Factors associated with contact tracing compliance among communities of color in the first year of the COVID-19 pandemic

Abstract

Rationale

The disproportionate impact of COVID-19 on communities of color has raised questions about the unique experiences within these communities not only in terms of becoming infected with COVID-19 but also mitigating its spread. The utility of contact tracing for managing community spread and supporting economic reopening is contingent upon, in part, compliance with contact tracer requests.

Objective

We investigated how trust in and knowledge of contact tracers influence intentions to comply with tracing requests and whether or not these relationships and associated antecedent factors differ between communities of color.

Method

Data were collected from a U.S. sample of 533 survey respondents from Fall (2020) to Spring 2021. Multi-group SEM tested quantitative study hypotheses separately for Black, AAPI, Latinx, and White sub-samples. Qualitative data were collected via open-ended questions to inform the roles of trust and knowledge in contact tracing compliance.

Results

Trust in contact tracers was associated with increased intentions to comply with tracing requests and significantly mediated the positive relationship between trust in healthcare professionals and government health officials with compliance intentions. Yet, the indirect effects of trust in government health officials on compliance intentions were significantly weaker for the Black, Latinx, and AAPI samples compared to Whites, suggesting this strategy for increasing compliance may not be as effective among communities of color. Health literacy and contact tracing knowledge played a more limited role in predicting compliance intentions directly or indirectly, and one that was inconsistent across racial groups. Qualitative results reinforce the importance of trust relative to knowledge for increasing tracing compliance intentions.

Conclusions

Building trust in contact tracers, more so than increasing knowledge, may be key to encouraging contact tracing compliance. Differences among communities of color and between these communities and Whites inform the policy recommendations provided for improving contact tracing success.

The COVID-19 pandemic introduced extraordinary challenges for public health. Despite the widespread, and seemingly indiscriminate toll COVID-19 took, it became evident that the virus disproportionately harms marginalized communities globally (Yancy et al., 2020). In the U.S., Black and Latinx communities are the two most affected groups therein exacerbating pre-existing health disparities (Selden and Berdahl, 2020). Similarly, the economic and educational impacts of COVID-19 have also disproportionately harmed communities of color (Yancy, 2020), and discrimination targeting the Asian American/Pacific Islander (AAPI) and Black communities has risen during the pandemic (Pew Research Center, 2020a; Tessler et al., 2020). Although reasons for COVID-19 disproportionately impacting communities of color are complex, one potential factor is systemic and historical discrimination impacting the social and structural determinants of health (e.g., health care access and quality, housing, educational and employment opportunities; Chowkwanyun and Reed, 2020; Hooper et al., 2020; Khazanchi et al., 2020). Therefore, governments' responses to the public health crisis must include efforts to ensure equity in the tools used to mitigate the spread of COVID-19—that efforts are successful for members of White and communities of color. We use the terms ‘people of color’ and ‘communities of color’, as opposed to ‘minority,’ to refer to non-White individuals, including three of the broadest categorizations of non-White communities in the U.S.–AAPI, Black, and Latinx–as the focus of this investigation.

One such tool in the fight to end the COVID-19 pandemic, as well as future health emergencies, is contact tracing—the process of interviewing infected individuals to identify their contacts, and then interviewing these contacts to determine if testing, self-isolation, or other follow-up measures are necessary. The efficacy of contact tracing depends, in part, on compliance with contact tracing requests. Here, we argue that race-based differences in trust of healthcare and government institutions, as well as lack of knowledge about contact tracing, can interfere with contact tracing compliance. Using a sample of Black, Latinx, AAPI, and White individuals, we demonstrate that intentions to comply with contact tracing requests are partially dependent on trust in contact tracers and accurate knowledge of contact tracers' roles. In addition, we identify and test a number of predictors of trust in and knowledge of contact tracers and evaluate their predictive power within the different racial groups. This approach of data disaggregation by racial/ethnic group answers calls to consider the unique experiences of distinct communities of color in the U.S., providing more nuanced evidence and recommendations to stop the spread of COVID-19 and eliminate minority health disparities (e.g., Bryan et al., 2021; Chen et al., 2020; Chowkwanyun and Reed, 2020). Our results provide important groundwork for the development of interventions to increase individuals’ willingness to comply with contact tracing requests to mitigate health disparities among communities of color related to COVID-19 and future public health crises.

1. Contact tracing

Contact tracing is a critical tool for mitigating the transmission of COVID-19 (World Health Organization [WHO], 2020). One explanation for the disproportionate impact of COVID-19 in communities of color is the overrepresentation of these individuals in living and working situations that precludes safe distancing (Hooper et al., 2020; Selden and Berdahl, 2020; Yancy, 2020); this suggests that contact tracing may be particularly critical in communities of color.

Contact tracing is the process by which public health officials talk to individuals infected with COVID-19 and collect information about other individuals with whom the infected individual has been in close proximity (Watson et al., 2020). With this information, contact tracers can inform these individuals that they have been near someone with COVID-19, and then provide guidance on the risk of exposure, symptomatology, testing, isolation, and medical care follow-up. Contact tracers also check in with individuals in quarantine to track symptomatology, assess resource needs, answer questions, and otherwise encourage compliance with isolation and/or quarantine orders (Watson et al., 2020). Although development of smartphone apps for certain contact tracing elements, such as exposure notification and information communication, can be effective, their use has been limited in the U.S. and other neo-liberal societies where they are voluntary due to privacy and surveillance concerns (Akinbi et al., 2021). Thus, although contact tracers use digital tools (e.g., automatic text and email reminders), tracing interviews and case work are predominately handled in-person in the U.S.

The spread of communicable diseases slows the most with a combination of expeditious testing and contact tracing compliance (Kretzschmar et al., 2020). Factors that impact the efficacy of contact tracing are: (1) the time that passes between an infected individual showing symptoms and being tested (i.e., testing delay); (2) delays in the time between testing positive and contact tracing beginning (i.e., tracing delay); (3) compliance with contact tracing requests (i.e., tracing compliance); and (4) the proportion of contacts identified and tested (i.e., tracing coverage), wherein 100% means all contacts are identified, spoken with, and tested/isolated therein stopping the spread along that transmission vector (Kretzschmar et al., 2020).

Here, we focus on tracing compliance and tracing coverage as factors affected more so by the behaviors of infected persons and their contacts. Indeed, contact tracing will only be successful if infected persons follow the instructions of the contact tracer, and if they communicate openly with contact tracers by sharing accurate and complete information. Obtaining this trust and compliance with contact tracers is not only difficult in general (Mooney, 2020), but may be especially difficult in communities of color due to historical mistreatment.

We propose a theoretical model, and derive hypotheses from it, that highlight trust and knowledge as two potential predictors of individuals' compliance with contact tracing requests. Additionally, we dig deeper to identify sources that might contribute to contact tracing trust and knowledge, and how these effects may differ for individuals in different ethnic/racial groups. We use intentions to comply with contact tracing requests as our primary dependent variable because behavioral intentions are better predictors of specific behaviors (e.g., following contact tracer recommendations for isolation or reporting) than general attitudes (e.g., keeping others safe; Ajzen and Fishbein, 1977). Moreover, contact tracers mostly do not track individuals' specific compliance behaviors (beyond text messages asking contacts to report symptoms), but communicate information with the expectation of compliance (Watson et al., 2020). Thus, individuals' willingness or intentions to comply with contact tracers’ requests is an important, accessible variable that needs to be better understood. As compliance intentions may be influenced by pressures outside the control of the person (e.g., needing to work for money), we also include some key control variables, detailed below.

