Factors associated with early receipt of COVID-19 vaccination and adherence to second dose in the Veterans Affairs healthcare system

Background We aimed to determine factors independently associated with early COVID-19 vaccination and adherence to two-dose regimens. Methods Among persons receiving care in the Veterans Affairs (VA) healthcare system (n = 5,766,638), we identified those who received at least one dose of COVID-19 vaccination through the VA, during the first ~3months following emergency use authorization, from December 11, 2020 to March 9, 2021 (n = 1,569,099, or 27.2%, including 880,200 (56.1%) Moderna, 676,279 (43.1%) Pfizer-BioNTech and 12,620 (0.8%) Janssen vaccines). Results Follow-up for receipt of vaccination began on December 11, 2020. After adjustment for baseline characteristics ascertained as of December 11, 2020, factors significantly associated with vaccination included older age, higher comorbidity burden, higher body mass index category, Black (vs. White) race (adjusted hazard ratio [AHR] 1.19, 95% CI 1.19–1.20), Hispanic (vs. non-Hispanic) ethnicity (AHR 1.12, 95% CI 1.11–1.13), urban (vs. rural) residence (AHR 1.31, 95% CI 1.31–1.31), and geographical region, while AI/AN race (vs. White), was associated with lower vaccination rate (AHR 0.85, 95% CI 0.84–0.87). Among persons who received both doses of Moderna or Pfizer-BioNTech vaccines, 95.3% received the second dose within ±4 days of the recommended date. Among persons who received the first vaccine dose, only 3.2% did not receive the second dose within 42 days for Pfizer versus 4.0% for Moderna (p<0.001). Factors independently associated with higher likelihood of missing the second dose included younger age (10.83% in 18–50 yo vs. 2.72% in 70–75 yo), AI/AN race, female sex, rural location, geographical region and prior positive test for SARS-CoV-2. Conclusions We identified sociodemographic and clinical factors that may be used to target vaccination efforts and to further improve adherence to second vaccine dosing.


Introduction
The Food and Drug Administration (FDA) issued Emergency Use Authorization (EUA) for COVID-19 vaccines by Pfizer-BioNTech on December 11, 2020, by Moderna on December 18, 2020, and by Janssen on February 27, 2021. The first two are administered as a 2-dose series, 3 weeks (Pfizer-BioNTech) or 4 weeks (Moderna) apart, while Janssen is a single-dose vaccine.
The Centers for Disease Control and Prevention (CDC) published recommendations on December 21, 2020 supporting a phased approach to vaccine allocation during the first few months after EUA: phase 1a (healthcare personnel and long-term care facility (LTCF) residents), phase 1b (frontline essential workers and persons aged �75 years), phase 1c (essential workers, persons aged 65-74 year and persons aged 16-64 years with high-risk medical conditions) and finally phase 2, which includes the remaining population [1]. These recommendations were adopted by the Veterans Affairs (VA) healthcare system, the largest comprehensive healthcare system in the United States, with some minor modifications to reflect the COVID-19 epidemiology of its population [2].
The extent to which older age-groups and persons with high-risk comorbid conditions were prioritized for vaccination in the first 3 months after EUA in real-world practice is unclear. Geographic variations in access to vaccination have been reported. Also, early reports suggested that racial/ethnic minority groups were under-represented among vaccinees in the United States, especially Hispanic, Black and American Indian/Alaska Native (AI/AN) groups [3], despite being disproportionately affected by COVID-19. Factors associated with receipt of different COVID-19 vaccines and adherence to the receipt and timing of the required second dose of Moderna and Pfizer-BioNTech vaccines are unclear. Differences in cold storage requirements of Pfizer-BioNTech (-70C) versus Moderna (-20C) or differences in perceived effectiveness or side effects of different vaccines may have resulted in associations of sociodemographic or clinical characteristics with use of a specific COVID-19 vaccine.
We aimed to describe factors associated with time-to-receipt of COVID-19 vaccination, type of vaccine used, and adherence to recommended two-dose regimens in the Veterans Affairs (VA) healthcare system, during the first 3 months of vaccine availability after EUA from December 11 to March 9, 2021.

