Evidence of leaky protection following COVID-19 vaccination and SARS-CoV-2 infection in an incarcerated population – Nature Communications

COVID-19 prevention and SARS-CoV-2 infections in the correctional system

The Connecticut DOC system is comprised of 13 facilities with a daily census of ~9300 residents²³. During the study (June 15, 2021 and May, 10, 2022), a total of 15,444 people spent at least one night housed in a DOC-operated facility, of which, 13,490 and 11,492 were residents during periods, respectively, of predominant Delta variant (June 15 to December 12, 2021) and Omicron variant (December 13, 2021 to May 10, 2022) transmission in Connecticut²⁴. As of the end of the study, 48% of currently incarcerated residents had completed their primary vaccine series and 27% had received a booster dose (Fig. 1A).

The DOC implemented a SARS-CoV-2 testing program consisting of testing of residents who were symptomatic, were contacts of confirmed cases, were due in court or had employment required testing, and residents who were newly incarcerated or transferred between facilities (rapid antigen testing). In addition, the DOC conducted voluntary, bi-weekly mass screening of 10% of residents (RT-PCR testing). Contact tracing included testing residents of (1) the same cell as an infected resident or (2) the same cellblock or facility as an infected resident if close contact (being within six feet for ≥15 min within a 24-h period) was reported by the infected resident (see Supplement DOC COVID-19 Testing). In total, 87,884 SARS-CoV-2 tests were performed during the study period, of which 20,794 were RT-PCRs and 67,090 were rapid antigen tests (Fig. 1B). Contact tracing among residents without reported symptoms comprised the largest proportion of testing (54%) followed by mass screening (24%; Fig. 1B). On average, the DOC tested 25% of residents every 2 weeks and 65% every 3 months during the study period.

Testing intensified from November 2021 to February 2022 (Fig. 1B) when Delta and Omicron BA.1 variant transmission contributed to an epidemic wave in Connecticut. During this period, the average proportion of residents tested in a 14-day period was 33.6% (red line, Fig. 1C). A total of 5079 SARS-CoV-2 infections were identified, of which 1598 and 3481 occurred during the Delta and Omicron periods, respectively. Among the 5079 infections, 57% and 38% were identified through contact tracing among residents without reported symptoms and testing in the presence of recorded symptoms, respectively (Fig. 1D).

Rolling matched cohort of residents exposed to SARS-CoV-2 infection

We conducted a rolling matched cohort study that compared the risk of SARS-CoV-2 infection and effectiveness of prior infection, vaccination, and hybrid immunity among residents with cell, cellblock, and without documented exposures to an infected case (Supplement Fig. 1). A cell exposure event was defined as having ≥1 cellmate test positive for SARS-CoV-2 in the absence of cellmates testing positive in the prior 14 days. A cellblock exposure event was defined as having ≥1 resident of the same cellblock (but different cell) test positive in the absence of a cellmate or resident of the cellblock testing positive in the prior 14 days. Events without documented exposures were defined as days when residents did not have a cell or cellblock exposure event in the prior 14 days. We prevented the inclusion of multiple events without documented exposures from the same person during a 14-day period through random selection. We selected a cohort of events by cluster matching on facility and calendar day and ascertained infection in the subsequent 14-day period during Delta and Omicron periods.

During the Delta period, we identified 290 cell and 5805 cellblock exposure events among the 7389 residents who were incarcerated for ≥14 days and spent ≥1 night in a cell with a roommate (Fig. 2A). Among the 584,629 events without documented exposures, we randomly selected 37,394 unique events. After matching, we identified a sample of 264 cell exposure events (258 residents), 5,616 cellblock exposure events (3745 residents), and 17,024 events without documented exposure (6073 residents).

Flowchart showing how people incarcerated within Connecticut Department of Correction facilities and who resided in cells between June 15, 2021, and December 12, 2021 (Delta Predominant Period [A]) and December 13, 2021, and May 10, 2022 (Omicron Predominant Period [B]), were included in the analysis. Residents were classified as having a cell exposure event (green) on the day their cellmate tested positive, having a cellblock exposure event (brown) the day a resident of their cellblock but not cell tested positive, and having an event without documented exposure (navy) if no one in their cellblock tested positive on a given day. Cell exposure events that occurred within 14 days following a prior cell exposure event were excluded. Cellblock exposure events and events without documented exposures that occurred in the 14 days following a cellblock or cell exposure event were excluded. A To prevent the inclusion of multiple events without documented exposures from the same person during a 14-day period, we randomly selected incarceration events without documented exposures and excluded all others within the prior or following 14 days. B We defined infections as a positive RT-PCR or rapid antigen test during the 14 days of follow-up.

