What Drives Poor Care for Child Diarrhea?: A Standardized Patient Experiment in India

Diarrhea is the second leading cause of death for children in low- and middle-income countries (LMICs). This is true despite the fact that nearly all such deaths could be prevented with a simple and inexpensive solution: oral rehydration salts (ORS). Private health care providers, who treat the majority of childhood illness in LMICs, are particularly unlikely to dispense ORS to children with diarrhea. Instead, providers often dispense antibiotics inappropriately. In order to improve prescribing practices, it is essential that we first have a good understanding of the key drivers of underprescription of ORS and overprescription of antibiotics in the private sector. In this study, we examine several leading explanations for poor quality of care for child diarrhea in the private sector. First, patient preferences for ORS alternatives (e.g., an antibiotic) could be driving underprescription of ORS. We will identify the causal effect of patient preferences by having anonymous standardized patients (SPs) pose as caretakers of children with diarrhea and express different (randomly assigned) preferences for treatment (ask for ORS, ask for antibiotics, or let provider decide). Second, private providers could be responding to financial incentives to sell more profitable alternatives to ORS (e.g., an antibiotic). To estimate the causal effect of financial incentives, we will instruct a subset of SPs to inform providers that they can get discounted treatments at a relative’s drug shop. This eliminates the provider’s financial incentive to recommend a given treatment and allows us to estimate the effect of such incentives. Finally, private providers might not directly distribute ORS or could have frequent stock-outs. To estimate the causal effect of stock-outs, we will randomly assign half of the providers to receive a three-month supply of ORS. This generates exogenous variation in stock outs and thus enables us to isolate the causal effect of stock outs on ORS and antibiotic prescribing. Combining, (a) causal estimates of the impact of each factor on prescribing, and (b) population estimates of the prevalence of each factor, will allow us to estimate the population level impact of implementing interventions that address each factor. The results of this study will inform the design of interventions aimed at increasing ORS dispensing and reducing antibiotic dispensing. If such interventions are targeted appropriately, millions of young lives could be saved.

External Link(s)

Registration Citation

Citation

Wagner, Zachary et al. 2022. "What Drives Poor Care for Child Diarrhea?: A Standardized Patient Experiment in India." AEA RCT Registry. July 08. https://doi.org/10.1257/rct.7276-2.1

Sponsors & Partners

Interventions

Intervention(s)

This study involves two experiments. In the standardized patient (SP) experiment we will randomly assign private providers (n=2000) to receive a visit from an anonymous SPs posing as a caretaker of 1 of 3 types: type-1 will request ORS to treat the diarrhea; type-2 will be uncertain and follow provider recommendations; type-3 will request an antibiotic. Comparison of these study arms will allow us to estimate the causal effect of patient preferences on ORS and antibiotic prescribing. To identify the causal effect of financial incentives, we will add a fourth type of SP that will be identical to type-2 but will inform providers that they can get discounted treatments at a relative’s drug shop. This eliminates the provider’s financial incentive to recommend a given treatment and thus enables us to isolate the causal effect of financial incentives on ORS and antibiotic prescribing.

In the second experiment we will randomly assign three-month's supply of ORS to half of the enrolled healthcare providers. Randomization will be clustered by village (roughly 400 villages) to avoid spillovers within a village. This generates exogenous variation in stock-outs and thus enables us to isolate the causal effect of stock-outs on ORS and antibiotic prescribing.

Intervention Start Date

2021-12-15

Intervention End Date

2022-04-15

Primary Outcomes

Primary Outcomes (end points)

ORS dispensation/prescription
Antibiotic dispensation/prescription

Primary Outcomes (explanation)

We will code an SP visit as having been dispensed/prescribed ORS if the provider either 1) gave the SP ORS directly or 2) prescribed ORS to be retrieved from a pharmacy. This is not conditional on other treatments

Secondary Outcomes

Secondary Outcomes (end points)

Zinc dispensation/prescription
ORS+Zinc dispensation/prescription
ORS without Antibiotic dispensation/prescription
ORS+Zinc without Antibiotic dispensation/prescription (Gold Standard)

Secondary Outcomes (explanation)

We will code an SP visit as having been dispensed/prescribed each of the secondary outcomes if the provider either 1) gave the SP an the secondary outcome directly or 2) prescribed the secondary outcome to be retrieved from a pharmacy.

