The HEROES-RECOVER studies were prospective cohorts of frontline workers participating in active surveillance for COVID-19 and influenza. The protocol and methods have been described previously [6, 7]. HEROES-RECOVER has documented the natural history of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and COVID-19 before vaccines became available, as well as VE of COVID-19 vaccines against infection and illness through active surveillance and systematic collection of respiratory and serum specimens [8–10]. Molecular testing for influenza began in November 2021; we used data from August 2022–May 2023 to assess influenza VE during the 2022–2023 influenza season.

Participants included primary healthcare personnel (physicians, physician assistants, nurse practitioners, dentists), allied healthcare personnel (eg, nurses, therapists, technicians, medical assistants, and orderlies), first responders (eg, law enforcement, firefighters, and other emergency services), and other frontline workers (eg, education, food production, and service occupations) [9]. During the 2022–2023 influenza season, participants were enrolled in 5 states: Arizona, Florida, Minnesota, Texas, and Utah. Participants from HEROES-RECOVER were eligible for inclusion in the current study if they were continuously enrolled during the 2022–2023 influenza season (September/October 2022–31 May 2023), participated in active surveillance, and provided complete data on vaccination status and date of vaccination. All participants provided informed consent, and the HEROES-RECOVER protocol was reviewed and approved by the institutional review boards at participating sites or under a reliance agreement.

Sociodemographic and health characteristics, including underlying medical conditions, were reported by participants in surveys completed at enrollment and updated quarterly. Active monitoring for illness symptoms was conducted through weekly surveillance questionnaires administered by text message and email, with a checklist where participants indicated whether they were experiencing any of the following: cough (wet or dry), nasal congestion, fatigue, febrile symptoms (fever or chills or both), headache, aches, sore throat, joint pain, shortness of breath, nausea, diarrhea, vomiting, abdominal pain, change in taste or smell, or rash. Self-reported mask use at work was collected monthly and averaged over the study period. Participants self-collected weekly mid-turbinate nasal swabs regardless of symptom status and provided additional nasal and saliva samples during the weeks in which they were experiencing any symptoms. Specimens were tested by reverse transcription–polymerase chain reaction (RT-PCR) for influenza virus and SARS-CoV-2; a subset of influenza A viruses were sub-typed. Influenza vaccination and date were ascertained by self-report through electronic surveys. Sites in Arizona, Minnesota, Texas, and Utah also had access to electronic medical records and state registries to verify vaccination dates; this date was used if it differed from self-reported vaccination history.

Each participant’s follow-up start date was the later of their enrollment date or the beginning of influenza virus circulation at their site (ie, the first week with an influenza-positive test among participants at that site). Participants who were vaccinated at least 14 days before their start date began contributing vaccinated person-time. Participants vaccinated 0–13 days before their start date began contributing vaccinated person-time on day 14 post-vaccination (ie, they were censored from their start date until day 14 post-vaccination). Participants who were unvaccinated on their start date began contributing unvaccinated person-time; if they received an influenza vaccination midway through follow-up, they were censored for days 0–13 post-vaccination and then contributed vaccinated person-time beginning 14 days post-vaccination. Follow-up ended at the earliest of the following: date of influenza virus infection, study withdrawal date, or the end of the influenza season (31 May 2023). Influenza virus infections occurring within 0–13 days post-vaccination were not included in VE analyses, as these participants were censored during that time.

The date of influenza virus infection was defined as the sample collection date of a specimen that tested positive by RT-PCR for influenza. Consecutive influenza detections within a 30-day period were considered a single infection. There were 3 infection outcomes of interest: (1) laboratory-confirmed influenza virus infection defined as a positive RT-PCR with or without associated symptoms, (2) asymptomatic infection, and (3) influenza illness, defined as a positive test with any associated symptom. We also explored another definition of symptomatic influenza illness: positive influenza PCR test plus febrile symptoms. For participants reporting symptoms through active surveillance, additional illness surveys were sent every 7 days until illness resolution was reported; asymptomatic participants did not receive illness surveys but confirmed absence of symptoms at the time RT-PCR test results were communicated. Clinical outcomes such as medical care-seeking were also self-reported via the illness survey.

