Qualitative assessment of the application of a discrete choice experiment with community health workers in Uganda: Aligning incentives with preferences

Abstract

Background: Maximizing the benefits of community health worker (CHW) programs requires strategies for improving motivation, performance, and retention. Discrete choice experiments (DCE) are increasingly used to inform policy response to health workforce shortages in rural areas, and may be of value in the context of CHW programs. Participants are presented with pairs of hypothetical jobs that are described by job attributes with varying levels and are asked what their preferred job is within each pair. Responses are then analyzed quantitatively to obtain information on what attributes are important to participants. We conducted a qualitative assessment to examine the appropriateness and validity of applying a DCE to a new population of CHWs with lower literacy. Methods: In 2011, we conducted a mixed-method study with CHWs in Uganda, consisting of 183 surveys and 43 in-depth interviews (IDIs). The DCE was administered to both survey and IDI participants. This article reports on the qualitative assessment of the implementation of the DCE. We compare DCE responses between survey and IDI participants to determine whether administering the DCE in a qualitative (IDI) context altered responses. We then present additional information collected on CHWs' decision-making processes and their experiences with the DCE in the IDIs. Results: Choices made by IDI participants were consistent with the choices made by survey participants. In-depth exploration of CHWs' observations in answering the DCE suggest that, overall, CHWs comprehended the DCE exercise and made reasoned choices. However, the data revealed some level of cognitive difficulty and highlighted some design and implementation challenges that are important to consider, particularly when applying a DCE to populations with lower literacy. These include the need to keep the number of attributes small; to choose levels that are realistic yet show sufficient range; and to clearly define attributes and their levels. Conclusion: DCEs can be an appropriate approach with CHWs but require careful design and implementation.