Effects of Return-to-Office, Public Schools Reopening, and Vaccination Mandates on COVID-19 Cases Among Municipal Employee Residents of New York City

A return-to-office mandate did not result in a relative increase in COVID-19 cases among New York City municipal employees. Reopening public schools resulted in a relative increase in cases among Department of Education employees. A vaccination mandate resulted in a relative decrease in cases and hospitalizations among municipal employees.

T o encourage social distancing while ensuring continuity of operations during the COVID-19 outbreak, the New York City (NYC) Department of Citywide Administrative Services (DCAS) issued a temporary telework policy on March 13, 2020, for eligible municipal employees of the City of New York (hereafter referred to as City employees). 1 After COVID-19 vaccination availability, the mayor directed all City employees to return to the office at least part-time by May 3, 2021 2 and full-time by September 13, 2021, 3 with protective measures in place including mandatory face coverings 4 and proof of full vaccination or weekly negative polymerase chain reaction (PCR) diagnostic tests. 5 As justification for withdrawing the telework policy, a mayor's spokesperson stated, "We know how to make workplaces safe, and public servants can deliver more for New Yorkers when they're working together." 3 During the 2020-2021 academic year, a remote learning option was available to families with children in public schools. 6 When the 2021-2022 academic year commenced on September 13, 2021, public schools returned to full-time, in-person instruction. 7 Staff and students were required to wear face coverings while on school property. 8 As with all City employees, unvaccinated teachers and staff were required to test weekly. 9 By order of the Commissioner of Health and Mental Hygiene, COVID-19 vaccination was required for Department of Education staff effective October 1, 2021, 10 and more generally for all City employees effective October 29, 2021. 11,12 This order cited a federal executive order stating, "It is essential that Federal employees take all available steps to protect themselves and avoid spreading COVID-19 to their co-workers and members of the public. The CDC has found that the best way to do so is to be vaccinated." 13 COVID-19 outbreaks have been documented across various worksite settings. [14][15][16] In California, where employer reporting of workplace COVID-19 outbreaks was mandated, the public administration sector (including correctional, police, and fire services) had the highest incidence of reported workplace outbreaks through August 2021. 17 Given the importance of evaluating the effectiveness of public health interventions for COVID-19, [18][19][20] we assessed whether the mandates achieved their objectives with respect to minimizing workplace transmission among NYC employees. NYC has the largest municipal workforce in the United States, 21 and the City of New York is the largest single employer in the New York metropolitan area. 22 Specifically, we used a quasi-experimental study design to compare COVID-19 case rates among City employees relative to other NYC residents of working age. We assessed whether the September 13, 2021 return-to-office mandate and reopening of public schools was not associated with a relative increase in cases and whether the October 29, 2021 vaccination mandate was associated with a relative decrease in cases.

Study Population
The study population was NYC residents of working age, 18 to 64 years old. We defined City employees as persons appearing on lists of municipal employees as of both July 8, 2021 and November 29, 2021, provided by DCAS under a data use agreement with the NYC Department of Health and Mental Hygiene (DOHMH). The DCAS lists were stored on a secure server with access limited to authorized DOHMH study personnel. For the denominator for City employee case rates (N = 212,953), we restricted to persons who, as of both dates, worked at the same City agency, were 18 to 64 years old, and did not live outside of NYC. Because NYC residency status could not be determined for 33,854 uniformed Police Department employees with a worksite instead of residential address on the DCAS lists, we assumed that 48% were NYC residents. 23 Non-City employees were defined as all other NYC residents 18 to 64 years old (N = 5,001,460), approximated as the 2020 intercensal population estimate 24 minus the number of 18-to 64-year-old NYC residents appearing on either DCAS list except the 52% of uniformed Police Department employees assumed to be non-NYC residents. Persons appearing on only one of the two DCAS lists were excluded, as the purpose was to assess the effects of mandates on COVID-19 case trends in closed cohorts.
The primary outcome of interest was diagnosis with a confirmed or probable case of COVID-19, per the national surveillance case definition. 25 Cases were ascertained primarily through electronic laboratory reporting through the New York State Electronic Clinical Laboratory Reporting System. 26 Patients whose address at the time of report indicated residence in a nursing home, adult care facility, jail, or prison were excluded. Symptom status was ascertained by routine interview, for example, for contact tracing, and we classified patients as symptomatic if they met the clinical criteria for COVID-19-like illness, per the surveillance case definition. 25 Case data were extracted from the DOHMH COVID-19 surveillance database (Maven® Disease Surveillance and Outbreak Management System; Conduent, Florham Park, NJ) on February 15, 2022.
We primarily assessed trends in cases, as opposed to hospitalizations or deaths, because the stated rationale for the mandates, as mentioned previously, emphasized making workplaces safe and avoiding spreading COVID-19 to co-workers; in addition, the study population excluded those 65 years or older, the group at highest risk for severe illness. However, in a secondary analysis to assess the first 3 months after the municipal employee vaccination mandate was implemented, we considered COVID-19 hospitalizations as an additional outcome of interest. These were ascertained by importing and matching data from supplemental systems, as previously described 27 and defined as patients having a positive severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) test result within 14 days before or 3 days after hospital admission. Because COVID-19 hospitalization risk increases with age, 28 we compared the age distribution of City and non-City employees as of November 27, 2021, and assessed hospitalization trends in the overall 18-to 64-year-old study population as well as restricting to 50-to 64-year-olds.