2. Trust in contact tracers

Although contact tracing is a critical tool to slow the spread of communicable diseases like COVID-19, most individuals have little to no experience with contact tracers. Therefore, willingness to comply with contact tracer's requests and recommendations may depend on the extent to which individuals trust this relatively unknown institution of contact tracing. Trust is typically defined as a willingness to be vulnerable to another based on the belief that the other party will fulfill task or role expectations competently, honestly, and reliably (Mayer et al., 1995; McAllister, 1995). As trust increases, individuals are willing to take larger behavioral risks in the relationship because they believe the trusted party has one's own best interests in mind (Mayer et al., 1995). This makes trust a critical lever for encouraging behavioral compliance in healthcare settings (Armstrong et al., 2013; Becker and Maiman, 1975). Accordingly, in the case of contact tracing, those who trust contact tracers, believing tracers are acting for their and their communities' health and safety, should be more likely to share information (i.e., be vulnerable) and more willing to comply with the tracers' requests for isolation or other measures (i.e., risk-taking).

Hypothesis 1

Trust in contact tracers is positively related to willingness to comply with contact tracer requests.

Yet, because contact tracers are unknown to the people they call, establishing trust in contact tracers may be more complicated for communities of color due to the history of discrimination these communities have experienced from similar government and health institutions. Typically, trust for an institution (e.g., contact tracing) or an agent of that institution (i.e., a contact tracer) develops over time as two parties learn about their intentions and behavioral patterns to better gauge trustworthiness (Mayer et al., 1995; McAllister, 1995; Robert et al., 2009). Such a series of interactions, however, is nonexistent in the context of one party trusting another party in a nonrepeating, short-duration interaction, such as when a COVID-19 infected person must decide whether to comply with a contact tracer. In such cases, people are more likely to rely on swift trust (Rafaeli et al., 2008). Swift trust is a willingness to be vulnerable to the actions of another that develops rapidly, prior to any significant exchange relationship between the two parties (Schilke and Huang, 2018). This changes the calculus of trustworthiness by encouraging people to transfer trust from a known, trusted source, to an unknown target (Stewart, 2003). In the absence of previous direct experience with contact tracers, therefore, individuals rely on who and what they do know to make swift trust determinations, with trust transfer more likely when the two parties share categorical features and operate in similar institutional contexts or environments (McKnight et al., 1998; Robert et al., 2009; Stewart, 2003). Trust transfer, therefore, enables the development of swift trust (Leung et al., 2022).

Although the institution of contact tracing may be largely unknown to most individuals until recently, its association with county health departments may invoke a connection between contact tracing and healthcare or government institutions. Therefore, we investigate the impact of (1) trust in healthcare providers, (2), trust in government officials, and (3) knowledge about COVID-19 and contact tracing, as important determinants of trust in contact tracers. Importantly, although the connection between contact tracers, healthcare, and government institutions may facilitate trust transfer to enable swift trust, it may also introduce problems for communities of color. The underlying distrust toward these institutions among communities of color is well-documented--particularly for Black and Latinx individuals (Armonstrong et al., 2013; Brown and Benedict, 2002; Kennedy et al., 2007; Marschall and Shah, 2007), but even among the AAPI community more recently as anxiety for targeted racist attacks and frustration with the government's response have increased during the COVID-19 outbreak (Chen et al., 2020; Pew Research Center, 2020a; Tessler et al., 2020). Consequently, compromised trust in government and healthcare institutions may impact the extent to which individuals trust contact tracers. Thus, although we expect these antecedent relations or sources of trust in contact tracers to be applicable across racial groups, the average levels of these variables may differ between groups given the history of systemic mistreatment toward communities of color.

Related recent research on contact tracing apps shows mixed evidence regarding the exact racial or cultural groups with the lowest levels of adoption, but in general, ethnic and cultural minority groups, including immigrants and non-citizens, are less likely to use digital contact tracing methods (Li et al., 2021; Villius Zetterholm et al., 2021). Reinforcing our focus here, reasons cited predominately include distrust with how data will be collected and used (Akintoye et al., 2021; Villius Zetterholm et al., 2021), so exploring these transferrable sources of distrust among racial/ethnic groups is important.

Trust in Healthcare Providers. Myriad factors have influenced people of colors’ negative experiences with healthcare systems from as early as how slave owners addressed the medical needs of slaves, to present-day disparities in healthcare access (Kennedy et al., 2007). Two infamous cases of mistreatment of Black Americans include the Tuskegee Syphilis studies, resulting in the death of 400 Black men (Gamble, 1993), and the harvesting of cells from Henrietta Lacks without her knowledge or consent, resulting in large financial gains without any compensation for the Lacks family (Truog et al., 2012). For the AAPI community, historical mistreatment such as internment camps during WWII, Islamophobia following 9/11, and even the current COVID-19-related racist attacks targeting this community as somehow responsible for the virus and its spread (Tessler et al., 2020) have been linked to increased chronic health conditions and may further discourage help-seeking from healthcare professionals (Chen et al., 2020). Finally, the Latinx community also struggles with healthcare systems as they battle lack of language translation services, discrimination and fear due to immigration status, and access to health insurance and quality care (Cristancho et al., 2008; Rhodes et al., 2015).

These factors permeate the healthcare landscape today leading to a lack of cultural competence, representation, and bilingualism among healthcare providers, poor patient-physician relationship, and limited access to care (Armstrong et al., 2013; Chen et al., 2020; Cristancho et al., 2008; Kennedy et al., 2007; Khazanchi et al., 2020). As a result, people of color are less likely to trust healthcare systems (Halbert et al., 2006; Rhodes et al., 2015; Whetten et al., 2006). To the extent that contact tracers are seen as representatives of the healthcare system, people of color may not trust the motives and/or intentions of these individuals, thereby reducing compliance (Whetten et al., 2006). Just as trust may transfer from known, similar sources, distrust due to repeated mistreatment of one's vulnerabilities may also transfer (Mayer et al., 1995; Stewart, 2003). Therefore, individuals may transfer trust from known healthcare providers to unknown contact tracers (Lewis and Weigert, 1985; McKnight et al., 1998).

Hypothesis 2

Trust in healthcare providers is positively related to trust in contact tracers.

Trust in Government Officials. Beyond healthcare systems, people of color may have elevated levels of distrust toward the government and government officials (Brown and Benedict, 2002; Marschall and Shah, 2007), and are less likely to turn to government employees, such as law enforcement, for help (Tessler et al., 2020). This lack of trust has been linked to neighborhood factors (Marschall and Stolle, 2004), increased legal cynicism (Nivette et al., 2015), and a fear of collusion between healthcare providers and immigration officials (Rhodes et al., 2015) among people of color. Furthermore, distrust in the government is associated with reduced use of healthcare services (Rhodes et al., 2015; Whetten et al., 2006). The sociopolitical milieu during the COVID-19 pandemic (e.g., government actions related to separating migrant families [Rose, 2020 October], continued police killings of people of color [Lett et al., 2021]), and targeted attacks against the AAPI community [Tessler et al., 2020]) may also heighten people of colors' distrust in government officials as previous negative interactions with the institution compromises one's willingness to be vulnerable (i.e., trust; Mayer et al., 1995). Thus, to the extent that contact tracers are viewed as similar to government officials, as most call on behalf of local governments' health departments, trust in government health officials should transfer to trust in contact tracers (Stewart, 2003).

Hypothesis 3

Trust in government health officials is positively related to trust in contact tracers.

Contact Tracing Knowledge. Knowledge is critical to making informed decisions but can also influence the extent to which we trust one another. Thus, although below we propose a direct effect of contact tracing knowledge on compliance, we also note its importance as a necessary precursor to trust. As mentioned previously, one key determinant of swift trust is accurate knowledge about the motives and goals of the trusted party (Lewis and Weigert, 1985; Robert et al., 2009). Indeed, if trust is a willingness to be vulnerable to the intentions/behaviors of another party, one must have some foundational knowledge about the institution and its motives in order for the party to be trusted (Mayer et al., 1995; McAllister, 1995; Rafaeli et al., 2008). Thus, one other critical factor underlying individuals’ trust in contact tracers and ultimate compliance with tracing requests may be the degree of accurate contact tracing knowledge they possess.

