This data contains individual patient data from a randomized intervention study collected at 60 CSComs (public health clinics) in Bamako, Mali. We enrolled all patients with acute symptoms related to malaria and surveyed them at the clinic. A random subsample were additionally interviewed at home. The data also contains doctor survey data from doctor trainings that were part of the intervention and from provider interviews collected at study endline in the participating CSComs.Finding ways to deliver high-quality health care to low-income populations in developing countries is a critical policy challenge. Our initial ESRC-funded project found that reducing user fees (by providing primary health care for free) does substantially increase Malian households' use of this care. However, we also find evidence that much of this care may be unnecessary or mis-targeted: our data suggest that children seeking care in government-run community clinics (CSCOMs) are frequently prescribed antimalarials and antibiotics when they do not need the treatment. This is particularly striking for malaria, since the Malian government has mandated that malaria diagnoses be confirmed by diagnostic testing. Our findings are consistent with a large body of economic literature, which on the one hand provides theoretical underpinnings for the problem of over-prescription and over-treatment, and on the other documents low levels of doctor effort and quality of care in both the public and private sectors across the developing world. Our implementing partner, Mali Health, has indicated that the increase in program costs due to over-prescription and the need for close monitoring and quality checks are a key barrier to scaling up the free-care intervention. We propose to conduct a follow-on project to identify the leading causes behind over-treatment, and test whether alternative incentive regimes can improve care outcomes without producing unnecessary costs. Our analytical framework is motivated by economic models of an "informed expert" selling "a credence good": the doctor has knowledge about the patient's illness and need for treatment that is not verifiable, and the patient must buy the treatment without knowing if it is truly what he or she needs. The model clarifies how doctor incentives, patient incentives, observability of diagnostic test results, and beliefs about test accuracy interact to produce care outcomes in this context. This analysis informs the design of a randomized controlled trial (RCT), which we will use to empirically test the model (as well as alternative theories for over-treatment) and identify promising strategies for improving care outcomes in the Malian public sector. Our primary application will be malaria, since high-quality, low-cost rapid diagnostic tests for the disease are readily available. However, given the striking rates of antibiotic use in our data, we also propose to use part of the new grant to conduct additional scoping work and expand the project to include bacterial illness if possible. The RCT will be conducted at 48 CSCOMs in the Bamako area and will allow us to evaluate the relative importance of test verifiability, provider beliefs about diagnostic test accuracy, and patient education about testing; provider incentives to diagnose and adhere to test results; and patient incentives to follow doctor advice and purchase medications. Over the course of the RCT we will construct a unique dataset that captures detailed information about patient demographic characteristics, symptoms, and treatment outcomes (tests and prescriptions given, medications purchased). We will also conduct home-based follow-up surveys to obtain information about patients' true malaria status, compliance with treatment, and provider satisfaction. This will allow us to estimate how alternative incentive and information regimes impact over-treatment and care outcomes in the public sector. We propose to forge a close collaboration with Malian health officials, to ensure that our project has maximal policy impact. Aside from its immediate relevance for the Malian public health system, this project will be of broad interest to researchers and policymakers working in the fields of economic development and public health.

At the outset of our study, field staff used administrative data for a list of all CSComs in the city of Bamako and in nearby Kati and Kalaban Coro in Koulikoro. After conducting a census of these CSComs, we dropped some CSComs that had closed, excluded CSComs that were more than 15 km away, and removed 21 CSComs that were working with a local NGO to offer subsidized and improved malaria care to patients – this yielded a final sample of 60 CSComs. Four care providers (doctors, nurses, and pharmacists) at each CSCom enrolled in the study were invited to attend a refresher training that covered Mali’s official malaria diagnosis and treatment guidelines. The basic refresher training included two sessions: one on Mali’s official diagnostic and treatment guidelines for malaria and one on how to administer an RDT. The training materials were prepared by the research team, and conducted by five trainers from the PNLP and one trainer from the regional health directorate (Direction Régionale de la Santé, or DRS) of Bamako. Doctor Information (Across-CSCom Randomization): Half the CSComs were randomly selected to receive the “Doctor Information” intervention. CSComs in this group received an enhanced refresher training that included the “basic information” referenced above and an additional session on the diagnostic accuracy of RDTs. Patient Information and ACT Subsidies (Within-CSCom Randomization): The other three experimental interventions were randomized within-CSCom across different days during a two-week observation period. The first intervention was designed to improve pa- tient and caregiver information about malaria treatment and diagnostic guidelines. The information was conveyed through a short narrative video, which depicted a mother taking her child to a CSCom for a suspected malaria case. The last two interventions involved distributing vouchers for free treatment for simple malaria (ACT tablets) at CSComs on selected days. In the “Patient Voucher” intervention, vouchers were distributed directly to patients when they first arrived at the CSCom. Patients and/or caregivers were informed that the voucher would pay for simple malaria treatment (ACT tablets) provided the doctor determined that this was the appropriate course of action (vouchers were not valid unless they received the doctor’s signature). Patients then went to consult the doctor and signed vouchers were processed at the CSCom pharmacy after the consultation was complete. In the “Doctor Voucher” intervention, the vouchers were instead left directly with the doctors, who could assign the vouchers to patients as the doctors saw fit. The field staff did not inform patients about voucher availability before the consultation. We divided the 60 CSComs into three 20-CSCom cohorts based on geography. Each of the three cohorts rotated through two weeks of data collection and experimental intervention. Field staff visited each CSCom six times over the two week observation period. Although all CSComs were informed of the upcoming study activities and interventions in advance, CSCom staff did not receive prior notice of the actual intervention schedule – rather, study staff informed them of the day’s intervention on the morning of an observation day. On observation days, two “survey team” enumerators were tasked with recording the details of each clinic visit for an acute illness. On all days except control days, we also stationed an “intervention officer” at the CSCom, who was charged with implementing the interventions (e.g. showing patients/caregivers the informational video, distributing vouchers to doctors or patients, and verifying all voucher redemptions). The survey team and intervention officers were stationed at different parts of the CSCom and intervention eligibility was not tied to survey consent. Patient Surveys: The survey team stationed at the CSCom was charged with recording the details of each acutely ill patient visiting the CSCom for care. We classified a patient as “acutely ill” if they were visiting the CSCom because they were feeling sick and exhibited any of the following symptoms: fever, chills, excessive sweating, nausea, vomiting, diarrhea, poor appetite, headache, cough, weakness, fatigue, or reduced consciousness. When the patient arrived at the clinic, the survey team collected basic demographic details, symptoms, and information on any prior treatment and/or diagnosis. After the patient’s consultation with the doctor was complete, the survey team recorded details of all blood tests performed and medications prescribed, as well as fees paid to the CSCom. We randomly selected a subset of patients for a more detailed home-based follow up survey, which was conducted the day after their CSCom visit. The home survey collected information on changes in the illness and any treatment and tests obtains after the CSCom visit. In addition, enumerators asked to perform an RDT on the patient. Doctor Surveys: We use data collected from health care providers at two points in time. First, we administered a post-training survey to doctors and other care providers who attended the refresher trainings that took place at the beginning of the study. The post-training survey tested providers’ knowledge of topics covered in the basic training (e.g. recommended malaria treatments, symptoms of severe malaria) and topics only covered in the extended “doctor information” treatment (e.g. sensitivity and specificity of RDTs). We also selected up to three care providers for a post-intervention endline survey. In addition to topics covered in the post-training survey, the endline asked caregivers about perceived patient knowledge and demand for drugs and personal preferences regarding malaria diagnosis and treatment.