Matching Employee and Case Data
Employee lists and cases were geocoded with version 22A of the NYC Department of City Planning's Geosupport geocoding software, 29 implemented in R through C++ using the Rcpp package. 30 Addresses that failed to geocode were then cleaned using a string searching algorithm performed against the Department of City Planning's Street Name Dictionary and Property Address Directory. 29 Addresses that still failed to geocode after cleaning were then run through a US Postal Service verification service 31 for further cleaning and to flag addresses outside of NYC.
We used three sets of geocoder outputs. First, the building classification code 32 was used to flag nonresidential addresses. Employees of certain City agencies (Police Department, Department of Corrections, and Department of Investigations) commonly self-reported to DCAS their work instead of home address because of security concerns. We matched cases diagnosed among such employees with laboratory reports of SARS-CoV-2 molecular and antigen tests, regardless of test date and result. This match used fuzzy matching on first and last name, exact matching on birth date, and the ordering facility name as available to confirm City employee status because Police Department employees frequently accessed testing through their employer. Cases diagnosed among employees with a valid NYC residential address on any laboratory report were considered NYC residents and study-eligible; otherwise, cases were excluded from the analysis because employee residency in NYC could not be assumed.
Second, geocoder outputs of the tax lot identification number (borough-block-lot) and building identification number were used in record matching, as described hereinafter. We prioritized these standardized geocoder outputs for matching over address fields to minimize problems with address data entry and missingness. Third, the US Census Bureau block was converted using Topologically Integrated Geographic Encoding and Referencing data 33 to 2010 ZIP Code Tabulation Areas, which were in turn used to determine United Hospital Fund (UHF) neighborhood (N = 42), a geography that aggregates adjoining ZIP code areas of similar characteristics to approximate community districts. 34,35 UHF neighborhoods were used to assess geographic representativeness of NYC employees compared with the general population, as described hereinafter.
To assign City employment status to each COVID-19 case, we used a multistep, deterministic, one-to-many hierarchical record matching algorithm. We linked records using a series of "keys" consisting of exact character correspondence between variables. 36 We constructed 17 matching keys in the DCAS and surveillance case linelists using combinations of first name, middle name, last name, date of birth, borough-block-lot, and building identification number; see Supplemental Digital Content 1, http://links.lww.com/JOM/ B247, which shows the steps for data set cleaning, standardization, and reformatting, as well as the matching keys.

Geographic Representativeness of NYC Employees
To assess support for the assumption that City employees and non-City employees residing in NYC were exposed to the same underlying epidemic trends, we compared the geographic distribution of City employees with that of the general population aged 18 to 64 years by UHF neighborhood of residence. This analysis included employees who had the same residential UHF neighborhood in DCAS lists from both July 8 and November 29, 2021, as well as employees whose residential address was not available in either DCAS list but was available through matching to case or laboratory testing data.

