Early Mortality After the First Dose of COVID-19 Vaccination: A Target Trial Emulation

Abstract

Background

Vaccine hesitancy persists alongside concerns about the safety of coronavirus disease 2019 (COVID-19) vaccines. We aimed to examine the effect of COVID-19 vaccination on risk of death among US veterans.

Methods

We conducted a target trial emulation to estimate and compare risk of death up to 60 days under two COVID-19 vaccination strategies: vaccination within 7 days of enrollment versus no vaccination through follow-up. The study cohort included individuals aged ≥18 years enrolled in the Veterans Health Administration system and eligible to receive a COVID-19 vaccination according to guideline recommendations from 1 March 2021 through 1 July 2021. The outcomes of interest included deaths from any cause and excluding a COVID-19 diagnosis. Observations were cloned to both treatment strategies, censored, and weighted to estimate per-protocol effects.

Results

We included 3 158 507 veterans. Under the vaccination strategy, 364 993 received vaccine within 7 days. At 60 days, there were 156 deaths per 100 000 veterans under the vaccination strategy versus 185 deaths under the no vaccination strategy, corresponding to an absolute risk difference of −25.9 (95% confidence limit [CL], −59.5 to 2.7) and relative risk of 0.86 (95% CL, .7 to 1.0). When those with a COVID-19 infection in the first 60 days were censored, the absolute risk difference was −20.6 (95% CL, −53.4 to 16.0) with a relative risk of 0.88 (95% CL, .7 to 1.1).

Conclusions

Vaccination against COVID-19 was associated with a lower but not statistically significantly different risk of death in the first 60 days. These results agree with prior scientific knowledge suggesting vaccination is safe with the potential for substantial health benefits.

SARS-CoV-2, COVID-19 vaccination, medication safety, vaccine safety, Risk of death

Concern regarding the safety of coronavirus disease 2019 (COVID-19) vaccines has hindered public vaccination campaigns, including among veterans [1–4]. Serious adverse events following immunization are rare and randomized clinical trials are typically underpowered to evaluate safety risks. Observational (“real-world data”) studies of COVID-19 vaccine safety after emergency use authorization are important for understanding risks of rare adverse events and long-term complications. Traditional safety surveillance approaches tend to focus on passive event monitoring or spontaneous reporting of events. However, the goal is often “signal detection,” where statistical associations are used to identify adverse events for further evaluation. These methods can be useful for identifying novel adverse events but are limited for estimating differences in risk, identifying reactions with many potential causes (eg, all-cause death), or where significant confounding may be present. Active surveillance methods and cohort studies have increased in recent years, especially with COVID-19 vaccines [5–7]. The Veterans Health Administration (VHA) population is ideal for studying rare adverse events due to vaccination. It is the largest integrated healthcare system in the United States, with dedicated infrastructure for studying vaccine safety, which allows in-depth observational studies among millions of veterans [8].

The early period (days 0 to 60) after vaccination represents an important window because the risk of adverse events is often higher close to vaccination time; in later periods, immunity and protection against COVID-19 infection may obscure underlying safety risks. Prior studies in the non-VHA population have suggested significant reductions in all-cause risk of death after vaccination but may be subject to residual confounding and have used unclear timeframes after vaccination [6, 7]. Those being vaccinated may be older and have multiple comorbidities, but vaccination is also strongly associated with employment, healthcare access, personal beliefs, and perceived or actual risk of disease [9, 10]. Target trial emulation approaches, which are increasingly used to study causal questions when randomized trials are not feasible [2–4, 11], are a promising approach for COVID-19 vaccine evaluations, where significant baseline confounding and immortal time bias threaten validity.

Here, we aimed to estimate the risk of death up to 60 days under a strategy of vaccination against COVID-19 versus a strategy of no vaccination. We anticipated that the risk of death for the population under the two treatment strategies (vaccination versus no vaccination up to 60 days) would be similar.

