of medical costs. Cost Surveys: A Handbook). PPA can be used to look at specific pathways in care delivery to a patient and determine the costs incurred at each step. There is a growing recognition of the need for more evidence-based design of patient pathways, with costs linked to exemption policies and insurance coverage for patients facing cost barriers to care-seeking. Module 5 provides additional information on how clinical pathway mapping can be used to measure costs generated along the client pathway within the health system. MODULE 5. HOW TO ASSESS PATIENT JOURNEYS AGAINST SPECIFIED CLINICAL PATHWAYS Clinical pathways for patient care Clinical pathways (CPW) represent a standardized, MODULE 5 – AT A GLANCE evidence-based path for patient care and the analytical focus here is on deviation from the CPW ⊲ A clinical pathway (CPW) represents in real-world care provision. An understanding of a standardized, evidence-based actual patient pathways can be useful in the path for patient care development or review of CPWs. This chapter ⊲ CPWs are part of the larger picture explains what CPWs are, how they have been of how people navigate the health assessed in relevant studies, and why they are part care system of the larger picture of how people navigate the ⊲ The analytical focus is on deviation from the CPW in actual patient care health care system. and some studies determine why CPWs are used all over the world for greater the deviation or variance occurred standardization of care processes to improve quality, ⊲ By assessing patients’ journeys, efficiency, and health outcomes, thus lowering more patient-centered CPWs can costs (Askari et al., 2021). There are a number of be developed definitions for CPWs, and the concept may differ 45 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS across stakeholders. Most commonly, the term CPWs is used to represent: the translation of clinical guidelines into practice through standardized diagnosis or treatment algorithms (stepwise, sequenced approaches) for specific conditions or distinct care episodes (Rotter et al., 2019) (see full definition, Box 7). BOX 7. DEFINITION OF CLINICAL PATHWAYS FROM A SYSTEMATIC REVIEW “Clinical pathways (CPWs) are tools used is used to translate guidelines or evidence to guide evidence-based healthcare. Their into local structures; (2) it details the steps aim is to translate clinical practice guideline in a course of treatment or care in a plan, recommendations into clinical processes pathway, algorithm, guideline, protocol or of care within the unique culture and other “inventory of actions;” and (3) it aims environment of a healthcare institution. A to standardize care for a specific clinical CPW is a structured multidisciplinary care problem, procedure or episode of healthcare plan with the following characteristics: (1) it in a specific population.” (Rotter et al., 2019) By translating clinical evidence to practice, the use of CPWs is intended to optimize patient outcomes and increase clinical efficiency, creating a standard of care that can be a reference for clinicians. CPWs have been widely used since the 1980s and increasingly in recent years (He et al., 2015). In 2012, the Chinese National Health and Family Planning Commission (NHFPC, previously called “Ministry of Health”) required every tertiary- and secondary-level hospital in China to implement at least 60 CPWs of the 400 established CPWs (although the hospitals may customize the pathways for their patients) (He et al., 2015). In Europe, CPWs are also referred to as care pathways, integrated care pathways, or care maps (Rotter et al., 2019; Askari et al., 2021; EPA, 2018). While the two terms (clinical pathways and patient pathways) are closely related, this guide will differentiate between the two. As previously described, in this guide patient pathway analysis means studying and mapping care-seeking journeys. Clinical pathways refer to standardized algorithmic paths that are the ideal standard of steps for the patient to progress (Rotter et al., 2019). CPWs are used for healthcare processes, define the different tasks that should be performed by providers, and are aimed at standardizing care to improve health outcomes and patient safety. CPWs range in scope from medication prescription to a comprehensive treatment plan. They are mostly used within health facilities, but some have been used for sequencing multi-disciplinary team steps outside of the hospital. At times, a clinical guideline or expert opinion of the standard of care may not exist and must be developed. The goal of a CPW is to sequence (with defined service steps) the care process, for instance, by the minute/hour in the emergency room, by the day in acute care visits, or by the diagnosis (as seen in surgical examples, over days in the hospital). This sequencing is similar to the touchpoints described in Part 1, but the goal of CPWs is to define pre-determined service steps or procedures 46 PART 3: Practical Examples of PPA Methods for healthcare staff along a patient’s pathway. Ideally, patients will progress along this CPW and obtain all the intended services within the specified time horizons. However, a patient’s pathway sometimes differs from the “ideal” predefined CPW. Every clinical scenario is personalized, and different patient or service delivery characteristics may lead to variations on the journey away from the predefined path. Understanding the deviation leads to valuable insights on potential challenges in providing high-quality, timely care. This is where patient pathway analysis comes in. THE CONNECTION WITH PPA METHODOLOGY Patient pathway analyses can be used to assess patient flow, similarly to process mapping, as detailed in Part 1. Process mapping refers to the identification of When used at a facility or network level instead of a different steps of a process taken to achieve a health system (macro) level, PPA can be used to test the particular aim. It is applicable to a wide range effectiveness (e.g., improvements in time, quality of care, of fields and is primarily intended to detail a process from start to finish and explore where or costs) of a specified CPW or to design a new CPW. improvements in efficiency can be made. It If a CPW is established, PPA can determine deviations therefore shares similarities with pathway analysis. from the specified CPW, thereby continuing to optimize the CPW to better align with patient preference and behaviors as well as quality of care issues. Steps in the CPW that are suboptimal can be identified and further understood, especially as they relate to health outcomes and patient safety. Finally, by assessing patients’ journeys, more patient-centered CPWs can be developed. Much of the reviewed PPA literature used process mapping (or mining) to examine patient flow (or pathways), which can be done in a variety of ways based on the data available. The analysis starts with a description of the actual pathways taken and is followed by a comparison (or deviation analysis). The next step is then to assess the causes of the observed deviation. This mirrors the core methods of PPA, which involve highlighting the steps of accessing care and how individuals progress through them. Methods to map a clinical pathway The broad analytical steps and where to find guidance include: ■ Characterize actual patient pathways experienced – this is presented in Module 1 of this guide. ■ Describe existing recommended CPWs (standard of care) / develop new CPWs – case studies in this chapter cover both the review of existing CPWs and the development of new CPWs ■ Analyze deviation/variance from recommended CPWs in practice – this is presented in this chapter through case studies which used a range of simple to advanced analytics. ■ Determine why the deviation/variance occurs – few studies formally include this step of analyzing the causes of deviation. 47 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS PPA methods can determine deviation from CPWs to inform: ■ Improvements of existing or development of new CPWs ■ Improved processes and determination of best practices ☐ Improved resource allocation, staffing, hospital flow, costs ☐ Reduced patient and physician time and delays ■ Improved patient outcomes (reduced complications, mortality, length of stay or hospitalization) ■ Improved client experience and and healthcare staff experience (care coordination, teamwork, reduction in health worker burn out) CASE STUDY EXAMPLES Case Study 6 in this module illustrates a great example of mapping two pathways within a clinic and comparing outcomes (costs and provider time). This can be done simply through process mapping and use of Excel. Case Study 7 demonstrates more advanced methods, using large datasets and machine learning. Case Study 8 also uses more advanced analytics to measure deviation from the CPW, including process mining and discrete event simulation. CASE STUDY 6: Developing a new clinical pathway, comparing costs and provider time – Adams et al., 2014 Mapping clinical pathways: CASE STUDY 6 HIGHLIGHTS • A simple method for mapping the clinical pathway (standard of care) and determining deviation/ variance (by comparing costs and time) to a newly proposed CPW. Background In “Mapping patient pathways and estimating resource use for point of care versus standard testing and treatment of chlamydia and gonorrhoea in genitourinary medicine clinics in the UK,” Adams et al. (2014) utilized CPWs to estimate and compare the cost of a proposed point of care (POC) nucleic acid amplification test (NAAT) to standard laboratory-based NAAT for chlamydia and gonorrhea in the United Kingdom (UK). The study developed a new CPW for POC NAAT based on the UK’s guidelines and clinical provider input. The study’s objective was to compare the new CPW to standard laboratory-based testing CPW. It also looked at costs and provider time along the pathway and uncovered that the new POC testing was less expensive, took less time than the clinic’s standard pathway, and reduced cost per patient by ∼ £16, as well as healthcare professionals’ time by ∼10 min per patient. The authors started by mapping both pathways within a sample of facilities and then compared costs and time associated with all pathways. Similarly to the CPWs discussed above, there was a standard of care that existed at each facility. However, in this case, that standard hadn’t been mapped or published. 