Study setting and data source
The VA provides care at 168 medical centers and 1,053 outpatient clinics throughout the country and employs a comprehensive, nationwide electronic health records system (EHR). We used data from the VA's Corporate Data Warehouse (CDW), a relational database of VA enrollees' comprehensive EHR [4]. The CDW includes the "COVID-19 Shared Data Resource [CSDR]", a set of analytic variables and datasets on all VA enrollees tested for or vaccinated against COVID-19 developed and maintained by the VA Informatics and Computing Infrastructure (VINCI) specifically to facilitate COVID-19 research and operations.
The study was approved by the VA Puget Sound Institutional Review Board. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study population and ascertainment of COVID-19 vaccination
We identified all VA enrollees who were alive as of December 11, 2020 and who had an inpatient or outpatient encounter in the VA healthcare system in the preceding 12 months or until March 9, 2021 (n = 5,766,638). We identified the subset of these VA enrollees who received at least one dose of an approved COVID-19 vaccine by the VA, on or after the EUA date of each vaccine, between December 11, 2020 (the date of EUA authorization for Pfizer-BioNTech) and March 9, 2021 (n = 1,569,101). Type and date of each vaccine dose were ascertained from VA pharmacy records. VA employees, including healthcare providers eligible for vaccination at the earliest phase 1a, were not included in the study.

Baseline characteristics
We ascertained sociodemographic characteristics and comorbid conditions included in the CDC prioritization criteria or known to be associated with SARS-CoV-2 infection or adverse outcomes [5][6][7][8][9][10][11][12][13][14][15][16][17] in order to determine whether they were associated with receipt of vaccination. Baseline characteristics were ascertained as of December 11, 2021, the beginning of the follow-up period. These included age, sex, self-reported race (White, Black, Asian, American Indian/Alaska Native [AI/AN], Pacific Islander/Native Hawaiian [PI/NH], Other) and ethnicity (Hispanic, non-Hispanic), urban vs rural residence (based on zip codes, using data from the VA Office of Rural Health [18], which uses the Secondary Rural-Urban Commuting Area [RUCA] for defining rurality), geographic location (divided into 10 standard U.S. Federal Regions [19]), body mass index (BMI, using measured weight and height recorded prior to December 11, 2021), and comorbid conditions including diabetes, congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD) and chronic kidney disease (CKD) defined by appropriate international classification of disease (ICD10) codes. These comorbid conditions were defined if the ICD10 codes were recorded in the 2-year period before December 11, 2020. We calculated the Charlson Comorbidity Index (CCI), using the Deyo modification [20][21][22], as a measure of comorbidity burden, using a 2-year window prior to December 11,2020. We identified persons who had a documented positive SARS-CoV-2 polymerase chain reaction (PCR) test in the VA healthcare system before December 11, 2020. In order to better capture relevant geographic variability related to institutional factors, we additionally categorized persons according to the 19 VA administrative regions, known as Veterans Integrated Service Networks or VISNs [23]. Also, we categorized VA medical centers according to the proportion of vaccines administered that were manufactured by Moderna vs. Pfizer-BioNTech. Data on VA enrollees' employment status (e.g. whether a person was a "frontline essential worker" or an "essential worker") were not available.
Missing values for BMI (11.6%) were multiply imputed using values of the other characteristics listed above; all missing values were successfully imputed. "Missing" values for race or ethnicity included persons who refused to declare their race/ethnicity or reported unknown or mixed race/ethnicity and did not self-identify as belonging to one of the pre-specified racial or ethnic group. For these persons we did not perform imputation of race/ethnicity, but rather included them in a "missing/unknown/refused" category.
at the time of death if earlier. The proportional hazards assumption was assessed graphically using log(-log) plots. We determined adherence to the timing of the recommended second dose of Pfizer-BioNTech (21 ± 4 days) or Moderna (28 ± 4 days) and the proportion who failed to receive the second dose within the maximum allowed time (42 days). Logistic regression was used to identify factors associated with a missed second dose (defined as no second dose within 42 days of the first dose among persons who had at least 42 days of follow-up time) using logistic regression adjusting for the aforementioned characteristics plus type of vaccine (Moderna vs. Pfizer-BioNTech) and the percentage of all vaccinees vaccinated with Moderna at each medical center. Logistic regression was also used to identify factors independently associated with receipt of Moderna (vs. Pfizer-BioNTech) vaccine after adjusting for these characteristics.
VISN was used rather than federal region in the multivariable models because VISN captures VA-specific variability in distribution of vaccination more accurately.

