Stigma, Trust, and procedural integrity: Covid-19 testing in Malawi

Abstract

An emerging consensus in public health views testing for Covid-19 as key to managing the pandemic. It is often assumed that citizens have a strong desire to know their Covid-19 status, and will therefore take advantage of testing opportunities. This may not be the case in all contexts, however, especially those where citizens perceive stigma associated with the Covid-19, have low trust in health institutions, and doubt the procedural integrity of the testing process. This article explores willingness to receive a free Covid-19 test via a vignette experiment (conjoint design) embedded in a phone survey conducted in Malawi in May 2020. The experiment varied test provider (public clinic versus international health organization), proximity to illness, and reassurance of confidentiality. We find that Malawians expect higher uptake of testing in their community when the international health organization offered the test rather than a public clinic, an effect we attribute to higher trust in the organization and/or perceptions of greater capacity to ensure procedural integrity. The confidentiality reassurance did not substantially alter beliefs about the privacy of results, but did increase doubts about the willingness of community members to get tested in a public health clinic. Our findings suggest the importance of considering the demand side of testing in addition to well-known challenges of supply.

1. Introduction

An emerging consensus in public health views testing for Covid-19 as key to managing the pandemic (Jha et al., 2020, Allen et al., 2020, Soy, 2020). Without knowledge of infection rates, governments and other health actors struggle to design public health interventions and citizens lack information about the risks of different behaviors. Most studies of Covid-19 testing focus on technical properties of the tests (Abbasi, 2020, Wang et al., 2020), roll out efforts (Jha et al., 2020), or low supply. We consider a different factor: willingness to get tested, which is a form of compliance – like washing hands, limiting social contacts, or adhering to lockdown measures – that shapes the course of the disease.

Implicit in the testing literature is an assumption that citizens will take advantage of testing opportunities when offered, a narrative driven by the epidemic in the ‘Global North.’ This may not be the case for all populations, however. We explore two factors identified in previous literature that could reduce willingness to be tested: anxiety about disease stigma and lack of trust in public health actors. We also consider a third factor: perceptions about the procedural integrity of testing, defined here as the confidentiality and accuracy of the testing process. Where citizens do not trust public health actors, doubt the integrity of the process, and fear stigma, they may see little value to getting tested. Low supply might hide these issues in the short term, only for them to emerge as significant later on.

We report results from a telephone survey of 4641 Malawians in May 2020 (Lust et al., 2020). At the time of fielding, incidence of Covid-19 was low in Malawi, but awareness of and anxiety about the disease were high. Covid-19 was fresh in the minds of our respondents, but very few had the opportunity to get tested. Malawi resembles other low-income countries experiencing the early stages of the pandemic and provides insights into barriers to widespread testing in these contexts.

The survey asked Malawians about their knowledge, fears, and perceptions about Covid-19 as well as their opinions about the ability of public health actors to handle the pandemic. It included a pre-registered experiment that explored receptivity to Covid-19 testing through a hypothetical vignette. (See pre-analysis plan in Appendix D.) The vignette randomized three factors: the identity of the agency offering the test (a local health clinic or the World Health Organization); an assurance about the confidentiality of test results; and the proximity of contagion. It then asked respondents how likely they would be, under the circumstances detailed in the vignette, to accept the test; whether they thought others in their community would accept it; and their perceptions about the procedural integrity of the process.

The Covid-19 pandemic played out in Malawi against a backdrop of political crisis, precipitated by a controversial election in May 2019 and the subsequent annulment of election results in February 2020. These events potentially reduced trust in the sitting government and, by association, the public health clinics it operates. International health organizations, like the WHO, play a significant role in public health interventions in Malawi, funding a substantial portion of the domestic health budget and engaging in active interventions. International actors bring resources and capacity, but do not always understand local preferences and priorities (Dionne, 2018) and, like domestic governments, do not always have the trust of the populations they serve. It was an open question when we designed the experiment whether Malawians would trust domestic or international public health actors more, but we expected feelings of trust to influence whether and from whom Malawians would accept a free test and their perceptions about the procedural integrity of the process.

