NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Community Health Care and COVID-19 Pandemic: Experimental evidence from Uganda
Initial registration date
May 21, 2020
May 21, 2020 1:34 PM EDT
Trinity College Dublin
Other Primary Investigator(s)
IIES, Stockholm University
Stockholm School of Economics
Stockholm School of Economics
Additional Trial Information
The limited capacity of the health systems in many low-income countries, especially in rural areas, suggests that the rapid spreading of the Covid-19 virus could have huge consequences in those areas. On top of the direct impact of the virus, a growing concern is that the shift in attention and resources towards the Covid-19 pandemic might crowd-out other essential care-seeking behavior and health services, leading to higher overall morbidity and mortality. This is a general concern, but may be a particularly acute in low-income countries given the higher incidence of deadly infectious diseases.
With this study we plan to do three things. First, by collecting novel data using mobile phone surveys we will document both the extent of (self-reported) incidence of Covid-19 and, importantly, the extent to which respondents adjust their health seeking behavior in response to the pandemic. This will allow us to estimate a more comprehensive measure of the impact of Covid-19 pandemic in rural Africa that embraces increases in morbidity and mortality from all conditions. Second, by exploiting the unique framework provided by an ongoing large-scale Randomized Controlled Trial (RCT), we will then test whether an innovative Community Health Worker (CHW) program can be effective in reducing this shift away from effective preventive and curative treatments, and possible misconceptions about Covid-19, cushioning the overall impact of the current pandemic. Finally, we will implement a field experiment, focusing on households in the treatment group of the larger trial mentioned above, where will test how different phone messages regarding Covid-19 and the importance of preventive and curative care more generally influence households health behavior and outcomes. The messages will have different behavioral framings (such as self-interest vs pro-sociality, profit motives vs community support etc) and we will be able to look at both demand (by focusing on the impact of messages sent to the households) and supply (by focusing on the impact of messages sent to CHWs) constraints. Registration Citation
Björkman-Nyqvist , Martina et al. 2020. "Community Health Care and COVID-19 Pandemic: Experimental evidence from Uganda." AEA RCT Registry. May 21.
See details reported in the "Experimental Design" section
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
1) prevalence of COVID-19 symptoms
2) non-Covid related health seeking behavior (vaccinations; malaria, diarrhea, pneumonia treatments; antenatal care; delivery)
3) All-cause Mortality
4) CHW activity levels
Primary Outcomes (explanation)
Secondary Outcomes (end points)
1) health knowledge
2) household interactions with CHWs
3) household self-reported covid-19 related behavior
Secondary Outcomes (explanation)
The study has multiple layers of randomization.
The study builds upon an on-going cluster randomized trial that began in 2016 with the objective of evaluating the scaling up of an incentivized community health workers (CHWs) program in rural Uganda. The program has already been found to be highly effective in improving basic health knowledge, health behavior, and health outcomes in a proof-of-concept study previously conducted by the research team (Björkman Nyqvist et al, 2019). The ongoing large-scale evaluation was designed to assess whether the program can retain its effectiveness once it runs at scale and to understand how other health actors might react to the scaling up. This study considers 250 villages that are part of the ongoing evaluation. The villages were randomly divided into a treatment group and a control group, using balanced randomization, back in 2016, at the onset of the main evaluation. Following randomization, at least one incentivized CHW was assigned to each one of the 125 villages in the treatment group. No incnetivized CHW was assigned to the 125 control villages. A rich baseline survey was collected in early 2016 from more than 6,250 households located across study villages (on average 25 households per village). Contact information of all participants was updated in 2018. We will now conduct a short phone survey with a representative sample of 8 households from each of the 250 village (for a total of 2,000 households), randomly selected from the original sample. The phone surveys will collect information on the households’ health behavior and outcomes; health knowledge; and beliefs related to the current pandemic. We will answer three main questions:
1) What are the direct and indirect effects of the Covid-19 pandemic on health seeking behavior and health outcomes in rural Uganda?
To answer this question we will use the new (and existing) survey data. These data will allow us to describe the change over time in health seeking behavior, both the medium run (2018 to 2020) and shorter run (in between survey rounds). This will primarily be descriptive work. 2) Does the presence of incentivized CHWs limits the negative impact of the pandemic (both in terms of direct and indirect effects)?
To answer question (2) we will use the survey data and exploit the experimental variation induced by the ongoing trial, whereby the incentivized CHW program was randomized across study villages. 3) Can messages about the importance of not reducing regular health-seeking behavior (preventive and curative care) be used to ensure that the Covid-19 pandemic is not crowding out the demand and/or supply of other essential health care? Do messages with different behavioral framings (such as self-interest vs pro-sociality, profit motives vs community support etc) have different impact? And, finally, do these messages help addressing demand (i.e. households) and/or supply (i.e. health workers) constraints?
To answer question (3) we will exploit new experimental variation in mobile text messaging. The messages intervention will cover the 125 treatment villages and it will be implemented by the NGO that is implementing the CHW program, as it has already the infrastructure in place to disseminate health messages among the CHWs as well as the communities they serve. The 125 villages will be randomly allocated in different groups that will receive different messages. While everyone will receive a health message, the content of the messages will be experimentally varied across different sub-groups. We will be able to speak to both demand and supply constraints, by focusing on the impact of different messages sent to the households and to the CHWs, respectively. We will also be able to complement the survey data with administrative data, collected on a regular basis by the NGO, on the activity of the CHWs in the study villages.
Experimental Design Details
Randomization done in office by a computer (using Stata software)
Randomization will always take place at the cluster (village) level.
Was the treatment clustered?
Sample size: planned number of clusters
Sample size: planned number of observations
~150 Community health Workers
Sample size (or number of clusters) by treatment arms
To answer question 2 we will compare outcomes from the 125 treatment villages (1,000 HHs) vs 125 control villages (1,000 HHs)
To answer question 3, we will divide the 125 treatment villages in different subgroups of 7 or 8 villages (56 or 64 HHs) each. Different versions of the messages will be sent to the households and / or CHWs within each group.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Assuming ICC=0.1, power=80%, significance level =5% the MDES is 0.16 SD for the comparison between T and C villages.
INSTITUTIONAL REVIEW BOARDS (IRBs)
Mildmay Uganda Research Ethics Committee (MUREC)
IRB Approval Date
IRB Approval Number