Health care experiences during the COVID-19 pandemic by race and social determinants of health among adults age ≥ 58 years in the REGARDS study

The study population was drawn from participants in the REGARDS study, details of which have been previously published [22]. In brief, REGARDS was designed to investigate the reasons for high rates of stroke mortality among Black Americans and individuals in the Southeastern US. In 2003–2007, 30,239 participants from the 48 contiguous states of the US were identified from commercial lists and recruited using mailings and telephone calls, with oversampling of Black individuals and individuals living in the Southeastern US. Individuals were eligible if they self-identified as Black or White race, non-Hispanic ethnicity, and age 45 years or older. The telephone response rate was 33% and the cooperation rate was 49% [23]. Participants completed a computer assisted telephone interview (CATI) and brief in-home physical exam at baseline and received $30 US as an incentive [22]. Approximately 10 years after baseline (2013–2016), 16,146 participants (63% of surviving Black participants and 72% of surviving White participants) completed a second computer assisted telephone interview and in-person assessment [24]; the incentive for the second in-person assessment was up to $100 US. The data collection instruments, available at https://​www.​uab.​edu/​soph/​regardsstudy/​, were developed for the REGARDS study, adapted from widely used surveys. Since recruitment, participants have been interviewed every 6 months for assessment of health events and cognitive function. Beginning July 6, 2020, a survey module was added to the 6-month follow-up calls to assess SARS-CoV-2 testing, COVID-19 illness, and experiences of health care during the pandemic (Additional file 1). At that time, there were approximately 11,000 participants still in active follow-up, mean age 76.9 years (range 58–105). The current analysis included participants who completed this survey module as of September 4, 2021. For participants who completed the survey module more than once, we selected the first survey completed.

The REGARDS study was approved by institutional review boards (IRB) at the participating institutions. All participants provided written informed consent. In order to abide by its obligations with the National Institutes of Health National Institute of Neurologic Disorders and Stroke (NIH/NINDS) and the IRB of the University of Alabama at Birmingham, REGARDS facilitates data sharing through formal data use agreements. Any investigator is welcome to request the REGARDS data and documentation through this process. Requests for data access may be sent to the REGARDS study at regardsadmin@uab.​edu.

As in previous publications in the REGARDS study population, we evaluated SDOH in the domains described in the Healthy People 2020 Framework [12, 25]. Specifically, we anticipated that less than high school education, being unmarried, annual household income <$35,000, residence in a ZIP code (US postal code) with > 25% of individuals below the federal poverty line, rural residence, living in a state without Medicare expansion or with limited public health infrastructure and greater residential segregation would be associated with less access to health care including SARS-CoV-2 testing. Education was reported by participants during the baseline telephone interview. Annual household income and marital status were self-reported during the baseline and second interviews. We used responses from the second interview, except for those participants missing data for whom the first interview responses were included (income: n = 341; marital status: n = 389). As in prior work, we included an income “not reported category” because a substantial proportion of participants declined to report income [25]. Participant residential addresses at baseline and the second assessment were geocoded using ArcGIS and a minimum match score of 90; the match rate was 97% [26]. The distribution of REGARDS participants’ addresses has been described in a previous publication [23]. We used the 2010 Rural Urban Commuting Area (RUCA) codes to identify those living in rural areas, except for individuals for whom address was not available at the second assessment [27]. For these 306 participants, rurality was defined from the first visit using the 2000 RUCA classification [27]. For measures of Medicaid expansion and public health infrastructure, addresses at the second in-home assessment were used, except for 417 participants who had available geocoded addresses at the first but not second assessments. At the state level, we examined whether the participant lived in a state that did not expand Medicaid (Alabama, Florida, Georgia, Kansas, Mississippi, North Carolina, South Carolina, South Dakota, Tennessee, Texas, Wisconsin, Wyoming) and whether the participant lived in a state with poor public health infrastructure (Arkansas, Florida, Louisiana, New Mexico, Mississippi, Nevada, South Carolina, Texas, Tennessee). Public health infrastructure was based on America’s Health Rankings, a state-level composite measure of public health status [28]. As in prior work, states in the bottom 20th percentile of the ranking for 8 or more years during the ten-year period 1993–2002 were considered to have poor public health infrastructure [25]. Residential segregation was assessed using the dissimilarity index from the geocoded address at baseline with information from the 2000 US Census [29]. The dissimilarity index ranges from 0 to 1, where larger values indicate a greater degree of residential segregation. For presentation, the dissimilarity index was dichotomized at the median value (0.50). ZIP code level poverty was also assessed using baseline participant address and information from the 2000 US Census.

