Browse the corpus

Covid-19 Surveillance Testing and Resident Outcomes in Nursing Homes. BACKGROUND: Despite widespread adoption of surveillance testing for coronavirus disease 2019 (Covid-19) among staff members in skilled nursing facilities, evidence is limited regarding its relationship with outcomes among facility residents. METHODS: Using data obtained from 2020 to 2022, we performed a retrospective cohort study of testing for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) among staff members in 13,424 skilled nursing facilities during three pandemic periods: before vaccine approval, before the B.1.1.529 (omicron) variant wave, and during the omicron wave. We assessed staff testing volumes during weeks without Covid-19 cases relative to other skilled nursing facilities in the same county, along with Covid-19 cases and deaths among residents during potential outbreaks (defined as the occurrence of a case after 2 weeks with no cases). We reported adjusted differences in outcomes between high-testing facilities (90th percentile of test volume) and low-testing facilities (10th percentile). The two primary outcomes were the weekly cumulative number of Covid-19 cases and related deaths among residents during potential outbreaks. RESULTS: During the overall study period, 519.7 cases of Covid-19 per 100 potential outbreaks were reported among residents of high-testing facilities as compared with 591.2 cases among residents of low-testing facilities (adjusted difference, -71.5; 95% confidence interval [CI], -91.3 to -51.6). During the same period, 42.7 deaths per 100 potential outbreaks occurred in high-testing facilities as compared with 49.8 deaths in low-testing facilities (adjusted difference, -7.1; 95% CI, -11.0 to -3.2). Before vaccine availability, high- and low-testing facilities had 759.9 cases and 1060.2 cases, respectively, per 100 potential outbreaks (adjusted difference, -300.3; 95% CI, -377.1 to -223.5), along with 125.2 and 166.8 deaths (adjusted difference, -41.6; 95% CI, -57.8 to -25.5). Before the omicron wave, the numbers of cases and deaths were similar in high- and low-testing facilities; during the omicron wave, high-testing facilities had fewer cases among residents, but deaths were similar in the two groups. CONCLUSIONS: Greater surveillance testing of staff members at skilled nursing facilities was associated with clinically meaningful reductions in Covid-19 cases and deaths among residents, particularly before vaccine availability.

No outside agencies were involved in the design or conduct of this study. In accordance with policies regarding the use of publicly available data adopted by Harvard University (the institution of the third [senior] author), the study was exempt from review by an institutional review board. Harvard played no other role in the study. The primary data source for this retrospective cohort study of U.S. skilled nursing facilities was the Covid-19 Nursing Home Database of the Centers for Medicare and Medicaid Services (CMS),20 a mandatory repository that contains data (including Covid-19 cases and deaths) that are submitted on a weekly basis by skilled nursing facilities to the National Healthcare Safety Network (NHSN) of the Centers for Disease Control and Prevention. Data were obtained from November 22, 2020, to May 15, 2022, which corresponds to the period in which all relevant testing questions were included in the NHSN Covid-19 modules. We obtained information about facility staffing from the CMS-mandated Payroll-Based Journal Employee Detail database. This database, which compiles individual-level staffing data from facility payroll records or other auditable sources, provides a record of daily individual staff shifts that can be used to determine staffing levels according to discipline (e.g., nursing and physical therapy), as well as the number of staff members who were working on a given day.21 Additional data sources are detailed in the Supplementary Appendix, available with the full text of this article at NEJM.org.