Hypothesis 4

Contact tracing knowledge is positively related to trust in contact tracers.

In summary, we view trust as necessary for the willingness to comply with contact tracing requests and identify three transferable sources that might impact trust in contact tracers. We further argue that these antecedents of trust in contact tracers may differ among communities of color. We hypothesize that trust in contact tracers will mediate the relations between these sources and individuals’ willingness to comply with tracing requests.

Hypothesis 5

Trust in contact tracers will fully mediate the relations between (a) trust in healthcare providers and (b) trust in government officials with individuals' willingness to comply with contact tracing requests. Trust in contact tracers will (c) partially mediate the relation between contact tracing knowledge and individuals' willingness to comply with contact tracing requests.

3. Knowledge about contact tracing

Misinformation about contact tracing, and COVID-19 more broadly, has spread rapidly across the internet (Shmerling, 2020), prompting the WHO to publish a COVID-19 “Mythbusters” page to combat the spread of inaccurate or misleading information (WHO, n.d.). Faulty or inadequate information may be problematic if it impedes the important work of contact tracing and interferes with individuals’ willingness to comply with contact tracing requests. Unfortunately, people of color are often targets for disinformation (the propagandistic spread of inaccurate information; Lewandowsky et al., 2012), due in part to systemic discrimination leading to differences in scientific literacy (Allum et al., 2018) and health literacy (Berkman et al., 2011). For example, inadequate language translation services (Cristancho et al., 2008), attempts to exploit any existing distrust and mistreatment among communities of color (e.g., Schumaker, 2019), and reduced access to both healthcare services and the credible healthcare information offered by providers and others (Boyd, 2021) may result in knowledge differences among communities of color. Therefore, in addition to decreasing trust in contact tracers, inadequate knowledge of contact tracing may also directly hinder contact tracing compliance. By comparison, people with more accurate information about contact tracing (e.g., contact tracers cannot report you to immigration services), would be more likely to comply with contact tracer requests.

Hypothesis 6

Contact tracing knowledge will be positively related to willingness to comply with contact tracing requests.

Although knowledge about contact tracing may come from a variety of sources, in this study we focus on health literacy as a key predictor.

Health Literacy. Health literacy is “the degree to which an individual has the capacity to obtain, communicate, process, and understand basic health information and services to make appropriate health decisions” (Centers for Disease Control and Prevention, 2020). Health literacy is a powerful determinant of health outcomes, and is a critical determinant of health disparities in healthcare access and making health-related decisions (Berkman et al., 2011). Individuals’ knowledge in a domain, such as health, is tied to their past experiences (in this case with health-related issues and services) and their typical intellectual engagement in the domain (e.g., seeking out health-related news; Beier and Ackerman, 2005). As noted above, seeming distrust in healthcare among communities of color may be at least partially attributable to negative experiences with healthcare providers (Armstrong et al., 2013; Boyd, 2021; Kennedy et al., 2007; Rhodes et al., 2015). These negative experiences likely decrease the opportunities and willingness of people of color to access and engage with healthcare information, therein reducing health literacy (Boyd, 2021). Nevertheless, we expect that health literacy may provide an important foundation of basic health information and experience with which to combat the spread of misinformation/inadequate information related to COVID-19 in communities of color, and to increase the likelihood of complying with contact tracing requests. So, we expect that as health literacy increases, knowledge about contact tracing will increase.

Hypothesis 7

Health literacy will be positively related to contact tracing knowledge.

As anticipated for the mediating role of trust in contact tracers on willingness to comply with tracing requests (Hypothesis 5 above), we also hypothesize that health literacy may influence contact tracing compliance indirectly via increased knowledge.

Hypothesis 8

Contact tracing knowledge will mediate the effects of health literacy on willingness to comply with contact tracing requests.

Additionally, because of the hypothesized connection between trust and knowledge, it is possible that health literacy may demonstrate serial indirect effects on compliance through knowledge and trust:

Hypothesis 9

Health literacy will have a serial indirect effect on willingness to comply with contact tracing requests through contact tracing knowledge and trust in contact tracers.

Our formal hypothesized model is presented in Fig. 1 . We include two important variables as statistical controls in this model because, as noted earlier, behavioral intentions may be superseded by factors outside one's control. First, given the polarized nature of American politics, there are well-documented partisan differences in the perceived threat of the COVID-19 virus and the preferred response (Gollwitzer et al., 2020; Pew Research Center, 2020b). Indeed, political partisanship is a much stronger predictor of COVID-19 safety behavior than other variables such as race, gender, age, or geography (Gollwitzer et al., 2020; Pew Research Center, 2020b). Individuals' ideological beliefs motivate behavior; thus, conservative individuals may view the threat caused by COVID-19 to be much lower than liberal individuals, therein seeking less information on mitigation strategies (Van Bavel et al., 2020). Because of this lower perceived threat, conservative individuals may engage with COVID-19 related information less and have less trust in contact tracing. As a result, these individuals may be less likely to comply with contact tracing requests. Second, financial insecurity was also included as a control variable because financial considerations differ between different racial/ethnic groups (Chowkwanyun and Reed, 2020) and may play a role in the willingness to comply with contact tracing requests. People of color are disproportionately represented in low-wage essential work, and thus are more likely to lack the privilege to reduce work hours or telecommute (Yancy, 2020); consequently, they may elect to continue working despite possible disease exposure to continue earning money, and not lose their job (Gould and Wilson, 2020). In short, financial insecurity may influence intentions to comply, outside of trust and knowledge, given the needs of individuals of color.

Fig. 1.

Model predicting trust in contact tracers, contact tracing knowledge, and likelihood of compliance with contact tracing requests for four racial groups. A = AAPI, B = Black, L = Latinx, W = White. *p < .05, **p < .001.

4. Overview of the research project aims

To investigate the influence of trust and knowledge on intentions to comply with contact tracers, we conducted a cross-sectional survey study of a sample of adults who live in New York State representing four major racial/ethnic groups: Black, Latinx, AAPI, and White. Relationships in the hypothesized model of tracing compliance predictors are expected to be similar for all racial groups; however, given expectations about race-based differences in the antecedents (e.g., trust in healthcare professionals and government officials, health literacy), we also conduct preliminary analyses to compare mean group levels on all study variables.

Although a cross-sectional design precludes strong claims of causation, this design is most appropriate given the dynamic nature of the COVID-19 pandemic and the need to understand contact tracing compliance quickly. We also collected qualitative responses to questions assessing individuals' knowledge and trust concerning contact tracers, and to solicit recommendations to overcome misinformation and distrust in order to increase compliance. This study contributes to our understanding of how to improve the willingness to comply with contact tracing requests among different racial/ethnic groups to help combat the disproportionate, negative impact of COVID-19 on communities of color. By isolating the potential causes of contact tracing noncompliance for different groups, policymakers can use these results to design interventions to increase individuals’ willingness to comply with contact tracing requests and customize these for different ethnic/racial groups, where necessary (Bryan et al., 2021). Such interventions will not only help with the COVID-19 response but likely future public health emergencies as well.