We explored stigma concerns and procedural integrity with the randomized confidentiality assurance. We expected the assurance to increase both willingness to get tested and expectations that test results would remain private. We also expected the confidentiality treatment to narrow differences between organizations offering the test. The final treatment, on proximity, specified risk of exposure, reducing chances that underlying variations in perception about risk would confound other treatments.

We find no treatment effects for personal willingness to be tested, in spite of high levels of reported anxiety about disease stigma. We suspect that social desirability bias induced nearly all respondents to report that they would accept the free test. Beliefs about whether others in the community would accept a free test were lower and did respond to treatments, however. Compared to the public health clinic, the WHO treatment significantly increased expectations of community uptake of testing. We also found a significant interaction between the confidentiality treatment and testing provider, albeit not the interaction expected. When reassured about confidentiality, respondents paradoxically reduced willingness to be tested by a public clinic (but not the WHO) and reduced estimates of the accuracy of a public clinic test (but not a WHO test). We do not believe these results reflect poor understanding of the concept of “confidentiality” in our study population as the term translates clearly into local languages and is familiar to respondents from their previous experience of the HIV-AIDS epidemic. Instead we speculate that the confidentiality reassurance primed respondents to think about the capacity of the provider to ensure the overall integrity of the testing process.

Our findings underscore the importance of considering the social and political aspects of testing and not simply assuming that ‘if you build it, they will come.’ Malawians expressed doubts about the willingness of their neighbors to get tested and these doubts seemed to center around the ability of public health clinics to offer confidential and accurate tests. We do not conclude from these findings that the WHO should take over testing in Malawi but rather that governments and public health agencies should pay careful attention to factors like stigma, trust, and beliefs about procedural integrity that shape responses to testing in their communities. We also echo the advice of Blair et al. (2017) that international organizations might best contribute to disease management in Africa by working to improve the capacity of domestic actors.

2. Stigma, institutional trust, and testing

Previous work on stigma suggests that disease can come to be negatively associated with groups (Parker and Aggleton, 2003, Mahajan et al., 2008, Lieberman, 2009, Adida et al., 2018). Studies have also noted that compliance with public health directives depends critically on trust in government and health care actors (Blair et al., 2017, Alsan and Wanamaker, 2017, Vinck et al., 2008, Wong et al., 2020). We speculate that these factors could also shape responses to Covid-19 testing. Stigma might generate perceived costs to learning one’s status, particularly in contexts where confidentiality is lacking. Lack of trust in public health actors could exacerbate fears if people believe testing agencies will not preserve their privacy. Where stigma is high and citizens lack faith in the public health system, costs of testing may outweigh benefits. A recent report on a Covid-19 outbreak in Nigeria, for example, noted that citizens believed a diagnosis is a ‘death sentence’ and did not want neighbors to know they were infected, ‘so they avoid being tested’ (Maclean, 2020).

Public health campaigns in Africa involve domestic and international actors. International actors fund public health initiatives, shape domestic priorities, and engage in health interventions. While they supply resources, they do not always understand the preferences of the local population and, like their domestic counterparts, may face challenges with distrust. Blair et al. (2017), studying the 2014–15 Ebola epidemic in Liberia, find that citizens had more trust in international non-governmental organizations (INGOs) than in their own government. The study, moreover, found that individual variations in government trust drove compliance with Ebola-related measures, while trust in INGOs bore no relation to such behaviors. We build on this study and those on stigma, to explore trust in domestic and international actors in Malawi, fears about stigma associated with Covid-19, and how these shape willingness to get tested for Covid-19.

3. Covid-19 testing in Malawi

In response to the Covid-19 global pandemic, Malawi declared a state of national disaster on 20 March 2020, before the first cases were recorded. Under this declaration, the government introduced public safety measures to curb the spread of the virus, including free testing for symptomatic individuals or those identified through contact tracing.1

Malawi established its first Covid-19 testing center in Lilongwe in late March 2020. Eleven Covid-19 testing centers operated in the country by the end of May and 4490 tests had been conducted. Notwithstanding the increased pace, the number of Covid-19 tests remained low during our survey fielding period. Unlike nearby South Africa, which used door-to-door testing, Malawi focused on suspected cases, their immediate contacts, and travelers entering the country.