48 PART 3: Practical Examples of PPA Methods Methods Two distinct tracks were mapped: 1. The current pathway of sexually transmitted disease testing 2. The new testing pathway (POC NAAT) at different sexual health clinics representing diverse models of care in the UK (New Clinical Pathway) The new pathway was developed based on expert opinion and was a result of workshops with the staff members of four sexual health clinics. To consider relevance to a greater range of services, clinics with diverse service delivery were selected. Differences in clinic costs and provider time were tracked for both scenarios. The selected clinics included a traditional genitourinary medicine (GUM) clinic, a fully integrated sexual health clinic with care services and contraception offered, and a clinic with a large proportion of patients from high-risk groups. A lead clinician was selected as a champion at each site and organized a workshop, encouraging participation from a wide range of staff. Workshops were attended by clinical and administrative staff (at a minimum, including one consultant, one nurse, one health advisor, and one administrator). During each workshop, open-ended questions were asked about the clinic’s current care pathways. A facilitator then proposed new patient pathways using a knowledge-based chlamydia/gonorrhea point of care test (POCT) example to prompt discussions at the workshop. The questions asked considered patient flow, time from test to treatment, and the total number of clinical steps or time which would be reduced by using POC NAAT. Figure 15 and Table 13 show the details of steps involved in the current and proposed testing/treatment pathways and the activities detailed for each step. FIGURE 15. STANDARD STEPS IN PATIENT TESTING IN ADAMS ET AL., 2014 Patient registration 1 Consultation 2 Clinical examination 3 Sample collection (blood, urine 4 and/or vaginal swab) Microscopic analysis of specimens in 8 On-site POC testing 7 Off-site laboratory- 6 based sample Health promotion counselling 5 the clinic processing Results management (data entry 9 Results 10 counselling Contacting patients who test positive to ensure 11 follow-up treatment and notifying patients of results) 49 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS TABLE 13. STEPS INVOLVED IN THE CURRENT AND PROPOSED TESTING/TREATMENT PATHWAYS As Primary As Additional Pathway Pathway Pathway Clinical steps Cost per Time Cost per Time patient (min) patient (min) A) Rapid sexual health screening for asymptomatic patients Asymptomatic Consultation / Sample collection → £62.16 21.2 £54.17 11.1 Self-service Off-site sample processing (1–2 weeks) (current) → Results management → Contact positives B) Current and proposed self-service pathways for asymptomatic patients Asymptomatic Prepare kits → Receive sample → £24.37 14.9 £23.70 14.2 Self-service Off-site sample processing (1–2 weeks) (current) → Results management → Contact positives Asymptomatic Prepare kits → Receive sample → £32.60 14.9 £31.94 14.2 Self-service POCT (90 min) → Results management POC → Contact positives (proposed) C) Gonorrhoea follow up visit for second line treatment after failure of initial treatment 2nd Exam → Treament → Health promotion £41.07 35.0 £33.97 25.0 Gonorrhoea / Partner notification treatment (current) The patients themselves were not recruited for this study as it was aimed at clinics and inter-facility patient flow of A complete list of the cost variables used in this sexually transmitted disease testing pathways, and the study can be found in the supplementary material published here. The categories outlined are drugs, analysis was done using a model built in Microsoft Excel consumables, pathology and staff. The appendix to replicate the flow in each CPW and calculate the costs also includes a list of definitions for activities incurred and provider time needed to fulfill each. identified for the CPW. Defining the CPWs was an iterative process. A follow-up workshop was convened to further refine the pathways generated at previous workshops, or create new pathways if care delivery varied significantly after learning from implementation. At the end of the study, the study group asked for consensus and final comments regarding the touchpoints for the CPWs. After the workshops, the study team outlined the pathways in an Excel spreadsheet, listing each step, activities within each step, and then summarizing time and costs required for each activity. 50 PART 3: Practical Examples of PPA Methods Analysis Estimated time and cost for the existing laboratory-based gonorrhea and chlamydia testing pathway were compared with expected time and cost for a new POC NAAT clinical pathway using absolute differences. No statistical analysis was completed to estimate uncertainty. The PPA of the diagnostic CPW found that: ■ the new POC NAAT pathway could save as much as 16 pounds ■ health care provider time could be reduced up to 10 minutes per patient The new POC NAAT testing could be implemented with confidence in the cost-savings for clinics and time-savings for providers. In this study (Case Study 6), a new CPW was developed and compared to the clinics’ typical standard of care. The new CPW was based on the UK’s GUM clinical guidelines. However, it was refined and developed through process mapping at the clinic level. This can be done with advanced analytics as well, using large data sets. Advanced methods to determine the clinical pathway and patient deviation Once a CPW is established, patient alignment to that CPW or their deviation can be studied through patient pathway analysis. Deviation or variation can be evaluated by mapping the client’s pathway compared to the CPW. Broadly speaking, this intends to describe how one flow of individuals in a CPW or any pathway is different from the expected flow; whether there is a deviation from the expected or pre- defined path. The measure of deviation can be thought There are certain analysis types that fit within of as variance analysis. the definition of variance analysis, such as: delta analysis method and gap analysis. Variance analysis is the quantitative difference between planned and actual behavior. In terms of ■ Delta analysis method is used in statistics patient pathways within the health system, variance is to determine variance around the mean. It seen at the individual level and at the group level can only be used in samples with normal (clustering) for certain sub-populations with similar distribution. However, the theory is similar characteristics. to variance analysis, and it is easily Despite the name of the method or the sector in which performed by comparing two pathways they are used, these are all forms of process mapping and determining the differences (either in and determining deviation from a planned/desired state. pathway variation or outcomes). Determining deviation can be done simply by comparing ■ Gap analysis aims to determine where a two pathways. Case Study 6 above is a great example process is from its goal state. In business of a simple process mapping model built in Microsoft process management, it is done by listing Excel to map the CPW touchpoints. Overall, the new out the organization’s current state, its POCT clinical pathway would save cost and time for the desired state, and determining the gap clinics in the UK. between these two states to correct With larger datasets, more advanced analytics such as course. Similarly, patient pathways can be process mapping can be used. As seen in the following compared to CPWs to determine the gaps case study process, mining can be used to study from the standard. deviation from the CPW. Phan et al. (2019) (below) uses 51 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS large electronic datasets (hospital databases) and process mining with simulation analysis to automatically generate CPWs for differing patient groups. Determine why the deviation occurred CPWs are highly focused on the physical journey a patient takes through the health system, utilizing mainly quantitative data. Quantitative or qualitative data may be analyzed to explore the individual’s experience at each touchpoint of their journey. Gaining the patient perspective through surveys (either paper-based or digital) as they traverse the CPW steps on their journey can bring insights into why delays or deviations occur. BOX 8. EXAMPLE OF ADVANCED ANALYTICS: PROCESS MINING Process mining to develop clinical pathways at the order “events” for TLH and 14,862 RCT events order-set level in South Korea – Cho et al., 2020 were extracted. Additionally, each event included Cho et al. (2020) compares a CPW developed by order attributes such as stages (e.g., pre- experts and a CPW developed by data mining. operation, operation, and post-operation) and The method, specifically called “matching rate- types (e.g., medicines and tests). based mining algorithm,” aims to produce the Put briefly, the matching rate is composed of the optimal set of clinical orders for specific timeframes application rate of orders in the CPW and the (each clinical stage) by employing matching rates. matched ratio of orders in the Electronic health This study used two different CPWs to validate record log data. Thus, it covers both how the the matching rate-based method for: (1) total orders included in the CPW are applied to the laparoscopic hysterectomy (TLH) and (2) rotator patients and how the orders used by the patients cuff tears (RCTs); utilizing two different datasets of are different from the CPW. The matching based- deidentified inpatient records from a tertiary hospital rate CP mining formula and detail algorithm can be in South Korea. They then compared “mined” found in the study. CPWs to knowledge-based models created by health experts. The schema for deriving CPWs and Once a new data-driven CP was developed comparing them with experts is seen in Figure 2. in this way, the team compared this matching based-rate CP to the knowledge-based pathway Data input included: 520 inpatient records and compared the number of orders between collected for patients who received the TLH the two pathways. Comparisons were both surgery from January 2012 to May 2014. Similarly, absolute (# of orders) and statistical, with t tests RCT inpatient data (a common disease managed performed. The data-driven pathway was found in orthopedics) from 360 patients between June to be superior. 2014 and 2015 were extracted to develop a data- driven CPW. To determine the CPW, 18,115 clinical Full study can be found here. 52 PART 3: Practical Examples of PPA Methods BOX 9. EXAMPLE OF ADVANCED ANALYTICS: PROCESS MINING AND DISCRETE EVENT SIMULATION Process mining and discrete event simulation method laparotomy surgery to occurrence or recurrence of to evaluate deviation from the clinical pathway incisional hernia (IH). Other studies have used similar and outcomes for obese vs. non-obese patients methods within a single hospital stay, from admission undergoing incisional hernia surgeries in France – to release (Günther & van der Aalst, 2007). Phan et al., 2019 Process mining was used to determine the standard Process mining and discrete-event simulation (DES) CPWs using medico-administrative data (ICD-10 was trialed as a method to determine CPW variance, codes, hospital touchpoints, outcomes, time and using a case study for incisional hernia (IH) surgeries costs). The figures shows the general methodology in France. The study compared outcomes of two used. Physician experts agreed upon diagnosis different incisional hernia surgery approaches (with related groups (DRG) for inclusion criteria. Data or without prophylactic mesh insertion) and between was collected from the French National Healthcare two different patient groups (obese and non-obese database, using these DRGs (found in ICD-10 and patients). This study focused on a macroscopic CPW – Common Classification of Medical Acts (CCAM) from first hospital stay to the most recent one, which codes). Data extraction consisted of querying the can last several years. It encompassed all events from database. Source: Phan et al., 2019 To be able to apply process mining, the original This study demonstrates an advanced case