5. Method

5.1. Participants and procedures

Participants were recruited through Prolific, an online participant management data collection site, between October 2020 and June 2021, and were restricted to New York State to minimize regional differences in health disparities and attitudes (Chowkwanyun and Reed, 2020; Pew Research Center, 2020b). Duplicate surveys were deployed with different eligibility requirements based on racial/ethnic identity (reported to Prolific, confirmed by self-report) to ensure adequate representation from Black, Latinx, AAPI, and White samples. The survey took approximately 30 min, and participants were paid $4.75. A total of 601 individuals participated in the online survey, however, 59 were removed from the dataset for failing more than one of four attention checks and nine for missing 95% or more of their data. Assumptions for data missing completely at random (MCAR) were met, supporting the removal of these nine people (Kang, 2013). The comparison of FIML and non-FIML estimation, and for including or removing those who failed attention checks revealed no differences in our hypothesis models. We analyzed a final sample of 533 individuals (Age: M = 30.74, SD = 10.69; 51.04% Female, 48.21% Male, 0.38% Other, 0.38% Prefer not to say). The sample was 27.58% AAPI (n = 147), 20.45% Black (n = 109), 19.70% Latinx (n = 105), 24.02% White (n = 128), and 8.26% other (n = 44). Additional demographic information is available in the supplemental materials. When completing the survey, participants first filled out the demographic information including age, gender, race/ethnicity, employment status, household size and income, their current zip code, if they are an essential worker, and if they received hazard pay. Next, assessments of all psychological variables (e.g., trust, knowledge, health literacy) and open-ended questions were presented in a randomized order. Political affiliation was assessed at the end of the survey.

Overwhelmingly, individuals reported that they had not been contacted by a contact tracer previously (91.44%), that they believed COVID-19 was not a hoax (98.05%), and that they or someone they knew personally had tested positive for COVID-19 (68.63%) at the time of the study.

5.2. Measures

Intentions to Comply with Contact Tracing. We created a four-item measure asking respondents how likely they were to respond to a contact tracer's requests (e.g., “If a contact tracer contacted you how likely would you be to give all asked information?“)—responses were on a five-point likelihood scale.

Trust in Contact Tracers, Healthcare Providers, and Government Health Officials. We measured trust, or the willingness to be vulnerable based on the belief that another party is reliable, competent, or concerned about one's own interests (Mayer et al., 1995; McAllister, 1995), for three targets: (1) contact tracers, (2) healthcare providers, and (3) government health officials. The same 16-item scale (Spreitzer and Mishra, 1999) was repeated for each target, but items were referent-shifted to refer to the different targets (e.g., “I trust that [Contact Tracers/Healthcare Providers/Government Health Officials] are completely honest with me”). Responses were made on a five-point agreement scale.

Knowledge about Contact Tracing. We constructed a 20-item contact tracing knowledge test based on materials provided by reliable sources (e.g., CDC; New York State Health). Questions represented basic facts and information about key tasks performed, information gathered, individuals’ rights and responsibilities, and the efficacy of contact tracing. The full test is available in the supplemental online materials (SOM). Higher scores represented more knowledge of contact tracing.

Health Literacy. We measured respondents' comfort with healthcare information using Chinn and McCarthy's (2013) ten-item measure (“When you need help, can you easily get hold of someone to assist you?“). Responses were made on a three-point frequency scale.

Liberalism-Conservatism. Respondents' political ideologies were measured with a single item asking: “How Liberal versus Conservative would you characterize yourself?” A seven-point response scale was used ranging from “far left liberal” to “far right conservative”—lower values indicated more liberal beliefs.

Financial Insecurity. Financial insecurity was measured using the Financial Chronic Stress Scale (Lantz et al., 2005) to determine perceptions of respondents' financial needs being met (e.g., “How satisfied are you with your/your family's present financial situation?“). Responses were made on a five-point satisfied to not at all satisfied scale.

Open-Ended Questions. To gauge respondents' thinking about contact tracing, we asked respondents five open-ended questions about contact tracing. We asked respondents: (1) “Who are contact tracers; ” (2) “Can contract tracers get you in trouble for not following public health guidelines? If so, how? ” (3) “Under what circumstances would you respond to a request for contact tracing? ” (4) “What would make it more likely that you would follow a contact tracer's guidance; ” and (5) “What do you think contact tracers could do or say to make it more likely that you would trust and listen to them?”

6. Results

Descriptive statistics and intercorrelations of study variables are presented in Table 1 collapsed across all groups (race-disaggregated correlations are presented in the SOM). Preliminary analyses tested for mean group differences on all study variables between the four main racial subgroups. These results, presented in Table 2 , reveal that for all variables but political affiliation, there were significant subgroup differences. As expected, White individuals appeared to be higher in trust in government health officials and health literacy; these were also the variables with the largest race differences as evidenced by effect sizes. However, Whites were not substantially higher than the other groups on their willingness to comply with tracing requests and other forms of trust. In general, the Latinx sample had the lowest scores for trust in contact tracers, healthcare professionals, and particularly government officials, and was also less willing to comply with tracing requests. Effect sizes identify the largest racial differences in trust in government health officials, health literacy, and contact tracing knowledge. Prior to hypothesis testing, a confirmatory factor analysis including all items for variables tested (loaded onto their respective factors) yielded marginally good fit: RMSEA = .055 (90% CI:,053, 0.056), SRMR = 0.067, CFI = 0.822, TLI = 0.816, likely due to correlated error terms among the trust items which shared stems but whose targets differed (e.g., contact tracer, healthcare provider).

Table 1.

Descriptive statistics and correlations of study variables for full sample (N = 533).

Variable	M	SD	1		2		3		4		5		6		7
1. Likelihood of Compliance	4.33	0.95	(.95)
2. Contact Tracing Knowledge	0.78	0.12	.28	**	–
3. Trust in Contact Tracers	3.84	0.78	.53	**	.19	**	(.96)
4. Trust in Healthcare Professionals	4.04	0.72	.37	**	.13	*	.62	**	(.96)
5. Trust in Government Health Officials	3.25	1.00	.34	**	.01		.54	**	.51	**	(.97)
6. Health Literacy	2.44	0.35	.10	*	.15	**	.09	*	.13	*	.07		(.74)
7. Financial Insecurity	2.41	0.94	−.02		−.03		−.06		−.08		−.11	*	−.10	*	(.86)
8. Political Conservatism	2.89	1.55	−.31	**	−.28	**	−.24	**	−.13	*	−.13	*	−.03		−.04

Note. Sample size (N) = 533. M = mean, SD = standard deviation. Cronbach's alpha reliability estimates are in italics and parentheses on the diagonal. Asterisks denote correlations that are significant at *p < .05 and **p < .001.

Table 2.

Mean group differences on all study variables by racial subgroup.

Variable	Black		Latinx		AAPI		White		ANOVA Results
	M	SD	M	SD	M	SD	M	SD	F	p	ηp2
Likelihood of Compliance	4.39	0.81	4.13	1.18	4.47	0.70	4.39	1.02	2.96	.032	0.02
Contact Tracing Knowledge	0.74	0.13	0.77	0.14	0.81	0.11	0.80	0.11	6.99	<.001	0.04
Trust in Contact Tracers	4.00	0.67	3.72	0.89	3.77	0.63	3.91	0.86	3.23	.022	0.02
Trust in Healthcare Professionals	4.15	0.66	3.91	0.84	4.03	0.59	4.13	0.71	2.86	.036	0.02
Trust in Government Health Officials	3.49	1.04	2.97	0.96	3.03	0.89	3.60	0.97	13.31	<.001	0.08
Health Literacy	2.44	0.35	2.46	0.32	2.34	0.38	2.55	0.30	8.61	<.001	0.05
Financial Insecurity	2.46	0.89	2.57	0.95	2.24	0.88	2.34	0.95	2.98	.031	0.02
Political Conservatism	3.08	1.71	2.88	1.54	2.82	1.27	2.87	1.71	0.64	.587	0.00

Note. N: Black = 109, Latinx = 104, AAPI = 146, White = 128. ANOVA results and eta squared effect sizes are from a one-way test (df[3, 485]) to compare all four group means on each variable.

