What drives poor quality of care for child diarrhea? Experimental evidence from India

Wagner, Zachary1 ; Mohanan, Manoj2 ; Zutshi, Rushil1 ; Mukherji, Arnab3 ; Sood, Neeraj4

Published Jan 25, 2024 on Dryad. https://doi.org/10.5061/dryad.3tx95x6nk

Data files

Jan 25, 2024 version files 4.91 MB

Abstract

Most healthcare providers in developing countries know that oral rehydration salts (ORS) is a lifesaving treatment for child diarrhea, yet few prescribe it. This know-do gap has puzzled experts for decades and has cost millions of lives. Using several randomized controlled trials among private providers in 253 towns in India, we estimated the extent to which ORS under-prescription is driven by financial incentives to sell more lucrative medicines, stock-outs of ORS, and provider perceptions that patients do not want ORS. We found that patients expressing a preference for ORS increased ORS prescribing by 27 percentage points. Eliminating stock-outs increased ORS provision by 6.8 percentage points. Eliminating financial incentives to sell medicines had no effect on average but increased ORS prescribing at pharmacies by 9 percentage points. Our findings, combined with patient exit surveys suggest that provider perceptions that patients do not want ORS explain 42% of under-prescribing, while stock-outs and financial incentives explain only 6% and 5% respectively.

README: What drives poor quality of care for child diarrhea? Experimental evidence from India

https://doi.org/10.5061/dryad.3tx95x6nk

These data were used for the paper "What drives poor quality of care for child diarrhea? Experimental evidence from India" by Wagner, Mohanan, Zutshi, Mukherji, and Sood. The data were collected in Bihar and Karnataka, India in 2022.

Description of the data and file structure

There are three different data sets included all in Stata (.dta) format:

1. ORS HH Main_noDCE: This is a survey of caretakers who had a child 10-years-old or younger with a case of diarrhea within 4 weeks prior to the survey. The key them of the survey is treatment seeking, what happened when they sought treatment, and what they did to treat their child's diarrhea. It also includes perceptions of and preferences for different child diarrhea treatments. This can be geo-linked to the other data sets via a town code but caretakers were not surveyed in every town for which the other surveys were conducted.

2. Provider Survey_noVig: This is a baseline provider survey that records information on provider characteristics and how providers treat child diarrhea. This survey was completed 2-3 weeks before the standardized

3. SP_Debrief_Provider: This is a standardized patient debrief survey that was completed within an hour of standardized patient visits that were presenting as fathers whose child had a case of diarrhea. This survey recorded information on what providers did during the standardized patient visits. Variables b57 to b66 are from a provider follow-up visit that occured after but on the same day as the standardized patients visit. This data set also includes all of the variables from the provider survey (merged in from Provider Survey_noVig) which label variables names starting with "ps_". The provider survey and standardized patient debrief can be linked with the variable capi_id.

These data were collected as part of a randomized controlled trial and provided free oral rehydration salts (ORS) for providers and varied the preferences of the standardized patients. The variable "treatment" indicates whether the provider was assigned free ORS and the variable sp_type_code indicates which SP role the provider received. All surveys used to create these data sets are including in this repository which can function as a codebook. Some of the variables from the surveys are missing from the publicly available data sets because of Dryad's de-identification criteria.

Code/Software

This repository includes the replication code for "What drives poor quality of care for child diarrhea? Experimental evidence from India" by Wagner, Mohanan, Zutshi, Mukherji, and Sood. Each Stata do file is labeled according to the table or figure from the paper that it produces. Regressions from Table 2 and S11 that included covariates are "commented out" from the code because Dryad's de-identification criteria restricted us from including them. The code for summary Tables S1, S2, S3, S6, S7, S22, and S23 also will not run because of missing variables.