scenario data is converted into event logs (similar to defining for comparing two CPWs. Health outcomes, costs, and touchpoints for a PPA). Existing machine algorithms hospital visits varied between clinical interventions were used, such as an extension of Fuzzy Miner (on (mesh or no mesh) and between two patient groups which the Disco software is based) for the analysis. The (obese or not obese). Large datasets allows for resulting causal net was converted into a statechart studying and comparing different populations by model that can be simulated and analyzed. This is the characteristics. This allows for improvements in clinical discrete-event simulation model, which performs practice. “what-if scenarios” of new clinical pathway inputs. While process mining and DES are advanced methods, This approach creates a model that generates events this could be done more simply. For instance, a study for statistical analysis. The analysis (combining could include comparing health outcomes, costs, time, process mining and simulation) showed that the use of length of stay, or hospital/clinic visits for two different prophylactic mesh (at time of laparotomy) significantly patient groups by the process mapping technique decreases the cost of medical care and enhances the shown in the Adams case study. CPW by reducing the average number of times that patients come to the hospital after a laparotomy. Full study can be found here. 53 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Key takeaways from module 5 ■ Patient pathway analysis can determine the alignment or deviation of patients from the planned CPWs by assessing the actual pathway patients take. Analyzing CPWs is one component of a larger analysis for health system redesign – and it’s important to combine it with patient experience and perspectives. ■ Most of the literature reviewed for CPWs didn’t explore why patients take a different pathway than the standard. Assumptions are typically made based on the findings, but patient perspective is rarely integrated. Patient surveys can provide much needed evidence on the reasons why their pathway deviates from the set-out CPW and inform service redesign. ■ While most CPWs are developed for in clinic or in hospital use, the use of process mapping can be extended outside the hospital, if data is collected on patients prior to them entering the facility based CPW. Touchpoints can be generated in surveys or digital health app data, instead of hospital databases. Similar methods could be used with these inputs. ■ The future of health systems redesign can use process mining and discrete event simulation and modelling to determine health systems resource needs and patient outcomes based on redesign decisions. Not only will simulations help systems plan (as it does for infectious disease simulation and modelling in public health), but it can also make service organization more predictable in terms of costs and results (Barros, 2021) MODULE 6. HOW TO INTEGRATE GEOSPATIAL MAPPING INTO A PPA MODULE 6 – AT A GLANCE ⊲ Geospatial analysis considers the geographical distribution of the population and the facilities ⊲ Considering where individuals start to seek care (origin) and where they go to seek care at each step (destination) opens the door for additional analyses of different complexities, all with useful insights ⊲ Data collection for geospatial analysis can be integrated into the same data collection tools used for previous modules, minimizing effort and allowing for additional analyses after data collection has been completed ⊲ Simple origin-destination straight-line distance analysis (Euclidean distance) can be helpful as a basic assessment of movements in care-seeking journeys ⊲ Additional analyses that consider the build environment, travel routes, and durations can offer additional insights into geographic access and distance to reach care Geospatial analysis enabled by Geographic Information Systems (GIS) systems can play a pivotal role in advancing health research by providing a spatial dimension to the analysis of health-related data. While previous modules have covered the analysis of care-seeking behaviors, provider preference, costs and delays of care-seeking, this module focuses on the spatial or geographic accessibility of care providers. 54 PART 3: Practical Examples of PPA Methods GIS technology enables researchers to integrate and visualize geographic “Geographic access to health locations, allowing for a comprehensive understanding of the spatial care refers to the difficulty or distribution of diseases, healthcare resources, population settlements, ease in moving from a place transport axes, and environmental factors. Through spatial analysis, where a need for health researchers can identify patterns, clusters, and trends that may otherwise not services is triggered to where be apparent. GIS is instrumental in epidemiological studies, helping to map the spread of diseases, assess the impact of environmental factors on health the health service provider is outcomes, and plan targeted interventions. Additionally, GIS can examine located. It addresses the disparities in accessibility of care and inform the optimization of healthcare complex interactions between resource allocation by identifying underserved areas and exploring populations’ population distribution, spatial access to sites of care delivery, with the goal of improving aspects of location of services and how efficiency, coverage, and equity of health service delivery. people move to the health services.” How geospatial analysis can enhance PPA —OUMA ET AL., 2021 Geospatial analysis extends the approach of mapping the patient journey of care-seeking and provider profiling by adding physical geographic distribution and distances to care. The emphasis of geospatial analysis here shifts to mapping the physical location of where care-seeking journeys start (the location of patients’ residence or referring facility) and where they end (the locations of accessed care providers). This allows for quantifying distances to care, visualizing care journeys spatially and identifying gaps in geographic access to healthcare. Geospatial analysis can offer additional layers of insights, ranging from an appreciation of how far patients choose to travel to exploring geographic access causes of delays in care-seeking or non- use of care providers. Integrating geospatial analysis in a PPA is not essential for the success of a PPA, but there are benefits that a geospatial component (no matter how simple or advanced) can add to a PPA that make it worth considering: Location of patients: When mapping the patient journey, an additional geospatial component can help map the geographic distribution of patients. This involves analyzing the physical location of patients’ homes or neighborhoods (where household surveys are conducted) and therefore identifies the areas where patient journeys commence. Location of care providers: This augments provider profiles and location mapping with an emphasis on the spatial distribution of healthcare providers, enabling visualization of where patients travel to access health services. Additionally, it can offer spatial information on bypassing patterns within the healthcare system in which individuals may visit higher-level and more distant facilities rather than primary care facilities closer to home. Distance to care: Geospatial analysis can provide valuable information on travel distance and time for individuals to reach healthcare facilities. The added value of such information related to access to care also links to delays as usefully defined in the Three Delay framework (Shah et al., 2020). Geographic access relates to delays in the decision to seek care, and delays in arrival at a health facility due to the travel distance, mode and cost of transport, and barriers to travel like infrastructure conditions (quality of roads) and natural barriers (mountains, rivers, or floods). This also applies to the time intervals of examining delays in care mentioned in Module 3, especially when considering 55 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS the reality of referrals and changes of provider mid-care. Additionally, longer travel times, especially in regions with significant transport challenges, can be costly and contribute to non-direct healthcare expenses (Delamater et al., 2012; Guagliardo 2004; Neutens 2015; Noor 2005; Ouma et al., 2021). Data collection for geospatial analysis How data for geospatial analysis are acquired varies across studies. The term ‘geospatial’ indicates that the data has a geographic component in the form of coordinates, address, postcode, etc. Such data can be collected along with other supply-side and demand-side data, whether through primary data collection in the field using GPS equipment, maps and addresses or by accessing data repositories, like national level master facility lists with location data included in the form of geo-coordinates. The use of GPS-enabled tablets, enumerators or even drones allows data to be collected real-time in the field. If interviews are conducted at households, in the community or at health facilities, geospatial information can be collected and recorded at the same time. Given the high precision of geolocation, there is a risk of identifying patients if the geocoordinates of their Geomasking is a measure that geospatial data household are collected. Also, publicly available data collection should consider on ethical grounds. sources such as Microsoft Bing Maps and Google Maps It is a technique that intends to protect privacy and confidentiality by modifying a geographic provide structure-level geocoding as part of their online location in a dataset in an unpredictable or random mapping services, making it relatively easy to combine way. This can be done through Aggregation GIS data with address and telephone directories which (i.e., assigning everyone in an area to a single can identify individuals in a given dataset. This makes coordinate, normally a central point in the area) it essential to implement geomasking to protect and Displacement (i.e., altering the coordinates iden­tification of individuals. randomly within set parameters). Donut masking Routine health information systems are also a main is another technique that assigns households and individuals within a dataset to “donut” shaped source of individual-level data, including residence and clusters that center on a specific point in close demographic information, names of health facilities proximity to their original locations. (PHIA 2021) individuals used, and medical information. An example of a platform with geospatial components used in many countries is District Health Information System version 2 (DHIS2), which has the event tracker module for tracking individual-level information (HISP 2017). This can be used to explore facility utilization patterns. However, in many countries, such systems still face challenges of poor reporting rates, low coverage and inadequate infrastructure. Nonetheless, once fully established, this resource has the potential to be an important source of data for PPAs and by extension, defining health facility catchment areas (Alegana et al., 2021; Mumo et al., 2023). A simple method of geospatial analysis of patient journeys