During the Omicron period, we identified 796 cell and 6408 cellblock exposure events and 259,320 events without documented exposures among 6161 residents who were incarcerated for ≥14 days and resided in a cell with a roommate for ≥1 day (Fig. 2B). We randomly selected 20,125 of 259,320 events without a documented exposure. Following matching, we selected 702 cell exposure events (671 residents), 5980 cellblock exposure events (4135 residents), and 13,464 events without documented exposures (5429 residents).

Characteristics of residents with and without exposure to SARS-CoV-2 infections

During the Delta period, events with and without documented exposures occurred among racially similar residents and residents with similar cell sizes (median: 2 residents). However, cellblock exposure events occurred more frequently among residents of larger median cellblock sizes (107.0 residents) than cell exposure events (74 residents) or events without documented exposures (88 residents). Cell exposure events occurred less frequently among people with recorded prior infections (32.2%), vaccination (41.3%), or hybrid immunity (17.1%) than events without documented exposures (infection, 38.8%; vaccination, 53.7%; hybrid, 25.2%; Table 1). Male residents were more likely to have had a prior, recorded SARS-CoV-2 infection than female residents regardless of their SARS-CoV-2 exposure status (Supplement Table 1). Among residents of the same age, race, room-size, cellblock, and inclusion time, the time since last prior infection and vaccination did not differ significantly between residents with and without documented exposures (Supplement Table 2).

During the Omicron period, events with and without documented exposures occurred among racially similar residents and residents with similar cell and cellblock sizes. Cell exposure events occurred with similar frequency among unvaccinated residents (46.0%) as cellblock exposure events (43.1%) and events without documented exposures (42.9%). Cell exposure events occurred less frequently among people with recorded prior infections (36.2%) or hybrid immunity (23.9%) than cellblock exposure events (prior infection: 43.6%; hybrid immunity: 28.9%) or events without documented exposures (prior infection: 47.0%; hybrid immunity: 30.4%; Table 1). Male residents were more likely to have had a prior, recorded SARS-CoV-2 infection than female residents regardless of their exposure status (Supplement Table 1). Among residents of the same age, race, room-size, cellblock, and inclusion time, the time since last prior infection and vaccination did not differ significantly between residents with and without documented exposures (Supplement Table 2).

High exposure settings in cells and cellblocks impart increased infection risk

During the Delta period, 122 residents tested positive following an event without documented exposure, 233 residents tested positive following a cellblock exposure event, and 53 residents tested positive following a cell exposure event (Fig. 2A). The hazard of infection was 2.67 (95% Confidence Interval [CI]: 1.84–3.88) and 9.70 (6.29–14.96) times higher following cellblock and cell exposure events than events without documented exposure, respectively (Fig. 3). The hazard of symptomatic infection, defined as a positive rapid antigen test collected from a symptomatic resident, was 2.21 (1.28–3.82) and 7.44 (3.87–14.30) times higher following cellblock and cell exposure events than events without documented exposure, respectively (Supplement Fig. 2; Unadjusted estimates: Supplement Table 4).

Forest plot depicting the association between documented close exposure to a SARS-CoV-2 infected resident and the risk of subsequent SARS-CoV-2 infection. Residents were classified as having a cell exposure event (green) on the day their cellmate tested positive, having a cellblock exposure event (brown) the day a resident of their cellblock but not cell tested positive, and having an event without documented exposure if no one in their cellblock tested positive on a given day. Cell exposure events that occurred within 14 days following a prior cell exposure event were excluded. Cellblock exposure events and events without documented exposures that occurred in the 14 days following a cellblock or cell exposure event were excluded. Facility exposures were stratified by periods of variant predominance (Delta [A]: June 15, 2021–December 12, 2021; Omicron [B]: December 13, 2021–May 10, 2022). The associations were estimated using a Cox Proportional Hazard Model stratified by facility and with robust standard errors. The model was adjusted for age, calendar date, race, room and cellblock size, vaccination, and prior infection status of the susceptible person. Boxes indicate estimated hazard ratio (HR) point values and whiskers indicate 95% confidence intervals (Delta Period: n = 22,904 facility events; Omicron Period: n = 20,146 facility events). Unadjusted results presented in Supplement Table 3.