Hypotheses were tested with a multigroup path analysis by fitting the model in Fig. 1 to the data using the lavaan package (Rosseel, 2012) in R. To control for political orientation, we residualized all variables with respect to political orientation; compliance intentions were also residualized with respect to financial insecurity. Three models tested all hypothesized relationships with increasing levels of specificity to allow for comparisons among racial/ethnic groups: Model 1 - overall/all groups combined; Model 2 - White individuals compared to all non-White individuals with unconstrained model parameters so that relations within each group were allowed to vary; Model 3 - four major racial/ethnic groups (i.e., AAPI, Black, Hispanic/Latinx, and White) compared similarly with unconstrained parameters.

The overall model across racial groups (Model 1) fit the data well, and all direct and indirect paths were significant, supporting all study hypotheses. Full results for Model 1 are presented in the SOM. Model 2, a two-group model (White, non-White), also fit the data well, but not significantly better than Model 1 (see SOM Table 6). Model 3 calculated separate path estimates for AAPI, Black, Hispanic/Latinx, and White respondents, and model fit was good, x ² (24) = 43.14, p = .010, CFI = 0.96, TLI = 0.92, RMSEA = 0.08 [90% CI: 0.04, 0.12], SRMR = 0.05. Additionally, a model freely estimating the paths among the groups fit better than the model constraining the paths to be equal across groups (△x ² (18) = 37.57, p < .05) suggesting subgroup differences in the relations among the variables in the hypothesized model. As Model 3 represents the most detailed view of the sample racial/ethnic divisions, we report results from this model to evaluate the hypotheses for each unique subgroup (see Table 3 , Table 4 , and Fig. 1).

Table 3.

SEM overall and within-group direct effects.

Relationship	Group	B		SE B	CI-L	CI–U
CT Trust → CT Compliance Intentions	Overall	.48	***	.04	.48	.66
	A	.46	***	.08	.32	.62
	B	.45	***	.10	.35	.75
	L	.33	***	.11	.20	.64
	W	.62	***	.08	.57	.87
CT Knowledge → CT Compliance Intentions	Overall	.17	***	.29	.80	1.94
	A	.03		.44	−.68	1.05
	B	.13		.10	−.26	2.17
	L	.31	***	.79	1.21	4.29
	W	.20	*	.59	.61	2.92
Health Trust → CT Trust	Overall	.44	***	.04	.39	.54
	A	.36	***	.08	.22	.54
	B	.61	***	.08	.48	.81
	L	.49	***	.09	.34	.67
	W	.27	***	.09	.14	.48
Gov Trust → CT Trust	Overall	.31	***	.03	.18	.29
	A	.24	***	.05	.07	.27
	B	.23	*	.05	.05	.25
	L	.25	*	.08	.08	.37
	W	.51	***	.07	.31	.57
CT Knowledge → CT Trust	Overall	.11	***	.22	.28	1.12
	A	.09		.42	−.33	1.33
	B	.05		.36	−.38	1.04
	L	.14		.52	−.003	2.03
	W	.11		1.71	−.12	1.77
Health Literacy → CT Knowledge	Overall	.15	***	.01	.02	.08
	A	.13		.02	−.01	.08
	B	.25	*	.03	.02	.15
	L	.13		.04	−.02	.12
	W	.20	*	.03	.01	.13

Note. A = AAPI, B = Black, L = Latinx, W = White, CT Trust = Trust in contact tracers, CT Compliance Intentions = intentions to comply with contact tracer requests, CT Knowledge = Knowledge about contact tracing, Health Trust = Trust in healthcare providers, Gov Trust = Trust in government officials. N: Overall = 487, AAPI = 146, Black = 109, Latinx = 104, White = 128. Parameter estimates are standardized. CI-L = lower bound of 95% confidence interval; CI–U = upper bound of 95% confidence interval. *p < .05, **p < .001.

Table 4.

SEM overall and within-group indirect effects.

Relationship	Group	B		SE B	CI-L	CI–U
Health Trust → CT Trust → CT Compliance Intentions	Overall	.21	***	.04	.19	.35
	A	.18	***	.05	.09	.27
	B	.35	***	.08	.20	.51
	L	.21	***	.07	.08	.34
	W	.22	***	.07	.09	.35
Gov Trust → CT Trust → CT Compliance Intentions	Overall	.15	***	.03	.09	.19
	A	.08	*	.03	.03	.13
	B	.08	*	.03	.02	.15
	L	.10	*	.04	.02	.17
	W	.32	***	.06	.21	.43
CT Knowledge → CT Trust → CT Compliance Intentions	Overall	.05	*	.13	.17	.67
	A	.23		.20	−.21	.57
	B	.18		.20	−.21	.57
	L	.42		.24	−.05	.90
	W	.59		.35	−.10	1.29
Health Literacy → CT Knowledge → CT Compliance Intentions	Overall	.03	*	.03	.03	.13
	A	.01		.02	−.02	.04
	B	.08		.06	−.04	.20
	L	.14		.11	−.07	.34
	W	.12		.07	−.01	.25
Health Literacy → CT Knowledge → CT Trust →CT Compliance Intentions	Overall	.01	*	.01	.01	.04
	A	.01		.01	−.01	.03
	B	.02		.02	−.02	.05
	L	.02		.02	−.02	.06
	W	.04		.03	−.02	.10

Note. A = AAPI, B = Black, L = Latinx, W = White, CT Trust = Trust in contact tracers, CT Compliance Intentions = intentions to comply with contact tracer requests, CT Knowledge = Knowledge about contact tracing, Health Trust = Trust in healthcare providers, Gov Trust = Trust in government officials. N: Overall = 487, AAPI = 146, Black = 109, Latinx = 104, White = 128. Parameter estimates are standardized. CI-L = lower bound of 95% confidence interval; CI–U = upper bound of 95% confidence interval. *p < .05, **p < .001.

In support of Hypothesis 1, trust in contact tracers was significantly positively associated with willingness to comply with contact tracing requests, and this effect was consistent across each of the four racial/ethnic subgroups (AAPI β = .46, Black β = 0.45, Hispanic β = .33, White β = 0.62). In support of Hypotheses 2–3, trust in healthcare providers (AAPI β = .36, Black β = 0.61, Hispanic β = .49, White β = 0.27) and trust in government health officials, (AAPI β = .24, Black β = 0.23, Hispanic β = .25, White β = 0.51) significantly predicted trust in contact tracers for each subgroup, although contact tracing knowledge did not (AAPI β = .09, Black β = 0.05, Hispanic β = .14, White β = 0.11), failing to support Hypothesis 4 for any racial/ethnic subgroup.

Further, in support of Hypothesis 5a, the results suggested that trust in contact tracers did mediate the association between trust in healthcare professionals and willingness to comply with tracing requests for all racial/ethnic groups (AAPI β = .16, Black β = 0.28, Hispanic β = .16, White β = 0.17). Similar support was found for Hypothesis 5 b, with trust in contact tracers mediating the association between trust in government health officials and tracing compliance willingness for all racial/ethnic groups (AAPI β = .11, Black β = 0.10, Hispanic β = .08, White β = 0.32). However, Hypothesis 5c, which anticipated a partial mediation of trust in contact tracers on the association between contact tracing knowledge and willingness to comply with tracing requests, was not supported for any subgroup (AAPI β = .04, Black β = 0.02, Hispanic β = .05, White β = 0.07). These results highlight the importance of trust for encouraging willingness to comply with contact tracing requests among individuals from different racial/ethnic backgrounds. The results also identify predictors of trust in contact tracers, suggesting that increased trust in healthcare providers and government health officials may also increase the likelihood of compliance with tracing requests by building trust in contact tracers.

