Influenza VE was assessed using a test-negative case-control design [15–17]. Cases were defined as persons with a medically attended ARI confirmed as influenza by RT-PCR. Controls were persons with a medically attended ARI who tested negative for influenza virus. The analysis was limited to patients who were enrolled during periods of influenza virus circulation at each site. We excluded controls with symptom onset before the first date or after the last date of symptom onset among influenza cases at each network site.

The primary exposure of interest was medical record/registry documented receipt of current-season influenza vaccine. Persons aged ≥9 years were considered vaccinated, given documented receipt of at least 1 dose of current-season (2012–2013) vaccine ≥14 days before illness onset. Children aged 6 months through 8 years were considered vaccinated if they received the number of doses recommended by the Advisory Committee on Immunization Practices (ACIP) for 2012–2013, based on their documented vaccination history [18]. Those who were vaccinated within 14 days of illness onset or vaccinated but did not receive all of the ACIP-recommended doses were excluded from the main analysis. Those without documented evidence of vaccine receipt and those vaccinated after illness onset were considered unvaccinated. For the 2012–2013 season, the northern hemisphere vaccine included the influenza A(H3N2) component A/Victoria/361/(2011), the influenza A(H1N1) component A/California/07/(2009), and the influenza B component B/Wisconsin/01/(2010)/Yamagata; the influenza A(H3N2) and influenza B components were different from those in the vaccine from the prior influenza season [19].

We also estimated VE for all combinations of vaccine exposure in the current and prior season (2012–2013 and 2011–2012, respectively) among patients aged ≥9 years who have been in their current health system for at least the past year. We excluded children aged <9 years from this analysis because the criteria for being considered appropriately vaccinated in this age group were based on influenza vaccination history in prior seasons. The 4 vaccine exposure groups were (1) vaccinated in both the current and prior season, (2) vaccinated in the current season only, (3) vaccinated in the prior season only, and (4) not vaccinated in either the current or prior season (reference group). Patients vaccinated with unknown vaccine types or vaccine types other than trivalent inactivated influenza vaccine (IIV3) in the current season were excluded from analyses that included prior season vaccination.

Logistic regression models were used to estimate odds ratios (ORs) comparing vaccinated (including joint current and prior vaccine exposure) and unvaccinated subjects; VE was calculated as 100% × [1 − OR]. We generated separate VE estimates for influenza overall, influenza A(H3N2), influenza B/Yamagata, and influenza B/Victoria outcomes. VE estimates were generated for all licensed vaccines, IIV3, and trivalent live attenuated influenza vaccine (LAIV3). VE was not calculated for 2009 pandemic influenza A(H1N1) strain (A[H1N1]pdm09) because of the low number of cases (n = 52). Estimates for LAIV3 VE were restricted to children aged 2–17 years because LAIV3 use was low in adults.

Age, network site, high-risk health status, and calendar time (modeled as a series of dichotomous variables representing 2-week intervals) were included in all adjusted VE models a priori. Age in years was modeled using linear tail-restricted cubic spline functions with 5 knots based on percentiles. Other potential confounders (eg, sex, race/ethnicity, reported general health, and number of children in the household) were assessed but were not included in the adjusted (simple) model because the covariate did not change the VE estimate by >5% [20]. A difference in VE of ≤5% was not considered meaningful. For age group–specific estimates, we used stratified models with data in only the range of the age group of interest (age 6 months–8 years, 9–17 years, 18–49 years, 50–64 years, and ≥65 years). Age was modeled as a linear term in the age-specific models. For comparison with prior season estimates, we also generated a fully adjusted VE model that included additional covariates (including those that did not change the VE estimate by >5%). These included age, sex, race/ethnicity, high-risk health status, reported general health, interval (in days) between illness onset and specimen collection, and calendar time [4].