We conducted a series of sensitivity analyses to address concerns regarding potential sources of bias (Supplement: Sensitivity Analyses). Of primary concern is the bias resulting from inequal testing following events with and without documented exposure (Fig. 1C; Supplement Fig. 4). To examine these biases, we performed sensitivity analyses (1) restricted to residents who were tested during follow-up and (2) restricted to residents tested during follow-up for non-symptomatic reasons. Following the restriction to tested residents, the hazard of infection was 1.89 (1.36–2.64) and 5.23 (3.50–7.82) times higher following cellblock and cell exposure events than events without documented exposures, respectively (Supplement Fig. 5; Unadjusted estimates: Supplement Table 9). Restricting to tests conducted for non-symptomatic reasons resulted in the exclusion of an additional 12 facility exposure events and the point estimates were within 0.01 of the sensitivity analysis restricting to tested residents (Supplement Fig. 7; Unadjusted estimates: Supplement Table 11).

Additionally, we did not have access to community infection data and were concerned about the bias introduced from prior infection misclassification. To reduce this bias, we conducted a sensitivity analysis limited to people incarcerated since the beginning of our study (June 15, 2021). The hazard of infection was 3.15 (2.01–4.92) and 12.96 (7.90–21.26) times higher following cellblock and cell exposure events than events without documented exposure, respectively (Supplement Fig. 9; Unadjusted estimates: Supplement Table 13). Further, since we may have overestimated the effect of facility exposures by including residents who were exposed to more than one index case on a given day, we conducted a sensitivity analysis limited to cellblock and cell exposure events with only one index case. To ensure including already infected residents did not drive our findings, we conducted an analysis restricted to residents who tested negative in the prior 5 days. To ensure our exposures were temporally linked to observed infections, we conducted two sensitivity analyses: one excluding the first 2 days of follow-up, and one limiting follow-up to 9 days. We found that cellblock and cell exposure events were significantly associated with the hazards of infection for each scenario (Supplement Figs. 11, 13, 14, 16; Unadjusted estimates: Supplement Tables 15, 17, 18, 20).

During the Omicron period, 167 residents tested positive following an event without documented exposures, 502 residents tested positive following a cellblock exposure event, and 109 residents tested positive following a cell exposure event (Fig. 2B). The hazard of infection was 3.34 (2.22–5.00) and 4.73 (3.05–7.36) times higher following cellblock or cell exposure events than events without documented exposure, respectively (Fig. 3). The hazard of symptomatic infection was 3.82 (2.08–7.00) and 7.00 (3.61–13.58) times higher following cellblock and cell exposure events than events without documented exposure, respectively (Supplement Fig. 2; Unadjusted estimates: Supplement Table 4).

We conducted the same sensitivity analyses as for the Delta period. Following the restriction to tested residents, the hazard of infection was 2.14 (1.62–2.82) and 2.23 (1.62–3.07) times higher following cellblock and cell exposure events than events without documented exposure, respectively (Supplement Fig. 5; Unadjusted estimates: Supplement Table 9). Following the restriction to residents incarcerated since the beginning of the study, the hazard of infection was 4.40 (2.84–6.82) and 6.17 (3.75–10.14) times higher following cellblock and cell exposure events than events without documented exposure (Supplement Fig. 9; Unadjusted estimates: Supplement Table 13). Cellblock and cell exposures were found to be significantly associated with an increased hazard of infection for each additional scenario (Supplement Figs. 7, 11, 13, 14, 16; Unadjusted estimates: Supplement Tables 11, 15, 17, 18, 20).

High exposure setting overcomes the protection afforded by infection, vaccination, and hybrid immunity

During the Delta period, the effectiveness of prior infection at reducing the hazard of SARS-CoV-2 infection was highest following events without documented exposure (Hazard Ratio [HR]: 0.21 [0.11–0.39]) and lowest following cell exposure events (HR: 0.59 [0.30–1.16]). Vaccine effectiveness was highest following events without documented exposure (HR: 0.32 [0.21–0.49]) and lowest following cell exposure events (HR: 0.74 [0.37–1.48]). The effectiveness of hybrid immunity was highest following events without documented exposure (HR: 0.05 [0.02–0.10]) and lowest following cell exposure events (HR: 0.29 [0.07–1.12]). The effectiveness of prior infection, vaccination, and hybrid immunity was significantly lower following cell exposure events than following events without documented exposure (P = 0.029, 0.033, 0.026, respectively; Fig. 4/Supplement Table 6). The effectiveness of prior infection and vaccination at reducing the hazard of symptomatic SARS-CoV-2 infection was highest following events without documented exposure (HR: infection, 0.18 [0.07–0.45], vaccination, 0.21 [0.11–0.41]) and lowest following cell exposure events (HR: infection, 0.42 [0.11–1.60], vaccination, 0.53 [0.17–1.64]). No residents with hybrid immunity had a symptomatic infection following a cell exposure event (Supplement Fig. 3; Unadjusted estimates: Supplement Table 5).