METHODS

Target Trial Emulation and Data Sources

The target trial framework is a structured approach for designing and conducting analyses that attempt to emulate hypothetical trials using observational data [12, 13]. We emulated a (hypothetical) target trial of COVID-19 vaccination using observational data collected in the VHA administrative and clinical records. The core elements of the target trial and its emulation are summarized in Table 1. The target trial included 5 index dates (a “time zero”), representing an enrollment window that started on the first day of each month, 1 March 2021 through 1 July 2021. This timeframe was selected to reflect the period when the general US population was being vaccinated, with multiple index dates to increase the precision of our results. Prior to 1 March, vaccination was mostly limited to special populations such as those aged ≥75 years [14]. This period represents the pre-Omicron phase of the pandemic. We compared COVID-19 vaccination within a grace period of 7 days of each index date versus no vaccination through 60 days of follow-up among all eligible veterans. The Hines VA Medical Center Institutional Review Board approved the study.

Table 1.

Parameters for the Target Trial and Its Emulation

	Target Trial	Emulation in Observational Data
Aim	Estimate and compare risk for death up to 60 d with two treatment strategies: vaccination within 7 d of eligibility assessment and no vaccination through follow-up
Eligibility	Inclusion Veteran, aged 18 y or older No prior COVID-19 vaccination Exclusion   Contraindication to vaccination: Monoclonal antibody treatment Active infection or COVID-19 diagnosis in prior 30 d Acute or life-threatening illness	Assessed on one of five index dates from 1 March 2021 through 1 July 2021 Inclusion:   Veteran, aged 18 y or older No prior COVID-19 vaccination Additional inclusion: At least one outpatient, inpatient, or pharmacy encounter in prior year Enrolled for at least 1 y in VHA Exclusion No contraindication to vaccination (Monoclonal antibody treatment [prior 90 d], CoVID-19 (+) diagnosis [prior 30 d]) Additional exclusion: Hospitalization (prior 30 d) Diagnosis of HIV, active cancer, dementia and/or palliative care treatment (prior year)
Treatment	Vaccination with a COVID-19 vaccine within 7 d of enrollment versus no vaccination	Same
Assignment	Randomly assigned	Veterans are randomly assigned within levels of observed confounders (eg, age, mortality risk score [CAN], homelessness, COVID-19 high-risk conditions, race/ethnicity, and recent hospitalization)
Follow-up	Followed from baseline until: Death 60 d of follow-up complete For outcome of non-COVID related mortality, observations censored when a new COVID-19 diagnosis occurs	Additionally: Censored if observed data inconsistent with emulated treatment assignment; No vaccination arm: censored if vaccinated at any point of follow-up, Vaccination arm: censored if unvaccinated at 7 d
Outcomes	Any cause deathDeath excluding recent COVID-19 infection	Same
Statistical analysis	Kaplan-Meier survival analysis	Pooled logistic regressions for parametric survival analysis, weighted by inverse probability of remaining uncensored

Eligibility Criteria

Eligibility criteria were chosen based on the completed randomized trials of the messenger RNA (mRNA) vaccines and routine clinical indications for vaccination [15, 16]. For each index date, veterans were considered eligible for inclusion if they had been enrolled in VHA for at least 1 year, had at least 1 healthcare encounter in that year, were aged ≥18 years, and had not yet received a COVID-19 vaccination. Veterans were excluded if they had a severe acute respiratory syndrome coronavirus 2 infection by positive laboratory result or diagnosis code in the 30 days prior to the index date; received COVID-19 monoclonal antibody treatment in the 90 days prior to the index date or a diagnosis code for a palliative care encounter, dementia, or cancer; or were hospitalized in the 30 days prior to the index date [17].

We examined 2 treatment strategies: vaccination within 7 days of the index date (first day of the month) or no vaccination through the end of follow-up at 60 days. For the first 7 days, veterans could be compatible with both treatment strategies, so each individual record was duplicated (cloned) and assigned to both strategies.

