Mental Health and Psychosocial Characteristics Associated With COVID-19 Among U.S. Adults

The COVID-19 pandemic has profoundly affected health care systems, the economy, and the health and well-being of people around the world. The potential effects of COVID-19 on mental health have raised grave concerns. For example, some observers have described public panic, depression, anxiety, and posttraumatic stress arising from concerns about infection, illness, and death (1). Others have pointed to the disruptive effects of social distancing on basic human needs for human connection, which may exacerbate underlying individual vulnerabilities for mental illness (2).

Historically, pandemics and disasters have disproportionately affected poor and vulnerable populations, including homeless individuals and those with severe mental illness (3). This may partly be due to disparities in economic resources, access to preventive health care, and psychosocial determinants of health (4). Through neurobiological pathways, poor mental health may weaken immune systems and increase the risk for developing infectious diseases (5). Thus, COVID-19 may negatively affect mental well-being, but, conversely, poor mental and psychosocial functioning may also increase the risk for COVID-19 infection.

We recruited a large national sample of middle- and low-income U.S. adults who reported that they had tested positive for COVID-19 (COVID-19+), had tested negative (COVID-19−), or had not been tested to understand the sociodemographic, psychosocial, and mental health factors that may put adults at risk for COVID-19 infection. We hypothesized that poor mental and psychosocial functioning would be associated with testing positive for COVID-19.

Methods

A national sample of 6,607 U.S. adults was recruited in May and June 2020 to examine health and social well-being during the COVID-19 pandemic. Eligible for the study were adults who were at least 22 years old, lived in the United States, and reported an annual personal gross income of ≤$75,000. Participants were recruited through Amazon Mechanical Turk (MTurk), which is an online labor market with >500,000 participants across 200 countries and has become a popular site for conducting surveys and online interventions. To ensure data quality, we invited only participants who had completed ≥50 approved Human Intelligence Tasks (HITs) and had a HIT approval rating of ≥50%. Results from cross-sample investigations have indicated that data obtained from MTurk have the same level of quality as or higher than data collected from traditional subject pools such as community samples, college students, and professional panels (6).

In total, 9,760 individuals agreed to participate in the study, of whom 6,762 (69.3%) met the eligibility criteria; 155 workers were removed because they failed a validity check (i.e., they failed on three items from the validity scales of the Minnesota Multiphasic Personality Inventory–2). The final study sample consisted of 6,607 participants (67.7% of the initial recruitment sample) from all 50 U.S. states and the District of Columbia. Using data from the 2018 American Community Survey for comparison, we found that the sample was comparable to the general U.S. population with respect to key demographic characteristics, including sex, race, ethnicity, and geographic region; the exception was age, with our sample being younger (mean age=37.9 years vs. general population age=49.5). (A table showing demographic data is available as an online supplement to this report.) All participants provided informed consent and were compensated for their participation through MTurk; study procedures were approved by the institutional review board at the University of Texas Health Science Center at Houston.

Sociodemographic information was based on self-reports. Veteran status was defined as “ever served on active duty in the U.S. military,” and history of homelessness was defined as “did not have a stable nighttime residence (such as staying on streets, in shelters, cars, etc.).” COVID-19 status was assessed by asking participants whether they had been tested for COVID-19 and if so, what the outcome was (i.e., positive or negative test result). They were also asked whether anyone close to them (e.g., friends or family) had tested positive for COVID-19.

Social connectedness was assessed with the Medical Outcomes Study Social Support Survey–Short Form (7), the short-form University of California, Los Angeles (UCLA), Loneliness Scale (8), and a question about the number of close friends and relatives that participants had. Health status was assessed by asking participants whether they had ever been diagnosed as having any of 22 different health conditions (e.g., cancer, heart disease, or arthritis) and summing the total number reported (9).

Psychiatric history was assessed by asking participants whether they had ever been diagnosed as having any of nine mental health conditions or substance use disorders. Current mental health and substance use were assessed with validated measures, including the Patient Health Questionnaire–4 (PHQ-4) (10), assessment of any past 2-week suicidal ideation, and the Alcohol Use Disorders Identification Test (11). Participants were also asked whether they currently smoked cigarettes, vaped, or were issued any illicit drugs during the past month.

Participants were divided into three groups—COVID-19+, COVID-19−, and not tested for COVID-19—whose sociodemographic, psychosocial, and clinical characteristics were compared in bivariate analyses using analysis of variance and chi-square tests. Post hoc pairwise tests were conducted with Tukey’s honestly significant difference test and chi-squared tests. A series of multinomial logistic regressions were conducted to examine sociodemographic, psychosocial, and clinical characteristics associated with COVID-19+ status, with separate regressions for past psychiatric diagnoses and for measures of current mental health and social support.

Results

In the total study sample (N=6,607), 5.4% (N=354) were COVID-19+, 27.5% (N=1,819) were COVID-19−, and 67.1% (N=4,434) had not been tested for COVID-19. Bivariate comparisons between groups revealed statistically significant group differences in nearly every sociodemographic category assessed (see online supplement to this report). Compared with the COVID-19− and untested groups, the COVID-19+ group was more likely to consist of veterans and have individuals with a history of homelessness or who reported a greater number of close friends or relatives, greater social support, and a greater sense of loneliness.

The COVID-19+ group also reported a significantly greater number of chronic medical conditions and was more likely to report having been given diagnoses of a range of psychiatric disorders, including schizophrenia-spectrum disorders, posttraumatic stress disorder (PTSD), bipolar disorder, anxiety disorder, major depression, and alcohol and drug use disorders as well as traumatic brain injury. The COVID-19+ group was also more likely to screen positive for current major depression, generalized anxiety disorder, and recent suicidal ideation and was also more likely to report recent use of illicit drugs, cigarettes, or vaping devices.