staff shifts that can be used to determine staffing levels according to discipline (e.g., nursing and physical therapy), as well as the number of staff members who were working on a given day.21 Additional data sources are detailed in the Supplementary Appendix, available with the full text of this article at NEJM.org. To examine the relationship between surveillance testing and Covid-19 outcomes, we constructed a study sample consisting of facility-weeks during which skilled nursing facilities were having a potential outbreak of Covid-19. To allow for adequate follow-up time, we included data from potential outbreaks from November 22, 2020, to March 20, 2022. We defined the start of a potential outbreak within a facility as the first week during which any Covid-19 cases among staff or residents were reported after 2 consecutive weeks of no such reports within the facility. The week that the initial case had been identified was defined as the start of the potential outbreak (i.e., week 1), and we followed outcomes over subsequent weeks until there were two consecutive weeks with no new staff or resident cases or 7 follow-up weeks, whichever came first. When a potential outbreak ended before 7 weeks of follow-up, we tracked Covid-19–related deaths among residents for an additional 2 weeks to account for the lag between the diagnosis of Covid-19 cases and deaths.

were two consecutive weeks with no new staff or resident cases or 7 follow-up weeks, whichever came first. When a potential outbreak ended before 7 weeks of follow-up, we tracked Covid-19–related deaths among residents for an additional 2 weeks to account for the lag between the diagnosis of Covid-19 cases and deaths. First, we quantified the volume of surveillance (as opposed to diagnostic) testing as the total number of Covid-19 tests — including point-of-care (POC) and non-POC tests — that were administered to staff members during weeks in which facilities had no Covid-19 cases among either staff or residents. We normalized this measure to the size of the staff, defined as the weekly average number of persons with any reported hours in the payroll-based data during the first quarter of 2021. We performed winsorization (i.e., the limitation of extreme values to reduce the effect of possibly spurious outliers) on weekly data regarding the number of Covid-19 tests that had been administered per staff member at the 1st and 99th percentiles.

rted hours in the payroll-based data during the first quarter of 2021. We performed winsorization (i.e., the limitation of extreme values to reduce the effect of possibly spurious outliers) on weekly data regarding the number of Covid-19 tests that had been administered per staff member at the 1st and 99th percentiles. To address confounding between staff test volume and resident outcomes (Fig. S1 in the Supplementary Appendix), we used a regression framework to estimate the propensity of each facility for surveillance testing as compared with other skilled nursing facilities in the same county, a value that we defined as the relative testing rate (our primary exposure variable) over the entire study window. To determine the relative testing rate, we restricted the sample to all facility-weeks with no Covid-19 cases among staff or residents and used linear regression to calculate the number of weekly tests administered per staff member after adjustment for facility fixed effects, county-week fixed effects, and time-varying weekly measures of staff and resident vaccination rates (Section S1.4). We used the facility fixed effects from this model to define the relative testing rate (i.e., the latent, time-invariant propensity for surveillance testing relative to other facilities in the same county-week) after adjustment for vaccination rates among staff and residents. This value reflects the variation in surveillance testing that is not attributable to community-level risk, regional testing regulations, secular time trends, or degree of vaccine coverage.

ce testing relative to other facilities in the same county-week) after adjustment for vaccination rates among staff and residents. This value reflects the variation in surveillance testing that is not attributable to community-level risk, regional testing regulations, secular time trends, or degree of vaccine coverage. Our two primary outcomes were the weekly cumulative number of Covid-19 cases and related deaths among residents, as reported by facilities in the NHSN database during a potential outbreak. Outcomes were reported as the number per 100 potential outbreaks. We focused on outcomes among residents because cases among staff members are, in part, a direct function of the key exposure — the number of tests that were performed among staff. In secondary analyses, we examined cumulative Covid-19 cases and related deaths among staff members.

the number per 100 potential outbreaks. We focused on outcomes among residents because cases among staff members are, in part, a direct function of the key exposure — the number of tests that were performed among staff. In secondary analyses, we examined cumulative Covid-19 cases and related deaths among staff members. We captured testing type and test turnaround time as facility-level variables. For testing type, we determined the share of staff tests that were POC (i.e., antigen tests) as opposed to non-POC (typically, PCR tests) during the week before the start of a potential outbreak. This information was missing for 14% of potential outbreak pairs (Section S1.5). We classified potential outbreaks into two groups: those in which 50% or more of tests at baseline were POC and those that were not (<50% POC) (Fig. S2). We defined test turnaround time on the basis of survey responses during the week before a potential outbreak and during the first week of the outbreak as 2 days or less in both weeks or 3 days or more in either week. The test turnaround time was defined only for skilled nursing facilities that predominantly used non-POC testing (i.e., in ≥80% of staff tests) before the start of a potential outbreak, since POC tests have immediate turnaround by definition. Covariates included other facility characteristics and measures of Covid-19 risk (Section S1.6).