is using the Euclidean Distance, and this is illustrated by Case Study 1C below as part of the authors’ PPA in urban Bangladesh. It illustrates the value added to a PPA by including origin-to-destination analysis. More complex methods for geospatial analysis can be found in Ouma et al., 2021, linked here. 56 PART 3: Practical Examples of PPA Methods CASE STUDY 1C: Use of the Euclidean Distance method to enhance a PPA in urban Bangladesh – (World Bank, icddr,b) CASE STUDY 1C HIGHLIGHTS: • Illustrates a geospatial component and the value it can add to a PPA. • Shows how a geospatial component can be integrated with other data collection activities, adding little extra data collection burden to the study. • Showcasing Euclidean distance as method of geospatial analysis in PPA. This example showcases another component of the study presented in Modules 1 and 2 (Case Study 1A and Euclidean distance is the simplest distance-based 1B respectively) that describes data collection activities in analysis method. It assumes travel distance is a the Bangladesh PPA. Data gathered from the household straight line from origin to destination (“as the crow survey on patient location (GPS) and self-reported health flies”). This analysis only requires an “origin” (either patient residence, or referring health facility) and a facility utilization for MNH and NCD care was integrated “destination” (health facility the patient used). with the geolocation data of these same facilities. Using these GPS points, we were able to map the spatial FIGURE 16. SUMMARY OF STUDY DESIGN FOR CASE STUDY 1 WITH AN INTEGRATED GEOSPATIAL COMPONENT Data category Primary data Analysis Insights collection Quantitative: Sankeymatic.com for Sankey Mapping HH questionnaire Visualized flow through the Socio-demographic diagrams explaining the flow of touchpoints by level & patient- Basic care-seeking patients between touchpoints sector pathway/ behavior journey Self-reported delays Discrete Choice in care Experiment Stata 16 for descriptive Choice of provider per (demand analysis and statistical touchpoint side Qualitative: inference tests data) Contextualizing the care-seeking pathway In-depth interviews Challenges to accessing Stata 16 for conditional Identifying factors Key informant care and factors logit regression model influencing the choice interviews influencing choice of of healthcare provider Focus group care provider To further understand discussions community experiences Additional context on Coding in Excel for thematic quantified care-seeking of obtaining NCD care analysis and triangulation insights Provider mapping Providers by type, Categorized providers by and sector and level Phone calls, site type level and sector profiling Health services visits, surveys, and Stata 16 for descriptive analysis (supply and their availability questionnaires side Provider location Availability of relevant data) data health services Origin and destination Geospatial ArcGIS for geospatial analysis Visual map showing of care-seeking analysis and origin-to-destination movements and outliers Provider locations mapping of care-seeking 57 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS distribution of patients in respect to the various medical facilities, allowing for calculating Euclidean distance (“as the crow flies”) from patient residences to health facilities. Combining data from the household survey and provider profiling with GPS points enabled the plotting of straight-line journeys on maps to visualize patient movement. Figure 17 shows this for maternal clients in urban Bangladesh. FIGURE 17. ORIGIN-TO-DESTINATION MAPS FOR MNH FIRST VISIT IN DHAKA AND CHATTOGRAM, BANGLADESH MNH FIRST VISIT DHAKA CHATTOGRAM Legend Legend Patient Patient Facility Service Availability Facility Service Availability Yes (64) Yes (33) No (1) No (0) Unknown (24) Unknown (11) Patient to Facility (Availability) Patient to Facility (Availability) Yes (558) Yes (365) No (4) No (0) Unknown (34) Unknown (12) Outliers (4) Outliers (4) The figure visualizes origin-to-destination Euclidean distances from patient residence to facility for the respondents’ first MNH visit. Some respondents had used far-away health facilities outside of the study area which had not been included in the provider profiling. Some of these longer journeys are likely due to travel to rural home areas for the pregnancy period. The information content of such maps can be enhanced by considering results from provider profiling: ■ Services of mapped providers: Whether the providers a PPA respondent reports using actually have the sought service available or not (e.g., diabetes monitoring for a diagnosed diabetes patient) ■ Type of mapped provider: Level, sector and type of provider (e.g., tertiary level, private sector, diabetes hospital) 58 PART 3: Practical Examples of PPA Methods Accessibility metrics were defined for each type of visit (touchpoints) by using distance bands (Figure 18). Given that the study areas were urban, the distance bands used in this study were narrow (1-kilometer increments). Geographical proximity of the services was generally high, with the analysis showing that most patients accessed MNH and NCD services close to home, with only a few outliers travelling far outside of Dhaka and Chattogram. The median and range of Euclidean distances from patient residences to the most popular facilities reported by patients were calculated. Travel distance, together with the household survey data on provider preferences and the provider profiling data on key services provided, cost, and staffing, provided a more nuanced understanding of: (1) how far patients were traveling for specific kinds of services and (2) the geographic reach of the health facility. For example, Figure 18 below shows that more than half of patients traveled less than one kilometer for every measured touchpoint across the MNH care continuum. FIGURE 18. DISTANCE DECAY FOR DIFFERENT TOUCHPOINTS ACROSS THE MATERNAL AND NEWBORN CARE CONTINUUM Distance travelled from MNH patients’ home to maternal care providers Postnatal Care Child Birth Referral after Complication Complication Referral after Checkup Checkup Referral after First Visit First Visit 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 100.0% 1km 2km 3km 4km 5km 6–10km 11–15km 16km Euclidean distance has several limitations, including the assumption that travel happens in a straight line, without considering other factors such as transportation infrastructure, land use, and elevation that may influence actual travel distance and time (Delameter et al., 2012; Guagliardo 2004; Neutens 2015; Noor 2005). Below are two examples of types of analyses that are more complex and can consider additional variables. 59 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS CASE STUDY 7: USING NETWORK ANALYSIS TO DEFINE GEOGRAPHIC ACCESSIBILITY TO HEALTH CARE IN NAMIBIA AND HAITI (TANSLEY ET AL., 2015) This GIS-based network analysis uses actual trans­ was integrated with a population layer to estimate the portation routes to compute travel time or distance to proportion of the population with spatial access to each the nearest healthcare provider. It therefore employed level of service. The findings revealed significant spatial a more realistic approach to travel distance than the disparities in access to emergency services, particularly Euclidean distance. However, its utility in rural areas at lower service levels. may be limited, as patients’ transport routes do not The national road network data of Namibia and Haiti always align with the road network. Accurate data on were sourced from the Google Map Maker project, a transportation routes and populated locations can crowd-sourced mapping initiative. This project relies on also be challenging to obtain, making this method contributions from the public to create digitized maps computationally intensive. by utilizing aerial imagery. Through collaborative efforts, This analysis requires patient locations, locations of individuals contribute to the mapping of roads and other health services, transport infrastructure, transport geographic features, making the information publicly barriers, and contextual information on utilization. available for various applications, including research and In this example, a Geographic Information System analysis. The full published study can be found here. (GIS)-based network analysis method was utilized to The analysis provided accessibility metrics adjusted define population residence areas within 5, 10, and for road networks, an improvement over the Euclidean 50 kilometers for facilities capable of providing 24-hour distance methodology. Figure 19 presents a heat map care, higher-level resuscitative services, and tertiary demonstrating the percentage of the population with care in both Haiti and Namibia. Catchment area data access to care within 50 kilometers. FIGURE 19. PERCENTAGE OF THE POPULATION WITH SPATIAL ACCESS TO EMERGENCY CARE SPATIAL ACCESS TO EMERGENCY CARE IN NAMIBIA SPATIAL ACCESS TO EMERGENCY CARE IN HAITI Source: Tansley et al., 2015 60 CASE STUDY 8: ACCOUNTING FOR THE EFFECT OF RAINFALL ON TRAVEL ROUTES IN UGANDA (OUMA ET AL., 2021) This examples highlights how climate can be a potential referencing publicly available road network data from variable in geospatial analyses. Some countries (such OpenStreetMaps, Google Map Maker, and NASA’s as Nigeria, Niger, Mozambique and Uganda) experience Shuttle Radar Topography Mission at the USGS Land a rainy season that renders a lot of unpaved roads Processes Distributed Active Archive Center (LP DAAC) unpassable for part of the year. The authors used a website. Additional data on water bodies and canals method of analysis called cost-distance analyses. It and the presence of bridges in relation to these bodies involves defining travel speeds over different terrains and of water was included in the analysis. This was sourced using this to determine the least time needed to reach from Global Lakes and Wetlands Database. Full details healthcare facilities from various population locations. of the study can be found here, along with data on the It has become more attractive with the availability of population distribution and where it was sourced. datasets for travel time calculations. Nevertheless, it The study found that those who were not able to access produces theoretical travel times, and it does not account healthcare within one or within two hours—possibly due for competition among facilities. to the impact of rainfall on roads—was 2% and 0.6%, Key variables used for facility access in this study respectively, at the national level. The maps below show included facility names, unique identifiers, location travel time (in minutes) in Uganda to the nearest public data (GPS), facility type, ownership and operational health facility for (a) dry season and (b) wet season. status. The dataset used to extract this information was Travel time in minutes ranges from <30 minutes a facility list (published here) created through cross- (dark blue) to >120 minutes (dark red). FIGURE 20. HEAT MAP SHOWING TRAVEL TIME TO THE NEAREST PUBLIC HEALTH FACILITY FOR THE DRY (LEFT) AND WET (RIGHT) SEASONS (OUMA ET AL., 2021) Travel time in minutes A < 30 B 30 to < 60 60 to < 90 90 to < 120 > 120 0 25 50 100 150 200 Km Accounting for the seasonality of climate (such as importance of considering the role of climate rainfall, floods, hurricanes etc.) and other natural as a highly influential determinant of access to disasters is becoming increasingly relevant due to healthcare. An analysis that takes into account changing climate patterns. While this guide does not climate may explore climate-related risks of seeking go into the details of conducting such an analysis, care in specific contexts, including potential travel this example is intended to shed light on the hazards. 