Quality of Care and Provider Decision-Making in Low- and Middle-Income Countries
Partner institution: Georgetown University
Funded by the 2024/25 LSE-Georgetown Joint Research Seed Fund
Start Date: 1 July 2025
End Date: 30 November 2026
Research team
- Jishnu Das, McCourt School of Public Policy, Georgetown
- Mylene Lagarde, Health Policy, LSE
- Harsh Gupta, Economics, LSE
Abstract
A well-established literature on variation in medical care in high-income countries has underlined that treatment decisions may reflect provider preferences more than patient needs. In low- and middle-income countries (LMICs), where the focus has traditionally been on improving access to care, our understanding of provider behavior and treatment choices remains limited. However, recent innovative studies, where trained mystery standardised patients (SPs) present identical cases to different providers, have begun to reveal important insights into clinical decision-making in LMICs.
Building on this progress, the proposed project will undertake two key tasks. First, it will collate and harmonise existing SP datasets from Georgetown and LSE researchers, creating the largest unified database of healthcare provider behaviour in LMICs, covering multiple countries, health conditions, and thousands of clinical interactions. Second, it will produce two major research outputs. The first is a descriptive analysis of how treatment strategies vary across providers and contexts, introducing the concept of a “clinical management strategy” as a set of actions taken by providers when faced with the same patient. This will highlight whether providers follow consistent protocols or adopt highly variable approaches, even at the diagnostic stage.
The second output will explore whether differences in provider behaviour reflect vertical differentiation (better vs. worse care) or horizontal differentiation (different but equally acceptable care). This distinction is critical for informing policy: vertical differences call for quality improvements, while horizontal differences suggest a need for better patient information and choice. The project will also contribute methodologically by applying machine learning tools, including natural language processing and similarity measures, to analyse complex treatment strategies, advancing the use of these techniques in health economics research.
Keywords: Quality of care, product differentiation, Health care markets
Activities
Funded through the LSE-NYU Joint Research Seed Fund, this project has involved a range of activities including:
- Producing a literature review on matrix completion, multi-task learning, transfer learning, and representation learning
- Developing a novel multi-task matrix completion framework that allows latent representations to vary across related tasks while retaining information-sharing benefits
- Establishing new statistical theory and error bounds using techniques from empirical processes, concentration inequalities, and random matrix theory
- Conducting simulation studies and computational experiments to compare the proposed methodology against existing single-task matrix completion approaches
- Applying methods to educational datasets, particularly student assessment and item-response data from online learning platforms
- Developing software tools and research code to support implementation and evaluation of the new methods
- Research collaboration visits between LSE and NYU investigators to advance theoretical development, address technical challenges, and plan future funding applications
- Presenting research findings at international conferences
- Preparing co-authored research publications and disseminating findings to academic audiences
Outputs
The activities above have contributed to the following outputs:
- A unified cross-country database of standardised-patient/provider-behaviour data
- A descriptive paper establishing new empirical benchmarks for treatment strategies and quality across provider types, countries and clinical cases
- A second technical/economics paper developing measures of differences in practice styles and assessing vertical and horizontal differentiation
- New analytical methods using machine learning, natural-language-processing approaches and similarity measures
- Exploration of future external funding opportunities
Outcomes
Recorded long-term outcomes, such as follow-on external funding and evidence of long-term collaboration, will be added here.