Difference-in-Difference Analysis
Difference-in-difference analyses are used to compare group means of an outcome before and after policy implementation. The counterfactual is based on a similar comparator group that was not subject to the policy intervention. 37 Described as a generalized linear model, the statistical components are: where Y is the outcome of interest, that is, COVID-19 cases per 100,000 person-days, accounting for preimplementation and postimplementation periods of unequal durations. Int is a dummy variable that is 1 if the observation occurred in the group that experienced the intervention (City employees) or 0 otherwise (non-City employees).
Post is a dummy variable that is 1 if the observation occurred temporally after the policy implementation and 0 otherwise. PostInt is a product of Int and Post and is a dummy variable that is 1 if the observation occurred temporally after the policy implementation in the intervention group or 0 otherwise.
The four coefficients can be interpreted as follows: β 0 is the mean outcome in the nonintervention group (non-City employees) in the preintervention period (premandate). β 1 is the difference in mean outcome in the intervention group (City employees) in the preintervention period compared with the mean outcome in the nonintervention group in the preintervention period. β 2 is the difference in the mean outcome in the nonintervention group in the postintervention period (postmandate) compared with the mean outcome in the nonintervention group in the preintervention period. β 3 it the measure of interest, or the differencein-difference, that is, the additional difference in the mean outcome in the intervention group in the postintervention period compared with the mean outcome in the intervention group in the preintervention period when compared with the difference defined for β 2 .
We used a Poisson link in the generalized linear model, with a log-transformed offset term for person-time at risk. Therefore, the coefficients of interest are interpreted as differences in the log incidence rates in the contrasts of interest described previously. We exponentiated these estimates and their respective confidence intervals (CIs) to calculate the incidence rate ratios of the contrasts described previously. For example, exponentiating β 2 yielded the ratio of the COVID-19 incidence rate in non-City employees after the mandate compared with before the mandate. To estimate the incidence rate in City employees after the mandate compared with before the mandate, we summed the β 2 and β 3 estimates before exponentiating them and extracted the variance-covariance matrices of these covariates to calculate the standard error and subsequent 95% CI.

Return-to-Office Mandate and Reopening of Public Schools (September 13, 2021)
To estimate the effect of the return-to-office mandate on COVID-19 transmission, we compared changes in COVID-19 case rates from premandate to postmandate implementation periods for office-based City employees relative to other working-age adults. Office-based City employees (N = 80,454) were defined by excluding City employees who were unlikely to have worked remotely during the preimplementation period, that is, uniformed employees of the Department of Corrections, Department of Sanitation, Fire Department, and Police Department; Department of Education employees; and employees with a civil service title description of city seasonal aide. To estimate the effect of reopening public schools, we restricted to Department of Education employees (N = 97,879).
We defined the preimplementation period as diagnoses from July 5 to September 12, 2021, that is, the 10 weeks before the mandate and coinciding with the increasing slope of NYC's third epidemic wave. 38 If the return-to-office mandate or reopening of public schools were associated with increased disease transmission, then it would take time for employees to become infected and develop symptoms. Assuming a median incubation period of 5 days, 39 we imposed a washout period (September 13-22, 2021) of two incubation periods for worksite transmission to occur. Laboratory-based testing availability in NYC during this period was widespread, so we did not build in extra time from symptom onset to testing. Before students returned on September 13, 2021, Department of Education staff began reporting to schools on August 30 or September 9, 2021, depending on their job title. 40 The postimplementation period was defined as September 23-October 28, 2021, ending before the vaccination mandate and coinciding with a citywide decline in cases. 38 The Delta variant predominated throughout these preimplementation and postimplementation periods. 41 We considered the interventions to be effective with respect to disease transmission if the change in COVID-19 case rates using a noninferiority test 42 from the preimplementation to postimplementation periods was not larger for office-based City employees or Department of Education employees than the comparison group by a prespecified margin. This margin is a policy decision as to what COVID-19 case rate is considered acceptable in the workplace as a trade-off for the benefits of in-person work. The policies did not define a "safe" level of transmission. At the time, the CDC defined a low community transmission level as <10 new cases per 100,000 persons in the past 7 days. 43 Thus, we a priori defined an acceptable margin consistent with "low" worksite transmission as <10 excess average weekly cases per 100,000 employees during the postimplementation period.