Date of birth was reported during the baseline telephone interview and used to calculate age at the time of survey administration. Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or self-report of use of antihypertensive medications at baseline or the 10-year follow-up visit. Diabetes was defined at baseline or 10-year follow-up as fasting blood glucose ≥126 mg/dl (or a glucose ≥200 mg/dl among those who did not fast) or self-report of use of glucose lowering medications. History of stroke, coronary heart disease, and heart failure were based on self-report of physician diagnoses at baseline or the 10-year follow-up or clinician-adjudicated events during follow-up [23, 30, 31]. Cigarette smoking and body mass index were assessed at the 10-year follow-up visit; baseline values were used for 428 participants missing follow-up visit information on cigarette smoking and 709 participants missing follow-up visit information on body mass index.

The survey module included early questions from the Centers for Disease Control and Prevention and the National Institute of Health recommended common data elements for COVID-19, adapted for telephone administration (Additional file 1). The time referent for the survey was the prior 6 months. Additional file 2 Fig. 1 illustrates the survey logic. Among participants who reported an illness with symptoms of COVID-19 (fever, cough sore throat, chills, muscle pain, loss of taste or smell, or shortness of breath or difficulty breathing), we examined the following outcomes: 1. Sought care or advice for the illness; 2. Received a SARS-CoV-2 test for the illness; and 3. Tested positive. Among participants who did not report an illness with symptoms of COVID-19, we examined the following outcomes: 1. Wanted a SARS-CoV-2 test; 2. Wanted and received a SARS-CoV-2 test; 3. Did not want but received a SARS-CoV-2 test; and 4. Tested positive. Among all participants, we examined participant characteristics by self-reported subjective COVID-19 and self-reported positive SARS-CoV-2 test. Additionally, we examined the impact of the pandemic on access to health care; survey options were “No change”, “Mild. Appointments moved to telehealth”, “Moderate. Delays or cancellations in appointments and/or delays in getting prescriptions; changes have minimal impact on health.”, and “Severe. Unable to access needed care resulting in moderate to severe impact on health.”

We calculated the percentage of participants experiencing each of the COVID-19 related outcomes overall and by race and each of the SDOH. We used chi-square tests for categorical variables and ANOVA for continuous variables to evaluate whether differences between groups were statistically significant. Participants missing data on a characteristic were excluded from the analysis for that characteristic. For each variable, data were missing for < 5% of participants. In secondary analyses, we standardized the population to the age, race, sex, and geographic distribution from the US Census 2020 estimated population age 58 years or older residing in the 48 contiguous states, who were not Hispanic, and who reported Black or White race [32]. Strata for calculating weights were defined by age (58–74, 75–84, and ≥ 85 years), sex, race, and Southeastern states (Alabama, Arkansas, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee) versus others. Statistical significance of differences between groups for standardized estimates were tested using bivariate logistic regression with inverse probability of sampling weights and robust standard errors. Additionally, we stratified the study population by calendar year of survey administration (2020 or 2021) and tested for differences across years using chi-square tests. Analyses were performed using SAS version 9.4 (Cary, NC) and Stata version 17.0 (College Station, TX). Two-sided p-values < 0.05 were considered statistically significant.

Among 9492 participants who completed the REGARDS COVID-19 survey module, the average age was 76.8 years (range 58–105 years), 92% were age 65 years or older, 36% were Black, and 57% were female (Table 1). Participants resided in 1279 counties across the 48 continental United States. Thirty percent of participants had income <$35,000 per year, 6% had less than high school education, 39% were unmarried, 16% lived in ZIP code areas where > 25% of residents were below the federal poverty limit, 17% lived in rural areas, 33% lived in states with poor public health infrastructure, and 54% lived in states that did not expand Medicaid. Illnesses with COVID-19 symptoms were reported by 697 participants (7%). Participants who reported illness with COVID-19 symptoms were slightly younger, had higher body mass index, and more commonly had diabetes and history of coronary heart disease compared to participants who did not report such an illness.

Among the 697 participants reporting an illness with COVID-19 symptoms, 514 (74%) sought care or advice for that illness, 349 (50%) received a SARS-CoV-2 test for that illness, and 173 participants had a positive SARS-CoV-2 test (25%) (Fig. 1 and Additional file 2 Table 1). Of participants tested, 50% (173 of 349) tested positive. Compared to urban residents, a higher percentage of participants residing in rural areas sought care or advice for an illness with COVID-19 symptoms (83% vs 72%, p = 0.03), received a SARS-CoV-2 test (62% vs 48%, p = 0.002), and tested positive (35% vs 23%, p = 0.004). A higher percentage of participants in states with poor public health infrastructure, compared to better, received SARS-CoV-2 testing (59% vs 45%, p = < 0.001) and tested positive for SARS-CoV-2 (30% vs 22%, p = 0.01). A higher percentage of participants living in areas with > 25% poverty tested positive than participants in lower poverty areas (32% vs 23%, p = 0.047). Differences by other characteristics were not statistically significant.

