Provider Decision-Making and Incentives in Resource-Poor Settings

Abstract

My dissertation investigates the incentives and decisions of healthcare providers in resource-poor settings and their critical role in shaping patient outcomes. Specifically, it is comprised of three studies in which I investigated the following policy and methodological questions: the impacts of widespread malaria test distribution in sub-Saharan Africa, the application of machine learning to predict malaria risks and evaluate provider decisions, and salary delay as a systemic inefficiency that potentially influences health workers’ motivation and performance. This first study evaluates the impact of distributing malaria diagnostic tests on provider management of pediatric febrile illnesses and child mortality. Malaria is a major health concern in sub-Saharan Africa. Leveraging quasi-experimental design and unique data linking test distribution to survey data, I found that malaria test distribution was associated with modest improvements in child survival, likely due to better malaria management. However, it was also associated with increases in already very high rates of antibiotic use without clear linkage to health benefits. These findings highlight the need to consider spillover effects of disease-specific (vertical) programs and to develop accessible diagnostic testing for a broader range of child illnesses in resource-poor settings. The second study leverages machine learning with large survey data to assess providers’ diagnostic accuracy in malaria care from direct clinical observations. By developing a machine learning model that predicts individual children’s malaria status, I assessed the alignment of healthcare provider clinical decisions with patient risks. I found that providers’ clinical decisions generally matched the predicted malaria risk, though inconsistencies in diagnostic accuracy were still prevalent across different countries. Facility routine practice of diagnostic tests and use of private care was associated with improved diagnostic accuracy and better patient satisfaction, despite higher out-of-pocket costs. These findings highlight the potential of machine learning in enhancing our understanding of provider behavior in settings where reliable clinical data is sparse. In the third study, I turned to salary delay, a prevalent yet neglected issue affecting healthcare workers’ motivation and performance in low- and middle-income countries. Using facility survey data collected from multiple countries, I found that salary delay is a widespread issue among primary care workforce in the countries studied and was not only associated with reduced workers’ motivation and satisfaction, but more importantly, an increase in suboptimal behaviors including absenteeism and outside employment. The findings emphasizes that interventions should be tailored to the specific and personnel management structures within each health systems, illustrating how financial instability could significantly affect healthcare provider motivation and performance in resource-poor settings.