Multivariate analyses were conducted to examine sociodemographic, psychosocial, and health characteristics independently associated with being in the COVID-19+ group. As shown in Table 1, a multinomial logistic regression found, in order of magnitude, that having any friends or family who were COVID-19+, being a veteran, having several close friends or relatives, having any history of homelessness, having an advanced degree, being a student, and being younger were each statistically significantly associated with COVID-19+ status. None of the psychiatric diagnoses were statistically significantly associated with COVID-19+ status (see online supplement for full table). When the regression was repeated with a veteran × history of homelessness interaction term, the interaction effect was not statistically significant (adjusted odds ratio [aOR]=0.68, 95% confidence interval [CI]=0.32–1.44), indicating that veteran status and homelessness history each independently increased risk for COVID-19+ status. When the regression was conducted without veteran status or history of homelessness in the model, among the psychiatric diagnoses only PTSD emerged as being significantly associated with COVID-19+ status (aOR=1.57, 95% CI= 1.07–2.31).

A second multinomial regression model was conducted, replacing psychiatric diagnoses with current measures of mental health and social support. The results were similar to the ones reported above; no measure of current social support and mental health was statistically significant, although both veteran status and history of homelessness remained significant (see online supplement). A veteran × history of homelessness interaction effect was not significant (aOR=0.69, 95% CI=0.32–1.46). When veteran status and history of homelessness were excluded, scores on the UCLA Loneliness Scale (aOR=1.12, 95% CI=1.00–1.25), any illicit drug use in the past month (aOR=1.55, 95% CI=1.03–2.32), and current vaping (aOR=1.73, 95% CI=1.17–2.58) were each significantly associated with COVID-19+ status.

Discussion

This national study of middle- and low-income adults during the COVID-19 pandemic confirmed some factors known to be associated with testing positive for COVID-19 infection. It is important to state some caveats of this study up front. COVID-19 has not spread uniformly across the United States and has varied geographically (12), which we tried to account for but without geographic specificity. Moreover, the study sample did not have equal access to COVID-19 testing, and most participants had not been tested. We focused on examining those who had been tested, so the study results need to be interpreted within the context of these important study limitations.

We found that those with close friends and family who had COVID-19 and those who had more close friends or family in general were at greater increased risk for COVID-19 infection. Being a part- or full-time student also increased the risk for testing positive for COVID-19, possibly because of greater exposure to others. These findings support the practice of social distancing and wearing personal protective equipment when interacting with others to reduce COVID-19 transmission. However, in terms of COVID-19–associated mental health and psychosocial factors, our hypothesis was only partially confirmed. Mental health factors did not emerge as significant risk factors for COVID-19 per se. However, adults who had served in the military or had histories of homelessness were significantly more likely to test positive for COVID-19, and both veteran and homeless populations are known to have higher rates of mental illness and poor social functioning (13, 14). Controlling for sociodemographic and other psychosocial factors, we found that veterans were more than twice as likely to test positive for COVID-19 than were nonveterans. Moreover, adults with any history of homelessness were nearly twice as likely to test positive for COVID-19 than were domiciled adults.

Veterans and adults with histories of homelessness are distinct populations that were found to have an increased risk for COVID-19. A possible common thread between these two populations with respect to COVID-19 may be that both veterans and those with histories of homelessness are likely to have lived in congregate settings, such as military bases or homeless shelters and transitional housing settings. However, we did not collect data on time since serving in the military or last episode of homelessness, but both of these populations may be more likely to congregate or interact socially because of shared identities and needs compared with those in the general population. Alternatively, they may share no common thread, and each population may have separate reasons for being at higher risk for COVID-19 than others. For example, veterans may be at increased risk because of vulnerabilities related to military service and combat, and those who are unstably housed may have increased COVID-19 exposure. However, we can only speculate on the reasons; it may also be that these characteristics are proxies for other factors more closely associated with COVID-19 risk. Certainly, these two populations deserve further study, and service models that encourage independent housing should be encouraged (15); special attention to the medical needs and clinical risk for COVID-19 among individuals in these populations may also be needed.

There were several important additional study limitations. This was a cross-sectional survey, so we could not infer the directionalities or causalities of the associations identified. COVID-19 test results were based on self-reports and were not confirmed by lab tests. Importantly, assessments have revealed variable validity of different COVID-19 tests, whose possible differences we were not equipped to examine. We also did not ask participants about any COVID-19 symptoms, and there is increasing evidence that many individuals who test positive for COVID-19 are asymptomatic and may have different transmission profiles and risks (16). Although we recruited a heterogenous sample from all four major geographic regions of the United States, no stratified sampling was used, and the sample may not be nationally representative, particularly of those at risk for homelessness who may be less likely to have online access. Prevalence and epidemiological values derived from the data should be interpreted with caution because they are likely not representative; however, the findings on associations between the different variables studied can be interpreted with greater confidence. These limitations notwithstanding, the study had several strengths, including the large national sample, the assessment of various psychosocial and mental health variables not readily available in medical records, and findings that may contribute to ongoing efforts to address the COVID-19 pandemic.

Conclusions

Having close personal contacts, being a veteran, and having a history of homelessness were three major factors associated with contracting COVID-19. These findings reinforce the importance of social distancing and highlight several population subgroups that may be at increased risk for contracting the virus.