Next, we examined the predicted positive direct effect of contact tracing knowledge on individuals' willingness to comply with contact tracing requests (Hypothesis 6). The results were mixed, with the positive association significant for the White (β = 0.20) and Hispanic (β = 0.31) respondents, but not for AAPI (β = 0.03) or Black (β = 0.13) respondents. Hypothesis 7, predicting health literacy as an antecedent to contact tracing knowledge, also received mixed support, with positive relations for the White (β = 0.20) and Black (β = 0.25) respondents, but not for the Hispanic (β = 0.13) or AAPI (β = 0.13) respondents. We found no support for Hypothesis 8, that contact tracing knowledge partially mediates the relationship between health literacy and willingness to comply with tracing requests, for any group (AAPI β < .01, Black β = 0.03, Hispanic β = .04, White β = 0.04). We similarly found no support for the hypothesized serial mediation, whereby health literacy influences compliance intentions indirectly through increased knowledge and trust in contact tracers. Thus, Hypothesis 9 was not supported (AAPI β = .01, Black β = 0.01, Hispanic β = .01, White β = 0.01). Altogether, knowledge of contact tracing appears to play a weaker role in predicting individuals' willingness to comply with tracing requests, relative to trust in contact tracers, and primarily for White respondents. Moreover, health literacy failed to demonstrate indirect effects on individuals’ willingness to comply with tracing requests through knowledge or trust. As noted next, qualitative results dovetail with these quantitative results insofar as respondents reported the importance of trust, over knowledge, in their willingness to comply.

6.1. Open-ended responses

To provide more nuance to these quantitative results, we categorized responses to five open-ended questions into similar content categories to provide a picture of contact tracing in communities of color. Categorization of open-ended responses proceeded in three phases. First, the second author extracted the unique ideas from each response for later coding. So, for example, the 533 raw responses to open-ended item four was separated into 667 unique ideas because some respondents included more than one unique explanation/rationale for their response. In the second phase, the same author created an initial list of content categories in which to categorize similar responses, including definitions for each category. The other two study authors reviewed and edited the categories and definitions, obtaining consensus before moving on to phase three.

In phase three, two trained research assistants (who were blind to study hypotheses) reviewed and categorized each response into the categories provided from phase two. Frame-of-reference training (Roch et al., 2012) taught coders about category definitions and distinctions and included practice on the same set of responses to calibrate shared understanding of the categories and definitions. Finally, coders were trained to categorize responses as “unclear” or “unclassified” if they did not fit cleanly into one content category to avoid false classifications. The inter-rater agreement rate between independent coders across all five open-ended questions was 67%, suggesting that the training was effective. Discrepancies in classification were resolved by the third author. This procedure was applied to each of the five open-ended questions individually. Here, we describe the overall results of the content coding, but full results across all content categories are available in the SOM. We also report results from chi-square tests for significant differences of the proportions of responses across categories within each racial group, with proportions broken apart by race presented in the SOM.

The first question asked respondents to describe, in their own words, who are contact tracers. Accuracy was judged based on whether any aspect of the response accurately described the work of a contact tracer, even if other aspects were not correct. This coding decision was made given the novelty of the position for most people, and recognizes that respondents may have had imperfect knowledge but generally understood the responsibilities of contact tracers. Based on this coding, most responses (74.11%) were accurate and very few were inaccurate (6.19%). Interestingly, 10.13% and 3.38% of the open-ended responses identified contact tracers as either public health/healthcare officials or government agencies, respectively, reinforcing the consideration of factors related to trust in healthcare and government as tied to contact tracing. On the whole, though, and in line with the quantitative results of the knowledge test, respondents have a correct understanding of the general role of contact tracers. Categorization of these responses did not differ between the racial groups.

Second, respondents answered whether they believed contact tracers could “get you in trouble for not following public health guidelines.” Any response that suggested punishment for lack of compliance was coded as a “yes”. For responses that provided more detail, we were able to code the nature of the punishment. Less than half of the respondents, 41.84%, correctly recognized that contact tracers could not “get you in trouble.” More concerning, around 52% of the respondents believed that contact tracers could report noncompliance for punishment. Results suggest differences in reports of punishment across races, $\chi$ ² (4) = 26.89, p < .001. Specifically, AAPI and Black individuals more frequently report the possibility of punishment compared to the other races. 11.82% of respondents reported that the punishment could be from government/law enforcement, whereas 11.26% identified a source other than law enforcement (e.g., employers). In general, the frequencies of punitive punishments did not differ significantly across racial groups suggesting that misperceptions about legal punishment does not vary by race. Misperceptions about punishment by nonlegal means, however, did differ by race such that Latinx, AAPI, and Black respondents reported higher frequencies of nonlegal punishment for noncompliance, compared to White and other racial respondents, $\chi$ ² (4) = 11.60, p = .02.

Interestingly 22.14% of the respondents reported being uncertain as to whether or not contact tracers could report noncompliance for punishment. Thus, respondents' knowledge of what happens when public health guidelines are violated is flawed and/or uncertain, with a large proportion of the respondents believing contact tracers can punish contacts’ noncompliance. Such inaccurate knowledge can be problematic for encouraging trust in contact tracers insofar as individuals may find it difficult to trust contact tracers if they are concerned their conversations are not confidential.

Third, respondents answered a question asking when they would be willing to comply with tracing requests. Two-thirds of respondents (67.92%) indicated that they would immediately comply with contact tracers, especially if it was necessary due to infection or exposure. Another 23.26% said they would comply if they were given a valid reason for the request. Although only 4.32% of respondents said they would never comply with a contact tracer, research has shown that the spread of COVID-19 can be exacerbated with only a few noncompliant individuals (Kretzschmar et al., 2020). Thus, even this small amount of noncompliance can exacerbate disease spread. These proportions did not differ by race.

Respondents also shared what would make it more likely that they would comply with contact tracer guidance. Although a few respondents indicated they would follow all requests no matter what (16.51%)—particularly Whites (W: 22.31%, B: 17.43%, L: 16.04%, A: 14.19%), others indicated that compliance would be more likely if it was necessary due to exposure or health/safety concerns (17.26%), although this reason was cited more often for people of color than Whites (B: 19.27%, L: 18.87%, A: 17.57%, W: 11.54%). Another reason frequently cited was if the tracer provided a valid rationale (12.57%), although this too differed by race, with AAPI (14.86%) and LatinX (10.38%) participants saying this would be more likely to encourage compliance compared to White (5.39%) and Black (4.59%) participants. The other most provided reasons for compliance included contact tracers clearly communicating guidelines or goals (7.50%), interacting politely (7.32%), and inspiring trust in their skills (6.94%)—though this last reason appeared more important for some groups than others (A: 7.43%, B: 3.67%, W: 3.08%, L: 0.00%). Importantly, 3.56% of the responses indicated that nothing could be done to make them want to comply with contact tracing requests, which is cause for concern (Kretzschmar et al., 2020). Ironically, since Whites were most likely to say they would follow requests no matter what, they were also most likely to say they would NOT follow requests no matter what (W: 6.15%, B: 2.75%, A: 1.35%, L: 0.94%). This suggests somewhat different underlying reasons for contact tracing compliance between racial groups, with Whites tending to use extreme responses of both always and never complying with contact tracers more frequently than communities of color, who more often cited the necessity due to exposure or health and safety concerns.

Finally, respondents indicated they would be more willing to trust contact tracers if they provide valid reasons for contacting the individual and/or information about the virus (16.32%), with this reason less compelling for Black respondents (8.26%) compared to others (A: 15.23%, W: 13.53%, L: 11.43%). The second most-common method suggested for increasing trust was respectful/kind communication (14.26%), although this was seen as less critical for the LatinX respondents (8.57%) compared to others (B: 18.35%, W: 15.04%, A: 13.25%). Clear communication was also mentioned as important for trust (10.51%), and this was largely similar across racial groups. Information security also seemed important: 9.94% of responses noted emphasizing confidentiality/data privacy would inspire trust and another 9.57% expressed wanting the tracer to prove their identity and authority, with this being a more common ask for Black respondents (14.68%) than others (L: 9.52%, A: 8.61%, W: 7.52%). Many others expressed an unequivocal willingness to trust no matter what (21.39%), although again, this was higher for Whites (15.79%) than communities of color (A: 9.94%, L: 9.52%, B: 5.50%).

