Effectiveness Associated With Vaccination After COVID-19 Recovery in Preventing Reinfection

Key Points Question How effective is vaccination against COVID-19 after recovery from prior SARS-CoV-2 infection? Findings In this cohort study of more than 95 000 Rhode Island residents from March 2020 to December 2021, including residents and employees of long-term congregate care (LTCC) facilities, completion of the primary vaccination series after recovery from COVID-19 was associated with 49% protection from reinfection among LTCC residents, 47% protection among LTCC employees, and 62% protection in the general population during periods when wild type, Alpha, and Delta strains of SARS-CoV-2 were predominant. Meaning These findings suggest that among people who have recovered from COVID-19, subsequent completion of the primary vaccination series reduced the risk of reinfection by approximately half.


Introduction
Current evidence suggests that individuals who experience a SARS-CoV-2 infection are susceptible to reinfection. 1,2While evidence suggests that rates of hospitalization and death are substantially lower in individuals who have been reinfected, 1 the dynamics of SARS-CoV-2 reinfection are still poorly understood.A prior infection of SARS-CoV-2 confers some immunity to individuals who are not vaccinated, 3,4 but this natural immunity wanes over time.[7] A prior case-control study of Kentucky residents found that individuals who had been previously infected with completion of the primary vaccination series (CPVS) were better protected against reinfection than individuals who remained unvaccinated. 8Similar results have been reported in larger cohort studies in Israel and United Kingdom. 9,10Our primary goal was to quantify the effectiveness associated with vaccination against reinfection among individuals who had been previously infected with SARS-CoV-2 prior to vaccination.Secondary goals were to estimate the probability of reinfection at 3, 6, and 9 months after recovery from primary infection and to assess the associations of measurable risk factors with reinfection.Our analysis has particular relevance to the ongoing discussion about vaccination requirements for individuals who have recovered from COVID-19.

Methods
This cohort study was classified as exempt from review and informed consent by the Rhode Island Department of Health (RIDOH) institutional review board because we used existing data that were deidentified.This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.

Study Design and Population
We conducted a retrospective population-based cohort study using statewide surveillance data collected by the RIDOH between March 1, 2020, and December 9, 2021.The study population included Rhode Island residents aged 12 years and older who were unvaccinated at the time of their first laboratory-confirmed SARS-CoV-2 infection and who remained unvaccinated at least 90 days after the initial positive test result date.We excluded people younger than age 12 years because they did not become eligible for vaccination until the final month of the analysis period.Example patient timelines are detailed in Figure 1.
The initial data set included 199 936 individuals.We excluded 11 262 non-Rhode Island residents, 18 468 individuals who were younger than age 12 years, 12 individuals who had missing or erroneous age information, and 51 044 individuals who were either vaccinated prior to their first laboratory-confirmed infection or before the 90-day recovery period for their first infection had ended (Figure 2).
We divided our sample into 3 separate subpopulations: (1) residents of long-term congregate care (LTCC) settings, which included nursing homes and assisted living facilities, (2) employees of LTCC settings, and (3) the general population (ie, individuals not associated with congregate settings).LTCC residents were analyzed separately owing to the high risk of morbidity and mortality

Data Management
The study database was assembled by linking data from 4 surveillance systems maintained by