61 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Key takeaways from module 6 ■ Geospatial analysis extends the approach of mapping the patient journey of care-seeking and provider profiling by adding physical geographic distribution and distances to care ■ Geospatial analysis is not essential for the success of a PPA, but it can help the interpretation of PPA findings by revealing more about patients’ physical movements within the health system ■ The analysis can produce origin-to-destination maps for visualizing care-seeking and gaps in geographic access to healthcare, or specific patterns such as facility bypassing ■ Geospatial analyses may provide additional context on changing geographic access conditions and the seasonality of travel barriers ■ Data collection for geospatial analysis often reflects the complexity of the intended analysis and can be integrated into primary data collection tools and supplemented by existing location data or GPS-enabled technology to capture locations METHODS SPOTLIGHT: CARBON ACCOUNTING IN PPA PPA patient and provider data have diverse uses in health system analytics. As health systems endeavor to become more resource efficient and decarbonize care models and care pathways, new types of assessments are required. This has led to methods development for carbon cost accounting. Akin to traditional costing studies, accounting for carbon needs to consider what resources it takes to deliver healthcare to patients. Similar to the methods in Module  6, case study 9 purposefully collected data on where patients go for care with the aim to assess carbon emissions of their travel. Globally, such assessments have gained interest as health systems need to become more resource efficient and decarbonize, including care pathways and models. The following study leveraged PPA data to carry out an explorative analysis of greenhouse gas emissions from diabetes care in Bangladesh districts. 62 CASE STUDY 9: CARBON ACCOUNTING IN PPA - EXPLORING THE CLIMATE COST OF CARE PATHWAYS IN BANGLADESH While the PPA method described in this guide the pathway into Care Delivery Components (CDC). conceptually breaks down pathways by touchpoint, the These can be seen as modular blocks which, strung method we are describing here does something different together, constitute the totality of a care pathway as for the purpose of assessing greenhouse gas (GHG) illustrated in Figure 21. emissions of care delivery: It conceptually breaks down FIGURE 21. CARE DELIVERY COMPONENTS COMPRISING A CARE PATHWAY FOR REDRAWING CDC 3 CDC 4 CDC 5 CDC 6 CDC 1 GP visits + CDC 2 Travel + ED department + Hospital bed days + Surgical procedures + management Self = Care pathway Here, the focus of the analysis is not on people’s countries. The surge of NCDs in LMICs is likely to navigation and touchpoints with the care system as cause drastic increases in the carbon footprints in the PPA concept put forward in this guide. Rather, of these health systems. Given the increasing it focuses on what health care services a patient trends in both the global NCD burden as well as consumes over a certain period and then translates atmospheric concentration of GHGs, it is critical this service consumption into GHG emissions. to ask questions like, how much is care-seeking for different diseases contributing to emissions, As part of a broader health system assessment in and why; and if efforts to improve outcomes Bangladesh, an exploratory GHG emission analysis for people living with NCDs can come without was conducted.2 The conceptual thinking on CDCs additional GHG emission costs. Hence the was based on the “Care Pathways: Guidance on importance of carbon accounting in healthcare Appraising Sustainability”.3 The analysis focused on (or GHG accounting), defined as the process of type-2 diabetes mellitus (T2DM) patients and used measuring the carbon dioxide equivalents (CO2e ) a newly developed mathematical model. The CDCs resulting directly and indirectly from care delivery were translated into sets of questions, which were for a specific condition. incorporated into questionnaires of the broader assessment. The calculations for each CDC were carried out separately and then aggregated to provide The necessary data a final overall picture of the GHG emissions associated Data on care consumption was collected from with T2DM care in the Bangladesh context. multiple sources: a) Physicians completed the CDC ‘Doctor consultation’ questions for NCD patients; Why consider this extension to PPA? b) Administrators and technical staff completed the The healthcare sector is estimated to contribute CDCs on Emergency department visits, Hospital about 5% of total global GHG emissions, with the inpatient/bed days, and Surgical departments using majority coming from health systems that rely on invoices as supportive evidence; c) Patients completed hospital-based service delivery in high-income the CDCs on travel and diabetes self-management (continues) Unpublished at the time of writing this guidance. 2 Coalition for Sustainable Pharmaceuticals and Medical Devices (CSPM). 2015. 3 63 CASE STUDY 9: CARBON ACCOUNTING IN PPA - EXPLORING THE CLIMATE COST OF CARE PATHWAYS IN BANGLADESH (CONTINUED) during exit interviews. Additional assumptions on the engines for generators and vehicles, and anesthetic frequency and intensity of use of services by CDC gases and accelerants used in inhalers). These (not collected in the survey) were informed by data are mainly sources owned and controlled by the sourced in publications and expert consultations. facility. Scope 2 - indirect emissions calculated as These sources also provided input into assumptions consumption of purchased electrical power from on medication regimens, adherence to medication the national grid, applying the power generation and resource usage (e.g. syringe reuse among mix in the facilities’ geographical areas. Scope 3 – patients on insulin). Furthermore, facility data on indirect emissions embodied in consumed energy sources and usage, staff numbers and staff medicines, consumables, as well as patient travel, equipment and consumables used in each and provider travel. CDC, water usage, and waste management. Analysis In this study, the health state of people living with Finding and allocating Emission Factors T2DM was assumed to be the determinant of the The research team used data from the tools published intensity and frequency of care consumption. Health by GHG Protocol.4 Emissions factors (EF) are often states were considered according to glucose control not country specific, or specific to LMICs. However, (controlled vs. uncontrolled) and progression of they enable an estimation process of the emissions T2DM disease (complicated vs. uncomplicated), with linked to patient care. Some EFs are available in the controlled uncomplicated patients assigned to lowest GHG Protocol’s Global Warming Potential Values5, and expected consumption of care, and patients with the United States Environmental Protection Agency’s complications expected to consume higher levels of (EPA) GHG Emission Factors Hub6 which contains resources (insulins, hemodialysis, hospital admissions, etc). information on energy use and consumables such as Emissions were calculated by a bespoke mathematical medications. These EFs can either be activity based, model that combines the annual numbers of people living such as the EF of a vehicle used to travel to the with T2DM by diagnosis and health states, tabulates the hospital, or cost based, which assigns an EF to every respective annual frequencies of CDCs, and uses the unit dollar spent on consumables. A GHG/care pathway estimates of GHG emissions per care pathway activity to analysis might need to utilize both, as finding EFs for calculate the total GHG emissions. This total represents the consumables like medications can be difficult, in which estimated number of mega tons of CO2 equivalents arising case cost data would need to be collected as well from T2DM care in the Bangladesh study population. to assign respective EF levels. Another challenge is Combining such an analysis with PPA methodology the apportioning of general emissions like a facility’s enables estimation of GHG emissions arising from care energy consumption or water usage to a specific care consumption. The GHG/care pathway analysis explored event. To make this estimation, additional data on floor how better care models may impact GHG emissions area, duration of event, and direct interaction with staff despite important data limitation on, for instance, patient might be necessary. behavior, disease progression and EFs. Undertaking GHG emissions can be classified into three this type of analysis (and advancing the methodology groups; Scope 1 – direct emissions from an action accordingly) can ultimately help the decarbonization of or consumable involving the release of CO2 and health systems through better evidence on the climate other Kyoto Protocol gases (such as combustion cost of care pathways. Greenhouse Gas Protocol, Calculation Tools and Guidance. 4 5 Greenhouse Gas Protocol, Global Warming Potential Values (2016 update). 6 United States Environmental Protection Agency - GHG Emission Factors Hub. 64 PART 4: Reflecting on Experience in the Field The authors implemented three PPA studies between 2021–2024 with the aim of mapping patient pathways in a quantitative way, assessing care-seeking behaviors, patient experiences and provider perspectives, as well as dimensions of care quality. There were many insights regarding service delivery challenges, and opportunities for improvements through redesign of service delivery were identified. The three PPA studies were conducted as part of health system research informing the design and implementation of ‘Program for Results’ projects with World Bank financing in urban Bangladesh, Ghana and Gujarat State in India (Table 14). TABLE 14. SUMMARY OF COMPLETED PPA STUDIES BY THE AUTHORS Urban Bangladesh Ghana districts Gujarat State (India) Study Bangladesh is strengthening Ghana’s Network of Practice The State intends to redesign context public sector service (NoP) initiative intends to the PHC system due to under- provision in cities for NCDs, scale up nationally as a key achievement in RMNCAH+N environmental health and strategy for universal health and the high burden of NCDs, other priority issues coverage among other reasons Purpose To inform the design and To assess the effect of NoPs To inform the design of health implementation of the Urban on patient pathways and system intervention pilots HNP Project on how slum providers, and learn about as part of the SRESTHA-G populations in Dhaka and network configurations, Program for Results Chattogram access and use as part of government health services research on early NoP implementation Research ■ How do low-income MNH ■ What do maternal and ■ How do individuals in need questions and NCD patients navigate hypertension care of care navigate the care (ital: PPA) the health systems of pathways look like? system? Dhaka and Chattogram? ■ How do networks ■ What is the quality of ■ What are the care-seeking operate? essential healthcare in the behaviours and experiences ■ What do findings mean targeted geographic areas of of these patients? for the NoP roll-out and Gujarat? ■ Are there opportunities human-centred service ■ What are the significant for reorganising care design? gaps, constraints and and promising service ■ What are concrete actions contextual factors that hinder delivery models to address for improving service improvement? bottlenecks in accessing access, quality and ■ How does the community quality PHC? effectiveness through engage in healthcare NoPs? decision-making? 