Vaccination Status

We used a combination of electronic health immunization records, health factors, claims, and outpatient pharmacy records to identify vaccination events. The VHA records include vaccinations that occur within the VHA system (see Supplementary Table 3), as well as documented vaccinations that occur outside the system (eg, pharmacy, clinic). We used the first documented vaccination as our exposure of interest.

Outcomes

The primary outcome was death from any cause, determined by a combination of veteran vital statistics (that include Social Security and Medicare death records), electronic health records, and Veteran Benefits Administration records [18]. These records for veteran deaths are validated and considered the reference standard for identifying death, but they do not reliably report cause of death [19]. In an exploratory analysis done to facilitate comparisons with prior publications, we evaluated the outcome of “non–COVID-19”–associated death. In analyses for this outcome, we censored veterans when they had a new COVID-19 diagnosis or positive test result, excluding those with death subsequent to a COVID-19 diagnosis.

Other Covariates

We ascertained resident covariates separately at each target trial date. Covariate data included Charlson index comorbidities documented up to 1 year prior, as well as an indicator for the total number of “high-risk” conditions as defined by the Centers for Disease Control and Prevention. Supplementary Table 1 includes International Classification of Diseases, Tenth Revision, Clinical Modification definitions. We also included baseline variables for veteran age (linear and quadratic), homelessness, hospitalization in the past year, duration of enrollment in VHA, and other demographics (eg, gender). We used the Care Assessment Need (CAN) risk score for 90-day probability of death as a summary score of time-varying confounders along with time-varying indicators for COVID-19 infection, diagnoses and healthcare exposure [20]. The CAN score is generated weekly for veterans assigned to a primary care patient-aligned care team, not hospitalized, and alive as of the date of calculation. The score is based on 36 elements including sociodemographics, clinical elements (eg, vital signs), medication use, comorbid conditions, and healthcare utilization. The baseline CAN score prior to time zero and nearest to the index date was used, along with weekly CAN scores in the follow-up period.

Statistical Analyses

The cumulative incidence of death is calculated as 1 - probability of survival on each day of follow-up under each strategy. To estimate the survival probabilities, we used a pooled logistic regression model for the outcomes where vaccination at each person-period (day of follow-up) was regressed on linear and quadratic polynomials for time, an interaction between time and high-risk condition, linear and quadratic terms for baseline CAN score, baseline indicators for recent hospitalization, outpatient visit or prescription fill in prior 90 days, baseline diagnoses of cardiovascular disease (myocardial infarction, congestive heart failure, or stroke), linear and quadratic terms for age, history of influenza vaccination in prior year, laboratory measurements (recent laboratory result or microbiology culture) in prior 30 days, homelessness, African-American race, recent history of weight loss, diabetes, renal disease, or neurologic condition.

Artificial Censoring and Inverse Probability of Censoring Weights

Clones were censored when veterans’ observed history was no longer compatible with the assigned treatment strategy. For example, if a clone assigned to the “no vaccination” strategy received the vaccine, we artificially censored that observation at the time point of nonadherence (date of vaccination). If a clone assigned to the vaccination group was not vaccinated by day 7, we censored that observation on day 7. This approach estimates a “per-protocol” effect (ie, a causal estimate that describes treatment effects if a particular treatment strategy is adhered to, versus an "intention to treat" effect ignoring adherence).

We used inverse probability of censoring weights to account for the artificial censoring [11, 21–23]. The method is discussed in greater detail in Supplementary materials. For the outcome of non–COVID-19–related deaths, veterans were censored for COVID-19 infection. The cumulative probabilities of remaining uncensored were estimated separately for each treatment strategy with pooled logistic regression models including the following terms: linear and quadratic polynomials for time, an interaction between time and high-risk condition, linear and quadratic terms for baseline and time-varying CAN score, baseline and time-varying indicators for recent hospitalization, outpatient visit or prescription fill, a baseline and time-varying diagnosis of cardiovascular disease (myocardial infarction, congestive heart failure, or stroke), linear and quadratic terms for age, history of influenza vaccination in prior year, laboratory measurement (recent blood draw or microbiology culture) in prior 30 days, homelessness, African-American race, recent history of weight loss, diabetes, renal disease, or neurologic condition.