To better understand the relationship between age and VE, we calculated VE by single year of age. In these models, we used data from ages 1 through 75 years and included an interaction term between the age and current-season vaccination status. Ages of <1 and >75 years were not included in this analysis because of the sparse data. Age was represented as a restricted cubic spline function with 5 knots based on percentiles.

VE was also estimated using 2 alternative means of representing vaccination status: by incorporating self-reported status with medical record/registry documentation and by using self-report only.

Statistical analyses were conducted using SAS statistical software (version 9.3; SAS Institute, Cary, North Carolina). P values of < .05 or a 95% confidence interval (CI) excluding 0% for VE was considered statistically significant. Adjusted estimates were not reported if the validity of the model fit was questionable.

From December 2012 through early April 2013, we enrolled 6766 patients with medically attended ARI; the peak enrollment of cases occurred in January. Among those enrolled, 314 (5%) were excluded from analysis of VE because of inconclusive RT-PCR test results (n = 16), negative results with illness onset outside the period when cases were identified at each site (n = 43), receipt of vaccine <14 days from onset (n = 68), receipt of an invalid vaccine dose (n = 1), and incomplete vaccination status, based on ACIP recommendations for children <9 years old (n = 186).

Of the 6452 patients included in the analysis, 2307 (36%) were influenza virus positive, including 1381 (60%) infected with influenza A virus and 933 (40%) infected with influenza B virus. Among cases where influenza virus subtype/lineage could be determined, 1292 were influenza A(H3N2), 52 were A/H1N1pdm09, 582 were influenza B/Yamagata, and 303 were influenza B/Victoria. Ten patients had virus coinfections: 7 were positive for both influenza A and B virus, and 3 were positive for both B virus lineages.

Median age was 28 years (interquartile range [IQR], 9–50 years) for case patients and 30 years (IQR, 10–51 years) for control patients (P = .2, by the Wilcoxon rank sum test). Among case patients, the median age was 34 years, 14 years, and 12 years among those infected with influenza A(H3N2), influenza B/Yamagata, and influenza B/Victoria, respectively. Compared with control patients, case patients were less likely to have a high-risk condition and more likely to report their general health as excellent; however, reported current health (at the time of enrollment) was rated lower by case patients (Table 1). There was a higher proportion of females among individuals with influenza, and those with influenza were more likely to live in the same household with ≥2 children (age <12 years), compared with controls. Patients with influenza and control patients did not differ significantly by race/ethnicity, obesity status, or household exposure to smoke. The median time from onset to enrollment was 3 days for both groups. Table 1.

Descriptive Characteristics of Enrolled Patients With Medically Attended Acute Respiratory Illness, by Case Status

There were 2877 patients (45%) with documented current-season influenza vaccine receipt ≥14 days before symptom onset. Of these, 83% received IIV3, 8% received LAIV3, <1% received intradermal or intramuscular high-dose vaccine, and 8% had an unknown vaccine type. LAIV3 accounted for 126 vaccines (19%) administered to enrolled children aged <9 years. Vaccine coverage was lowest among adolescents and among adults aged <50 years old and highest among adults aged ≥65 years (Table 2). Compared with unvaccinated participants, vaccinated individuals were more likely to be female; to be white, non-Hispanic; to have a high-risk condition; to have no children aged <12 years in the household; and to have household exposure to cigarette smoke. Vaccinated individuals were less likely to report their general health status as excellent. The median time to enrollment was 3 days from onset for both vaccinated and unvaccinated patients. Table 2.

Descriptive Characteristics of Enrolled Patients With Medically Attended Acute Respiratory Illness, by Vaccination Status for the Current (2012–2013) and Previous (2011–2012) Influenza Seasons

Current and prior season vaccination statuses were highly correlated. Among patients who received current-season vaccination, 70% were also vaccinated during the prior season. Among patients who did not receive current-season vaccination, 81% were not vaccinated during the prior season. Approximately 60% of adults aged ≥65 years were vaccinated during both the current and prior seasons. In contrast, 22% of patients aged 9–17 years and 21% aged 18–49 years were vaccinated during both seasons.