time was defined only for skilled nursing facilities that predominantly used non-POC testing (i.e., in ≥80% of staff tests) before the start of a potential outbreak, since POC tests have immediate turnaround by definition. Covariates included other facility characteristics and measures of Covid-19 risk (Section S1.6). We estimated the association between testing patterns and weekly resident Covid-19 outcomes within a potential outbreak using linear regression models. We estimated separate models for each outcome in which the key quantities of interest were interaction terms between relative testing rate (a continuous variable) and indicators for each potential outbreak week (i.e., weeks 1 to 8 during a potential outbreak). Models included facility- and county-level covariates, as well as county fixed effects and fixed effects for the calendar week in which the potential outbreak began. We performed multiple imputation for missing values in facility covariates and used the last nonmissing values to impute missing values in time-varying characteristics (i.e., the availability of personal protective equipment) (Sections S1.6 and S1.7). Standard errors were clustered at the facility level. All 95% confidence intervals were not adjusted for multiplicity and should not be used in place of hypothesis testing. Because the key quantities of interest (relative testing rate according to potential outbreak week) were continuous variables, we report weekly estimates at two points in the relative testing rate distribution: the 90th percentile, representing high-testing facilities, and the 10th percentile, representing low-testing facilities.

key quantities of interest (relative testing rate according to potential outbreak week) were continuous variables, we report weekly estimates at two points in the relative testing rate distribution: the 90th percentile, representing high-testing facilities, and the 10th percentile, representing low-testing facilities. To examine whether test volume had different effects during different phases of the pandemic, we stratified our analyses into three periods: before vaccine availability (November 22, 2020, to January 17, 2021), before the wave of the B.1.1.529 (omicron) variant (January 18 to October 31, 2021), and during the omicron wave (November 1, 2021, to March 20, 2022). We performed a series of sensitivity analyses to examine the robustness of our findings, including the use of alternate specifications of the covariates, indicators to account for missing covariate information, imputed data for missing information regarding the type of testing used, the estimation of Poisson rather than linear-regression models, the extension of follow-up time after the start of a potential outbreak, the restriction of analyses to actual outbreaks (i.e., two or more Covid-19 cases among residents), and the use of an alternate analytic framework that estimates the association between the surveillance test volume and resident outcomes according to variation over time within the same facility and estimation models that contain facility fixed effects (Section S1.7).

To examine the relationship between surveillance testing and Covid-19 outcomes, we constructed a study sample consisting of facility-weeks during which skilled nursing facilities were having a potential outbreak of Covid-19. To allow for adequate follow-up time, we included data from potential outbreaks from November 22, 2020, to March 20, 2022. We defined the start of a potential outbreak within a facility as the first week during which any Covid-19 cases among staff or residents were reported after 2 consecutive weeks of no such reports within the facility. The week that the initial case had been identified was defined as the start of the potential outbreak (i.e., week 1), and we followed outcomes over subsequent weeks until there were two consecutive weeks with no new staff or resident cases or 7 follow-up weeks, whichever came first. When a potential outbreak ended before 7 weeks of follow-up, we tracked Covid-19–related deaths among residents for an additional 2 weeks to account for the lag between the diagnosis of Covid-19 cases and deaths.