Vaccination Mandate (October 29, 2021)
The vaccination mandate was layered on top of the return-tooffice mandate. We defined the preimplementation period for the vaccination mandate to be the same as the postimplementation period for the return-to-office mandate (September 23-October 28, 2021). In a sensitivity analysis because this preimplementation period was brief, and we found as below that the return-to-office mandate did not affect case trends, we extended the preimplementation period for the vaccination mandate to July 5-October 28, 2021. On October 29, 2021, the weekly testing option was rescinded, and City employees were required to have received at least one COVID-19 vaccine dose. 11,44 The postimplementation period for the vaccination mandate was defined as October 29-November 30, 2021, ending before the identification of the first confirmed US case of infection with the Omicron SARS-CoV-2 variant on December 1, 2021. 45 Because there was little time for the vaccination mandate to reduce worksite transmission before Omicron emergence, we did not further shorten the postimplementation period by imposing a washout period.
Cases increased rapidly in NYC throughout December, 38 with Omicron constituting 75% of sequencing results by week ending December 18, 2021. 41 Given the reduced vaccine effectiveness against infection with the Omicron variant, 46 we did not necessarily expect the vaccination mandate (which included no requirement for a booster dose) to strongly reduce disease transmission after Omicron emergence. In addition, COVID-19 vaccination was required for the general workforce in NYC effective December 27, 2021, 47 after which effects of the vaccination mandate on COVID-19 cases among City employees could be biased toward the null. Nevertheless, in a secondary analysis, we defined an extended postimplementation period as October 29, 2021-January 31, 2022, to assess COVID-19 cases and hospitalizations diagnosed during the first 3 months after the municipal employee vaccination mandate was implemented, including during the large Omicron (BA.1) epidemic wave.
We compared changes in COVID-19 case rates over time between City employees and other working-age adults. We considered the vaccination mandate to be effective with respect to disease transmission if the change in COVID-19 case rates from the preimplementation to postimplementation periods was statistically significantly lower in City employees than in the comparison group, controlling for differences between groups in the preimplementation period.
Compliance with the vaccination mandate at the time of implementation varied across agencies. The reported percentage of an agency's workforce (including City employees who were not NYC residents) with at least one COVID-19 vaccine dose as of October 30, 2021, ranged from 60% at the Department of Corrections to 100% at the Landmarks Preservation Commission. 48 We grouped employees of agencies with higher versus lower vaccination coverage, defined as ≥96% versus <90% with at least one vaccine dose as of October 30, 2021. 48 Employees of agencies with intermediate vaccination coverage (90%-95%) were excluded from this subanalysis. If the vaccination mandate had been effective in reducing disease transmission, then in a difference-in-difference-in-difference (triple difference) analysis, 37 employees of agencies with higher vaccination coverage might be expected to have experienced greater reductions in COVID-19 case rates from the preimplementation to postimplementation periods relative to other working-age adults than employees of agencies with lower vaccination coverage.
The statistical components of the triple-difference analysis are: where High is a dummy variable that is 1 if the observation occurred in employees of agencies with higher vaccination coverage or 0 otherwise (non-City employees or employees of agencies with lower vaccination coverage). PostHighInt is a product of Post and High and Int and is a dummy variable that is 1 if the observation occurred temporally after the policy implementation in employees of agencies with higher vaccination coverage or 0 otherwise. β 7 is the measure of interest, or the triple difference, that is, the contrast in the difference-in-difference estimates of two comparisons: (1) the difference-in-difference of the mean outcome in employees of agencies with higher vaccination coverage and non-City employees when comparing the postvaccination mandate period with prevaccination mandate period, and (2) the difference-in-difference of the mean outcome in employees of agencies with lower vaccination coverage and non-City employees when comparing the postvaccination mandate period with prevaccination mandate period.

RESULTS
Of 921,057 confirmed and probable COVID-19 cases diagnosed among community-dwelling NYC residents 18 to 64 years old during July 5, 2021-January 31, 2022, 45,291 (4.9%) were among eligible City employees and 863,459 (93.7%) were among non-City employees. The remaining 12,307 (1.3%) were excluded as ineligible (diagnosed in a person present on only one of the two DCAS lists; present on both DCAS lists but employed at different agencies; having a non-NYC address on one of the DCAS lists; or Police Department, Department of Corrections, or Department of Investigations employees using their work address and for whom NYC residency could not be assumed).
Within municipal employment categories, the percentage of case-patients known to have had COVID-19-like illness was similar across diagnosis periods ( Table 1), suggesting that any changes in testing rates over time did not lead to large differences in the proportion of infections ascertained as cases. The percentage of case-patients known to have COVID-19-like illness was similar between City employees (87.1%) and non-City employees (84.9%).
Of NYC residents of working age, 33.5% of City employees were 50 to 64 years old, a slightly higher proportion than the 28.3% of non-City employees (see Figure, Supplemental Digital Content 2, http://links.lww.com/JOM/B247, which illustrates that among NYC residents 18-64 years old, City employees generally skewed older than non-City employees).