Experimental Design

We will enroll 2,000 private providers that care for children with diarrhea (sampling details below). Most providers will be single provider establishments, the most common type of facility from which Indian patients seek care for child health services. Each provider will receive one visit from an actor posing as a caretaker for a child with a case of diarrhea (i.e., standardized patients or SP). We will randomly assign providers to receive a visit from one of four different SP-types. The types are designed to isolate the role of patient preferences from supply-side issues. Type-1 SPs will request ORS to treat the child’s diarrhea and will purchase ORS if available (unless the provider recommends against ORS) plus whatever else the provider recommends. Type-2 SPs will request antibiotics, and purchase antibiotics if available (unless the provider recommends against it) plus whatever else the provider recommends. Type-3 and Type-4 SPs will both have no treatment preferences and will ask the provider for their recommendation. Type-3 will purchase whatever treatments the provider recommends. However, Type-4 will inform the provider that they only want a treatment recommendation (i.e., they will not purchase any treatment) because they have an uncle with a drug shop where they can get drugs for free. They will claim the uncle is not very knowledgeable about treatment and they want a recommendation from a professional. This eliminates any influence of financial incentives in the provider’s recommendation for Type-4 SPs. All SPs (including Type-4) will pay any fees required to see the doctor. We will also vary the severity of the case of diarrhea presented by the SPs. This serves two purposes: 1) it limits the possibility that providers catch on to the experiment and 2) it gives our results more external validity, because our estimates will apply to a variety of cases instead of one specific case. Comparing ORS and antibiotic dispensing/prescription across the different types addresses several research questions.

1. Type-1 vs. Type-3: To what extent is ORS dispensing/prescribing sensitive to patient demand?
2. Type-2 vs. Type-3: To what extent does patient demand for antibiotics crowd out ORS dispensing/prescribing?
3. Type-3 vs. Type-4: To what extent is ORS dispensing/prescribing sensitive to financial incentives?

In addition to the SP experiment, we will layer on an experiment where we randomly assign increased supply of ORS to half of the enrolled providers. Randomization will be clustered by Gram Panchayat (a small group of villages) so that all providers in a village have the same assignment. This will avoid the potential for providers to give the free ORS away to other providers. This experiment builds on our previous work which highlighted lack of ORS availability in the private sector as a potential barrier to use. This work documented that only 35% of providers visited in Gujarat had ORS available on site. Moreover, according to a 2012 market analysis, 50% of 300 private providers surveyed in Uttar Pradesh had ORS stocked, 12% were out of stock, and 38% never had ORS stocked. Common reasons for not stocking ORS were low patient demand (35%) and low profit margin (15%). Our intervention is expected to create an exogenous increase in the share of providers that have ORS stocked. Comparison of the providers that received increase supply relative to the provider that did not will identify the intention-treat-effect of increased supply. We will distribute three month’s supply of ORS to providers assigned to received increased supply and we will ensure that all SP visits are conducted before this supply is expected to run out. Supply will be distributed free of charge. We will base supply quantities for each facility on diarrhea case load and provide enough ORS so that each case could receive two sachets of ORS (the recommended treatment quantity). All standardized patient visits will be conducted within three months of the roll-out of the supply intervention to ensure that facilities assigned to receive free ORS still have it in stock.

Experimental Design Details

Randomization Method

Randomization will be done by computer

Randomization Unit

The ORS supply intervention will be cluster randomized at the Gram Panchayat level (Gram Panchayat is a body governing a few villages). The SP experiment will be randomly assigned at the provider level.

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

260 Towns

Sample size: planned number of observations

2,000 healthcare providers

Sample size (or number of clusters) by treatment arms

ORS Supply intervention: 130 towns (1,000 providers) in treatment and 130 towns (1,000 providers) in control
SP Experiment: 500 providers will receive a visit from each of the four types

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

For our power calculations, we focus on detecting differences in ORS dispensing between the four SP-types. We will have 500 SP visits per SP-type. This allows us to detect a difference of 8.8 percentage points off of a base of 50% (power of 0.8). In other words, our sample size will allow us to detect effect sizes that are plausible and to rule out large effects if our results are not significant. We will have more power for our analysis of the impact of increasing ORS supply because there are only two arms instead of four. For this analysis, we will be able to detect a difference in ORS dispensing of 6.2 percentage points.

Supporting Documents and Materials

IRB