First, we quantified the volume of surveillance (as opposed to diagnostic) testing as the total number of Covid-19 tests — including point-of-care (POC) and non-POC tests — that were administered to staff members during weeks in which facilities had no Covid-19 cases among either staff or residents. We normalized this measure to the size of the staff, defined as the weekly average number of persons with any reported hours in the payroll-based data during the first quarter of 2021. We performed winsorization (i.e., the limitation of extreme values to reduce the effect of possibly spurious outliers) on weekly data regarding the number of Covid-19 tests that had been administered per staff member at the 1st and 99th percentiles.

Our two primary outcomes were the weekly cumulative number of Covid-19 cases and related deaths among residents, as reported by facilities in the NHSN database during a potential outbreak. Outcomes were reported as the number per 100 potential outbreaks. We focused on outcomes among residents because cases among staff members are, in part, a direct function of the key exposure — the number of tests that were performed among staff. In secondary analyses, we examined cumulative Covid-19 cases and related deaths among staff members.

We captured testing type and test turnaround time as facility-level variables. For testing type, we determined the share of staff tests that were POC (i.e., antigen tests) as opposed to non-POC (typically, PCR tests) during the week before the start of a potential outbreak. This information was missing for 14% of potential outbreak pairs (Section S1.5). We classified potential outbreaks into two groups: those in which 50% or more of tests at baseline were POC and those that were not (<50% POC) (Fig. S2). We defined test turnaround time on the basis of survey responses during the week before a potential outbreak and during the first week of the outbreak as 2 days or less in both weeks or 3 days or more in either week. The test turnaround time was defined only for skilled nursing facilities that predominantly used non-POC testing (i.e., in ≥80% of staff tests) before the start of a potential outbreak, since POC tests have immediate turnaround by definition. Covariates included other facility characteristics and measures of Covid-19 risk (Section S1.6).

We estimated the association between testing patterns and weekly resident Covid-19 outcomes within a potential outbreak using linear regression models. We estimated separate models for each outcome in which the key quantities of interest were interaction terms between relative testing rate (a continuous variable) and indicators for each potential outbreak week (i.e., weeks 1 to 8 during a potential outbreak). Models included facility- and county-level covariates, as well as county fixed effects and fixed effects for the calendar week in which the potential outbreak began. We performed multiple imputation for missing values in facility covariates and used the last nonmissing values to impute missing values in time-varying characteristics (i.e., the availability of personal protective equipment) (Sections S1.6 and S1.7). Standard errors were clustered at the facility level. All 95% confidence intervals were not adjusted for multiplicity and should not be used in place of hypothesis testing. Because the key quantities of interest (relative testing rate according to potential outbreak week) were continuous variables, we report weekly estimates at two points in the relative testing rate distribution: the 90th percentile, representing high-testing facilities, and the 10th percentile, representing low-testing facilities.

During the 77-week study window, we identified 13,424 study-eligible skilled nursing facilities and 66,900 potential outbreaks for an average of 5.0 potential outbreaks per facility. On average, sample facilities performed 1.1 surveillance tests per staff member per week (interquartile range, 0.4 to 1.5) (Figure 1). Estimated relative rates of surveillance testing ranged from −0.8 to 1.6 tests per staff-week relative to the facility county-by-week average, such that high-testing facilities in the 90th percentile performed an additional 0.6 surveillance tests per staff-week relative to the county-week average, whereas low-testing facilities in the 10th percentile performed 0.5 fewer surveillance tests than the county-week average. Relative testing rates were predictive of actual test volumes (correlation coefficient, 0.51). As compared with facilities in the bottom quartile of the relative testing rate, facilities in the top quartile were more likely to be nonprofit and to have a five-star quality score; in addition, top-quartile facilities had lower percentage of non-White residents, smaller staff sizes, older residents, and lower rates of reporting a shortage of personal protective equipment (Table 1). Facilities across all quartiles had a similar number of CMS-defined infection-control deficiencies during the most recent inspection. Missing data for facility characteristics ranged from 0.04 to 8.7% of sample observations (Table S1). The characteristics of the facilities in the top and bottom deciles of relative testing rates are shown in Table S3.

