Published In

Plos Medicine

Document Type


Publication Date



Mental illness -- Social aspects, Mentally ill -- Public opinion, Stigma (Social psychology)



Mental illness stigma is a fundamental barrier to improving mental health worldwide, but little is known about how to durably reduce it. Understanding of mental illness as a treatable medical condition may influence stigmatizing beliefs, but available evidence to inform this hypothesis has been derived solely from high-income countries. We embedded a randomized survey experiment within a whole-population cohort study in rural southwestern Uganda to assess the extent to which portrayals of mental illness treatment effectiveness influence personal beliefs and perceived norms about mental illness and about persons with mental illness.

Methods and findings

Study participants were randomly assigned to receive a vignette describing a typical woman (control condition) or one of nine variants describing a different symptom presentation (suggestive of schizophrenia, bipolar, or major depression) and treatment course (no treatment, treatment with remission, or treatment with remission followed by subsequent relapse). Participants then answered questions about personal beliefs and perceived norms in three domains of stigma: willingness to have the woman marry into their family, belief that she is receiving divine punishment, and belief that she brings shame on her family. We used multivariable Poisson and ordered logit regression models to estimate the causal effect of vignette treatment assignment on each stigma-related outcome. Of the participants randomized, 1,355 were successfully interviewed (76%) from November 2016 to June 2018. Roughly half of respondents were women (56%), half had completed primary school (57%), and two-thirds were married or cohabiting (64%). The mean age was 42 years. Across all types of mental illness and treatment scenarios, relative to the control vignette (22%–30%), substantially more study participants believed the woman in the vignette was receiving divine punishment (31%–54%) or believed she brought shame on her family (51%–73%), and most were unwilling to have her marry into their families (80%–88%). In multivariable Poisson regression models, vignette portrayals of untreated mental illness, relative to the control condition, increased the risk that study participants endorsed stigmatizing personal beliefs about mental illness and about persons with mental illness, irrespective of mental illness type (adjusted risk ratios [ARRs] varied from 1.7–3.1, all p < 0.001). Portrayals of effectively treated mental illness or treatment followed by subsequent relapse also increased the risk of responses indicating stigmatizing personal beliefs relative to control (ARRs varied from 1.5–3.0, all p < 0.001). The magnitudes of the estimates suggested that portrayals of initially effective treatment (whether followed by relapse or not) had little moderating influence on stigmatizing responses relative to vignettes portraying untreated mental illness. Responses to questions about perceived norms followed similar patterns. The primary limitations of this study are that the vignettes may have omitted context that could have influenced stigma and that generalizability beyond rural Uganda may be limited.


In a population-based, randomized survey experiment conducted in rural southwestern Uganda, portrayals of effectively treated mental illness did not appear to reduce endorsement of stigmatizing beliefs about mental illness or about persons with mental illness. These findings run counter to evidence from the United States. Further research is necessary to understand the relationship between mental illness treatment and stigmatizing attitudes in Uganda and other countries worldwide.


Copyright: © 2019 Rasmussen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.



Persistent Identifier


Plos One