Statistical Analysis
We conducted a survival analysis to characterize the risk of reinfection by subpopulation and to estimate the effectiveness associated with vaccination in preventing reinfections among people who have recovered from prior COVID-19.We assumed that it takes 90 days to recover from a previous infection, 11 starting from the date of the positive test results for the first infection.After the 90-day recovery period, individuals who have remained unvaccinated were considered eligible to enter the risk cohort for COVID-19 reinfection.Individuals were then followed from entry time to time of second infection.Individuals who died after entry or did not experience a second infection by December 9, 2021, were considered censored (Figure 1).Vaccination status was recorded each day and categorized as (1) CPVS (2 doses of mRNA-1273 or BNT162b2 vaccine or a single dose of JNJ-78436735 vaccine), (2) partially CPVS (single dose of mRNA-1273 or BNT162b2), or unvaccinated.Individuals were classified as CPVS or partially CPVS 14 days after administration of any dosage of a vaccine. 12r each subpopulation, we calculated the 3-, 6-, and 9-month probability of reinfection by vaccination status using the Kaplan-Meier method.We also calculated incidence rates by dividing the number of reinfections during follow-up by the number of person-days spent in each vaccination category.For this calculation, a person with CPVS with either mRNA-1273 or BNT162b2 contributed person-time to all 3 categories (Figure 1).
We previous hospitalization during the first infection, symptom status during the first infection (ie, asymptomatic, symptomatic, or unknown), and time of entry into the risk cohort for reinfection relative to the beginning of the pandemic in Rhode Island (in weeks).Race and ethnicity were included because of previously documented disparities in infection risk. 13We also included a zip code-based 3-tier community risk classification created by RIDOH to help guide COVID-19 surveillance and response efforts (low-, moderate-, and high-risk tiers).The tier classification is based on community characteristics, such as population density, sociodemographic characteristics, and COVID-19 burden.Within each model, the interaction of age and time of entry into the risk cohort for reinfection were captured using cubic splines with df = 4. Cubic splines were used to capture nonlinear associations.Of note, case investigations for residents of LTCC settings generally occurred through a single point of contact at each facility, rather than individual discussions with residents who were diagnosed with COVID-19, leading to heterogeneity in the level of detail on sociodemographic and clinical characteristics.For each of the 3 subpopulations, we estimated vaccine effectiveness (VE) associated with preventing COVID-19 reinfection as VE = 1 -AHR, 7,14,15 where AHR is the adjusted hazard ratio estimated by the proportional hazards model.

Results
Between March 1, 2020, and December 9, 2021, 100 517 Rhode Island residents aged 12 years and older had recovered from SARS-COV-2 infection and remained unvaccinated at the end of their 90-day recovery period.During the subsequent study follow-up period, a total of 56 888 people  1).Among LTCC employees who were reinfected, the median (IQR) time between the positive tests associated with their first and second SARS-CoV-2 infections was 6.9 (4.8-8.7)months (Table 1).

JAMA Network Open | Infectious Diseases
For the general population, the median (IQR) age was 35 (24-52) years, 45 030 (47.6%) were female, and 66 030 (69.9%) were confirmed symptomatic during their first infection.Among members of the general population who were reinfected, median (IQR) time between the positive test results for with their first and second SARS-CoV-2 infections was 7.9 (4.8-10.7)months.Among the general population, 31 895 individuals (33.7%) were residents of high-risk communities and 22 963 individuals (24.3%) were residents of moderate-risk communities, per RIDOH's zip code-based community COVID-19 risk classification (Table 1).
The cumulative probabilities of reinfection as a function of time since recovery from the primary infection for each subpopulation are shown in eTable 1 in the Supplement.At 9 months after primary infection, the probability of reinfection among those who remained unvaccinated in the general population was 1.9% (95% CI, 1.8%-2.0%).The 9-month probability of reinfection among LTCC residents who remained unvaccinated was 13.0% (95% CI, 12.0%-14.0%)and among LTCC employees who remained unvaccinated was 10.0% (95% CI, 8.8%-11.5%).Overall cumulative reinfection probability by vaccination status appears in Figure 3, which shows substantially lower reinfection probabilities across time.Corresponding incidence rates for cases appear in eTable 2 in the Supplement, which also reports information on hospitalizations and deaths.
The findings of the proportional hazards regression model fit to each subpopulation are presented in eTable 3 in the Supplement, and the interactions of the continuous age and calendar time variables are show in the eFigure in the Supplement.The vaccination status HRs from this model were used to calculate VE estimates (Table 2).VE associated with CPVS (2 doses of mRNA-1273 or BNT162b2 or 1 dose of JNJ-78436735) after recovery from prior SARS-CoV-2 infection was 62% (95% CI, 56%-67%) for the general population , 49% (95% CI, 23% to 66%) for LTCC employees, and 49% (95% CI 26%-65%) for LTCC residents.
The proportional hazards model was also used to identify factors associated with risk of reinfection (eTable 3 and eFigure in the Supplement).In the general population, individuals who had been hospitalized as a result of their primary SARS-CoV-2 infection had almost 2-fold higher risk of reinfection (AHR, 1.91; 95% CI, 1.56-2.33).Risk of reinfection was lower among men than women (AHR, 0.74; 95% CI, 0.67-0.83);higher among individuals reporting White race or ethnicity compared with Asian, Black, and Hispanic individuals and individuals of other or unknown race; and lower among individuals who were confirmed symptomatic, compared with asymptomatic, during their primary infection (AHR, 0.58; 95% CI, 0.51-0.66).We did not find evidence of similar associations among LTCC residents or employees.
Details on sensitivity analyses are presented in the eAppendix in the Supplement.To assess sensitivity to our definition of CPVS, we restricted the definition of CPVS to receiving both doses at most 42 days apart (eTable 4 in the Supplement).To assess sensitivity to timing of first infection, we repeated the analysis on a sample restricted to individuals whose primary infection occurred during or after December 2020 (eTable 5 in the Supplement).In both cases, VE was similar.A comparison of VE by vaccine type (mRNA-1273 or BNT162b2 vs JNJ-78436735) did not generate strong evidence that estimated effectiveness differed between CVPS via JNJ-78436735 (48%; 95% CI, 26%-63%) was lower than VE associated with mRNA-1273 or BNT162b2 (64%; 95% CI, 58%-69%) in the general population (difference, 16%; 95% CI, −19% to 51%) (eTable 8 in the Supplement).Additionally, there was a higher case rate among individuals aged 20 to 30 years, likely resulting from surveillance testing at colleges and universities (eFigure in the Supplement).