Clones under the vaccination strategy were assigned were assigned a weight of 1 on days less than 7 and if they were vaccinated on days prior to day 7 [24]. Model selection for outcome and censoring probabilities was driven by author expertise and observed imbalances between the censored and uncensored risk set. Weights were truncated at the upper 99.9th percentile [25].

If veterans met eligibility on more than 1 index date, we randomly selected 1 index date per veteran.

Stability Analyses

We evaluated differing grace period durations, as well as treatment effect heterogeneity by different subgroups including age and presence of chronic health conditions.

Statistical Inference

We used sampling with replacement by veteran with 250 replications to generate 95% confidence intervals with the percentile method. This bootstrap approach is used to reflect uncertainty in the estimated weights [12].

Initial data preparation was done with Microsoft SQL Server Management Studio 18 and SAS Version 9.4 (Cary, NC). We conducted a complete case analysis (excluded 6.3% of veterans missing covariate data; see Figure 1). All analyses were performed using R statistical software, Version 4.1.2 (Vienna, Austria).

Figure 1.

Flowchart of eligible individuals, Veterans Health Administration 2021. The figure depicts the workflow that led to the final sample size. Abbreviations: COVID-19, coronavirus disease 2019; MAB, monoclonal antibody therapy.

RESULTS

Study Cohort

We included 3 158 507 veterans in the analysis (see Figure 1) with a median (quartiles, 1– 3) follow-up of 60 days (18–60) under the no vaccination strategy. Under the vaccination strategy, 2 793 251 (88%) were censored at day 7 for not receiving vaccine. For those observations under the vaccination strategy, 364 993 received vaccine within 7 days; 155 408 (42.6%) received the Pfizer BioNTech vaccine, 169 170 (46.3%) the Moderna mRNA-1273 vaccine, and 35 067 (9.6%) the Johnson & Johnson vaccine. Of these, 335 515 (92%) received a second vaccination. The most prevalent index date was 1 March 2021 (1 456 972; 46.1%). Supplementary Table 2 summarizes inclusion by index date, and Supplementary Table 3 provides the source of vaccine records.

Table 2 summarizes the entire cohort at baseline and information from observations remaining uncensored in the vaccinated and no vaccination strategies at the end of the 7 day grace period. Individuals who received the vaccine within 7 days (adherent to assigned strategy) were older (69.5 versus 62.8 years) and had a higher prevalence of factors associated with a high mortality risk, high CAN scores (5.7% versus 4.9%), more COVID-19 high-risk conditions (78.5% versus 71.5%), and chronic conditions such as diabetes, hypertension, and depression. Individuals who did not receive the vaccine within 7 days were more likely to be homeless (4.8% versus 3.5%). Supplementary Figure 1 demonstrates covariate balance between groups after applying weights, where all observed covariates had a weighted standardized mean difference of <0.2, and only 1 covariate (age) had a difference >0.1.

Table 2.

Characteristics at Baseline and End of Grace Period, Veterans From 1 March 2021 through 1 July 2021