Forest plot depicting the association between prior infection, vaccination, and hybrid immunity and the risk of SARS-CoV-2 infection by facility exposure type. Residents were classified as having a cell exposure event (green) the day their cellmate tested positive, having a cellblock exposure event (brown) the day a resident of their cellblock but not cell tested positive, and having an event without documented exposure (navy) if no one in their cellblock tested positive. Cell exposure events that occurred within 14 days following a prior cell exposure event were excluded. Cellblock exposure events and events without documented exposure that occurred in the 14 days following a cellblock or cell exposure event were excluded. Associations were examined using Cox Proportional Hazard Models stratified by cellblock with robust standard errors. Each model was adjusted for age, date of exposure, race, room size, and model (a) was adjusted for vaccination status and model (b) was adjusted for prior infection status. Model (c) was limited to residents with hybrid immunity or residents without a record of prior infection or vaccination. Prior infection was defined as a recorded positive SARS-CoV-2 test ≥90 days before the event and vaccination was defined as the receipt of ≥1 dose before the event. Hybrid immunity was defined as a record of both a prior infection and ≥1 vaccine dose. Boxes indicate estimated hazard ratio (HR) point values and whiskers indicate 95% confidence intervals (Delta: n = 17,024 no exposure events, 5616 cellblock exposure events, 264 cell exposure events; Omicron: n = 13,464 no exposure events, 5980 cellblock exposure events, 702 cell exposure events). The ratio of HRs refer to the p-value comparing the HR following cellblock or cell exposure events to the HR following events without documented exposures, estimated using two-sided z-tests. No multiple testing adjustment was performed. Unadjusted results in Supplement Table 7.

We performed sensitivity analyses that paralleled those described above (see Supplement: Sensitivity Analyses). The effectiveness of prior infection and hybrid immunity was highest following events without documented exposure and lowest following cell exposure events under all scenarios (Supplement Figs. 6, 8, 10, 12, 15, 17; Unadjusted estimates: Supplement Tables 10, 12, 14, 16, 19, 21). Vaccine effectiveness was highest following events without documented exposures for all scenarios, except when we limited follow-up to 9 days (Supplement Fig. 17; Unadjusted estimates: Supplement Table 21). Under this scenario, vaccination reduced the hazard of infection by 0.31 (0.19–0.51) times following events without documented exposure, 0.31 (0.21–0.46) times following cellblock exposure events, and 0.87 (0.45–1.69) times following cell exposure events. When we restricted to residents tested during follow-up, the effectiveness of prior infection, vaccination, and hybrid immunity was highest following events without documented exposure (HR: infection, 0.23 [0.12–0.42]; vaccination, 0.34 [0.22–0.52]; hybrid, 0.05 [0.02–0.11]) and lowest following cell exposure events (HR: infection, 0.50 [0.25–0.98]; vaccination, 0.72 [0.37–1.41]; hybrid, 0.33 [0.12–0.91]; Supplement Fig. 6; Unadjusted estimates: Supplement Table 10). Following the restriction to people incarcerated since the beginning of the study, the effectiveness of prior infection, vaccination, and hybrid immunity was highest following events without documented exposure (HR: infection, 0.25 [0.13–0.48]; vaccination, 0.32 [0.19–0.54]; hybrid, 0.07 [0.02–0.18]) and lowest following cell exposure events (HR: infection, 0.51 [0.25–1.03]; vaccination, 0.77 [0.38–1.58]; hybrid, 0.31 [0.10–0.99]; Supplement Fig. 10; unadjusted estimates: Supplement Table 14).

During the Omicron period, the effectiveness of prior infection was highest following events without documented exposure (HR: 0.36 [0.25–0.54]) and lowest following cell exposure events (HR: 0.89 [0.58–1.35]). Vaccine effectiveness was highest following events without documented exposure (HR: 0.57 [0.42–0.78]) and lowest following cell exposure events (HR: 0.96 [0.64–1.46]). The effectiveness of hybrid immunity was highest following events without documented exposure (HR: 0.24 [0.15–0.39]) and lowest following cell exposure events (HR: 0.80 [0.46–1.39]). The effectiveness of prior infection, vaccination, and hybrid immunity was significantly lower following cell exposure events than events without documented exposure (P = 0.002, 0.041, and 0.001, respectively; Fig. 4/Supplement Table 6). The effectiveness of prior infection, vaccination, and hybrid immunity at reducing the hazard of symptomatic SARS-CoV-2 infection was highest following events without documented exposure (HR: infection, 0.35 [0.21–0.59]; vaccination, 0.33 [0.21–0.53]; hybrid, 0.13 [0.06–0.28]) and lowest following cell exposure events (HR: infection, 0.77 [0.45–1.31]; vaccination, 0.62 [0.35–1.10]; hybrid, 0.53 [0.26–1.11]; Supplement Fig. 3; Unadjusted estimates: Supplement Table 5).