Discussion
This cohort study is among the first, to our knowledge, to examine VE among people who recovered from COVID-19 and were unvaccinated at the time of their first infection; that is, we examined the effectiveness associated with vaccination after a primary infection.This analysis provides important information for guideline development, particularly relating to residents of LTCC facilities and employees who work in these settings.Previous infection is now taken into account in the Using a statewide population cohort, we found that probability of reinfection within 9 months of recovery from primary infection with SARS-CoV-2 was approximately 2% in the unvaccinated general population, but 10% in LTCC employees and 13% among LTCC residents.Using a prospective event-history analysis with vaccine exposure as a time-varying covariate, we found that the VE associated with CPVS after recovery from prior SARS-CoV-2 infection was 62% for the general population and 49% for both residents and employees of LTCC facilities.These estimates adjusted for age, sex, race and ethnicity, community COVID-19 risk, symptom and hospitalization status during the primary SARS-CoV-2 infection, and timing of entry into the risk cohort relative to the onset of the pandemic (calendar time).The findings of our statewide cohort study suggest that vaccination after recovery from prior COVID-19 was associated with substantial benefit for preventing subsequent reinfection, with risk reduced by nearly half for all subpopulations.Among LTCC residents and employees, for whom the probability of reinfection at 9 months after primary infection among those who remained unvaccinated was at least 5-fold greater than in the general population (10%-13% vs 2%), this is a particularly notable finding.
Our findings are consistent with a case-control study by Cavenaugh et al 8 using surveillance data from Kentucky, which found that being unvaccinated was associated with approximately 2-fold higher odds of reinfection compared with being fully vaccinated.A cohort study by Hammerman et al 9 using electronic medical record data from Israel estimated VE among people who had recovered from prior COVID-19 to be higher (approximately 60% for those aged Ն65 years and 82% for those aged 16-64 years).Another study of health care workers from the United Kingdom found that infection-acquired immunity waned after a year but was consistently higher than 90% for individuals who were subsequently vaccinated. 10Given that most hospitalizations and fatalities in our cohort were associated with the first infections, we were not able to conduct formal statistical inference about VE associated with preventing more severe reinfections; however, the observed incidence rates of reinfection-associated COVID-19 hospitalization and death were lower among individuals who were vaccinated compared with individuals who were unvaccinated.
Our findings highlight the continued importance of vaccination, including after recovery from prior COVID-19, among LTCC residents and employees, who are likely at increased risk of reinfection owing to their older age and/or greater exposure to the virus in LTCC settings.Employees and some residents or clients of other settings with greater COVID-19 exposure (eg, health care or correctional environments) may similarly experience high risk of reinfection and benefit from vaccination after recovery from COVID-19.