	Baseline	At End of Day 7 (Grace Period)
Characteristic	Total (n = 6 317 014)	Unvaccinated (n = 2 793 251)	Vaccinated (n = 364 993)
White, no. (%)	4 419 420 (70.0%)	1 958 473 (70.1%)	251 044 (68.8%)
Male, no. (%)	5 546 840 (87.8%)	2 446 613 (87.6%)	326 558 (89.5%)
Age, mean (SD), y	63.6 (18.2)	62.8 (18.4)	69.5 (15.6)
18–49	1 645 582 (26.0%)	775 578 (27.8%)	47 204 (12.9%)
50–64	1 402 508 (22.2%)	631 761 (22.6%)	69 477 (19.0%)
65–84	2 603 408 (41.2%)	1 097 819 (39.3%)	203 774 (55.8%)
≥85+	665 510 (10.5%)	288 090 (10.3%)	44 538 (12.2%)
High-risk condition,a 1+, no. (%)	4 567 048 (72.3%)	1 996 651 (71.5%)	286 648 (78.5%)
Diabetes, no. (%)	1 295 972 (20.5%)	548 232 (19.6%)	99 659 (27.3%)
Hypertension, no. (%)	2 667 994 (42.2%)	1 141 233 (40.9%)	192 615 (52.8%)
Depression, no. (%)	1 339 550 (21.2%)	597 225 (21.4%)	72 508 (19.9%)
Neurologic, no. (%)	274 722 (4.3%)	119 604 (4.3%)	17 726 (4.9%)
Major adverse cardiovascular event (eg, stroke, myocardial infarction), no. (%)	293 830 (4.7%)	125 551 (4.5%)	21 289 (5.8%)
Homeless, no. (%)	294 942 (4.7%)	134 657 (4.8%)	12 791 (3.5%)
Prescription filled, prior 30 days	5 470 614 (86.6%)	2 410 103 (86.3%)	324 964 (89.0%)
Outpatient encounter,b prior 30 days	6 231 366 (98.6%)	2 754 102 (98.6%)	361324 (99.0%)
Hospitalization, prior 30 days	276 954 (4.4%)	120 758 (4.3%)	17 688 (4.8%)
Care Assessment Needs score in top 5th %c	315 852 (5.0%)	136 920 (4.9%)	20 883 (5.7%)
Veterans Health Administration enrollment, years mean (SD)	6.9 (6.4)	6.8 (6.4)	7.3 (6.6)

Abbreviation: SD, standard deviation.

^aIncludes Centers for Disease Control and Prevention–defined comorbid conditions that increase the risk of illness or death due to coronavirus disease 2019 infection (see Supplementary Table 1).

^bOutpatient encounter includes clinic visit, laboratory result, telephone call, or any noninstitutionalized clinical interaction.

^cA probability model for risk of death or hospitalization in 90 days; variable describes if person has a score in the top 5% (higher risk for death).

Comparison of the Strategies

Figure 2 presents estimated cumulative incidences for death from any cause; results are summarized in Table 3. The risk for death was similar in absolute terms (difference of <5 per 100 000) between the 2 groups up to approximately day 28. At 60 days, for the outcome of risk of death from any cause, there were 156 (95% confidence limit [CL], 126 to 185) deaths per 100 000 veterans among those under the vaccination strategy versus 185 (95% CL, 178 to 191) deaths per 100 000 veterans under the no vaccination strategy, representing an absolute risk difference of −25.9 (95% CL, −59.5 to 2.7) and relative risk of 0.86 (95% CL, 0.7 to 1.0). In the analysis that excluded deaths with COVID-19 infections in the 30 days prior, at 60 days of follow-up, there were 155 (95% CL, 123 to 194) non–COVID-19 deaths per 100 000 veterans under the vaccination strategy and 175 (95% CL, 168 to 181) non–COVID-19 deaths per 100 000 veterans under the no vaccination strategy with an absolute risk difference of −20.6 (95% CL, −53.4 to 16.0) and a relative risk of 0.88 (95% CL, 0.7 to 1.1).

Figure 2.

Risk of death among US veterans included in a target trial emulation comparing vaccination within 7 days of the index date versus no vaccination from 1 March 2021 through 1 July 2021. A, Cumulative incidence in deaths per 100 000 for each outcome (death, solid line; deaths excluding coronavirus disease 2019, dashed line). B, Difference in deaths per 100 000 by treatment strategy (no vaccination or vaccination within 7 days). Abbreviation: COVID-19, coronavirus disease 2019.