Geographic Representativeness of NYC Employees
The residential geographic distribution of City employees was moderately well correlated with that of the general population (see Figure, Supplemental Digital Content 3, http://links.lww.com/JOM/ B247, which illustrates that UHF neighborhoods where larger percentages of NYC employees resided generally also had larger percentages

Return-to-Office Mandate and Reopening of Public Schools
The Delta epidemic wave was waning in NYC when the returnto-office mandate was implemented and public schools reopened on September 13, 2021 38 (Fig. 1). Accordingly, the case rate per 100,000 person-days decreased during the postperiod compared with the preperiod for all groups, that is, office-based City employees, Department of Education employees, and non-City employees ( Table 2).

Office-Based City Employees
Among non-City employees, we found a 32.1% (95% CI, 31.1%-33.0%) reduction in the COVID-19 case rate during the postperiod compared with the preperiod; that is, the incidence rate ratio was e β2 = e −0.3868 = 0.679. Among office-based City employees, we found a 29.0% (95% CI, 20.5%-36.6%) reduction in the case rate during the postperiod compared with the preperiod; that is, the incidence rate ratio was e β2 + β3 = e −0.3868 + 0.0445 = 0.710. The incidence rate ratio among office-based City employees comparing the postmandate to premandate periods was approximately the same (4.6% increase, ie, e β3 = e 0.0445 = 1.046; 95% CI, 6.8% decrease to 17.0% increase), as it would have been had the office-based City employees followed the same trend as non-City employees. Epidemic trends were similar between the two groups (Fig. 1).

Department of Education Employees
Among Department of Education employees, we found only a 12.8% (95% CI, 4.8%-20.2%) reduction in the case rate during the postperiod compared with the preperiod. The incidence rate ratio among Department of Education employees comparing the postmandate to premandate periods was 28.4% (95% CI, 17.3%-40.3%) larger than it would have been had Department of Education employees followed the same trend as non-City employees.
Among Department of Education employees, the proportion of COVID-19 cases diagnosed among employees with a civil service title of teacher increased from 44.7% (710 of 1590) during the preimplementation period to 57.5% (410 of 713) during the postimplementation period. The proportion of Department of Education case-patients who were known to be symptomatic with COVID-19like illness also increased from the preimplementation period (87.0%) to the postimplementation period (93.1%) ( Table 1). During the washout period to account for worksite transmission to occur, case FIGURE 1. Weekly COVID-19 case rates among New York City residents 18 to 64 years old, by municipal employment status, July 5 to October 28, 2021. The first and last weeks depicted are partial weeks: week ending July 10 reflects diagnoses during July 5 to 10; week ending October 30 reflects diagnoses during October 24 to 28. rates decreased among non-City employees and increased among Department of Education employees (Fig. 1), possibly reflecting transmission stemming from staff returning to schools before students.

Noninferiority Tests
During the postimplementation period, the excess average weekly cases per 100,000 office-based City employees was 4.4 (95% CI, 3.0-5.6), which was within the a priori threshold consistent with low worksite transmission of <10. In contrast, the excess average weekly cases per 100,000 Department of Education employees was 31.3 (95% CI, 29.7-32.8), exceeding the threshold consistent with low worksite transmission (see Table, Supplemental Digital Content 4, http://links.lww.com/JOM/B247, which shows the logic for the noninferiority tests).

Vaccination Mandate
The Delta wave had plateaued in NYC when the vaccination mandate was implemented on October 29, 2021 38 (Fig. 2). Accordingly, the case rate per 100,000 person-days was similar during the postimplementation period for the primary analysis compared with the preimplementation period among all groups, that is, all City employees, employees of agencies with higher and lower vaccination rates, and non-City employees (Table 3A). Dramatic increases in COVID-19 case rates (Table 3A)

Primary Analysis
Among non-City employees, there was a 19.7% (95% CI, 17.7%-21.7%) increase in the COVID-19 case rate during October 29-November 30, 2021 (after implementation of the vaccination mandate and before Omicron emergence) compared with the preimplementation period. Among City employees, there was little change in the case rate during the postperiod compared with the preperiod (4.3% decrease; 95% CI, 11.2% decrease to 3.1% increase). Among City employees, the incidence rate ratio comparing the postperiod with the preperiod was 20.1% (95% CI, 13.7%-26.0%) smaller than it would have been had City employees followed the same trend as non-City employees. See Supplemental Digital Content 7, http:// links.lww.com/JOM/B247, which shows findings from a sensitivity analysis using an extended prevaccination mandate implementation period of July 5-October 28, 2021.