Limitations
Our analysis has some limitations.First, our analysis was based on diagnosed reinfections; consequently, we are likely underestimating infection rates, particularly in the general population.
Our adjustment for age partially accounts for differential testing patterns induced by programs, such as surveillance testing (eg, a higher case rate among individuals aged 20-30 years, likely resulting from surveillance testing at colleges and universities).We may have incomplete information on reinfections identified through home-based antigen testing, but this is expected to be minimal because home-based testing was uncommon until the final weeks of our analysis period.
Additionally, our VE estimates could be biased to the extent that testing propensity differs between individuals who are vaccinated and those who are not; for example, individuals who have been previously vaccinated may be less likely to seek testing for a mild infection, leading to an overestimate of VE.Another limitation is that RIDOH's immunization database does not include most vaccination doses administered by federal entities or out-of-state, which may lead to some misclassification of vaccination status.However, we expect this bias to be conservative, as erroneously classifying some individuals who had been vaccinated as unvaccinated should attenuate our VE estimates.Additionally, children ages 5 to 11 years were not included in our study because they were not eligible for vaccination until the final month of the analysis period.This is an important subgroup for future research.Furthermore, a substantial percentage of LTCC residents were missing data on race and ethnicity (65%) and symptom status during the first infection (80%), owing to heterogeneity in the level of detail reported by each facility, which limited our ability to account for potential confounding by these characteristics.

Conclusions
The findings of this cohort study suggest that risk of reinfection after recovery from COVID-19 was relatively high among individuals who remained unvaccinated, including the general population and LTCC residents and employees.Vaccination after recovery from COVID-19 was associated with substantial benefit, reducing risk of reinfection by approximately half.Individuals who have recovered from COVID-19 and remain unvaccinated should be encouraged to complete vaccinations, as they are eligible, to reduce their risk of reinfection.

Figure 1 .
Figure 1.Sample of Observation Patterns

Figure 2 .
Figure 2. Flowchart for the Analysis Cohort 199 936 New SARS-CoV-2 infections with test results between March 1, 2020, and December 9, 2021 March 2020 to December 2021, meaning that the VE estimates partially depend on data from a period during which vaccination was not yet available.To assess the robustness of our findings to sample selection, we conducted a supplementary sensitivity analysis in which we restricted to individuals whose first infection occurred during or after December 2020, when vaccines first became available.Statistical analyses were conducted with R statistical software version 4.1.2(R Project for Statistical Computing) using the 'survival' package for Kaplan-Meier and Cox proportional hazards model.Data were analyzed from October 2021 to January 2022.
, and LTCC employees were analyzed separately because they worked for extended periods of time in these high-transmission environments.We did not include group home residents or employees (2040 individuals), other congregate setting residents or employees (1397 individuals), and individuals with unknown congregate status in this study.Other congregate settings included predominantly correctional facilities.After exclusions, our analysis sample included 3124 LTCC residents, 2877 LTCC employees, and 94 516 members of the general population.

Table 2 .
Summary of Reinfections, Hospitalizations, and Deaths, and Vaccine Effectiveness Associated With Protection From COVID-19 Reinfection, by SubpopulationOur study has several key strengths.The database contains statewide information on positive PCR test results, most positive antigen test results, and nearly all vaccine doses administered in Rhode Island during the study period.The database also has information about whether the individual with positive test results is a resident or employee of a LTCC, a critical subpopulation with respect to SARS-COV-2 infection risk.A second major strength is the prospective nature of the analysis: we were able to follow individuals prospectively in time and calculate VE in terms of relative incidence (1 − AHR).