had a similar number of CMS-defined infection-control deficiencies during the most recent inspection. Missing data for facility characteristics ranged from 0.04 to 8.7% of sample observations (Table S1). The characteristics of the facilities in the top and bottom deciles of relative testing rates are shown in Table S3. During a potential outbreak, high-testing facilities had fewer adjusted Covid-19 cases and deaths among residents than low-testing facilities (Figure 2, Table 2, and Table S4). During the full study period, high-testing facilities had 519.7 cumulative resident Covid-19 cases per 100 potential outbreaks, as compared with 591.2 cases per 100 potential outbreaks in low-testing facilities, an adjusted difference of −71.5 cases (95% confidence interval [CI], −91.3 to −51.6). High-testing facilities had 42.7 cumulative resident Covid-19 deaths per 100 potential outbreaks, as compared with 49.8 deaths in low-testing facilities (adjusted difference, −7.1; 95% CI, −11.0 to −3.2). High-testing facilities had approximately 15% more Covid-19 cases among staff than low-testing facilities, with no appreciable differences in the number of deaths (Fig. S4).

-19 deaths per 100 potential outbreaks, as compared with 49.8 deaths in low-testing facilities (adjusted difference, −7.1; 95% CI, −11.0 to −3.2). High-testing facilities had approximately 15% more Covid-19 cases among staff than low-testing facilities, with no appreciable differences in the number of deaths (Fig. S4). The difference between high- and low-testing facilities was larger during the prevaccine period than after vaccines had become available. High-testing facilities had 759.9 cases as compared with 1060.2 cases per 100 potential outbreaks in low-testing facilities (adjusted difference, −300.3; 95% CI, −377.1 to −223.5). During this period, high-testing facilities had 125.2 deaths per 100 potential outbreaks as compared with 166.8 deaths in low-testing facilities (adjusted difference, −41.6; 95% CI, −57.8 to −25.5). In an extrapolation of observed relative risks during the prevaccine period, the performance of 1 additional test per staff member per week across all facilities (i.e., approximately 1.1 million additional tests per week) was associated with 3079 fewer resident cases (30.5% reduction) and 427 fewer resident deaths (26.4% reduction) per week of heightened testing (Table S5).

e prevaccine period, the performance of 1 additional test per staff member per week across all facilities (i.e., approximately 1.1 million additional tests per week) was associated with 3079 fewer resident cases (30.5% reduction) and 427 fewer resident deaths (26.4% reduction) per week of heightened testing (Table S5). After the availability of vaccines, the outcomes between high- and low-testing facilities were similar during the period before the omicron wave. During the omicron wave, high-testing facilities had 1038.2 resident cases per 100 potential outbreaks as compared with 1092.9 cases in low-testing facilities (adjusted difference, −54.7; 95% CI, −102.2 to −7.2). No appreciable between-group difference in Covid-19–related deaths among residents occurred during this period. After adjustment for surveillance-testing volume, facilities that used POC tests had 537.6 resident Covid-19 cases per 100 potential outbreaks as compared with 509.4 residents in facilities using non-POC tests (adjusted difference, 28.2; 95% CI, 9.4 to 47.1) (Table 2, Table S6, and Fig. S5). The numbers of Covid-19–related deaths among residents were similar regardless of the type of testing that was performed during the full study window, as were the numbers of cases or deaths during the prevaccine phase (Fig. S6).