Factors associated with vaccination
As of March 9, 2021, 1,569,101 out of 5,766,638 (27.2%) VA enrollees received at least one vaccination dose through the VA, including 880,200 (56.1%) with Moderna, 676,279 (43.1%) with Pfizer-BioNTech and 12,620 (0.8%) with Janssen vaccines. Vaccination rates were very low for the first~30 days after December 11 ( Fig 1A and 1B), corresponding to CDC Phase 1a, when vaccination was restricted almost exclusively to healthcare workers not captured in our data and long-term care facility residents, but increased rapidly thereafter.
Graphical assessment of our Cox multivariable model using log(-log) plots suggested good adherence to the proportional hazards assumption as evidenced by near parallel plots (representative plots for CCI categories and CKD are shown in S1A and S1B Fig).
Vaccination rate increased dramatically with increasing age from 2.04 per 100 personmonths in 18-50 year-olds to 19.24 per 100 person-months in 75-80 year-olds, and then declined slightly for older age groups. These associations persisted after adjustment.
Compared to White persons, Black and Asian persons had slightly lower vaccination rates, but were more likely to be vaccinated after adjustment for baseline characteristics (AHR 1.19, 95% CI 1.17-1.19 for Black and AHR 1.20, 95% CI 1.18-1.22 for Asian persons). Conversely AI/AN persons had lower vaccination rates than White persons (7.13 vs. 10.42 per 100 personmonths), which persisted after adjustment (AHR 0.85, 95% CI 0.84-0.87). Compared to non-Hispanic persons, Hispanic persons had lower vaccination rates (8.5 vs. 10.25 per 100 personmonths) but were more likely to receive vaccination after adjustment for baseline characteristics (AHR 1.12, 95% CI 1.11-1.13).
Vaccination rates differed by geographical region ranging from 8.3 per 100 person-months in region 6 (AR, LA, NM, OK, TX) to 13.37 per 100 person-months in region 2 (NY, NJ, PR); substantial associations between geographical region and vaccination persisted after  ), persons in overweight and obese categories were slightly more likely to receive vaccination after adjustment for baseline characteristics. Higher CCI category was associated with significantly higher likelihood of vaccination, which was attenuated somewhat but persisted after adjustment for baseline characteristics. Diabetes, CKD, CHF and COPD were all significantly associated with vaccination. Persons who previously tested positive for SARS-CoV-2 had similar vaccination rates to those without a prior positive test and were slightly more likely to be vaccinated after adjustment for baseline characteristics (AOR 1.05, 95% CI 1.04-1.06).  (Table 2). An additional, 4.1% received the Pfizer-BioNTech second dose from 26-42 days and 3.6% received the Moderna second dose from 33-42 days. Only a very small proportion (0.1% for Pfizer-BioNTech and 0.3% for Moderna) received the second dose "late",   i.e. beyond the allowable interval of 42 days. Also, a very small proportion (0.1%) received a second dose that was a different vaccine type from the first. Among persons who had at least 42 days of follow-up time, i.e. sufficient to receive the second dose within the maximum allowable interval, only 3.2% missed the second dose in that interval for Pfizer-BioNTech versus 4.0% for Moderna (p-value <0.001) ( Table 2). Factors independently associated with lower likelihood of missing the second dose included older age (e.g. missed rate 2.72% in 70 to <75 year-old vs. 10.83% in 18 to <50 yo, adjusted odds ratio (AOR) 0.24, 95% CI 0.23-0.26) male sex, Black race, and urban location (Table 3). Factors independently associated with higher likelihood of missing the second dose included receiving Moderna (vs. Pfizer-BioNTech) (AOR 1.19, 95% CI 1.15-1.23), AI/AN race (AOR 1.34, 95% CI 1.14-1.57), having a prior positive SARS-CoV-2 test (AOR 1.39, 95% CI 1.27-1.53) and being at a medical center with high percentage Moderna use. There were also significant differences in second dose adherence by federal region or VISN ( Table 3).

Predictors of vaccine type: Moderna versus Pfizer-BioNTech
Older age was significantly and progressively associated with higher likelihood of receiving the Moderna rather than the Pfizer-BioNTech vaccine (  VISN 19. There was even greater variability by medical center ranging from 0% to 100%. Individual comorbidities or the CCI did not have high magnitude associations with type of vaccine.

