Distance, price and healthcare use in rural areas

Working alongside a local NGO, we will test how two key factors influence the demand for healthcare:
(1) distance— here we test the impact of improving physical access to healthcare. Working with the NGO, we will randomise the timing of the opening of new clinics in remote, underserved villages, located 1-2 hours away from the nearest existing clinic. The clinics set up by the NGO provide nurse-led primary care, offering treatment for around 30 common conditions, and basic diagnostic tests such as malaria rapid tests. Nurses can also deliver urgent treatments—such as injectable medications or intravenous fluids—before referring severe cases to higher-level facilities. Clinics maintain a reliable stock of medicines and conduct monthly quality audits to ensure consistent, safe care. Patients pay a small flat fee (approximately $2 or UgX5,000), which covers consultation, diagnostic tests, and treatment.

(2) cost—this second intervention tests the role of financial barriers. In communities where a clinic is opened, we will cross-randomise access to care for free for some target children. Half of the participating households will be randomly assigned to receive a card granting free consultations for one child under five for the duration of the study. This allows the study to measure how reducing out-of-pocket costs affects demand for care.

2025-07-01

2026-07-31

Probability of underuse of formal medical care by children under 5

Probability of underuse is calculated at the individual level as the ratio of the number of serious illness episodes for which care is recommended and sought over the total number of illness episodes for which care is recommended. To determine whether care is recommended in an illness episode, we will use the information about illness symptoms and apply WHO standards for community management of child illness.

Probability of overuse of healthcare for target child (seeking care during mild illness episode)
Medical expenditures (fees + drugs) for target child
Probability of using informal providers for target child
Health-related transport expenditures (eg spending on local transport to reach clinics) for target child
Formal and informal healthcare use for household members (conditional on being sick in past 4 weeks)
Proportion of household members testing positive for malaria at endline
Number of days of work (adult) or school (child) missed due to illness

The study uses a paired-matched cluster randomisation design. Communities identified as eligible by the NGO to receive a clinic will be grouped into pairs based on similarity in need and feasibility. Within each pair, one community is randomly assigned to receive a new clinic, while the other serves as a control. This produces 20 control communities and 20 treated communities where a clinic will open.

In the 20 clinic communities, a second randomisation is conducted at the household level: among the 20 study households per village, half are randomly allocated to receive free care for one child, while the remaining half pay the usual clinic fees. This creates three experimental groups: pure control communities (no clinic), households in clinic communities with standard priced care, and households in clinic communities receiving free care.

Not available

Randomisation is done by a computer programme (Stata)

We will have 2 levels of randomisation: Villages for intervention 1 (clinic opening) and individuals for intervention 2 (free card)

Yes

40 villages

800 individuals (20 in each village)

N=400 in the control arm, N=400 in treated villages (with clinics) ; within treated village, n=200 children will have access to free care, and the rest will not

● MDE of reduced distance: With 20 clusters per arm, assuming an intra-cluster correlation coefficient of 0.10 and that 80% of children have at least one serious illness episode over the duration of the study, we will have 16 individuals in control clusters and 8 in treated clusters, the trial is powered to detect a reduction of the probability of underuse by 18.6 percentage points from 50% in the control group (note that the proportion of 50% is chosen because it provides the most conservative detectable effect size; the MDE would be 16.9pp if the probability of underuse in control areas is 80%). ● MDE of reduced cost: In the treated clusters, we will follow a total of 400 households, half of whom will receive access to free care for their child for the duration of the study. Assuming that 80% of children have at least one serious illness episode over the duration of the study, the trial is powered to detect a reduction of 13.5 percentage points in the probability of underuse from 32% for households without free cards.