OC tests (adjusted difference, 28.2; 95% CI, 9.4 to 47.1) (Table 2, Table S6, and Fig. S5). The numbers of Covid-19–related deaths among residents were similar regardless of the type of testing that was performed during the full study window, as were the numbers of cases or deaths during the prevaccine phase (Fig. S6). Among facilities that predominantly used non-POC tests before the start of a potential outbreak, faster test turnaround time was associated with fewer resident deaths. Facilities that had an average turnaround time of 2 days or less reported 41.6 resident deaths per 100 potential outbreaks, as compared with 59.1 resident deaths in facilities that received results in 3 days or more (adjusted difference, −17.5; 95% CI, −25.2 to −9.8) (Figure 3 and Table S7). During the prevaccine phase, these differences were more marked; facilities with turnaround times of 2 days or less had 112.0 fewer cases (95% CI, −221.0 to −3.0) and 44.1 fewer deaths (95% CI, −68.6 to −19.7) than facilities with turnaround times of 3 days or more (Fig. S7). We performed multiple sensitivity analyses to confirm the main study findings. The results of these analyses were similar to the main results (Fig. S8 and Tables S8 through S11).

During a potential outbreak, high-testing facilities had fewer adjusted Covid-19 cases and deaths among residents than low-testing facilities (Figure 2, Table 2, and Table S4). During the full study period, high-testing facilities had 519.7 cumulative resident Covid-19 cases per 100 potential outbreaks, as compared with 591.2 cases per 100 potential outbreaks in low-testing facilities, an adjusted difference of −71.5 cases (95% confidence interval [CI], −91.3 to −51.6). High-testing facilities had 42.7 cumulative resident Covid-19 deaths per 100 potential outbreaks, as compared with 49.8 deaths in low-testing facilities (adjusted difference, −7.1; 95% CI, −11.0 to −3.2). High-testing facilities had approximately 15% more Covid-19 cases among staff than low-testing facilities, with no appreciable differences in the number of deaths (Fig. S4).

e prevaccine period, the performance of 1 additional test per staff member per week across all facilities (i.e., approximately 1.1 million additional tests per week) was associated with 3079 fewer resident cases (30.5% reduction) and 427 fewer resident deaths (26.4% reduction) per week of heightened testing (Table S5). After the availability of vaccines, the outcomes between high- and low-testing facilities were similar during the period before the omicron wave. During the omicron wave, high-testing facilities had 1038.2 resident cases per 100 potential outbreaks as compared with 1092.9 cases in low-testing facilities (adjusted difference, −54.7; 95% CI, −102.2 to −7.2). No appreciable between-group difference in Covid-19–related deaths among residents occurred during this period.

After adjustment for surveillance-testing volume, facilities that used POC tests had 537.6 resident Covid-19 cases per 100 potential outbreaks as compared with 509.4 residents in facilities using non-POC tests (adjusted difference, 28.2; 95% CI, 9.4 to 47.1) (Table 2, Table S6, and Fig. S5). The numbers of Covid-19–related deaths among residents were similar regardless of the type of testing that was performed during the full study window, as were the numbers of cases or deaths during the prevaccine phase (Fig. S6). Among facilities that predominantly used non-POC tests before the start of a potential outbreak, faster test turnaround time was associated with fewer resident deaths. Facilities that had an average turnaround time of 2 days or less reported 41.6 resident deaths per 100 potential outbreaks, as compared with 59.1 resident deaths in facilities that received results in 3 days or more (adjusted difference, −17.5; 95% CI, −25.2 to −9.8) (Figure 3 and Table S7). During the prevaccine phase, these differences were more marked; facilities with turnaround times of 2 days or less had 112.0 fewer cases (95% CI, −221.0 to −3.0) and 44.1 fewer deaths (95% CI, −68.6 to −19.7) than facilities with turnaround times of 3 days or more (Fig. S7). We performed multiple sensitivity analyses to confirm the main study findings. The results of these analyses were similar to the main results (Fig. S8 and Tables S8 through S11).