Following the restriction to residents tested during follow-up, the effectiveness of prior infection, vaccination, and hybrid immunity was highest following events without documented exposures (HR: infection, 0.44 [0.30–0.63]; vaccination, 0.49 [0.36–0.68]; hybrid 0.32 [0.19–0.53]) and lowest following cell exposure events (HR: infection, 0.69 [0.47–1.02]; vaccination, 0.81 [0.53–1.22]; hybrid, 0.67 [0.44–1.02]; Supplement Fig. 6; unadjusted estimates: Supplement Table 10). Following the restriction to people incarcerated since the study began, the effectiveness of prior infection, vaccination, and hybrid immunity was highest following events without documented exposures (HR: infection, 0.38 [0.24–0.58]; vaccination, 0.61 [0.43–0.86]; hybrid, 0.23 [0.14–0.38]) and lowest following cell exposure events (HR: infection, 0.88 [0.56–1.37]; vaccination, 0.79 [0.49–1.28]; hybrid, 0.64 [0.39–1.05]; Supplement Fig. 10; Unadjusted estimates: Supplement Table 14). The effectiveness of prior infection, vaccination, and hybrid immunity was highest following events without documented exposures and lowest following cell exposure events for each additional sensitivity analysis (Supplement Figs. 8, 12, 15, 17; Unadjusted estimates: Supplement Tables 12, 16, 19, 21).

SARS-CoV-2 exposure specific effects of prior infection and vaccination on Infectiousness

As a secondary analysis, we hypothesized that the prior infection and vaccination status of the index cases may influence transmission. We examined this by restricting our sample to cellblock and cell exposure events and comparing the hazards of infection when the index case had and did not have the immunizing event of interest. During the Delta period, the prior infection history of the index case was associated with a non-significantly higher hazard of SARS-CoV-2 transmission following cellblock exposure events (HR: 1.96 [0.93–4.12]) and a non-significantly lower hazard following cell exposure events (HR: 0.91 [95 CI: 0.20–4.18]). The vaccination status of the index case was associated with a non-significantly lower hazard of SARS-CoV-2 transmission among cellblock exposure events (HR: 0.75 [0.18–3.12]) and cell exposure events (HR: 0.71 [0.26–1.93]; Fig. 5).

Forest plot depicting the association between vaccination and prior infection and the risk of subsequent SARS-CoV-2 infection by documented SARS-CoV-2 exposure status. Residents were classified as having a cell exposure event (green) on the day their cellmate tested positive, having a cellblock exposure event (brown) the day a resident of their cellblock but not cell tested positive, and having an event without documented exposure if no one in their cellblock tested positive on a given day. Cell exposure events that occurred within 14 days following a prior cell exposure event were excluded. Cellblock exposure events and events without documented exposures that occurred in the 14 days following a cellblock or cell exposure event were excluded. Residents were classified as being vaccinated if they had received at least one vaccine dose. Associations were examined using Cox Proportional Hazard Models stratified by housing cellblock with robust standard errors. The models were adjusted for (a) age, date of exposure, race, room size, vaccination and prior infection status of the susceptible resident, and vaccinated status of the index case (limited to exposed residents), (b) age, date of exposure, race, room size, vaccination and prior infection status of the susceptible resident, and prior infection status of the index case (limited to exposed residents). Prior infections were defined as a recorded positive SARS-CoV-2 test at least 90 days prior to the event and vaccination was defined as the receipt of at least one dose prior to the event. Boxes indicate estimated hazard ratio (HR) point values and whiskers indicate 95% confidence intervals (Delta: n = 4407 cellblock exposure events, 127 cell exposure events; Omicron: n = 3831 cellblock exposure events, 250 cell exposure events). Unadjusted results in Supplement Table 8.

During the Omicron period, the prior infection status of the index case was associated with a non-significantly lower hazard of SARS-CoV-2 transmission following cellblock exposure events (HR: 0.52 [0.27–1.03]) and cell exposure events (HR: 0.72 [95 CI: 0.25–2.03]). The vaccination history of the index case was associated with a non-significantly lower hazard of SARS-CoV-2 transmission among cellblock exposure events (HR: 0.55 [0.24–1.24]) and cell exposure events (HR: 0.52 [0.20–1.36]; Fig. 5).