First, we compared participant vaccination status by study site, race and ethnicity, gender, age group, occupation, pre-existing conditions, smoking status, and mask-wearing practices at work. We calculated rates of all laboratory-confirmed influenza virus infection, asymptomatic influenza virus infection, and illness per 100 000 vaccinated versus unvaccinated person-days of follow-up. We summarized the symptomatic fraction of influenza virus infections, frequency of specific symptoms among participants with influenza illness, and medical care-seeking.

Next, we estimated adjusted hazard ratios (aHRs) comparing rates of laboratory-confirmed influenza, asymptomatic infection, and influenza illness among vaccinated versus unvaccinated participants. For the outcome of laboratory-confirmed influenza, aHRs were also stratified by influenza A subtype [A(H1N1)pdm09 and A(H3N2)] (described in the Supplementary Methods). We used the Andersen-Gill extension of the Cox proportional hazards regression model [11], controlling for participant occupational category (primary healthcare personnel, allied healthcare personnel, first responder, or frontline worker) and site. Asymptomatic infections were included as a competing hazard in the model of VE against influenza illness [12]. We also assessed whether aHRs differed between asymptomatic and symptomatic cases using a similar model including an interaction term between the binary vaccination status measure and a binary measure for presence of symptoms. A statistically significant aHR for this interaction term would be interpreted as rejecting the null hypothesis that aHRs were equivalent across symptomatic and asymptomatic cases. Vaccine effectiveness was calculated as (1 − aHR) × 100%.

We conducted 3 sensitivity analyses for potential bias in VE estimates. First, we calculated VE against laboratory-confirmed influenza virus infection excluding SARS-CoV-2 co-positive illness episodes from the analysis. Second, we calculated VE excluding participants who had less than 80% adherence to weekly swabbing. Third, we calculated influenza VE against non-influenza, SARS-CoV-2–negative illness episodes (ie, participants reporting symptoms without an influenza or SARS-CoV-2–positive test result) to assess potential under-ascertainment of influenza in the cohort or residual bias in VE estimates.

Of 3990 HEROES-RECOVER participants who were enrolled during the 2022–2023 influenza season, we excluded 10 (0.3%) who were not continuously enrolled, 136 (3.4%) with missing influenza vaccination status in 2022–2023, and 52 (1.3%) who indicated they were vaccinated but did not report a date, resulting in 3785 eligible participants. These participants had high adherence to weekly surveillance; the median (interquartile range [IQR]) percentage of weeks during the study period when participants submitted respiratory specimens was 91% (74%–97%).

In all, 2465 (65%) participants received an influenza vaccine and 1320 (35%) remained unvaccinated for the entire 2022–2023 season. Influenza vaccination rates varied by gender (72% among females vs 54% among males), race and ethnicity (50% among Hispanic participants vs 69% among White, non-Hispanic and Asian, non-Hispanic participants), and occupation (36% among first responders vs 90% among primary healthcare personnel). Vaccination rates also varied by study site, with the highest vaccine coverage in Minnesota (85%) and the lowest in Florida (27%); Florida had a higher proportion of participants who were first responders (58%) than did sites in other states (5%–22%) (Table 1). Vaccinated participants were more likely than unvaccinated participants to wear a mask at work (median percentage of time wearing a mask at work was 50% vs 7%).