Table 3.

Mortality Rates Under 2 COVID-19 Vaccination Strategies for Veterans From 1 March 2021 to 1 July 2021

Outcome	Follow-up (days)	No Vaccination	Vaccination Within 7 Days	Risk Ratio (95% CL)	Risk Difference (95% CL)
Deaths due to any cause	7	6.2 (5.5 to 7.0)	5.6 (4.5 to 8.2)	0.89 (0.8 to 1.4)	−0.7 (−1.5 to 2.1)
	28	47 (45 to 51)	45 (31 to 64)	0.95 (0.6 to 1.3)	−2.3 (−17.6 to 15.7)
	60	185 (178 to 191)	159 (126 to 185)	0.86 (0.7 to 1.0)	−25.9 (−59.5 to 2.7)
Deaths, excluding coronavirus disease 2019 infections	7	6.2 (5.6 to 7.0)	5.6 (4.9 to 8.7)	0.89 (0.8 to 1.4)	−0.7 (−1.2 to 2.3)
	28	46 (45 to 49)	45 (29 to 60)	0.97 (0.6 to 1.3)	−1.4 (−18.1 to 13.1)
	60	175 (168 to 181)	155 (123 to 194)	0.88 (0.7 to 1.1)	−20.6 (−53.4 to 16.0)

Event rates per 100 000 (95% CL) are reported along with a relative and absolute risk difference. The outcome regression models are weighted for censoring and included covariates for a restricted cubic spline for time (days of follow-up), Care Assessment Need score, age, influenza vaccination, number of high-risk conditions, African-American race, and indicators for recent hospitalization, outpatient, or pharmacy visits. The estimates are also adjusted with inverse probability weights for remaining uncensored.

Abbreviations: COVID-19, coronavirus disease 2019; CL, confidence limit.

Supplementary Figure 2 describes the impact of widening the “grace” period or enrollment window from 7 days to 21 days. Longer intervals (21+ days) lead to larger observed differences at 60 days between vaccine strategies, but any grace period of 3 to 21 days favors the vaccine strategy by day 60. Supplementary Figure 3 provides death risk differences by different subgroups with no increased risk under the vaccination strategy within strata of age, chronic conditions, or race/ethnicity.

DISCUSSION

We used target trial emulation methods to compare the risk of death for a strategy of vaccination versus no vaccination in a cohort of US veterans. Our analysis estimated that the absolute risk difference of death between those with a vaccination versus no vaccination strategy is small in the 28 days after the index date (<5 per 100 000). The 60-day estimated risk of death under the vaccination strategy was 14% lower than under the unvaccinated strategy, with 26 fewer deaths per 100 000 veterans. However, 95% CLs suggested that our results are compatible with a hypothesis of no difference. The results of this study can support the communication of the risks and benefits of vaccination against COVID-19 infection [26].

This study uniquely evaluates all-cause risk of death for those vaccinated versus not using target trial emulation methods. Our hypothesis was that there would be little to no increased risk of death but possibly a modest reduction in risk of death in those vaccinated. Our results are generally compatible with this finding. Prior work has explored this question but with limitations due to either not focusing on periods soon after vaccination or describing population-level trends instead of attempting to estimate average causal effects of vaccination [6]. For example, Xu et al reported a 61% reduction in non-COVID-19 mortality from an observational cohort of Medicare beneficiaries but did not report risk relative to time after vaccination [7]. Large reductions in risk occurring soon after vaccination may be implausible since current scientific understanding suggests the available vaccines do not confer immunity for at least 10–14 days post-vaccination [15, 16]. If the available vaccines do not increase risk of death and do not protect against COVID-19 for the first 2 weeks, then the risk difference for death in that early period should be close to zero between the vaccination and no vaccination strategies. We estimated differences in the risk of death close to zero with confidence intervals consistent with a no-difference (null) hypothesis up to day 60.