We found that greater surveillance testing of staff members in skilled nursing facilities was associated with clinically meaningful reductions in Covid-19 cases and deaths among residents, particularly before the availability of highly effective vaccines. Greater surveillance testing was also associated with more Covid-19 cases among staff during potential outbreaks, findings that were consistent with the protection of residents through the increased detection of Covid-19 among staff members. In addition, such detection may have occurred earlier in the disease course, thereby disrupting potential viral transmission chains. Little difference was detected between facilities that predominantly used POC testing and those that used non-POC testing, but faster turnaround times for testing results were associated with reductions in resident deaths among facilities that were highly reliant on non-POC tests that required laboratory processing. Previous simulation-based studies have suggested important features of an effective surveillance testing system, including testing at least twice weekly in high-risk settings3,4,7 and immediately providing test results.3,7 Such studies have also suggested limited benefit from using tests with greater sensitivity for detecting SARS-CoV-2.7,10 The findings of our current real-world study largely support the conclusions of the modeling studies.

sting at least twice weekly in high-risk settings3,4,7 and immediately providing test results.3,7 Such studies have also suggested limited benefit from using tests with greater sensitivity for detecting SARS-CoV-2.7,10 The findings of our current real-world study largely support the conclusions of the modeling studies. During the prevaccine phase of the pandemic, we found that the performance of one additional test per staff member per week was associated with a 30% reduction in resident Covid-19 cases and a 26% reduction in related resident deaths. In addition, a shorter test turnaround time (≤2 days as compared with ≥3 days) was associated with fewer cases and deaths in facilities that predominantly performed non-POC tests. During the full study period, the use of non-POC tests was associated with only modest reductions in the number of cases and no reduction in deaths. Because POC tests are less expensive than non-POC tests ($5 vs. $100 per test),22,23 frequent testing may be more financially feasible with POC testing. In addition, the use of POC tests avoids delays in laboratory turnaround time.

ssociated with only modest reductions in the number of cases and no reduction in deaths. Because POC tests are less expensive than non-POC tests ($5 vs. $100 per test),22,23 frequent testing may be more financially feasible with POC testing. In addition, the use of POC tests avoids delays in laboratory turnaround time. Surveillance testing was not strongly associated with differences in the numbers of cases or deaths during the preomicron phase, findings that are consistent with data showing that the available Covid-19 vaccines were highly effective at preventing both SARS-CoV-2 infections and severe disease.24-26 Similarly, the finding that more frequent surveillance testing was associated with fewer resident cases but no difference in resident Covid-19 deaths during the omicron wave is consistent with documented reductions in vaccine efficacy for preventing Covid-19 infections but maintenance in efficacy for preventing hospitalizations and deaths.27,28

t more frequent surveillance testing was associated with fewer resident cases but no difference in resident Covid-19 deaths during the omicron wave is consistent with documented reductions in vaccine efficacy for preventing Covid-19 infections but maintenance in efficacy for preventing hospitalizations and deaths.27,28 Our study has several limitations. Our observational study design precludes a conclusion of causality. Although our analyses were adjusted for several facility- and county-level covariates, we cannot rule out the possibility of unmeasured confounding as a driver of differences among facilities with higher as compared with lower relative testing rates. However, the finding that greater testing volume was associated with fewer resident cases but more staff cases suggests that our measure of testing rate is not simply a proxy for facility quality or underlying risk of Covid-19 outbreaks, both of which should have affected staff and resident cases in the same direction. We could not account for the potential contribution of outside visitors to infection rates. However, during the pandemic, visitors were largely banned from skilled nursing facilities before 2021 for infection control, so this factor would be unlikely to bias our prevaccine estimates. In addition, we lack data on resident and staff vaccination rates before June 2021, when these data were first reported in the NHSN system, as well as data regarding booster vaccinations.

rom skilled nursing facilities before 2021 for infection control, so this factor would be unlikely to bias our prevaccine estimates. In addition, we lack data on resident and staff vaccination rates before June 2021, when these data were first reported in the NHSN system, as well as data regarding booster vaccinations. We found that higher rates of surveillance testing of staff in U.S. skilled nursing facilities were associated with fewer resident Covid-19 cases and deaths regardless of testing method (POC or non-POC). These effects were most pronounced during the prevaccine time period and among facilities with faster turnaround times for non-POC testing.