Between 1 July 2022 and 31 May 2023, 279 RT-PCR–confirmed influenza virus infections and 1044 PCR-confirmed SARS-CoV-2 infections were identified. There were 12 coinfections of influenza virus and SARS-CoV-2; all were symptomatic. Ten participants experienced 2 laboratory-confirmed influenza virus infections more than 30 days apart during the study period, resulting in 269 first influenza virus infections during 703 326 person-days of follow-up (38.2 per 100 000 person-days). Among these 269 infections, 68 (25%) were asymptomatic. The presence of symptoms did not vary significantly by age group (range: 74%–75%) or occupation (range: 68% among primary healthcare personnel to 83% among allied healthcare personnel; P = .22). Of 201 (75%) infections associated with at least 1 symptom, 81% of participants reported cough, 66% reported febrile symptoms, and 55% reported sore throat. In addition, 76% of participants reported nasal congestion, 72% reported fatigue, and 57% reported body aches (Supplementary Table 1). Among 185 symptomatic participants with confirmed infection who reported the number of days they were sick in bed, the median (IQR) was 2 (0–3) days. Percentages of participants with RT-PCR–confirmed influenza virus infection during the 2022–2023 season were similar by study site, gender, occupation, presence of a chronic condition, and current smoking status (Table 1). Similar proportions of vaccinated (19%) and unvaccinated (21%) participants with an influenza virus infection sought medical care for illness, with 1 hospitalization in an unvaccinated participant.

Ten participants’ influenza virus infections occurred within 14 days of vaccination; these participants were censored at their vaccination date, resulting in 259 infections being included in VE analyses. The incidence of any influenza virus infection per 100 000 person-days was 31.6 among vaccinated participants and 44.8 among unvaccinated participants (aHR: .68; 95% CI: .51–.90) (Figure 1, Table 2); influenza vaccination was associated with a 32% reduced risk of any influenza virus infection (VE: 32%; 95% CI: 10%–49%). The incidence of symptomatic illness per 100 000 person-days was 23.7 among vaccinated participants and 33.2 among unvaccinated participants (aHR: .62; 95% CI: .45–.85); influenza vaccination was associated with a 38% reduced risk of symptomatic illness (VE: 38%; 95% CI: 15%–55%). The aHR was similar when using the alternate definition of symptomatic illness requiring febrile symptoms (aHR: .72; 95% CI: .49–1.05). The incidence of asymptomatic infection per 100 000 person-days was 8.0 among vaccinated and 11.6 among unvaccinated participants (aHR: .87; 95% CI: .51–1.47); VE against asymptomatic influenza was not statistically significant (VE: 13%; 95% CI: −47%, 49%) (Table 2). In the interaction model, aHRs were statistically significantly different between symptomatic and asymptomatic infections (interaction term coefficient: .60; 95% CI: .03–1.16). Vaccine effectiveness was similar against A(H1N1)pdm09-associated (VE: 47%; 95% CI: 9%–69%) and A(H3N2)-associated (VE: 52%; 95% CI: 26%–69%) infections, although the 95% CI for VE against A(H1N1)pdm09 was wider (Table 2 and Supplementary Table 2). Only 3 influenza B infections occurred in this population, and therefore VE against influenza B was not estimated.

In sensitivity analyses, VE against symptomatic illness was similar when excluding 12 coinfections of influenza and SARS-CoV-2 (VE: 35%; 95% CI: 12%–53%). When we excluded 1142 (29%) participants with less than 80% adherence to swabbing, VE was slightly higher against symptomatic illness (VE: 44%; 95% CI: 24%–59%), and VE against asymptomatic infection remained not statistically significant (VE: 29%; 95% CI: −23%, 60%). Influenza vaccination was not associated with rates of non-influenza, SARS-CoV-2–negative illness (237 per 100 000 vaccinated person-days vs 252 per 100 000 unvaccinated person-days; aHR: 1.01; 95% CI: .92–1.10).

During the 2022–2023 influenza season, influenza vaccination provided a significant reduction in the risk of any symptomatic influenza illness, but not asymptomatic infection, in a prospective occupational cohort. Protection was similar against infection with influenza A(H3N2) as with influenza A(H1N1) pdm09. Asymptomatic infections represented a substantial proportion of all influenza virus infections, highlighting the importance of further research to understand VE and transmission of influenza among those with asymptomatic infections.