Insurance Terms and Access to Health Services: Experimental Evidence from Uganda

How to cite

Casaburi, Lorenzo; Cramer, Kim Fe; Willis, Jack. 2024. "Insurance Terms and Access to Health Services: Experimental Evidence from Uganda." AEA RCT Registry. AEARCTR-0012956. https://doi.org/10.1257/rct.12956-1.0

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

Lorenzo Casaburi* , Kim Fe Cramer , Jack Willis

* Primary investigator

Registry ID
AEARCTR-0012956
DOI
https://doi.org/10.1257/rct.12956-1.0
First registered
Last updated
Status
In development
Trial dates
2024-02-12 – 2026-01-01
Countries
Private
JEL codes
I12, I13, I15, O1
Registry record
http://www.socialscienceregistry.org/trials/12956

Abstract

In the absence of formal health insurance, low-income households often rely on costly coping strategies to finance health expenses, or simply forgo healthcare altogether. Health insurance can thus improve access to healthcare and reduce financial stress, but only if people buy it. In this trial, we collaborate with a health insurance provider in rural Uganda and study experimentally how premiums and co-payment levels affect insurance take-up, utilization of health services, and health outcomes.

Intervention

We partner with a local health insurance provider to study how premiums and co-payment levels affect insurance take-up, utilization of health services, and health outcomes. To obtain the study sample, the insurance provider first gathers information on groups in the geographical area of the hospitals. These groups were formed for purposes unrelated to our project, for instance savings groups. In a second step, the survey company contacts the group leaders to obtain a full list of group members. The survey company arranges in-person surveys with listed group members. In the baseline survey, we elicit participants’ valuation for health insurance at 3 different levels of co-payment ($0, $1, $2). Next, we randomly assign the study participants to receive either cash (equivalent to the normal insurance premium) or free insurance at one of the 3 different levels of co-payment ($0, $1, $2). Under the insurance plan, participants are required to pay a predetermined fee for outpatient services based on their specific co-payment level. Each insurance policy is valid for one year and extends coverage to a maximum of four household members. We will monitor administrative data on the participants collected by the hospitals, and we will conduct an endline survey after one year.

Experimental design

In the baseline survey, we elicit how much study participants value health insurance at 3 different levels of co-payment ($0, $1, $2). Specifically, for each copay level, we run a multiple price listing exercise with the choice between receiving money vs insurance for free. To ensure incentive compatibility, a small randomly selected subset of participants is assigned to receive one of their choices, chosen at random. These subjects do not form part of our subsequent experimental sample. The rest of the subjects are assigned at random to one of four treatment groups: T0, T1, T2, or T3. Those assigned to T0 will receive cash equivalent to the insurance premium. T1, T2, and T3 will receive the insurance product for free (i.e., with no premium cost) at three different copayment levels ($0, $1, $2).

Comparisons across treatment groups will identify the impact of insurance on the study outcomes and how it varies based on the copay (to gain power, for some of the analysis, we may pool various copayment groups together in the comparison between insurance and control). Comparing outcomes, both baseline and endline, by insurance valuation will allow us to study selection into insurance and how it varies by copay and by premium. Combining individuals’ valuations and usage of insurance (for a given copay) will allow us to simulate outcomes under different premium and copay combinations.

Design details

Randomization method
Randomization on a computer using Stata
Randomization unit
The main level of randomization is the household level. Additionally, to decide who in the household obtains insurance, we select up to two adults and two children. For the adults, the survey respondent and the household head are always selected. If the person responding to the survey is the household head and there are more than two additional adults in the household, then the remaining spot is filled with a remaining adult at random. For children, two children are selected at random out of the pool of children in the household. Outcomes for household members with insurance is of primary interest to us, but we will also look at spillovers to others in the household.
Clusters
Approx. 3,000 households.
Observations
Approx. 3,000 households.
Arms
T0: approx. 750 households, T1: approx. 750 households, T2: approx. 750 households, T3: approx. 750 households
Primary outcomes
Health insurance valuation (valuation of health insurance at different copay levels; correlation between baseline characteristics and valuation) Health care usage (visits, tests, drugs, expenditures (when applicable); at insured providers and at all providers; preventative and curative healthcare; chronic and non-chronic illnesses; inpatient and outpatient) Health status (chronic and non-chronic illnesses; subjective well-being, quality of life, and mental and behavioral health) Risk coping strategies (skipping meals; informal medical loans; support from friends and family; selling assets; labor supply; school attendance)
Secondary outcomes
Health insurance usage (successful or failed usage of insurance at hospital) Health care usage (vaccinations for children; childbirth) Health status (mortality) Economic well-being (income; assets; food consumption; finances (savings, loans))

Cited by

No citation data available for this registration yet.

Keywords