Supply side challenges therefore posed the most immediate and obvious barrier to Covid-19 testing in Malawi in the early stages of the pandemic. Supply challenges do not imply the absence of demand side barriers to testing, however. Low supply may in fact conceal testing reluctance on the part of some citizens. Our experiment was designed to probe whether such reluctance might exist in the population and factors that might alleviate it.

4. Sampling, design, & data

We conducted a phone survey in Malawi in May 2020 (Lust et al., 2020) with 4641 respondents from 27 districts across all three regions. We obtained phone numbers from surveys we conducted in 2016 (Lust et al., 2016) and 2019 (Lust et al., 2019). About half (54%) of the respondents came from the 2019 study, which drew random samples from the capital city, Lilongwe, and a region along the Zambian border. Because the 2019 study did not include the Southern Region, we contacted Southern respondents from the 2016 study. (Response rate was 62%. See Appendix A for more details.)

The resulting sample is not a random sample of the Malawian population but does include respondents in all three regions and is broadly representative. The proportions of most ethnic groups are roughly on target with national demographics from the Malawi 2018 census. In addition, 41% report being employed in agriculture compared with World Bank estimates of 44%.

Our sample is 56% female and the average age is 37. Less than 40% of respondents had schooling beyond primary school. Three quarters had lived in their current neighborhood for at least 10 years. Most said the household had enough water (94%) and soap or hand sanitizer (74%) for everyone to wash hands frequently. A staggering 83% feared going hungry due to the pandemic, and 65% reported that they had already experienced loss of income.

To investigate citizens’ willingness to get tested for Covid-19, we embedded a vignette experiment (also known as a single–profile conjoint experiment) in the survey. The experiment presented each respondent with a hypothetical scenario that described an organization offering free Covid-19 tests. The experiment randomly varied the proximity of the illness, the organization offering the test, and whether or not the results of the test would be kept confidential. Each respondent viewed a single scenario. The vignettes read as follows (items in bold are randomly-assigned treatment factors):

“Testing for the Novel Coronavirus and Covid-19 involves inserting a long swab into the nostril for 15 seconds and rotating the swab several times. This swabbing is then repeated for the other nostril. If [you/someone in your household/someone in your neighborhood or village] fell ill with Covid-19 symptoms, and [a public health clinic/the World Health Organization] offered free Covid-19 testing, [with results that would remain confidential/BLANK]…”

We then asked four follow-up questions (Yes/No responses):

• 1.Would you agree to be tested?
• 2.Do you think most others in your community would agree to be tested?
• 3.Do you think that others would find out the results?
• 4.Do you think the test would be accurate?

The experiment occurred in the middle of the survey after questions about knowledge and attitudes toward Covid-19. Respondents would have been primed on Covid-19, but not about testing or any of the organizations involved in it.

Using this experiment, we test three pre-registered hypotheses:

Hypothesis 1

Willingness to get tested should be higher under the WHO treatment (relative to the local clinic).

Hypothesis 2

Willingness to get tested should be higher under the confidentiality reassurance treatment (relative to the ‘blank’ treatment).

Hypothesis 3

Willingness to get tested is not likely to differ across local and WHO clinics if local clinics can guarantee confidentiality.

Additional factors that could drive testing not manipulated in the experiment include test cost, insufficient knowledge about how to get tested, fear of getting infected while being tested, and anxiety about the physical discomfort of the test. The design of the experiment and its timing mitigate these as confounds. We resolve the first two by stipulating that the test was free and offered by a specific organization. The timing of the survey – before extensive viral spread – should have minimized fear of exposure during testing. And we began the experiment with a description of the process, ensuring that all respondents were equally aware of the discomforts involved regardless of which treatments they received. We focused on confidentiality of results rather than the confidentiality of testing itself (i.e. whether others would know that an individual had been tested) because agencies might not have leeway to keep the act of testing itself private; much Covid-19 testing happens in public venues, for example. Additional dimensions of confidentiality might be a useful avenue for future work.