Extended Postvaccination Mandate Implementation Period (October 29, 2021-January 31, 2022)
In a secondary analysis assessing the first 3 months of the vaccination mandate extending through the Omicron (BA.1) wave, non-City employees had an 11.2-fold (95% CI, 11.0-11.3) increase in the case rate, whereas City employees had a 10.2-fold (95% CI, 9.7-10.8) increase in the case rate during the postperiod compared with the preperiod. Among City employees, the incidence rate ratio comparing the extended postperiod with the preperiod was 8.5% (95% CI, 3.4%-13.2%) smaller than it would have been had City employees followed the same trend as non-City employees.
As for COVID-19 hospitalizations among the full study population of 18-to 64-year-olds, non-City employees had a 5.9-fold (95% CI, 5.5-6.3) increase in the hospitalization rate, whereas City employees had a 3.3-fold (95% CI, 2.4-4.7) increase in the hospitalization rate during the postperiod compared with the preperiod. Among City employees, the hospitalization incidence rate ratio comparing the extended postperiod with the preperiod was 42.9% (95% CI, 17.1%-59.4%) smaller than it would have been had City employees followed the same trend as non-City employees.
We repeated this analysis, restricting to 50-to 64-year-olds as a subpopulation with greater underlying hospitalization risk. Older non-City employees had a 5.5-fold (95% CI, 5.0-6.0) increase in the hospitalization rate, whereas older City employees had a 4.1-fold (95% CI, 2.3-7.5) increase in the hospitalization rate during the postperiod compared with the preperiod. Among older City employees, the hospitalization incidence rate ratio comparing the extended postperiod with the preperiod was not statistically significantly different (24.2% smaller [95% CI, 56.6% smaller to 45.5% larger]) than it would have been had older City employees followed the same trend as older non-City employees. This estimate should be interpreted with caution, as only 12 hospitalizations were observed among 50-to 64-year-old City employees during the 5-week premandate period (Table 3C), leading to wide uncertainty. In addition, weekly hospitalization rates during the brief premandate period seemed to be stable among older non-City employees but decreasing among older City employees (Supplemental Digital Content 6B, http://links.lww.com/JOM/B247), possibly violating the common trends assumption.

City Employees Stratified by Agency Vaccination Coverage
Another secondary analysis grouped employees of agencies with higher and with lower vaccination coverage and compared each group with non-City employees using a triple-difference analysis. During the primary postvaccination mandate period (October 29-November 30, 2021), City employees experienced decreased case rates compared with the preperiod: the decrease was 2.1% (95% CI, −8.5% to 11.7%) among employees of agencies with higher vaccination coverage and 11.5% (95% CI, −0.5% to 22.0%) among employees of agencies with lower vaccination coverage; as discussed previously, non-City employees experienced a 19.7% (95% CI, 17.7%-21.7%) increase. There was no statistically significant difference in case rates (10.5% increase; ie, the incidence rate ratio was e β7 = e 0.1003 = 1.105; 95% CI, 6.3% decrease to 30.4% increase) from the preperiod to postperiod when comparing employees of agencies with higher than lower vaccination coverage and controlling for the trend among non-City employees.
We repeated this analysis using the extended postvaccination mandate implementation period that included the Omicron (BA.1) wave (October 29, 2021-January 31, 2022). The increase in case rates was 9.6-fold (95% CI, 8.9-10.3) among employees of agencies with higher vaccination coverage and 11.6-fold (95% CI, 10.7-12.7) among employees of agencies with lower vaccination coverage; as discussed previously, non-City employees experienced an 11.2-fold (95% CI, 11.0-11.3) increase. Controlling for the trend among non-City employees, the increase in case rates for employees of agencies with higher vaccination coverage was 17.4% (95% CI, 7.5%-26.3%) lower than the increase in case rates for employees of agencies ‡Analyses to assess cases diagnosed during the first 3 months of the vaccination mandate (which included no requirement for a booster dose) were considered secondary because vaccine effectiveness against infection with the Omicron variant was reduced relative to prior variants, whereas vaccine effectiveness against severe illness remained high.
with lower vaccination coverage. The triple-difference analysis was not conducted for the COVID-19 hospitalizations outcome because <10 hospitalizations were observed among employees of agencies with higher vaccination coverage during the prevaccination mandate implementation period (Table 3B), leading to unstable rates.

