Quality signals in health care: a vignette experiment in South Africa

How to cite

Stacey, Nicholas; Lagarde, Mylene; Blaauw, Duane. 2021. "Quality signals in health care: a vignette experiment in South Africa." AEA RCT Registry. AEARCTR-0008443. https://doi.org/10.1257/rct.8443-1.2000000000000002

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Nicholas Stacey* , Mylene Lagarde , Duane Blaauw

* Primary investigator

Registry ID
AEARCTR-0008443
DOI
https://doi.org/10.1257/rct.8443-1.2000000000000002
First registered
Last updated
Status
In development
Trial dates
2021-10-26 – 2022-02-28
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/8443

Abstract

Overtreatment is widespread in health care, including in low- and middle-income settings. In primary care, providers often recommend unnecessary drugs, with potential dire consequences for patients, health systems and public health. Recent evidence suggests that providers recommend unnecessary treatment even in the absence of direct financial incentives and when they know what the appropriate treatment is. Providers have sometimes explained their decisions to overtreat as a response to an actual or perceived demand from patients, who, in a relationship characterized by asymmetry of information, may interpret treatment characteristics as a signal of quality. We design a simple vignette experiment to test (1) whether reduced overtreatment is perceived as a signal of worse quality of care and (2) the relevance of the treatment signal against another typical signal of quality care (thoroughness of the consultation).

Intervention

In a survey experiment respondents see and rate one clinical vignette depicting the story of a patient suffering from a minor lower-respiratory tract infection (acute bronchitis) who decides to consult a doctor. Our design cross-randomizes variation in the type of treatment recommended by the doctor with information about the thoroughness of the consultation, in a 5x2 design. Along the treatment dimension, each respondent sees one of five possible treatment outcomes:
- No drugs prescribed
- Two Over-the-Counter (OTC) drugs prescribed
- Four OTC drugs prescribed
- Two drugs prescribed: one prescription drug (antibiotic) and one OTC drug.
- Four drugs prescribed: two prescription drugs (antibiotic and steroids) and two OTC drugs.

Along the second dimension, we randomize the amount of information about the technical quality of the consultation process. For respondents assigned to receiving some information about the consultation, the vignette includes a summary of the questions asked by the providers and physical examinations undertaken.

Experimental design

This survey experiment is embedded in a broader survey looking at provider preferences and the demand for healthcare. Survey respondents are part of an online panel in South Africa, aged 18-65, who do not have a medical health insurance. Respondents see and rate one clinical vignette depicting the story of a patient suffering from a minor lower-respiratory tract infection (acute bronchitis) who decides to consult a doctor. The design cross-randomizes variation in the type of treatment recommended by the doctor with information about the thoroughness of the consultation, in a 5x2 design. Along the treatment dimension, each respondent sees one of five possible treatment outcomes:
- No drugs prescribed
- Two Over-the-Counter (OTC) drugs prescribed
- Four OTC drugs prescribed
- Two drugs prescribed: one prescription drug (antibiotic) and one OTC drug.
- Four drugs prescribed: two prescription drugs (antibiotic and steroids) and two OTC drugs.
Along the second dimension, we randomize the amount of information about the technical quality of the consultation process. For respondents assigned to receiving some information about the consultation, the vignette includes a summary of the questions asked by the providers and physical examinations undertaken.

Design details

Randomization method
Survey platform random number generator.
Randomization unit
Individual
Clusters
1800
Observations
1800
Arms
180 individuals per clinical vignette presentation.
Primary outcomes
The main outcomes of interest are the first-order beliefs of respondent, judging different aspects of quality of care. The main outcome is a rating of the provider’s competence on a scale of 1-4. Other outcomes of interest include rating of the effectiveness of care received, the treatment safety, and satisfaction with the provider (likelihood to us the same doctor again).
Secondary outcomes
Secondary outcomes include second-order beliefs of the same four dimensions quality of care, which are incentivised.

Cited by

No citation data available for this registration yet.

Keywords