65 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Each of these PPAs focused on maternal and NCD care continua due to the importance of accessible and interlinked primary care services for these conditions. These care continua encompass preventative (ANC, NCD screening), diagnostic and treatment aspects, and, in some cases, acute and emergency care. Continuity of care is particularly important for maternal and cardiometabolic NCDs, as complications that arise must be followed and managed appropriately. These conditions are therefore well suited for pathway analysis which studies how people navigate part of a multi-tier health system. They can reveal a host of health system issues, such as bypassing the primary care level by patients and poor gatekeeping by facilities, access barriers to care, insufficient up- and down-referral, weak service decentralization, and others. The Ghana case study also illustrates PPA use for the evaluation of a system-level intervention which was at the early roll-out stage for future national coverage. BOX 10. A PRACTICAL APPLICATION: PPA AS A METHODOLOGY FOR EVALUATION OF A NATIONAL PRIORITY INITIATIVE IN GHANA, THE NETWORKS OF PRACTICE – (WORLD BANK AND IQVIA/GHANA PUBLIC SCHOOL FOR HEALTH/NTTDATA) To address the underperformance of its PHC system, the collection from participants—although it utilized semi- government of Ghana through its Ministry of Health and structured pathway interviews similarly to other PPAs, Health Service (GHS) have started to introduce Networks of the Ghana pathway interviews were carried out at facility Practice (NoP) among sub-district health facilities to improve exits, as well as within communities. There were: the quality of services, while also upgrading health centers ■ 780 pathway interviews at facility exits, to “Model Health Centers.” A network of practice (NoP) is a implemented at district hospitals (14% of sample), hub-and-spoke care model based around health centres, health centers (23%), clinics (21%), CHPS posts/ a highly scalable, efficient design with satellite care sites compounds (32%), and pharmacies/drug added as needed. The vision behind Ghana’s NoP model sellers (9%). of care is to establish a long-term PHC model and financing ■ 351 pathway interviews at the community level, system capable of delivering equitable, affordable and completed with MNH and NCD cases identified in high-quality PHC services by 2030 (GHS, 2024). markets (32%), bus stops (35%), lorry parks (11%), The Ghana PPA was a mixed methods study with secondary among traditional healers/practitioners (8%), parks data from routine systems integrated with primary data (6%) and restaurants etc. (7%). collected in the six districts from June 2023 to March 2024. Quantitative data was collected via pathway interviews and The shadowing of 156 MNH and HTN clients provided patient shadowing, while qualitative data came from focus insight into how clients navigated a health facility, how group discussions and key informant interviews with care services were delivered to them and how they perceived providers and managers. the care obtained. Although this study focuses on pathways for NCD and The study findings indicated that NoP implementation MNH care (similarly to Case Study 1), it is a unique has had multiple positive impacts including infrastructure application of PPA methods as it had a comparative development, service upgrades and insurance component (two matched district pairs) and a accreditation, as well as stronger collaboration and formative component (two urban districts) to inform coordination between facilities, knowledge-sharing and NoP strategy and roll-out. It also differs in the data trust-building. 66 PART 4: Reflecting on Experience in the Field BOX 10. A PRACTICAL APPLICATION: PPA AS A METHODOLOGY FOR EVALUATION OF A NATIONAL PRIORITY INITIATIVE IN GHANA, THE NETWORKS OF PRACTICE – (WORLD BANK AND IQVIA/GHANA PUBLIC SCHOOL FOR HEALTH/NTTDATA) (CONTINUED) NoP implementation also positively impacted patient delivery in the NoP district Hohoe and HTN treatment pathways by ‘moving the needle’ from service utilization initiation in the NoP district Dormaa Central were at hospitals to health centres. PPA was therefore used ascribed to referrals and the presence of a tertiary to evaluate the effect of NoPs on pathways. Although hospital in proximity. While this increased use of health hospitals were the preferred provider in NoP districts, centres cannot directly be linked to NoP implementation, health centres were more commonly used than in their when considered alongside other quantitative and paired non-NoP districts across most MNH and HTN qualitative findings, this suggests a positive impact of touchpoints (Figure 22). Minor exceptions noted for the NoP intervention on the decentralization of care. FIGURE 22. GHANA: EVALUATING THE NOP EFFECT ON HEALTH CENTRE USE THROUGH PPA Maternal care pathway: District 1st ANC visit Regular ANC Onset of labour Delivery PNC Dormaa Central (NoP) 31% 39% 35% 35% 30% Bono +4 % +11 % +15 % +13 % +15 % pair points points points points points Tain 27% 28% 20% 22% 25% Hohoe 48% 55% 33% 22% 33% (NoP) Volta +19 % +11 % +5 % –6 % +1 % pair points points points points points Ketu North 29% 44% 28% 28% 32% Hypertension care pathway: Treatment Treatment Treatment District 1st Contact Diagnosis initiation maintenance monitoring Dormaa Central (NoP) 27% 27% 24% 25% 30% Bono +6 % +6 % –1 % +1 % +5 % pair points points points points points Tain 21% 21% 16% 24% 25% Hohoe 22% 20% 18% 24% 40% (NoP) Volta +9 % +6 % +5 % +11 % +20 % pair points points points points points Ketu North 13% 14% 13% 13% 19% 67 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS The three applications have helped to test our PPA methodology and provided several insights and conclusions: 1. PPA methodology enables collection of longitudinal data on service use and can therefore complement cross-sectional datasets. Most data sources on care for MNH and NCD conditions – such as the Demographic and Health Surveys, Multiple Indicators Cluster Surveys and STEPS Surveys – provide important data on health service use but lack detail on service continuity given their cross-sectional nature. Longitudinal data is relatively scarce, yet information on continuity of care is critical for service delivery improvements and PHC redesign. Although our PPA methodology is not able to track all care-seeking across the maternal and NCD continuum, it records critical care contacts. Recall data on referrals were collected and analyzed, with results shown in the individual country reports (in urban Bangladesh the MNH and NCD referrals were also spatially mapped). 2. Identification of respondents in the right locations is crucial for the validity of pathway results. We preferred respondent enrollment at the household/community level; at the facility level we focused on a mix of facility types. For the PPAs enrolling respondents at the household level (Bangladesh, Gujarat), there was a necessary screening to identify eligible respondents. We applied carefully developed eligibility criteria across households in the sampling frame. In Bangladesh, a total of 17,923 households needed to be screened for eligible participants across 10 slum areas to interview a total of 5,829 MNH and NCD pathway respondents (number needed to screen = 4–5 households). In Gujarat, 22,679 households were screened across the study areas in four districts to deliver 2,212 pathway interviews. 3. Clear inclusion and exclusion criteria provide well-defined study populations for assessing their flow through parts of the health system. For MNH, we generally included pregnant women and women who had given birth in the past 12 or 24 months, but excluded respondents who had experienced stillbirths, miscarriages, pregnancy terminations, or neonatal deaths. For NCDs, individuals were eligible based on having been diagnosed with or receiving treatment for the NCDs of interest, or having a documented history of strokes, heart attacks or other heart diseases. For the chronic respiratory illness cases in Bangladesh, people needed a diagnosis of COPD, asthma, chronic cough or breathlessness since at least two years, and individuals with a diagnosis of tuberculosis were excluded. 4. A balance needs to be struck between recall capacity and data collected from individual participants. The pathway interviews collect retrospective data, and the studies demonstrated that respondents can recall their care journey with some level of detail. It makes sense that major health care events, such as ANC, care at onset of labour, child delivery and NCD diagnosis are well remembered by people. To maximize both the number of responses and their quality, we found it useful not to apply a recall window for basic touchpoint questions of who and where the care provider was, or whether there was any referral. In contrast, we did use a 1–2 year recall window for additional pathway questions, such as reasons for provider choice, barriers experienced, delays, medications and diagnostics prescribed at a touchpoint, as well as costs incurred. Respondents always had an answer option of “don’t know.” 5. The start of the care pathway needs to be defined in each case, using health cards and appropriate questions. While in some contexts (e.g., Ghana) the maternal pathway interview could easily ascertain whether the first pregnancy contact represented ANC1, this was less clear 68 PART 4: Reflecting on Experience in the Field elsewhere. By using mother and child health cards and questioning, the Bangladesh and Gujarat PPAs determined whether a provider visit represented ANC1 (e.g., Were you provided basic checks like blood test, BP measurement, weight measurement, urine tested, abdomen check? Were you provided iron and folic acid tablets? Were you provided counselling for pregnancy and related risks?). The start of the NCD pathway was generally easier to determine, and the focus was on the circumstances of first contact (such as part of an emergency visit, a routine vital signs check, community-based screening, etc.), which gives some insight into NCD case-finding strategies in the study locations. 