DISCUSSION
The return-to-office mandate enacted on September 13, 2021, was not associated with a relative increase in COVID-19 cases diagnosed during September 23-October 28, 2021, among office-based municipal employees residing in NYC. The excess average weekly cases per 100,000 office-based City employees during the postimplementation period was low, at 4.4 (95% CI, 3.0-5.6). This finding supports that the safety measures in place during that period, including mandatory face coverings and proof of full vaccination or weekly negative PCR diagnostic tests, were effective in keeping worksite transmission low while the Delta variant predominated. This finding cannot be generalized to later periods because of the predominance of the more transmissible Omicron variant 49 and the revised DCAS directive (March 7, 2022) authorizing City employees to remove their face coverings in the workplace. 50 In contrast, the reopening of public schools for the 2021-2022 academic year was associated with an increase in COVID-19 cases among Department of Education employees, controlling for the trend among non-City employees. The excess average weekly cases per 100,000 Department of Education employees during September 23-October 28, 2021, was 31.3 (95% CI, 29.7-32.8), consistent with the CDC's definition at the time of "moderate" community transmission (ie, 10.00-49.99 new cases per 100,000 persons in the past 7 days). 43 This finding was consistent with contemporaneous local news reports of outbreaks affecting staff, related to congregating indoors with insufficient masking and social distancing. 51,52 In addition, unlike office-based City employees, Department of Education employees worked with children. Asymptomatic SARS-CoV-2 infections are more common among children, 53 so untested students could have unknowingly transmitted to staff. The finding of a relative short-term increase in COVID-19 cases among Department of Education employees cannot be generalized to later periods, as the CDC expanded COVID-19 vaccination eligibility to 5-to 11-year-olds on November 2, 2021 54 and to ≥6-month-olds to <5-year-olds on June 18, 2022, 55 and school ventilation systems were further upgraded. 56,57 COVID-19 vaccination mandates were previously demonstrated to increase vaccination coverage. 58,59 Our analysis further demonstrated that the vaccination mandate for NYC municipal employees was associated with a relative decrease in COVID-19 cases. Case rates increased from the premandate to primary postmandate implementation period through November 30, 2021 among non-City employees but not for City employees. In addition, despite reduced vaccine effectiveness against infection with the Omicron variant 46 and no mandate for a booster dose, increasing trends among City employees from the premandate to secondary postmandate implementation period through January 31, 2022 and extending through the Omicron (BA.1) wave were approximately 9% smaller for case rates and 43% smaller for hospitalization rates than they would have been had City employees followed the same trends as non-City employees. These findings support the effectiveness of the vaccination mandate in modestly reducing workplace transmission when Omicron (BA.1) predominated.