Overall adjusted VE was 49% (95% CI, 43%–55%) against all outpatient influenza illness. VE was 39% against influenza A(H3N2) (95% CI, 29%–47%), 66% against the vaccine lineage influenza B/Yamagata (95% CI, 58%–73%), and 51% against the nonvaccine lineage influenza B/Victoria (95% CI, 36%–63%; Tables 3 and 4). There was significant protection against influenza A(H3N2) among patients aged 6 months–8 years, 18–49 years, and 50–64 years (Table 3); VE CIs among children aged 9–17 years and adults aged ≥65 years included 0. VE estimates from the simple and fully adjusted models were similar overall and within age strata (Supplementary Table 1). Examination of VE as a function of continuous age (Figure 1A and 1B) demonstrated a gradual decline in children and adolescents, and a relatively stable VE among working-age adults. These results were generally consistent with the age-stratified measures, but the CIs were wide when VE was assessed by year of age. The effectiveness of IIV3 and LAIV3 did not differ significantly among children aged 2–17 years. VE against influenza A(H3N2) was 36% (95% CI, 15%–51%) for children who received IIV3 and 46% (95% CI, 13%–66%) for those who received LAIV3. VE against the B vaccine lineage (Yamagata) was 68% (95% CI, 54%–77%) for children who received IIV3 and 53% (95% CI, 20%–73%) for those who received LAIV3. Data were too sparse to estimate VE against influenza B/Victoria (nonvaccine lineage) by vaccine type. Table 3.

Overall and Influenza A(H3N2)–Specific Unadjusted and Adjusted Vaccine Effectiveness (VE), by Age Group and Vaccination Status

In analyses incorporating self-reported vaccination status, VE estimates were similar to estimates determined using medical record/registry data only (Supplementary Table 2).

The effectiveness of IIV3 was calculated with persons who were unvaccinated in both the current (2012–2013) season and the prior (2011–2012) season as the reference group. Protection against influenza A(H3N2) in the current season was similar for participants aged ≥9 years old who received vaccine in both the current and previous seasons, the current season only, and the previous season only (Figure 2). The adjusted VE against influenza A(H3N2) was 35% (95% CI, 21%–47%), 37% (95% CI, 19%–51%), and 33% (95% CI, 12%–49%), respectively, for these 3 groups. The pattern of current- and prior-season VE against influenza A(H3N2) varied by age, but CIs were wide. Among children aged 9–17 years, significant protection against influenza A(H3N2) was conferred by receipt of current-season vaccine alone, and residual protection was evident among those vaccinated with prior-season vaccine alone. However, children aged 9–17 years who were vaccinated in both seasons had significantly lower VE point estimates against influenza A(H3N2), compared with those who were vaccinated in the current season only (P = .01). Among those aged 18–49 years, all 3 vaccination histories provided significant protection against influenza A(H3N2). Adults aged 50–64 years had significant protection if they received both current- and prior-season vaccine, while VE was lower and CIs overlapped with 0 for receipt of prior- or current-season vaccine alone. VE was close to 0 for all vaccine combinations among adults aged ≥65 years. Figure 2.

Age-specific trivalent inactivated influenza vaccine effectiveness (VE) of current and prior vaccination against influenza A(H3N2) among persons age ≥9 years, in which persons not vaccinated in either the current or prior season served as the reference group. VE data are adjusted for network site, subject age, presence of high-risk health conditions, and calendar time (2-week intervals). *P = .01 for comparison between the 2 vaccine groups. Abbreviation: CI, confidence interval.

Antigenic characterization demonstrated that influenza A(H3N2) viruses from 226 participants were antigenically similar to the cell-propagated A/Victoria/361/2011 reference virus (data not shown). Among the 44 viruses sequenced, 42 (95%) belonged to HA clade 3C and were similar to the majority of circulating influenza A(H3N2) viruses in the United States (Supplementary Figure 1).