6. The pathway interview can collect a rich set of data on diverse aspects of care seeking: ■ Health indicator data on the specific study population (e.g., slum dwellers) such as ANC1 and ANC4+ coverage, institutional and assisted delivery, and C-section. ■ Characteristics of respondents with specific outcomes, such as home delivery, referral, care- seeking delays, non-retention in care. ■ Data on travel mode and travel time to reach care, which can be used for geospatial mapping, to access metrics or carbon footprint estimation. ■ Data on provider preferences and barriers, as well as direct and indirect expenditure data for accessing care. ■ Data on most popular providers, including pharmacies, which can help prioritize capacity strengthening and partnership building activities. ■ Data on possible solutions as proposed by the respondents themselves, based on their care experiences. 7. Our literature review on PPA methods showed there are diverse approaches to tracing patient journeys and that research objectives and data availability determine the choice of PPA methodology. Some studies use routine data drawing on individual-level electronic health records, although these studies are mostly focused on narrower assessments of clinical pathways. Often, PPA does require new data collection and respective resources to undertake field work. In our case, we opted for retrospective pathways through recall by study respondents who are MNH and NCD clients. Prospective data collection on care journeys is also being implemented. The MNH e-Cohorts in Ethiopia, India, Kenya and South Africa offer a great example of an innovative PPA methodology collecting prospective data on user experience, processes of care, outcomes and health system quality. Women are enrolled at ANC1 and followed during pregnancy and postpartum until three months after delivery. The methodology combines in-person surveys at ANC1 and at endline, and repeated phone surveys in between with trained data collectors administering an interview. Similarly, to our PPA methodology, it establishes a sequence of events and gives insight into care continuity and timeliness of care. 8. Patient pathways are a useful and relevant measure for tracking PHC transformation activities. However, any PPA research needs to carefully review the strengths and weaknesses of the available methodological approaches and understand the resource needs for implementation. Given the high prevalence of mobile phone ownership and the willingness of most health system users to share their experiences, hybrid approaches using in-person and phone-based interviews, combined with any available routine service data, hold great promise for gaining a greater understanding of care-seeking journeys in changing health systems. 69 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS CONCLUSION The Patient Pathway Analytics guide provides an overview of PPA studies to inspire and explore the available study designs and methods to measure the patient journey. The guide is meant to be a companion for exploring possible methods and for inspiration. While PPA analyses share common goals, methods vary. There is no one way to do a PPA, and as methods evolve and become more diverse, there will be alternative ways to pursue the same aim of mapping patient journeys and better understanding their experiences within the healthcare system. The methods presented here are not the first of their kind, but they can be deployed in complementary ways to generate new insights. Above all, this guide seeks to support readers in exploring how patient pathway analysis can generate meaningful evidence to improve health system design and performance. The studies featured in the guide are non-exhaustive. They were chosen to offer examples and lessons from research carried out in different health systems. By examining the design options of different study types, implementers and decision-makers can incorporate valuable knowledge and experience into their own assessments. The selected studies encompass different world regions, target populations, health programs and implementation settings, underscoring the importance of tailoring research approaches to the specific contexts in which they are implemented. With a renewed focus on PHC and patient-centered care, the drive to meet UHC targets with constrained resources, the epidemiological transition, and emerging digital solutions, it is more important than ever to understand how patients navigate the health care system. The knowledge gained through patient pathway analytics can then be used to intentionally reorganize services to serve people better, address barriers to care utilization and harness facilitators. Using study findings on patient journeys and challenges with access to timely care, service delivery redesign can then plan where, when, and by whom healthcare services should be provided. Patient pathway studies can help health service planners, managers, and financial partners to “stand in the shoes of patients” and see the patient perspective. Conducting interviews with people with low health service use or poor care outcomes is especially important as it allows for the gathering of first-hand information about their experiences and perceptions regarding their healthcare journey. This can directly inform priority actions to improve patients’ care experience. Patients can also help us reimagine how we might deliver care in the future. By employing a combination of data sources and methods, healthcare professionals can gain a comprehensive understanding of patient pathways, enabling them to prioritize actions for greater access, quality and efficiency leading to better patient experiences. Through the creation of visual representations of patient journeys, stakeholders can be engaged in a different way to reflect about health service access and utilization. PPA findings can then directly support the policy dialogue on ‘right-place care’ given the improved understanding of how patients flow through the health system. The availability of this guide empowers different audiences in the development of patient pathway studies. By leveraging the experiences shared within this resource, stakeholders can identify and adapt suitable study designs and methodologies that align with their local needs and capacity for decision support, and use well-designed patient pathway studies to contribute to health system design, PHC restructuring and patient-centered service delivery models 70 References Aalst, W. (2011). Process Mining: Discovery, Conformance and Enhancement of Business Processes. In Media (Vol. 136). https://doi.org/10.1007/978-3-642-19345-3 Adams EJ, Ehrlich A, Turner KME, et al. (2014). Mapping patient pathways and estimating resource use for point of care versus standard testing and treatment of chlamydia and gonorrhoea in genitourinary medicine clinics in the UKBMJ Open, 4, e005322. doi: 10.1136/bmjopen-2014 -005322 Alegana, Victor A. et al. (2021). “Malaria Micro-Stratification Using Routine Surveillance Data in Western Kenya.” Malaria Journal, 20(1): 1–9. https://doi.org/10.1186/s12936-020-03529-6 Anbari, Z., Mohammadbeigi, A., Mohammadsalehi, N., & Ebrazeh, A. (2014). Health Expenditure and Catastrophic Costs for Inpatient- and Out-patient Care in Iran. International Journal of Preventive Medicine, 5(8), 1023–1028. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4258666/ Augusto, V., Xie, X., Prodel, M., Jouaneton, B., & Lamarsalle, L. (2016). Evaluation of discovered clinical pathways using process mining and joint agent-based discrete-event simulation. 2016 Winter Simulation Conference (WSC), Washington, DC, USA, 2016, pp. 2135–2146, doi: 10.1109/ WSC.2016.7822256. https://doi.org/10.1109/WSC.2016.7822256 Barros, S. (2021, April 12). Measuring the impact of clinical pathways. https://uphillhealth.com/blog/ measuring-the-impact-of-clinical-pathways Bermudez-Tamayo, C., Besançon, S., Johri, M., Assa, S., Brown, J. B., & Ramaiya, K. (2017). Direct and indirect costs of diabetes mellitus in Mali: A case-control study. PLOS ONE, 12(5), e0176128. https:// doi.org/10.1371/journal.pone.0176128 Bhosai, S. J., Sinthusake, T., Miwa, S., & Bradley, E. (2011). Factors affecting patient access in Thailand: understanding delay in care seeking for patients with cancer. Global Public Health, 6(4), 385–397. https://doi.org/10.1080/17441692.2010.508750 Chen, C. C., Chiang, C. Y., Pan, S. C., Wang, J. Y., & Lin, H. H. (2015). Health system delay among patients with tuberculosis in Taiwan: 2003–2010. BMC Infectious Diseases, 15, 491. https://doi.org/​ 10.1186/s12879-015-1228-x Cho, M., Kim, K., Lim, J., Baek, H., Kim, S., Hwang, H., Song, M., & Yoo, S. (2020). Developing Data- driven Clinical Pathways using Electronic Health Records: The Cases of Total Laparoscopic Hysterectomy and Rotator Cuff Tears. International Journal of Medical Informatics, 133, 104015. https://doi.org/10.1016/j.ijmedinf.2019.104015 Delamater, Paul L, Joseph P Messina, Ashton M Shortridge, & Sue C Grady. (2012). “Measuring Geographic Access to Health Care: Raster and Network-Based Methods.” International Journal of Health Geographics, 11(1): 15. http://www.ij-healthgeographics.com/content/11/1/15 Esgin, E., & Senkul, P. (2011). Delta Analysis: A Hybrid Quantitative Approach for Measuring Discrepancies between Business Process Models. In E. Corchado, M. Kurzyński, & M. Woźniak 71 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS (Eds.), Hybrid Artificial Intelligent Systems (pp. 296–304). Springer. https://doi.org/10.1007/978-​ 3-642-21219-2_38 Evans, D., van Rensburg, C., Govathson, C., Ivanova, O., Rieß, F., Siroka, A., Sillah, A. K., Ntinginya, N. E., Jani, I., Sathar, F., Rosen, S., Sanne, I., Rachow, A., & Lönnroth, K. (2020). Adaptation of WHO’s generic tuberculosis patient cost instrument for a longitudinal study in Africa. Global Health Action, 14(1), 1865625. https://doi.org/10.1080/16549716.2020.1865625 Eze, P., Lawani, L. O., Agu, U. J., & Acharya, Y. (2022). Catastrophic health expenditure in sub-Saharan Africa: Systematic review and meta-analysis. Bulletin of the World Health Organization, 100(5), 337–351J. https://doi.org/10.2471/BLT.21.287673 Fatima, R., Haq, M. U., Yaqoob, A., Mahmood, N., Ahmad, K. L., Osberg, M., Makayova, J., Hymoff, A., & Hanson, C. (2017). Delivering Patient-Centered Care in a Fragile State: Using Patient-Pathway Analysis to Understand Tuberculosis-Related Care Seeking in Pakistan. The Journal of Infectious Diseases, 216(suppl_7), S733–S739. https://doi.org/10.1093/infdis/jix380 Fatima, R., Qadeer, E., Enarson, D. A., Hinderaker, S. G., Harris, R., Yaqoob, A., & Bassili, A. (2014). Investigation of presumptive tuberculosis cases by private health providers: lessons learnt from a survey in Pakistan. Public Health Action, 4(2), 110–112. https://doi.org/10.5588/pha.14.0010 Fox MP, Pascoe SJ, Huber AN, et al. (2018). Assessing the impact of the National Department of Health’s National Adherence Guidelines for Chronic Diseases in South Africa using routinely collected data: a cluster-randomised evaluation. BMJ Open, 8, e019680. doi:10.1136/ bmjopen-2017-019680 Fraser-Hurt, N., Shubber, Z., & Ward, K. (2022). Improving Health Services and Redesigning Health Systems: Using Care Cascade Analytics to Identify Challenges and Solutions, Volume 1. Population- level Cascade Analytics. © World Bank, Washington, DC. http://hdl.handle.net/10986/36993 License: CC BY 3.0 IGO