Limitations
Difference-in-difference analyses rely on the common trends assumption that important unmeasured variables are either time-invariant group attributes or time-varying factors that are group-invariant. 37 Specifically, we assumed that COVID-19 case ascertainment either (1) differed between City and non-City employees (due to, for example, differences in age distribution, prior SARS-CoV-2 infection history, or testing rates), but this difference was consistent over the study period, or (2) differed over the study period (due to, for example, epidemic trends 60 ), but this difference was consistent for City and non-City employees. Figure 1 provides empirical support that during the 10 weeks before the first mandate, weekly trends in COVID-19 case rates for non-City employees, office-based City employees, and Department of Education employees were similar, as each group experienced an increase in cases during the Delta epidemic wave. City employees were also geographically representative of the general population (Supplemental Digital Content 3, http://links.lww.com/JOM/B247) and thus likely exposed to the same underlying epidemic trends. Other policies implemented during the study period likely affected City and non-City employees consistently, including on September 13, 2021, the start of enforcement of the Key to NYC vaccination mandate applying to patrons of gyms, restaurants, and indoor entertainment venues. 61 Similarly, we would not expect that use of at-home rapid antigen tests, which became widely available in NYC starting mid-December 2021 62 and were not reportable to DOHMH, would substantially differ by City employment status.
Differential testing requirements for City and non-City employees were the main threat to study validity. A weekly PCR diagnostic testing requirement for unvaccinated City employees 5 (including unvaccinated Department of Education employees 9 ) was in effect during the postimplementation period for the return-to-office mandate and reopening of public schools. Similarly, the federal Occupational Safety and Health Administration issued an emergency temporary standard effective November 5, 2021 to January 26, 2022, for weekly testing of employees who were not fully vaccinated, applying to employers with at least 100 employees. 63 Mandatory routine testing for subsets of City and non-City employees at different times might have led to differential case ascertainment and biased our findings. Reason for testing (workplace requirement, pretravel clearance, etc) was incomplete for reported cases. However, the proportion of case-patients by municipal employment status who were known to be symptomatic with COVID-19-like illness did not vary markedly by timing of diagnosis ( Table 1), suggesting that the magnitude of this potential bias is likely small. Furthermore, among Department of Education employees, the proportion of case-patients who were known to be symptomatic with COVID-19-like illness increased after reopening of public schools, indicating that the relative increase in case rates cannot be explained by increased ascertainment of asymptomatic infections. We are unaware of other reasons that differences in COVID-19 case trends between City and non-City employees before and after the mandates could not be causally attributable 64 to the mandates.
There were at least three additional limitations. First, as with any analysis of matched data sets, overmatching and undermatching are possible, 36 resulting in misclassification of City employment status for some case-patients and likely biasing findings toward the null. Undermatching was more likely for employees of agencies for whom home address was frequently unavailable because of security concerns, which included two of the agencies (Police Department, Department of Corrections) that also had lower vaccination coverage. However, 2 of the 12 primary matching keys and all 5 secondary matching keys used identifiers other than address information (Supplemental Digital Content 1, http://links.lww.com/JOM/B247).
Second, the interventions were gradually and partially adopted. Before the full-time return-to-office mandate on September 13, 2021, some office-based City employees had already returned to the office on at least a part-time basis. Under the City of New York's Equal Employment Opportunity Policy, eligible employees were entitled to reasonable accommodations to continue to telework full-time (eg, due to a disability), to be exempted from the vaccination requirement due to medical or religious reasons, or both. 65 Furthermore, as of October 28, 2021, the day before the vaccination mandate, 74.3% of NYC residents had already received at least one COVID-19 vaccine dose. 66 After the vaccination mandate, thousands of City employees who were noncompliant were placed on unpaid administrative leave, 67 so would not have reported to office settings or contributed to worksite transmission. Publicization of the municipal employee mandates might have influenced telework reductions and vaccination uptake for employees in other sectors. These phenomena, as well as the lack of a washout period to allow for the vaccination mandate to reduce worksite transmission, would reduce differences between the intervention and comparator groups and thus likely bias findings toward the null.
Third, we assessed only the direct effects of municipal employee mandates on workplace transmission among NYC residents of working age. The mandates also might have indirectly affected transmission among the group at highest risk for severe COVID-19 illness, ≥65year-olds, for example, through employees residing in multigenerational housing. 68,69 Indirect effects, such as increases or decreases in transmission from City employees to their household contacts relative to non-City employees to their household contacts, were not assessed.

CONCLUSIONS
We found short-term evidence that the return-to-office and vaccination mandates for NYC municipal employees were successful with respect to minimizing workplace transmission among City employees and that reopening public schools was associated with a relative increase in COVID-19 cases among Department of Education employees. These findings should be interpreted in broader context with other health, social, educational, and economic effects of the mandates, including municipal workforce attrition. 21 As of July 2022, 33% of employed US adults (53% of education workers, 33% of white-collar/office-based workers, and 24% of blue-collar workers) remained moderately or very concerned about being exposed to COVID-19 at work. 70 The ability to assess longer-term effects of the NYC mandates on COVID-19 case trends is limited in part by wide adoption of at-home rapid antigen tests, which are not reportable to the DOHMH. 71 Guidance to prevent and reduce transmission in workplaces is available from public health authorities, including the DOHMH 72 and CDC. 73