Serious Mental Illness Diagnosis and COVID-19 Vaccine Uptake in the Veterans Health Administration

We used retrospective observational data from the VHA Corporate Data Warehouse, a national repository of VHA electronic health record (EHR) data, for all analyses. We used data from all veterans ages ≥18 years who had one or more VHA outpatient visits from March 1, 2018, through February 29, 2020. We identified patients with a serious mental illness diagnosis (0=no, 1=yes) on the basis of the presence of serious mental illness diagnoses (one or more inpatient visits or two or more outpatient visits) during the aforementioned period, as indicated by VHA-defined ICD-10 codes. (A list of the ICD-10 diagnostic codes and a table with frequency counts are included in an online supplement to this report.) Therefore, the comparison group comprised all veterans in the cohort without a serious mental illness diagnosis.

We identified VHA COVID-19 vaccinations through EHR procedure codes and immunization records for appointments from December 1, 2020, to June 1, 2021. This study was approved by the VA Greater Los Angeles Institutional Review Board.

We collected demographic and clinical characteristics hypothesized to predict both serious mental illness diagnosis and COVID-19 vaccination status for use in statistical modeling. Variable selection was informed by Andersen’s behavioral model of health services use (9)—specifically, the primary determinants (population characteristics) and health behavior (personal health practices and use of health service) domains. To measure primary determinants, we used EHR tables listing patient demographic characteristics to obtain age, gender, marital status, and race-ethnicity. We determined use of health insurance not provided by the U.S. Department of Veterans Affairs (VA) on the basis of billing information from patients’ most recent VHA appointment before the VHA vaccine rollout in December 2020. We assessed history of posttraumatic stress disorder (PTSD) and substance use disorder on the basis of the presence of VHA outpatient visits (one or more inpatient or two or more outpatient visits) with corresponding ICD-10 codes from March 1, 2018, through February 29, 2020. To measure health behaviors, we assessed smoking status on the basis of VHA visits during the same period, and we assessed pre–COVID-19 influenza vaccination history on the basis of VHA visits from October 1, 2018, through May 31, 2019.

Finally, we included two measures of medical morbidity: the Gagne comorbidity index score, which assesses probability of death on the basis of medical and psychiatric diagnoses during the previous 12 months, and the VHA Care Assessment Needs (CAN) score, an internal VHA measure that estimates 1-year probability of hospitalization or death on the basis of an algorithm that includes both clinical and sociodemographic variables. Veterans are then assigned a percentile based on their probability relative to the general VA population. For the Gagne comorbidity index, possible scores range from −2 to 26, with higher scores indicating increased mortality risk. For the CAN, possible scores range from 0 to 100, with higher scores indicating higher risk of death or hospitalization. To provide a descriptive overview of sample characteristics, we calculated means and standard deviations for veterans with serious mental illness and those without serious mental illness and conducted independent-samples t tests for continuous variables. For categorical variables, we calculated percentages by group and conducted chi-square tests of independence.

To test our hypothesis, we estimated the average treatment effect (ATE) of serious mental illness on COVID-19 vaccination, where patients with serious mental illness were the “treated” population and patients without serious mental illness diagnoses were the “control” population. ATE estimation is commonly used in observational designs and, consequently, includes methods for controlling confounding such as inverse-probability weighting (IPW) (10). IPW is a method for equalizing control and treatment groups on a series of covariates before estimating the ATE. Thus, the ATE is estimated in two steps. First, we equalized the serious mental illness and control groups on all aforementioned variables by deriving inverse-probability weights from a treatment model with serious mental illness diagnosis as the dependent variable and baseline covariates as independent variables. Second, the ATE on COVID-19 vaccination was estimated. All analyses were carried out with Stata, version 15.1.

We identified 4,890,693 veterans with available data for all model covariates who met our inclusion criteria; 4% had an EHR serious mental illness diagnosis. (See the online supplement to this report for descriptive statistics and tests of differences between the groups with and without serious mental illness on all variables.)

Overall, 48% of veterans diagnosed as having serious mental illness received a COVID-19 vaccination, compared with 46% of those without serious mental illness (χ²=299.5, df=1, p<0.001). Veterans with serious mental illness were on average younger (mean±SD=55.5±14.3 years) than veterans without serious mental illness (62.0±16.4 years) (t=173.9, df=4,890,691, p<0.001). A larger proportion of veterans with serious mental illness were women, compared with veterans without serious mental illness (17% [N= 32,932] vs. 9% [N=424,446]; χ²=13,019.2, df=1, p<0.001). Patients diagnosed as having serious mental illness were less likely to be White, compared with veterans without serious mental illness (63% [N=124,764] vs. 71% [N=3,334,393]; χ²=7,150.1, df=1, p<0.001), were less likely to have non-VA insurance coverage (56% [N= 110,475] vs. 68% [N=3,189,017]; χ²=12,454.7, df=1, p<0.001), and were more likely to be single or to have never married (27% [N=52,876] vs. 14% [N=636,308]; χ²=55,391.3, df=1, p<0.001).