After excluding COVID-19 infections, we still observed lower rates of death under the vaccination strategy. Evaluations which exclude deaths after COVID-19 infection may be subject to bias due to collider stratification, but we include an exploratory analysis for reference with other's work. A report from the private healthcare US population observed a 54% reduction in mortality after the first COVID-19 vaccine dose with a similar non–COVID-19 mortality outcome [27]. We estimate a much smaller 12% benefit that is compatible with chance. In addition to excluding COVID-19 infection, it is difficult to make claims about vaccination preventing particular causes of death because an explicit determination is not available in our data, and we lacked the resources to infer causes using the available data. Also, COVID-19 infection may not be captured if the veteran was never tested or the positive result was not recorded; counting those COVID-19 deaths could erroneously increase the estimated impact of vaccination on non–COVID-19 deaths. Deaths due to other causes such as cardiovascular events could be due to the stress of a COVID-19 infection.

Our study has important limitations. The foremost limitation is that this observational study may not have accounted for all relevant factors in the receipt of vaccination. However, the design ensures baseline factors are the same because at time zero veterans are allocated to both treatment strategies (ie, cloned). The impact of differential loss to follow-up may also be small in the shorter time frame studied here (60 days). Additionally, veterans’ vaccination status may be misclassified due to receipt of vaccines outside the VHA system. The VHA documents outside vaccinations, but there is no way to verify or validate if all outside vaccinations were recorded in this period. Because of this, it is possible that we may have misclassified some individuals as unvaccinated or as eligible to receive the vaccine who had already received it. This could bias results toward the null. Misspecification of the weights or residual confounding could lead to biased results. Last, some covariates included in our analyses were missing for some veterans. Those veterans excluded for missing data primarily represent veterans enrolled in VHA but who did not receive care at the VHA and so had missing clinical data.

Vaccination against COVID-19 was associated with a 12–14% lower but not statistically significant different risk of death. This finding represents a lower mortality benefit than reported in some prior studies, but it is more consistent with prior scientific knowledge. Our findings suggest vaccination is safe without an apparent increased risk of death in the first 60 days.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author Contributions. K. W. M. is responsible for the overall study design, primary analysis, and drafting of the manuscript. K. H., I. J. D., D. R. M., A. R. Z., W. F. G., C. B., P. G., S. G., and F. C. contributed to study design, interpretation of results, and manuscript writing. R. J. and J. P. performed data preparation and analysis.

Disclaimer. The views expressed here are those of the authors and do not necessarily reflect the position or policy of the Department of Veteran Affairs or the US government. This article does not necessarily reflect the views of the Patient-Centered Outcomes Research Institute (PCORI), National Library of Medicine (NLM), National Institute on Aging (NIA), or National Institute of General Medical Sciences (NIGMS). The funders had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; and decision to submit the manuscript for publication. K. W. M. had full access to all data from the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial support. K. W. M. was supported by a VA Health Services Research and Development Researchers and Evaluators Residency to complete this work. I. J. G. was supported in part by award ME-2021C2-22365 from the Patient-Centered Outcome Research Institute and grant R01 LM013616 from the National Library of Medicine. A. R. Z. was supported by grants RF1AG061221, R01AG065722, and R01AG077620 from the National Institute on Aging.

Author notes

Potential conflicts of interest. K. W. M., S. G., and A. R. Z. report investigator-initiated support from Sanofi, Seqirus, and Pfizer for other non–coronavirus disease 2019 (COVID-19) vaccine-related work. I. J. D. is principal investigator of a research agreement between Sanofi and Harvard University to develop methods for transportability analyses for use of influenza vaccination trials and reports consulting fees from ModernaTX for work unrelated to COVID-19. S.G. reports vaccine related consulting or speaker fees from GlaxoSmithKline, Janssen, Moderna, Novavax, Pfizer, Sanofi, Seqirus. All remaining authors: No reported conflicts of interest.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.