When standardized to the age, sex, and geographic distribution of the US population, a higher percentage of Black compared to White participants received SARS-CoV-2 tests (58% vs 44%, p = 0.02) and tested positive (34% vs 20%, p = 0.007) (Additional file 2 Table 1a). Compared to participants responding to the survey in 2020, the percentage of participants reporting seeking care or advice in 2021 was slightly higher (77% vs 73%, p = 0.03), and the percentage receiving SARS-COV-2 tests (67% vs 44%, p < 0.001) and testing positive (47% vs 16%, p < 0.001) was substantially higher (Additional file 2 Table 1b and c). Compared to White participants, Black participants reported higher proportion of positive tests in 2020 (22% vs 14%, p = 0.01) and lower proportion of positive tests in 2021, though the difference between Black and White participants was not statistically significant in 2021 (40% vs 51%, p = 0.14).

Of the 8795 participants who did not report an illness with COVID-19 symptoms, 2525 (29%) reported wanting SARS-CoV-2 testing, 1972 (22%) wanted and received a test, and 744 (8%) had a SARS-CoV-2 test despite not reporting wanting one (Fig. 2 and Additional file 2 Table 2). Among the participants who did not report COVID-19 symptoms, 85 had a positive SARS-CoV-2 test (1% of participants not reporting COVID-19 symptoms; 3% of those tested). A higher percentage of Black compared to White participants reported wanting a SARS-CoV-2 test (38% vs 23%, p < 0.001), receiving a SARS-CoV-2 test (30% vs 18%, p < 0.001), and testing positive compared to White participants (1.4% vs 0.8%, p = 0.005). A greater percentage of lower-income participants reported testing positive (1.4% vs 0.8%, p = 0.03) compared to higher income participants. Being unmarried, compared to married, was associated with a greater percentage of participants reporting wanting a SARS-CoV2 test (30% vs 28%, p = 0.003) and receiving a test (24% vs 22%, p = 0.03). Individuals with less than high school education reported testing positive more commonly than individuals with high school education or greater (2.3% vs 0.9%, p = 0.001). Rural compared to urban residence was associated with a lower percentage desiring testing (26% vs 29%, p = 0.008). Area-level poverty > 25% and greater residential segregation were associated with higher percentage reporting desire and receipt of testing.

Patterns of reporting wanting SARS-CoV-2 tests, receiving tests, and testing positive were similar when standardized to the age, sex, and geographic distribution of the US population (Additional file 2 Table 2a). Compared to participants who completed the survey in 2020, a higher percentage of participants completing the survey in 2021 reported wanting a SARS-CoV-2 test (37% vs 26%, p < 0.001), wanting and receiving a test (30% vs 20%, p < 0.001), receiving a test despite not wanting one (10% vs 8%, p < 0.001), and testing positive (1.5% vs 0.8%, p = 0.002) (Additional file 2 Table 2b and c).

Of the 279 total participants reporting a positive SARS-CoV-2 test, 74 (27%) reported that they did not think they had COVID-19, of whom 23 reported an illness with symptoms of COVID-19. Compared to participants testing positive who thought they had COVID-19 (n = 205), participants who did not think they had COVID-19 were more commonly Black, low income, unmarried, and female (Table 2). Among the 9213 participants without a positive SARS-CoV-2 test, 297 reported that they thought they had COVID-19; 100 (34%) of those participants reported an illness with symptoms of COVID-19 and 140 (47%) received SARS-CoV-2 testing.

Overall, 51% of participants reported no change in access to care during the COVID-19 pandemic, 22% reported mild changes, 25% reported moderate changes, and 1% reported severe changes (Fig. 3 and Additional file 2 Table 3). Reporting no change in access to care was more common in Black than White participants (53% vs 50%, p = 0.01), those with income <$35,000 compared to ≥$35,000 (57% vs 46%, p < 0.001), unmarried compared to married participants (53% vs 50%, p = 0.005), individuals with less than high school education compared to high school or greater (70% vs 50%, p < 0.001), residence with area-level poverty > 25% compared to ≤25% (55% vs 50%, p = 0.006), rural compared to urban (57% vs 50%, p < 0.001), those in states with poor public health infrastructure compared to better public health infrastructure (54% vs 50%, p < 0.001), and Medicaid non-expansion states compared to Medicare expansion states (53% vs 49%, p = 0.002). Patterns were similar when standardized to the age, race, sex, and geographic distribution of the US (Additional file 2 Table 3a). A greater percentage of participants who completed the survey in 2021 reported no change (56%) compared to those who completed the survey in 2020 (49%, p < 0.001) (Additional file 2 Table 3b and c).