Synthesizing across these qualitative results, responses emphasize the importance of factors related to trust in contact tracers (e.g., verified identity; protecting anonymity) for willingness to comply, and the importance of professionalism, data security, confidentiality, and strong communication skills for improving trust (see SOM for full results). Thus, in alignment between the qualitative and quantitative results, willingness to comply with contact tracing requests among communities of color appears to be influenced more strongly by factors related to trust in the contact tracer than by knowledge about contact tracing alone. Although there were many similarities across racial groups’ statements regarding their determinants of trust and compliance, non-trivial differences speak to factors that may be more important to some groups than others. For example, Black respondents more frequently stated their desire to have contact tracers verify their identity to establish trust. Communities of color were just as likely to correctly describe who contact tracers are but were more likely than Whites to attribute to them the ability to punish noncompliance (see SOM for full results). This speaks to the need to consider the unique needs of different racial/ethnic groups when trying to encourage contact tracing trust and compliance.

7. Discussion

The results of this study detail the relative importance of trust, compared to knowledge, as influential for peoples' willingness to comply with contact tracing requests, along with some key differences between racial groups. Our results point to four conclusions. First, based on the knowledge test scores and coding of open-ended responses, the sample demonstrated a relatively good understanding of some, but not all, of the basic facts and information about contact tracing. For instance, they generally understood who contact tracers are and what they do, but misconceptions persisted about data privacy and tracers’ ability to punish people for noncompliance. Additionally, trust in contact tracers was neutral on average, but with a wide degree of variability, suggesting that, although many of the people sampled here did somewhat trust contact tracers, a sizable number did not.

A second and important conclusion is that there were significant differences in measures of trust, knowledge, and willingness to comply with contact tracing requests across racial/ethnic subgroups. The Latinx sample reported the lowest levels of trust (in government health officials, healthcare professionals, and contact tracers) and willingness to comply with tracing requests across all groups compared (see Table 2). The Black sample, despite having the lowest levels of contact tracing knowledge, had the highest levels of trust in contact tracers and were relatively similar to Whites in their compliance intentions and trust in healthcare professionals. Finally, although the AAPI sample had lower levels of trust (in contact tracers, healthcare professionals, and government officials) compared to the White sample, they also had the highest levels of knowledge and willingness to comply with tracing requests. These results suggest that trust, knowledge, and willingness to comply with health guidelines are not uniformly lower among communities of color compared to Whites, nor are different racial/ethnic groups all alike. These racial differences are reflected in the responses to open-ended questions. Most notably, communities of color were more likely than Whites to believe that contact tracers can punish people for noncompliance, which is not true. Interestingly, Whites seemed more unequivocal in their support for but also resistance to contact tracing, while communities of color were more likely to see compliance as necessary for health and safety. When it came to building trust, Black respondents were more concerned about tracers verifying their identity than other groups. In sum, broad statements about communities of color as less trusting or more resistant to public health guidelines may be an inaccurate oversimplification—one that could have damaging effects on public health outreach and policy (Boyd, 2021). Instead, those in the field ought to acknowledge that different racial/ethnic groups may have unique and multifaceted concerns about trusting and complying with public health directives.

Third, as expected, trust in contact tracers was a strong predictor of willingness to comply with tracing requests. This suggests that to be vulnerable and share information with contact tracers, the public may need to see contact tracers as reliable, competent, and concerned in their own interests to be willing to follow their guidance (Mayer et al., 1995; McAllister, 1995). We also identified several useful factors that seem likely to build trust in contact tracers among different racial/ethnic groups, including known sources where trust may be able to transfer to unknown contact tracers (Stewart, 2003). Specifically, trust in healthcare providers and government health officials are both associated with increase in compliance intentions mediated through their positive relations with trust in contact tracers. Stated differently, evidence of distrust toward healthcare providers (Chen et al., 2020; Cristancho et al., 2008; Whetten et al., 2006) and government officials (Marschall and Shah, 2007; Rhodes et al., 2015) may also extend to contact tracers, thereby serving as a roadblock to individuals’ willingness to comply with tracing requests. Therefore, healthcare providers and public health officials may be able to influence intentions to adhere to contact tracing guidance, with the results suggesting this would be effective for all racial/ethnic groups evaluated. This is consistent with research showing that trust transfer from known sources can influence swift trust determinations for unknown, but categorically related parties (Leung et al., 2022). Indeed, past pandemics have shown that trust in both the government and the media (Ali et al., 2020) is associated with preventive health behaviors (Liao et al., 2011). Thus, it is critical that the sources to which people turn for health-related care and information build trust in contact tracers by conveying positive messages concerning the dependability and competence of contact tracers. However, our results also suggest that willingness to comply with tracing requests is indirectly more positively related to trust in government health officials for Whites than for the Black, Latinx, or AAPI samples (see non-overlapping 95% confidence intervals for this indirect effect in Table 4). Thus, building trust in contact tracers and increasing compliance may be accomplished more effectively by healthcare professionals than government officials among communities of color. This may be due, in part, to lower levels of trust in government health officials among communities of color compared to Whites (see Table 2), with trust less likely to transfer from institutions with whom these communities have had more negative experiences in the past (Mayer et al., 1995; Stewart, 2003).

Fourth and finally, knowledge about contact tracing was a relatively less potent predictor of individuals' willingness to comply with tracing requests compared to trust, and its association varied somewhat across racial/ethnic groups. The direct effect of contact tracing knowledge on willingness to comply with tracing requests was significant only for the White and Hispanic respondents, not AAPI or Black respondents. Similarly, health literacy was only positively associated with contact tracing knowledge for the White and Black respondents, not AAPI or Hispanic respondents. Moreover, health literacy showed no evidence of indirect associations with compliance intentions through either knowledge or trust. Thus, knowledge about tracing may not be a strong enough predictor of willingness to comply with tracing requests to adequately motivate compliance intentions. These results also suggest that distal efforts to improve health literacy may not have an appreciable effect on more immediate tracing compliance intentions and may not have a similarly positive effect for all racial groups (although improving health literacy has other important benefits). This coincides with research showing that willingness to comply with the requests of others depends, in large part, on a willingness to be vulnerable to (i.e., trust) the requestor (e.g., Gilles et al., 2011; Mayer et al., 1995), but challenges theories of swift trust which propose that individuals must have some basic knowledge of another (i.e., contact tracer) upon which to build trust (Lewis and Weigert, 1985; Rafaeli et al., 2018; Schilke and Huang, 2018). Instead, these judgments may be largely based on stereotypes about the unknown source's group membership or identity (Robert et al., 2009). These mixed findings seem to be consistent with our qualitative results. For example, although a majority of respondents mistakenly believed that contact tracers could punish individuals for noncompliance, most still intended to comply with tracers' requests, suggesting accurate knowledge may not be as critical of a determinant. Furthermore, these results suggest that even with accurate knowledge, one may not comply with requests if they fear consequences. In short, these results seem to point to the general importance of trust as the determinant of a willingness to comply with the requests of others. Drawing on our conclusions, Table 5 presents policy recommendations to increase the effectiveness of contact tracing, particularly among communities of color.

Table 5.

Policy recommendations to improve contact tracing compliance in communities of color.