Patients with serious mental illness showed evidence of higher medical morbidity than patients without serious mental illness: they had higher Gagne comorbidity index scores than those without serious mental illness (1.3±1.8 vs. 0.5±1.5; t=216.2, df=4,890,691, p<0.001), and a higher proportion had CAN scores greater or equal to the 75th percentile of VHA users (56% [N= 109,583] vs. 24% [N=1,103,923]; χ²=124,322.9, df=1, p<0.001). Patients with serious mental illness also had higher rates of PTSD compared with those without serious mental illness (40% [N=78,695] vs. 19% [N=909,080]; χ²= 49,304, df=1, p<0.001), as well as substance use disorders (40% [N=79,640] vs. 11% [N=533,379]; χ²= 144,987.7, df=1, p<0.001). Regarding health behaviors, patients with serious mental illness were more likely than patients without serious mental illness to be current smokers (35% [N=69,462] vs. 20% [N= 919,779]; χ²=29,078.2, df=1, p<0.001) and were more likely to have a history of influenza vaccination (55% [N=109,395] vs. 47% [N= 12,188,181]; χ²=5,849.1, df=1, p<0.001).

The effect of serious mental illness on the likelihood of COVID-19 vaccination was not statistically significant. The estimated vaccination rate among patients with a diagnosis of serious mental illness (45.8%) was effectively the same as the rate among patients without serious mental illness (46.2%), and serious mental illness had only a minor effect on the vaccination rate (ATE=–0.4%, 95% CI=−0.8% to 0.1%). Thus, after weighting based on demographic and clinical characteristics, we found that patients diagnosed as having serious mental illness and those diagnosed as having no serious mental illness were equally likely to receive a VHA COVID-19 vaccination.

It was necessary to evaluate covariate balance as a check on the extent to which the IPW had minimized confounding by equalizing the serious mental illness and control groups on the covariates. We did so by calculating weighted standardized differences between serious mental illness and control groups and ratios of serious mental illness group variances on each covariate. In these analyses, we observed standardized differences that fluctuated around the zero and homogeneous group variances. Thus, the IPW apparently created homogeneous covariate distributions between serious mental illness groups.

We found that COVID-19 vaccine receipt was comparable among VHA patients with serious mental illness and those without serious mental illness. Although we know of no previous study reporting on associations between serious mental illness and COVID-19 vaccination, previous studies in the VHA population have reported higher use of general medical and mental health services among persons with psychiatric diagnoses. In addition, early studies of associations between mental illness diagnoses and COVID-19 testing reported higher rates among persons with psychiatric diagnoses in the VHA population and other populations (11, 12). Our finding of similar vaccine uptake by persons with serious mental illness and those without serious mental illness supports past research reporting similar or higher access to COVID-19 health services in closed or hybrid health care systems.

VHA outreach activities may have helped address COVID-19 vaccine access barriers that are common in the population of individuals with serious mental illness. VHA staff called eligible patients directly to schedule appointments, which reduced logistic barriers. The VHA disseminated educational information about the vaccine by using various media channels, addressing lack of medical knowledge and mistrust among some individuals with serious mental illness. VHA provided both technological and nontechnological options for scheduling appointments, an important measure given reports that some patients with serious mental illness report difficulty using health technology (13). Similar approaches might improve vaccine uptake for individuals outside the VHA.

This study had some limitations. The data were observational and captured COVID-19 vaccinations during a 6-month period; therefore, we cannot draw conclusions about longer-term trends. We determined serious mental illness diagnosis on the basis of EHR data, but it is possible that not all veterans meeting criteria for this diagnosis were identified in the health record. In addition, our findings may not generalize to veterans who do not utilize VHA services. including those who obtained non-VHA vaccinations.

Persons with serious mental illness are known to be highly vulnerable to the psychological and physiological sequelae of COVID-19. Reducing COVID-19 incidence among persons with serious mental illness via widespread vaccination is an important step toward preventing longer-term harms. Our findings show that despite previous evidence of logistic and psychosocial barriers to COVID-19 vaccine uptake among persons with serious mental illness, VHA has thus far provided equitable distribution of the vaccine among patients with serious mental illness and those without serious mental illness. Continued monitoring of vaccine uptake among patients with serious mental illness and identification of strategies to maximize COVID-19 vaccine uptake are important steps toward mitigating the negative impacts of COVID-19.