Garfin, C., Mantala, M., Yadav, R., Hanson, C. L., Osberg, M., Hymoff, A., & Makayova, J. (2017). Using Patient Pathway Analysis to Design Patient-centered Referral Networks for Diagnosis and Treatment of Tuberculosis: The Case of the Philippines. The Journal of Infectious Diseases, 216(suppl_7), S740–S747. https://doi.org/10.1093/infdis/jix391 Ghana Health Service. (2024). Implementation guidelines for networks of practice Ghosh, P. K., Kane, S., Kumar, S., Ali, Shayequa Z., Sahu, S., Sanyal, S., Chowdhury, Madhura, McPake, Barbara, & Mahal, Ajay. (2021). Healthcare seeking pathways in Uttar Pradesh and Odisha, India. NCAER. https://www.ncaer.org/publication/healthcare-seeking-pathways-in-uttar-pradesh-and -odisha-india Guagliardo, Mark F. (2004). “Spatial Accessibility of Primary Care: Concepts, Methods and Challenges.” International Journal of Health Geographics, 3(1), 3. http://www.ij-healthgeographics. com/content/3/1/3. Günther, C. W., & van der Aalst, W. M. P. (2007). Fuzzy Mining – Adaptive Process Simplification Based on Multi-perspective Metrics. In G. Alonso, P. Dadam, & M. Rosemann (Eds.), Business Process Management (Vol. 4714, pp. 328–343). Springer Berlin Heidelberg. https://doi.org/ 10.1007/978-3-540-75183-0_24 72 REFERENCES Hanson, C. L., Osberg, M., Brown, J., Durham, G., & Chin, D. P. (2017). Conducting Patient-Pathway Analysis to Inform Programming of Tuberculosis Services: Methods. The Journal of Infectious Diseases, 216(suppl_7), S679–S685. https://doi.org/10.1093/infdis/jix387 HISP. (2017). “District Health Information Software (DHIS2).” HISP website. https://www.dhis2.org/ inaction. International Diabetes Federation. (2021). IDF Diabetes Atlas | Tenth Edition. IDF. https:// diabetesatlas.org/ Kane, J., Landes, M., Carroll, C., Nolen, A., & Sodhi, S. (2017). A systematic review of primary care models for non-communicable disease interventions in Sub-Saharan Africa. BMC Family Practice, 18(1), 46. https://doi.org/10.1186/s12875-017-0613-5 Knaul, F. M., Bhadelia, A., Atun, R., & Frenk, J. (2015). Achieving Effective Universal Health Coverage And Diagonal Approaches To Care For Chronic Illnesses. Health Affairs (Project Hope), 34(9), 1514–1522. https://doi.org/10.1377/hlthaff.2015.0514 Kreindler, S. A. (2017). The three paradoxes of patient flow: An explanatory case study. BMC Health Services Research, 17, 481. https://doi.org/10.1186/s12913-017-2416-8 Kreindler, S., Aboud, Z., Hastings, S., Winters, S., Johnson, K., Mallinson, S., & Brierley, M. (2021). How Do Health Systems Address Patient Flow When Services Are Misaligned With Population Needs? A Qualitative Study. International Journal of Health Policy and Management, 11(8), 1362–1372. https://doi.org/10.34172/ijhpm.2021.36 Lehmann, J., Machira, Y., Schneidman, M., & Chuma, J. (2020). Economic and Social Consequences of Cancer in Kenya: Case Studies of Selected Households (Report No: AUS0001478). Washington, DC: World Bank. https://documents.worldbank.org/en/publication/documents-reports/ documentdetail/099930106092217126/P1638530bf035f0709da90b525153ad54c MacCarthy, S., Hoffmann, M., Ferguson, L., Nunn, A., Irvin, R., Bangsberg, D., Gruskin, S., & Dourado, I. (2015). The HIV care cascade: models, measures and moving forward. Journal of the International AIDS Society, 18(1), 19395. https://doi.org/10.7448/IAS.18.1.19395 Misra, S., Huddy, J., Hanna, G., & Oliver, N. (2017). Validation and regulation of point of care devices for medical applications (pp. 27–44). https://doi.org/10.1016/B978-0-08-100072-4.00002-2 Mitchell, I. (2021). Blockchain Medicine Administration Records (BMAR): Reflections and Modelling Blockchain with UML. In: Jahankhani, H., Jamal, A., Lawson, S. (Eds.), Cybersecurity, Privacy and Freedom Protection in the Connected World. Advanced Sciences and Technologies for Security Applications. Springer, Cham. https://doi.org/10.1007/978-3-030-68534-8_2 Moser, A., & Korstjens, I. (2017). Series: Practical guidance to qualitative research. Part 3: Sampling, data collection and analysis. European Journal of General Practice, 24(1), 9–18. https://doi.org/ 10.1080/13814788.2017.1375091 Mumo, Eda et al. (2023). “Geographic Accessibility and Hospital Competition for Emergency Blood Transfusion Services in Bungoma, Western Kenya.” International Journal of Health Geographics, 22(1), 6. https://doi.org/10.1186/s12942-023-00327-6. 73 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Neutens, Tijs. (2015). “Accessibility, Equity and Health Care: Review and Research Directions for Transport Geographers.” Journal of Transport Geography, 43, 14–27. http://linkinghub.elsevier.com/ retrieve/pii/S0966692315000034. Noor, Abdisalan Mohamed. (2005). “Developing Spatial Models of Health Service Access and Utilisation to Define Health Equity in Kenya By Kenya Medical Research Institute / Wellcome in Collaboration with The Department of Zoology, University of Oxford, United Kingdom. Thesis Submitted T.” (July). Osman, O. (2016). Meet the Kenyans too poor to afford cancer treatment. https://www.aljazeera.com/ features/2016/2/4/meet-the-kenyans-too-poor-to-afford-cancer-treatment Ouma, Paul, Peter M Macharia, Emelda Okiro, & Victor Alegana. (2021). “Methods of Measuring Spatial Accessibility to Health Care in Uganda.” In Practicing Health Geography: The African Context, ed. Prestige Tatenda Makanga. Cham: Springer International Publishing, 77–90. https:// doi.org/10.1007/978-3-030-63471-1_6. Population-based HIV Impact Assessment (PHIA) Geospatial Data Use Manual. New York, NY. April 2021 https://phia-data.icap.columbia.edu/ Rachow, A., Ivanova, O., Wallis, R., Charalambous, S., Jani, I., Bhatt, N., Kampmann, B., Sutherland, J., Ntinginya, N. E., Evans, D., Lönnroth, K., Niemann, S., Schaible, U. E., Geldmacher, C., Sanne, I., Hoelscher, M., & Churchyard, G. (2019). TB sequel: incidence, pathogenesis and risk factors of long-term medical and social sequelae of pulmonary TB - a study protocol. BMC Pulmonary Medicine, 19(1), 4. https://doi.org/10.1186/s12890-018-0777-3 Rojas, E., Munoz-Gama, J., Sepúlveda, M., & Capurro, D. (2016). Process mining in healthcare: A literature review. Journal of Biomedical Informatics, 61, 224–236. https://doi.org/10.1016/ j.jbi.2016.04.007 Rotter, T., Jong, R. B. de, Lacko, S. E., Ronellenfitsch, U., & Kinsman, L. (2019). Clinical pathways as a quality strategy. In Improving healthcare quality in Europe: Characteristics, effectiveness and implementation of different strategies [Internet]. European Observatory on Health Systems and Policies. https://www.ncbi.nlm.nih.gov/books/NBK549262/ Roux, J., Grimaud, O., & Leray, E. (2018). Use of state sequence analysis for care pathway analysis: The example of multiple sclerosis: Statistical Methods in Medical Research. https://doi.org/10.1177/ 0962280218772068 Shah, B., Krishnan, N., Kodish, S. R., Yenokyan, G., Fatema, K., Burhan Uddin, K., Rahman, A., & Razzak, J. (2020). Applying the Three Delays Model to understand emergency care seeking and delivery in rural Bangladesh: a qualitative study. BMJ Open, 10(12), e042690. https://doi.org/10.1136/ bmjopen-2020-042690 Shewade, H. D., Govindarajan, S., Thekkur, P., Palanivel, C., Muthaiah, M., Kumar, A. M., Gupta, V., Sharath, B. N., Tripathy, J. P., Vivekananda, K., & Roy, G. (2016). MDR-TB in Puducherry, India: reduction in attrition and turnaround time in the diagnosis and treatment pathway. Public Health Action, 6(4), 242–246. https://doi.org/10.5588/pha.16.0075 Špacírová, Z., Epstein, D., García-Mochón, L., Rovira, J., Olry de Labry Lima, A., & Espín, J. (2020). A general framework for classifying costing methods for economic evaluation of health care. 74 REFERENCES The European Journal of Health Economics, 21(4), 529–542. https://doi.org/10.1007/ s10198-019-01157-9 Srivastava, S., & Dubey, S. (2020). Knowledge-based systems in medical applications (pp. 189–215). https://doi.org/10.1016/B978-0-12-820604-1.00013-3 Statechart and Activity Diagrams. (n.d.). Retrieved November 30, 2022, from https://www.cs.sjsu.edu/ ~pearce/modules/lectures/uml/behavior/behavior.htm Tansley, G., Schuurman, N., Amram, O., & Yanchar, N. (2015). Spatial Access to Emergency Services in Low- and Middle-Income Countries: A GIS-Based Analysis. PloS One, 10(11), e0141113. https://doi.org/ 10.1371/journal.pone.0141113 Trends in maternal mortality 2000 to 2020: estimates by WHO, UNICEF, UNFPA, World Bank Group and UNDESA/Population Division. Geneva: World Health Organization. (2023). Licence: CC BY-NC-SA 3.0 IGO https://www.who.int/publications/i/item/9789240068759 Turner, H. C., Lauer, J. A., Tran, B. X., Teerawattananon, Y., & Jit, M. (2019). Adjusting for Inflation and Currency Changes Within Health Economic Studies. Value in Health, 22(9), 1026–1032. https:// doi.org/10.1016/j.jval.2019.03.021 van der Aalst, W. M. P., & Weijters, A. J. M. M. (2004). Process mining: A research agenda. Computers in Industry, 53(3), 231–244. https://doi.org/10.1016/j.compind.2003.10.001 Vesga JF, Hallett TB, Reid MJA, Sachdeva KS, Rao R, Khaparde S, et al. (2019). Assessing tuberculosis control priorities in high-burden settings: a modelling approach. Lancet Glob Health, 7, e585–e595. https://doi.org/10.1016/S2214-109X(19)30037-3 What is a state diagram? Knowledge Base. (n.d.). MicroTOOL. Retrieved November 30, 2022, from https://www.microtool.de/en/knowledge-base/what-is-a-state-diagram/ World Health Organization, & Organisation for Economic Co-operation and Development & International Bank for Reconstruction and Development. (2018). Delivering quality health services: A global imperative for universal health coverage. World Health Organization. https://apps.who.int/ iris/handle/10665/272465 World Health Organization, & UNICEF. (2018). A vision for primary health care in the 21st century: Towards universal health coverage and the Sustainable Development Goals (WHO/HIS/ SDS/2018.15). World Health Organization. https://apps.who.int/iris/handle/10665/328065 World Health Organization. (2017). Tuberculosis patient cost surveys: A handbook. World Health Organization. https://apps.who.int/iris/handle/10665/259701 Yang, Q., Tong, Y., Yin, X., Qiu, L., Sun, N., Zhao, Y., Li, D., Li, X., & Gong, Y. (2020). Delays in care seeking, diagnosis and treatment of patients with pulmonary tuberculosis in Hubei, China. International Health, 12(2), 101–106. https://doi.org/10.1093/inthealth/ihz036 Zhang L et al. (2021). Patient pathway analysis of tuberculosis diagnostic delay: a multicentre retrospective cohort study in China, Clinical Microbiology and Infection, https://doi.org/10.1016/ j.cmi.2020.12.031 Zhang, Y., & Padman, R. (2016). Data-Driven Clinical and Cost Pathways for Chronic Care Delivery. Am J Manag Care, 22(12), 816–820. https://www.ajmc.com/view/data-driven-clinical-and-cost -pathways-for-chronic-care-delivery 75 USING PATIENT PATHWAY ANALYTICS FOR PERSON-CENTERED HEALTH SYSTEM ASSESSMENTS Zhang, Y., Padman, R., & Patel, N. (2015). Paving the COWpath: Learning and visualizing clinical pathways from electronic health record data. Journal of Biomedical Informatics, 58, 186–197. https://doi.org/10.1016/j.jbi.2015.09.009 Zhang, Y., Padman, R., Wasserman, L., Patel, N., Teredesai, P., & Xie, Q. (2015). On Clinical Pathway Discovery from Electronic Health Record Data. IEEE Intelligent Systems, 30, 70–75. https:// doi.org/10.1109/MIS.2015.14 76