Policy Recommendation	Supporting Results
1. Build trust in CT by clearly detailing the motives and authority of CT, confidentiality of the information gathered, necessity of tracing to limit disease spread, and validity of the information/guidance provided.	SEM results identify trust as an antecedent of willingness to comply with CT requests. Qualitative results identify these strategies as likely to increase trust and compliance.
2. Involve healthcare professionals and government health officials in the process of building trust in contact tracers. Recognize, however, that government health officials may be more effective for increasing trust for Whites relative to communities of color.	SEM mediation results identify trust in healthcare providers and government health officials as significant predictors of willingness to comply by increasing trust in CT. Yet, the indirect effect of government health officials is significantly stronger for the White sample than for the Black, AAPI, or Latinx samples. This is not the case for healthcare providers; indirect effects are similar across groups.
3. Maintain and clearly communicate the separation of CT from government authorities, particularly law enforcement and immigration agencies.	SEM results suggest that lower trust in the government would indirectly reduce CT compliance. Qualitative results reveal that many respondents mistakenly believed that CT could share their information with law enforcement agencies.
4. Target knowledge-based interventions on potential sources of mistrust in CT to debunk misinformation (see Lewandowsky et al., 2012).	SEM results suggest that CT knowledge and health literacy play a more limited role as predictors of willingness to comply with tracing requests. Knowledge test and qualitative results suggest people generally understand CT basics, but some ignorance persists that may erode trust.
5. Ensure CT have the skills and knowledge needed to communicate clearly and respectfully. This can be aided by effective personnel selection and training.	Survey respondents indicated that they would be more willing to trust and comply with contact tracers if they were respectful, compassionate, clear, and knowledgeable.
6. Researchers and policymakers should consider the nuanced effects of race in data collection and analysis to inform more targeted recommendations.	Collapsing all respondents into one category without considering the effects of race would have produced different conclusions than we obtained with multi-group analysis.

Note. CT = contact tracing.

7.1. Study limitations

Our study has some limitations that future research can address. First, we acknowledge that our main dependent variable of interest is behavioral intention to comply with contact tracer requests, not actual compliance behavior. Although compliance behavior would be ideal, these data were unrealistic to acquire in the middle of the COVID-19 global pandemic (e.g., accurately recording compliance behaviors amid social distancing requirements). Since behavioral intentions are the most proximal predictor of actual behavior (Ajzen and Fishbein, 1977), we measured this outcome; however, some questions regarding the extent to which intentions translate to behavior remain (Fiske and Taylor, 2017; Webb and Sheeran, 2006). We believe that intentions to comply with contact tracing requests would translate to actual behaviors because aspects of intentions to comply with contact tracing requests make the link to actual behavior stronger. Specifically, intentions to comply represent a cognition that is accessible, related to a vested interest (i.e., remaining safe during the pandemic), and an important behavior (limiting disease spread)—all three of these factors are related to stronger intentions-behavior links (Fiske and Taylor, 2017). Furthermore, intentions to behave are particularly strong predictors of health behaviors (Taylor, 2006) and have been utilized in other studies on contact tracing (e.g., Li et al., 2021). Nevertheless, behavioral intentions do not always produce a strong causal effect on individuals' health behaviors; factors beyond one's control can attenuate this link (Webb and Sheeran, 2006). Thus, future research can investigate actual compliance behaviors, if possible, while controlling for credible constraints.

A second limitation of this study is that all data were single-source, single time point, and came from an online Prolific panel. To partially address concerns of common-method bias we intentionally used short scales with clear, concise, items, and participation in the survey was voluntary and personally relevant (Podsakoff et al., 2012). Additionally, we believe that self-report data was appropriate for this study given the nature of the constructs measured (Chan, 2009). It would not be appropriate to ask anyone else but study participants about the focal constructs of trust, knowledge, and intentions to comply with contact tracing requests. The source of CMV that is of most concern for our study is that same time-point measurement may inflate scores (odsakoff et al., 2012). However, if that were the case, we should see uniformly high correlations among the variables (Lance and Siminovsky, 2014), which we do not. Therefore, although there may yet be some concern about CMV and a longitudinal design would be better to test mediation, we can still learn important things about the relationships between the variables included in this study. Use of Prolific, an online recruitment platform designed for researchers (Palan and Schitter, 2018), may present some limitations for external validity, as such samples tend to be younger and more educated. This is somewhat represented in the data as the average age was low 30.74 (but the variability was wide; SD = 10.69, range = 18–75) and 61.16% of the sample had a bachelor's degree or higher. However, using Prolific allowed us to pre-screen participants based on eligibility criteria and to oversample from different racial/ethnic groups to recruit participants who fit the population of interest; therefore, it is an appropriate sample to use for research (Highhouse, 2009). Nonetheless, future research may look at alternative recruitment methods to recruit individuals who vary more in terms of age, education level, and region. This may be done in concert with technological tools contact tracers use for case management included automated notifications of exposure risk. However, we note with caution that such apps in the U.S. and other neo-liberal societies have been used sparingly and voluntarily, despite their utility, due to concerns of privacy and surveillance (Akinbi et al., 2021), and there is a concern that use of such apps will exacerbate health inequities given that individuals from socio-culturally disadvantaged communities are less likely to adopt and utilize them (Villius Zetterholm et al., 2021).

Finally, it may be noted that the effect sizes predicting compliance intentions in this study are moderate, altogether explaining about one-third of the variance in willingness to comply with contact tracing requests for communities of color in our sample (AAPI r ² = 0.21; Black r ² = 0.23; Hispanic/Latinx r ² = 0.23), but almost one-half of the variance in White respondents’ intentions to comply (r ² = 0.45). Thus, there appear to be other predictors of compliance intentions besides trust and knowledge, particularly for respondents of color, suggesting the need for future research in communities of color. These effects are quite meaningful, however, when considering the severity of the COVID-19 disease among communities of color and the dire need for compliance with health guidelines because any degree of noncompliance with contact tracers will enable disease spread (Kretzschmar et al., 2020). Future research should continue to investigate alternative socio-behavioral determinants of healthcare compliance (Becker and Maiman, 1975).

8. Conclusions

Our results highlight the importance of trust, relative to knowledge, in increasing the potential effectiveness of contact tracing, and introduce some key race-based differences in these relationships. Of the communities of color investigated in this study—AAPI, Black, and Latinx—trust, knowledge, and intentions to comply with contact tracers were not uniformly lower compared to Whites. Instead, contact tracing knowledge and compliance intentions were highest among the AAPI sample, and trust in contact tracers was highest among the Black sample. Furthermore, the different racial/ethnic group were distinguishable in other unique ways, as well (e.g., Blacks were more concerned than others about contact tracers proving their identity to encourage trust; Whites were more unequivocal in their support for and against contact tracing under any circumstances). Similarities among communities of color, compared to Whites, included the more frequent misconception that contact tracers can punish people for noncompliance, and the relatively weaker influence of trust in government health officials as an antecedent of contact tracer trust. Thus, although generalizations about communities of color as less trusting or more resistant to public health guidelines may be an oversimplification, care should still be taken to understand the unique perspectives and needs of different racial/ethnic communities. To this aim, we present a set of policy recommendations that will hopefully help improve contact tracing for all, including for communities of color, to ensure a more equitable response to future public health emergencies.

Author note

This research project was funded by the New York State COVID-19 Minority Health Disparities Award at the University at Albany, SUNY (SUNY Research Foundation Award #8826), Co-Principal Investigators: Jason G. Randall and Dev K. Dalal. The sponsor had no direct involvement in the research project or preparation of this manuscript. This study was pre-registered at the Open Science Framework: https://osf.io/tdw2v. Declarations of interest: none.

Data statement

Data are available from the corresponding author upon request, including during the peer review process. We would also be willing to share the data publicly pending article acceptance.

Author contribution

Jason G. Randall: Conceptualization, methodology, data analysis, and writing. Dev K. Dalal: Conceptualization, methodology, data analysis, and writing. Aileen Dowden: Methodology, data analysis, and writing.

Handling Editor: Blair T. Johnson

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

Data will be made available on request.