5. Results

Our survey respondents reported significant anxieties about stigma: 81% felt they would be treated poorly if they contracted Covid-19 (Fig. 1 (a)). Roughly one third had concerns about the ability of the government (which would be voted out in June 2020) to manage the pandemic, and 63% feared it would be more difficult to access healthcare due to the pandemic (See Fig.  1(b)).2 They also professed more trust in the WHO (53%) than their own government (48%) or the Covid-19 Special Committee (29%; see Fig. 2 ). With concerns about stigma, healthcare access, and low trust in their government’s Covid-19 response, Malawians might have reasonably questioned the personal utility of getting tested.

Fig. 1.

Stigma and Anxiety over Covid-19.

Fig. 2.

Trust in Institutions.

We follow the procedure for causal conjoint analysis outlined by Hainmueller et al. (2013). We estimate the Average Marginal Component Effect (AMCE), i.e. the marginal effect of a specific factor, and the Average Component Interaction Effect (ACIE) between two factors. Ordinary Least Squares (OLS) regression provides consistent estimates of both quantities of interest. We use the following model to obtain AMCE estimates:

$$\mathrm{y}=\alpha +{\mathrm{\beta }}_1\text{Household}+{\mathrm{\beta }}_2\text{Neighbor}+{\mathrm{\beta }}_3\text{WHO}+{\mathrm{\beta }}_4\text{Confidential}+\mathit{\beta }\mathbf{d}+\epsilon$$ (1)

The outcome, y, is the perceived willingness of others to seek a test. Each β_i in Eq. (1) represents the AMCE for the associated factor. For example, β₃ estimates the effect of a test being offered by the WHO (relative to a local clinic; a test of Hypothesis 1) while averaging across the remaining factors (confidentiality and who has symptoms) in the experiment. The βd term represents a vector of district fixed effects. We focus on the AMCE rather than the ATE (average treatment effect, which estimates the effects of each of the 12 treatments/individual vignettes) because we seek to understand how each experimental factor influences the likelihood of seeking a test, while controlling for all other factors.

To estimate the ACIEs, we use the same model as for the AMCE, adding the interactions between all factors:

$$\begin{array}{c}\text{y}=\alpha +{\mathrm{\beta }}_1\text{Household}+{\mathrm{\beta }}_2\text{Neighbor}+{\mathrm{\beta }}_3\text{WHO}+{\mathrm{\beta }}_4\text{Confidential}\\ +{\mathrm{\beta }}_5\text{WHO}*\text{Confidential}+{\mathrm{\beta }}_6\text{Household}\ast \text{WHO}\\ +{\mathrm{\beta }}_7\text{Household}*\text{Confidential}+{\mathrm{\beta }}_8\text{Neighbor}*\text{WHO}\\ +{\mathrm{\beta }}_9\text{Neighbor}*\text{Confidential}\\ +\mathit{\beta }\mathit{d}+\epsilon \end{array}$$ (2)

This ‘long’ model (Muralidharan, Romero, & Wuthrich, 2020) allows us to estimate the effect of each factor conditional on each of the other factors presented in the experiment. We are particularly interested in β₅ as it provides a test of Hypothesis 3. For this model, y is either the likelihood of others to seek a test or the expected accuracy of the test. We estimate robust standard errors for all models.

Fig. 3 reports ACME estimates for the outcome “Do you think most others in your community would agree to be tested.” The reference categories in all models are ‘you fell ill’, ‘a public health clinic’, and no mention of the confidentiality of the test. Estimates for the other outcomes are reported in Appendix B. As noted above, social desirability is likely limiting variation in respondent’s own willingness to get tested explaining the null effects for that outcome.

Fig. 3.

Effect on response to 'Do you think most others in your community would agree to be tested?’ (Yes/No response, Average = 0.85). The figure plots coefficient estimates along with their 95% confidence intervals from estimating Eq. (1).

We find support for Hypothesis 1. Respondents expect modestly higher (around 3 percentage points) community uptake of testing when the WHO offers the test. We find a significant effect for the confidentiality treatment (Hypothesis 2), but not in the expected direction: when reassured about confidentiality, respondents expect reduced community uptake of testing. We also find surprising results for the interaction between confidentiality and WHO treatments. Hypothesis 3 anticipated a negative interaction, i.e. that the confidentiality treatment would narrow differences between the public health clinic and WHO. The results reported in Fig. 4 suggest that the confidentiality treatment actually accentuated differences.