Discussion
In this study of 5,766,638 Veterans who received care in the VA healthcare system, 1,569,101 (27.2%) were vaccinated from December 11, 2020 to March 9, 2021. Factors independently associated with receipt of vaccination during this early period included older age, higher comorbidity burden, a number of individual comorbid conditions (CHF, CKD, COPD and diabetes), higher body mass index (BMI) category, Black or Asian (vs. White) race, Hispanic (vs. non-Hispanic ethnicity), urban (vs. rural) residence, and geographical region in the US, while AI/AN race (vs. White) was associated with lower vaccination rate. Adherence to the receipt and recommended timing of the second vaccine dose was remarkably high for both Moderna and Pfizer-BioNTech vaccines. Factors associated with higher likelihood of missing the second dose included younger age, AI/AN race, female sex, rural location, geographical region and receiving the Moderna rather than the Pfizer-BioNTech vaccine.
The CDC phases of COVID-19 vaccination prioritization strongly emphasized age (�75 years in phase 1b and 65-74 years in phase 1c) and high-risk medical conditions (allocated to phase 1c in those aged 16-64 years) [1]. Consistent with these recommendations, our results demonstrate the highest vaccination rate during the observation period in persons �75 years of age. Among this age group, the highest vaccination rate was in those 75 to <80 years of age (45.74%) with a slight decline in even older age groups. Vaccination rates were also very high in those 65 to <70 (35.97%) and 70 to <75 (38.33%) years old, who correspond to phase 1c. Individual comorbid conditions that have been associated with risk of adverse outcomes in SARS-CoV-2, such as diabetes, CHF, COPD and CKD, as well as overall comorbidity burden were also strongly associated with vaccination.
Studies based on national US vaccination data suggested that Black and Hispanic persons have received smaller shares of vaccinations compared to their shares of the population in most states of the US [24,25]. However, interpretation of these results is limited by the high percentage of records with unknown or missing race/ethnicity information (almost 50%), the unknown proportions of healthcare personnel among early vaccine recipients, and the lack of adjustment for comorbid conditions and other potential confounders. Our results among VA enrollees excluding VA employees (i.e. excluding most healthcare workers) show that although Black, Asian and Hispanic persons have slightly lower unadjusted rates of vaccination than White or non-Hispanic persons, they were actually more likely to receive vaccination after adjustment for other sociodemographic factors and comorbid conditions. In contrast, AI/AN persons had the lowest vaccination rates of any racial/ethnic group, an association that persisted after adjustment for baseline characteristics (AHR 0.83, 95% 0.82-0.85).
Person living in urban (vs. rural) areas were significantly more likely to undergo vaccination in both unadjusted and adjusted analyses. Cold storage requirements for Pfizer and Moderna vaccines and distance to vaccination centers may contribute to lower vaccination rates in  rural areas. Additionally, there was substantial variation in vaccination by both federal region and VISN, which likely reflects regional logistical limitations in access to vaccines and ability to use them quickly. Only 3.6% of Moderna or Pfizer-BioNTech first vaccine dose recipients failed to receive the second dose within 42 days. This is likely an even higher completion rate than was reported in study based on the US population as of February 14, 2021 (88% completion, 8.6% no second dose but still within allowable interval and 3.4% missed dose), although results in that study were not limited to those with at least 42 days of follow-up [26]. Our results suggest that the most profound variation is related to age. Older age was associated with a dramatic reduction in the likelihood of missing the second dose from 10.83% in 18 to <50 year-olds to 2.72% in 75 to <80 year-olds (AOR 0.24, 95% CI 0.23-0.26), with a slight increase in missed second doses in even older age groups. Future efforts should focus on understanding the reasons for higher rates of missed second doses in younger persons and implementing interventions to improve adherence. Racial/ethnic minority groups did not have lower adherence to second dose vaccination, with the exception of AI/AN persons who were significantly more likely to miss the second dose. Thus, AI/AN persons were both less likely to get vaccinated and less likely to receive the required second dose. There was also substantial geographical variability in adherence to second vaccine dosing, much more pronounced at the VISN rather than the federal region level. This suggests that institutional as well as geographical issues likely affect second dose adherence. The fact that vaccination with Moderna (vs. Pfizer-BioNTech) was associated with slightly higher likelihood of missed second dose, requires further confirmation: we did not explicitly apply methods of clinical trial emulation and causative inference in our study that was not specifically designed to compare the two vaccines.
Geographical differences in use of Moderna versus Pfizer vaccines likely reflect the numbers of each vaccine that were distributed to each VA VISN and medical center. The fact that that rural areas were associated with lower use of Pfizer-BIoNTech may be related to the more stringent cold storage requirements for Pfizer-BioNTech (-70C) versus Moderna (-20C). The very strong association of older age with Moderna use is hard to explain, may be due to unmeasured confounding and requires additional confirmation. There was little or no flexibility for VA enrollees to select their preferred vaccine during the study period.
Persons previously testing positive with SARS-CoV-2 were slightly more likely to be vaccinated but significantly more likely to miss the second vaccine dose. The requirement for a second dose should be emphasized in persons with prior infection.
Limitations of our study include lack of data on any additional vaccinations that might have been performed among VA enrollees outside of the VA healthcare system. Our results apply to VA enrollees who are predominantly male and tend to have more comorbid conditions and adverse social determinants of health, but also have access to comprehensive, high quality healthcare that has been shown to attenuate some racial and ethnic disparities [27]. We did not have access to any side effects or adverse reactions experienced after vaccination to test whether they were associated with second dose adherence.
We described sociodemographic and clinical factors that may be used to target vaccination efforts and to further improve adherence to second vaccine dosing. Continued monitoring of vaccination uptake, type of vaccine used and adherence to dosing requirements will be necessary as vaccine efforts expand even further. 28. World Health Organization. Body Mass Index. Available at: https://www.euro.who.int/en/health-topics/ disease-prevention/nutrition/a-healthy-lifestyle/body-mass-index-bmi Last accessed on 03/21/2021.