Fig. 4.

Interaction Model: Effect on response to 'Do you think most others in your community would agree to be tested?’ (Yes/No response, Average = 0.85). The figure plots coefficient estimates along with their 95% confidence intervals from estimating Eq. (2). Only the interaction of interest is reported here; see Appendix B for full regression output.

To investigate this surprising interaction effect, we estimate its marginal effects. We find that the positive WHO-confidentiality interaction effect is driven by a backlash from respondents receiving the public clinic and confidentiality assurance treatments. When respondents do not receive an assurance of confidentiality (the “blank” treatment), the test provider has no discernible effect. The WHO and public health providers lead to an 89.8 and 89.3 percent likelihood of seeking a test, respectively, when there is no confidentiality assurance. By contrast, a confidential test from the WHO is associated with an 89.6 percent likelihood of seeking a test compared to only an 84.4 percent likelihood when the confidential test is offered by the public health clinic (a 5% drop). When no reassurance of confidentiality was given, we observe identical outcomes for the public health and WHO treatments (roughly 89 percent likelihood). With the reassurance, however, respondents reduced their estimation of community uptake for the public health clinic (but not the WHO).

These results seem counter-intuitive at first glance. Why would a confidentiality reassurance induce respondents to believe others in their community would eschew testing from a public clinic? We do not believe the results reflect a problem with comprehension or translation of “confidentiality.” We allowed respondents to take the survey in English, Chichewa, or Chitumbuka. “Confidentiality” did not present a problem for translation, as it has clear and commonly understood analogs in the local languages.3 Moreover, Malawians are familiar with the concept of medical confidentiality from their experience with HIV-AIDS. The term confidentiality is also used in the consent agreements for surveys without problems.

We speculate instead that the confidentiality assurance primed respondents to think about the overall process of testing and the capacity of the provider to ensure credible results. Where respondents had lower trust in the testing provider to begin with (e.g. public clinics linked to the government), this assessment led to a downgrading of expectations about test integrity. Bolstering this interpretation, when we examine beliefs about the accuracy of results, we find a similar interaction effect of the testing provider and the confidentiality treatments: respondents who received the confidentiality and public health clinic treatments were substantially less likely to think the test would be accurate than respondents who received the confidentiality and WHO treatments (Fig. 5 ).4

Fig. 5.

Interaction Model: Effect on response to 'Do you think the test would be accurate?’ (Yes/No response, Average = 0.90). The figure plots coefficient estimates along with their 95% confidence intervals from estimating Eq. (2). Only the interaction of interest is reported here; see Appendix BB for full regression output.

In summary, Malawians expect higher community uptake of testing when the agency offering the tests is the WHO rather than a public health clinic. The effect may reflect greater trust in the WHO and its ability to ensure the procedural integrity of the process. Previous work argued that trust in institutions shaped behavioral responses to public health crises (Blair et al., 2017, Alsan and Wanamaker, 2017, Vinck et al., 2008, Wong et al., 2020). Our findings indicate the value of moving beyond undifferentiated conceptualizations of trust to studying related or component factors like technical capacity and procedural integrity.

6. Conclusion

This study provides timely information about public willingness to comply with Covid-19 testing in Malawi. Testing is not widespread there, but as the pandemic proceeds, it could become an important aspect of Malawi’s public health response.

Malawi provides insight into other cases where populations might have concerns about stigma, lack trust in government public health efforts, and doubt the procedural integrity of the testing process. We believe these are not uncommon conditions. In a companion survey in Zambia (n = 1900), we find comparable levels of stigma associated with Covid-19 (56%) and doubts about confidentiality of tests (49%). One limitation of our study is that it was fielded early in the epidemic. Our findings might not generalize to later stages, when experience with Covid-19 become more widespread. It is nonetheless important to understand these early stages for their own sake, as they represent a crucial period of time for intervention. Moreover, the factors we identify – stigma, trust, and beliefs about procedural integrity – likely remain significant over the course of the pandemic.

Appendix A. Supplementary data

The following are the Supplementary data to this article: