Summary
Frailty has important impacts on influenza vaccine effectiveness (VE). Here, frailty was the most important confounder of VE, and not accounting for frailty underestimated VE. VE was high in nonfrail older adults, but diminished with increasing frailty.
Keywords: influenza, vaccine effectiveness, elderly, frailty, hospitalization
Abstract
Background
Influenza is an important cause of morbidity and mortality among older adults. Even so, effectiveness of influenza vaccine for older adults has been reported to be lower than for younger adults, and the impact of frailty on vaccine effectiveness (VE) and outcomes is uncertain. We aimed to study VE against influenza hospitalization in older adults, focusing on the impact of frailty.
Methods
We report VE of trivalent influenza vaccine (TIV) in people ≥65 years of age hospitalized during the 2011–2012 influenza season using a multicenter, prospective, test-negative case-control design. A validated frailty index (FI) was used to measure frailty.
Results
Three hundred twenty cases and 564 controls (mean age, 80.6 and 78.7 years, respectively) were enrolled. Cases had higher baseline frailty than controls (P = .006). In the fully adjusted model, VE against influenza hospitalization was 58.0% (95% confidence interval [CI], 34.2%–73.2%). The contribution of frailty was important; adjusting for frailty alone yielded a VE estimate of 58.7% (95% CI, 36.2%–73.2%). VE was 77.6% among nonfrail older adults and declined as frailty increased.
Conclusions
Despite commonly held views that VE is poor in older adults, we found that TIV provided good protection against influenza hospitalization in older adults who were not frail, though VE diminished as frailty increased.
Clinical Trials Registration
(See the editorial commentary by Neuzil and Chen, on pages 397–8.)
Many jurisdictions offer vaccination programs for older people, aiming to reduce the burden of serious influenza-related outcomes in this vulnerable population. While there is evidence that influenza vaccination is effective in reducing medically attended influenza, hospitalization, and mortality [1], vaccine effectiveness (VE) in elderly populations appears modest [2–5]; estimates of VE against laboratory-confirmed influenza hospitalization range from 33% to 86% [6–16]. This variability has led to controversy about the value of influenza vaccines in protecting elderly people against hospitalization and serious outcomes. Concern has also been raised about the potential for residual bias in observational studies arising from incomplete adjustment for confounding from important factors, most notably health status [17]. It is therefore important to carefully consider how to best design VE studies to account for potential confounders as robustly as possible.
Frailty and functional status are important candidates in this endeavor, as they are integrative measures of older adults’ health status. Functional status (measured using tools such as the Barthel index) has been considered in some prior studies of influenza VE [18]. While function captures older adults’ overall health status to some degree, a more holistic measure of health status may provide additional benefit.
Frailty is a holistic measure of health, functioning, and vulnerability that is more strongly predictive of health outcomes than age [19]. Moreover, frailty may be associated with age- related declines in innate and adaptive humoral and cell-mediated immunity, which impairs the ability to resist influenza infection and respond to vaccination in a process known as immunosenescence [19–22]. Frailty is therefore an important candidate measure to consider in VE studies, although it is currently insufficiently accounted for in prospective influenza vaccine studies. Information on frailty is generally incompletely captured in health records, which also limits retrospective studies [14, 19, 23].
The Serious Outcomes Surveillance (SOS) Network of the Public Health Agency of Canada/Canadian Institutes of Health Research Influenza Research Network (PCIRN) was established in 2009 to monitor severe complications of seasonal influenza among hospitalized adults in Canada. A broad clinical screening definition was used to improve sensitivity, and cases were confirmed using polymerase chain reaction (PCR). Data relevant to elderly people were prospectively collected using several standardized measures of underlying health status, including a validated frailty index (FI) [24].
Here we aimed to measure VE against influenza hospitalization and serious outcomes in people aged ≥65 years, with particular focus on assessing the impact of frailty on VE estimates and an exploratory analysis of VE stratified by level of baseline frailty.
METHODS
This prospective, multicenter, test-negative case-control study was conducted by the PCIRN SOS Network in collaboration with the Toronto Invasive Bacterial Diseases Network (TIBDN). Active influenza surveillance was performed across approximately 16000 adult acute care beds in 38 academic and community sentinel hospitals in Nova Scotia (2 hospitals), New Brunswick (1), Quebec (4), Ontario (29), Manitoba (1), and British Columbia (1) between 1 November 2011 and 25 May 2012. Here we describe VE against influenza hospitalization for subjects aged ≥65 years, calculated using a test-negative design (comparing the odds of having been vaccinated in cases vs in controls, as described in more detail below). VE results for younger adults and overall burden of illness data, including strain-specific VE and outcomes, will be reported elsewhere.
The protocol was approved by the research ethics boards of participating institutions (ClinicalTrials.gov identifier NCT01517191). Patients provided informed written consent for data collection and medical record screening in accordance with the local research ethics board requirements. At some sites, research ethics board approval allowed for medical record review with data abstraction in the absence of consent.
Patients
SOS Network surveillance monitors reviewed daily admissions to medical and coronary intensive care units (ICUs) and medical wards. Patients with community-acquired pneumonia, acute exacerbation of chronic obstructive pulmonary disease or asthma, unexplained sepsis, any other respiratory infection or diagnosis, or any respiratory or influenza-like symptom (eg, dyspnea, cough, sore throat, myalgia, arthralgia, fever) were screened within 5 days of admission. Nasopharyngeal swabs were collected and tested for influenza viruses either as part of clinical care or by SOS Network monitors.
To capture influenza cases presenting atypically, we conducted enhanced surveillance as follows. One day per week, beginning when the local laboratory reported ≥2 positive influenza tests in a single week or 1 or more positive influenza tests in 2 consecutive weeks, patients who were admitted with a triage temperature ≥37.5°C associated with acute coronary syndrome (eg, myocardial infarction, unstable angina), any other cardiac diagnosis (eg, atrial fibrillation, other arrhythmia, myocarditis), or stroke were screened. Enhanced surveillance was stopped when the local laboratory reported no positive influenza results for 2 consecutive weeks. In hospitals associated with TIBDN, enhanced influenza testing was performed 7 days per week as routine clinical practice.
Patients were deemed to be influenza cases if they fulfilled the eligibility criteria and tested positive for influenza at any time during their hospitalization (hereafter, “cases”), and deemed to be controls if they fulfilled the eligibility criteria and tested negative for influenza within 7 days of hospital admission (hereafter, “controls”). Each case was age-matched (65–75 years and >75 years) with the next 1 or more sequentially enrolled control(s) admitted to the same site within 14 days before or after admission.
Vaccination
Patients were defined as vaccinated if they reported receipt of a 2011–2012 seasonal influenza vaccine >14 days before symptom onset. Patients who had received the current seasonal influenza vaccine but whose date was unknown were initially defined as “status unknown” until after 14 January 2012, when they were defined as “vaccinated” (the vast majority of older adults in Canadian immunization programs are vaccinated before the end of December).
There were 8 seasonal trivalent influenza vaccines (TIVs) approved for use in Canadian adults containing the influenza strains recommended by the World Health Organization for the 2011–2012 Northern Hemisphere influenza season: A/California/7/2009 (H1N1)-like virus; A/Perth/16/2009 (H3N2)-like virus; B/Brisbane/60/2008-like virus (Victoria lineage) [25, 26]. No quadrivalent vaccines were available in Canada during the study period.
Data Collection
Data were collected by SOS Network monitors by patient interview and medical record review during the hospital admission. Thirty-day postdischarge data were collected by telephone. Standardized case report forms were used to collect demographic information, medical and surgical history, details of presenting illness, medical management, healthcare use, and outcomes. Seasonal influenza vaccination status was collected by interview with the patient or their caregiver; self-reported immunization history was verified with the immunization provider or an immunization registry (where one existed) to collect product, lot number, and vaccination date. The VE analysis included patients with known influenza vaccination status.
In an effort to account for confounding, several standardized assessment tools were used to quantify patients’ underlying health status. Frailty was assessed using a validated FI based on a comprehensive geriatric assessment of 39 health and functional impairments (Supplementary Data 1). Each item was scored as 0 (no deficit) or 1 (deficit present) with intermediate values assigned for ordered response variables [24]. The FI was generated by summing each participant’s deficits and dividing by the total number of deficits included (39). Frailty was graded by validated cutoffs: FI 0–0.1, nonfrail; >0.1–0.21, prefrail; >0.21–0.45, frail; >0.45, most frail [27]. The Barthel index of activities of daily living was used to assess functional impairment [28], and the Hierarchical Assessment of Balance and Mobility (HABAM) was used to assess mobility and balance [29]. The FI, Barthel index, and HABAM were used to assess health and functional status 2 weeks before admission to hospital (baseline), at enrollment, and at 30 days postdischarge.
All-cause and influenza-attributable mortality were assessed at 30 days postdischarge (±7 days). When day 30 data were not available, patients who were not readmitted to the same hospital were considered to have survived.
Laboratory Methods
Influenza testing was performed according to usual local protocols at the participating hospital’s virology laboratory or provincial public health laboratories. All SOS Network sites used reverse-transcription PCR (RT-PCR) for influenza virus typing and subtyping, apart from 1 site that used viral culture. After local testing, samples were transported to the SOS Network central laboratory at the Canadian Center for Vaccinology where they were tested for influenza using RT-PCR to confirm local laboratory results and for further influenza subtyping. Twenty patients (1.4%) with discrepant influenza positive/negative results were excluded from VE analyses.
Ribonucleic acid was extracted from nasopharyngeal swab material using an automated system. A primary screening RT-PCR assay targeting the Matrix gene differentiated the influenza strains as A or B, thereby determining which of 2 secondary assays directed at the hemagglutinin antigen gene were used for subtyping. Subtyping identified pH1N1 and sH3N2 strains, or B/Yamagata vs B/Victoria lineage.
Statistical Analysis
Using a test-negative case-control design, VE was calculated as 1 minus the odds ratio (OR) of vaccination in cases compared with controls multiplied by 100. ORs were estimated using conditional logistic regression. The sample size was determined for the null hypothesis that VE was 0, against the alternative that VE was >0. Calculations were based on standard results for the asymptotic distribution of the estimated log OR. Sample size calculations showed that with 1:2 matching of cases:controls, 201 cases aged ≥65 years are needed to demonstrate a VE >0 assuming a true VE of 40%, probability of vaccination in controls of 60%, power of 90%, and a 2-sided significance level of 10%. VE was assessed against influenza-related hospitalization due to any influenza strain and by influenza type, subtype, and lineage. VE for the prevention of severe disease (ICU admission or mechanical ventilation) and death was also assessed.
Characteristics of cases vs controls and vaccinated vs unvaccinated cases were described and assessed using Mantel-Haenszel methods for discrete variables and linear mixed model methods for continuous variables with matched sets as random effect. No adjustment was made for multiple comparisons.
VE estimates were adjusted for age (65–75, >75 years), FI before admission to hospital (baseline) and antiviral use before admission. VE estimates in the final model were further adjusted in a multivariate conditional logistic regression with stepwise backward selection of covariates with P values of <.1 in univariate analysis. All matched sets with at least 1 case and 1 control without missing data for the final set of covariates were considered in the estimation of the final adjusted VE. Unadjusted and adjusted VE estimates were provided with a 95% confidence interval (CI). All analyses were performed using SAS software version 9.2 or later (SAS Institute, Cary, North Carolina).
To explore the impact of missing frailty data, multivariate logistic regression analysis was repeated with and without inclusion of the baseline FI in the cohort for whom FI had been measured; VE was also assessed after adjustment for baseline frailty only. A post hoc analysis was performed to assess VE stratified by level of baseline frailty (nonfrail, prefrail, frail, most frail). Baseline frailty data were missing in 44% of cases (Table 1); based on prior experience in the frailty literature, those missing frailty data tend to be more frail and vulnerable than those for whom frailty can be measured [30]. Reasons for missing frailty data included patients being too sick or confused to participate in the interview, inability to locate frailty-defining data on chart review, and retrospective enrollment (eg, when a positive PCR result was reported after a patient had already been discharged home, transferred, or died). Cases were more often missing frailty than controls because they were more likely to have been enrolled retrospectively following a positive influenza test flagged by the laboratory, whereas controls would not have been identified in this way.
Table 1.
Demographic and Clinical Characteristics
| Characteristic | Cases (n = 320) |
Controls (n = 564) |
P Value |
|---|---|---|---|
| Age, y | |||
| Mean (SD) | 80.63 (9.03) | 78.77 (7.91) | .001 |
| Median (range) | 81 (65–104) | 79 (65–99) | |
| Age subgroups, No. (%) | |||
| 65–75 y | 101 (31.6) | 206 (36.5) | .118 |
| >75 y | 219 (68.4) | 358 (63.5) | |
| Female sex, No. (%) | 176 (55.0) | 321 (56.9) | .635 |
| BMI, kg/m2 | |||
| Mean (SD) | 25.71 (5.69) | 26.24 (6.36) | .227 |
| Median (range) | 25.06 (6.64–48.28) | 25.33 (10.7–53.04) | |
| Obese (BMI ≥30 kg/m2), No. (%) | 56 (17.5) | 147 (26.1) | .025 |
| ≥1 comorbidity, No. (%) | |||
| Cardiac disease | 174 (54.4) | 345 (61.2) | .073 |
| Vascular disease | 243 (75.9) | 440 (78.0) | .548 |
| Pulmonary disease | 151 (47.2) | 296 (52.5) | .184 |
| Smoking, No. (%) | |||
| Current | 28 (8.8) | 60 (10.6) | .413 |
| Past | 113 (35.5) | 264 (46.8) | .010 |
| Frailty index prior to illness | |||
| Mean (SD) | 0.22 (0.13) | 0.2 (0.11) | .006 |
| Median (range) | 0.21 (0–0.62) | 0.18 (0–0.63) | |
| Unknown, No. (%) | 142 (44.4) | 20 (3.5) | |
| Mean HABAM score prior to illness | |||
| Mean (SD) | 2.56 (0.68) | 2.65 (0.54) | .056 |
| Median (range) | 2.96 (0–3) | 2.96 (0–3) | |
| Unknown, No. (%) | 150 (46.9) | 31 (5.5) | |
| Mean Barthel index prior to illness | |||
| Mean (SD) | 81.73 (28.83) | 88.08 (21.04) | .003 |
| Median (range) | 100 (0–100) | 100 (0–100) | |
| Unknown, No. (%) | 167 (52.2) | 80 (14.2) | |
| Received 2011–2012 seasonal influenza vaccine, No. (%) | 193 (60.3) | 417 (73.9) | .000 |
| Received 2010–2011 seasonal influenza vaccine, No. (%) | 181 (56.6) | 404 (71.6) | .001 |
Abbreviations: BMI, body mass index; HABAM, Hierarchical Assessment of Balance and Mobility; SD, standard deviation.
RESULTS
Patients
A total of 7044 patients were screened, of whom 1474 were enrolled in the overall ≥16 years eligible population (reported elsewhere); the first patient was enrolled on 20 December 2011 and the last contact was on 15 July 2012. The analysis of VE in the elderly cohort included 884 patients aged ≥65 years (320 cases and 564 controls; Figure 1). Of the 96 influenza A cases, 43 (44.8%) were A/H3N2, 34 (35.4%) were A/H1N1, and 19 (19.79%) were influenza A subtype unknown. Of the 224 influenza B cases, 135 (60.3%) were B/Yamagata lineage, 44 (19.6%) were B/Victoria lineage, and 45 (20.09%) were influenza B lineage unknown. The weekly distribution of influenza infections detected suggested that both influenza A subtypes and both influenza B lineages were circulating in Canada during the 2011–2012 influenza season (Figure 2).
Figure 1.
Patient disposition. Abbreviations: N, number of subjects in group; n, number of subjects fulfilling case definition; VE, vaccine effectiveness.
Figure 2.
Laboratory-confirmed influenza cases and test-negative controls by week and virus subtype, 1 November 2011 to 25 May 2012. Includes all laboratory-confirmed cases enrolled by the Serious Outcomes Surveillance Network of the Public Health Agency of Canada/Canadian Institutes of Health Research Influenza Research Network over the study period. Of the 588 cases enrolled, 528 cases were included in the vaccine effectiveness analysis.
Mean age was 80.6 years for cases and 78.7 years for controls. Table 1 summarizes baseline demographic and clinical characteristics. Cases were more frail than controls, with higher baseline FI, lower baseline mobility on the HABAM, and lower baseline function on the Barthel index. Cases were less likely than controls to have received the current and the preceding season’s influenza vaccine (Table 1).
Six hundred one (67.9%) patients had received 2011–2012 seasonal influenza vaccine. Baseline characteristics of vaccinated and unvaccinated patients are shown in Table 2. Vaccinated patients were older, more frail, and were more likely to have underlying comorbidities, be admitted from a long-term-care facility, be a current or past smoker, and take 0–4 prescribed medications before the onset of illness, and had a higher number of children aged ≤5 years living in the dwelling (Table 2).
Table 2.
Demographics and Clinical Characteristics of Unvaccinated and Vaccinated Patients
| Characteristic | Unvaccinated (n = 283) |
TIV (n = 601) |
P Value |
|---|---|---|---|
| Age, y | |||
| Mean (SD) | 78.24 (8.09) | 80.01 (8.46) | .003 |
| Median (range) | 79 (65–98) | 80 (65–104) | |
| Age subgroups, No. (%) | |||
| 65–75 y | 112 (39.6) | 195 (32.4) | .041 |
| >75 y | 171 (60.4) | 406 (67.6) | |
| Female, No. (%) | 168 (59.4) | 329 (54.7) | .217 |
| BMI, kg/m2 | |||
| Mean (SD) | 25.89 (5.66) | 26.12 (6.35) | .596 |
| Median (range) | 24.8 (11.71–44.17) | 25.39 (9.64–53.04) | |
| Obese (BMI ≥ 30 kg/m2), No. (%) | 63 (22.3) | 140 (23.3) | .863 |
| ≥1 comorbidity, No. (%) | |||
| Cardiac disease | 150 (53.0) | 368 (61.2) | .023 |
| Vascular disease | 200 (70.7) | 480 (79.9) | .004 |
| Pulmonary disease | 126 (44.5) | 310 (51.6) | .052 |
| Smoking, No. (%) | |||
| Current | 31 (11.0) | 57 (9.5) | .549 |
| Past | 104 (36.7) | 273 (45.4) | .007 |
| Frailty index prior to illness | |||
| Mean (SD) | 0.17 (0.11) | 0.2 (0.11) | .000 |
| Median (range) | 0.14 (0.01–0.58) | 0.18 (0–0.62) | |
| Unknown, No. (%) | 64 (22.6) | 98 (16.3) | |
| Admitted from an LTCF, No. (%) | 15 (5.3) | 65 (10.8) | .058 |
| 0–4 prescribed medications before admission, No. (%) | 91 (32.2) | 108 (18.0) | |
Abbreviations: BMI, body mass index; LTCF, long-term-care facility; SD, standard deviation; TIV, trivalent influenza vaccine.
The 2011–2012 influenza vaccine brand was ascertained in 69 (35.8%) vaccinated cases and 133 (32.1%) vaccinated controls. Of these, 8.3% and 12.2%, respectively, had received Fluviral (GSK Vaccines), 13.5% and 10.1% had received Agriflu (Novartis Vaccines), 11.4% and 8.9% had received Vaxigrip (Sanofi Pasteur), and 2.1% and 0.7% had received Fluad (Novartis Vaccines). One vaccinated control had received FluMist (MedImmune), and 1 vaccinated case had received another approved TIV.
Vaccine Effectiveness
The unadjusted VE for the prevention of influenza hospitalization due to any strain was 45.0% (95% CI, 25.7%–59.3%); the adjusted VE estimate was 58.0% (95% CI, 34.2%–73.2%). The adjusted VE against influenza A was 62.6% (95% CI, 2.4%–85.7%), and against influenza B was 58.3% (95% CI, 29.6%–75.3%) (Table 3).
Table 3.
Unadjusted and Adjusted Vaccine Effectiveness Estimates for Trivalent Influenza Vaccine Against Influenza Hospitalization
| Strain | Cases | Controls | Unadjusted | Adjusted | ||
|---|---|---|---|---|---|---|
| No. | No. | VE, % | (95% CI) | VE, % | (95% CI) | |
| All strains | 320 | 564 | 45.0 | (25.7–59.3) | 58.0a 43.3b |
(34.2–73.2) (22.1–58.9) |
| Influenza A | 96 | 191 | 51.7 | (15.1–72.6) | 62.6c | (2.4–85.7) |
| A/H1N1 | 34 | 74 | 88.7 | (48.1–97.5) | … | … |
| A/H3N2 | 43 | 87 | 30.3 | (–53.7 to 68.4) | 27.6d | (–227.3 to 84.0) |
| Influenza B | 224 | 373 | 37.3 | (16.0–53.2) | 58.3e | (29.6–75.3) |
| B/Victoria | 44 | 78 | 59.3 | (8.6–81.9) | 64.4f | (–0.7 to 87.4) |
| B/Yamagata | 135 | 222 | 32.4 | (–6.2 to 56.9) | 54.7g | (8.6–77.5) |
Blank values represent categories for which logistic regression could not be performed due to a low number of subjects represented.
Abbreviations: CI, confidence interval; VE, vaccine effectiveness.
Covariate (P value in model):
aInfluenza vaccination (0.000), age (0.267), antiviral use before onset of illness (0.775), frailty index before onset of illness (0.007), number of children aged ≤5 years living in the dwelling (0.034), medications before onset of illness (0.042).
bInfluenza vaccination (0.000), age (0.267), antiviral use before onset of illness (0.775), number of children aged ≤5 years living in the dwelling (0.034), medications before onset of illness (0.042).
cInfluenza vaccination (0.044), age (0.329), frailty index before onset of illness (0.581), number of children aged ≤5 years living in the dwelling (0.063), medications before onset of illness (0.917).
dInfluenza vaccination (0.675), age (0.092), frailty index before onset of illness (0.288), number of children aged ≤5 years living in the dwelling (0.466), medications before onset of illness (0.441).
eInfluenza vaccination (0.001), age (0.527), antiviral use before onset of illness (0.878), frailty index before onset of illness (0.002), number of children aged ≤5 years living in the dwelling (0.294), medications before onset of illness (0.025).
fInfluenza vaccination (0.051), age (0.434), frailty index before onset of illness (0.336), number of children aged ≤5 years living in the dwelling (0.995), medications before onset of illness (0.431).
gInfluenza vaccination (0.027), age (0.162), antiviral use before onset of illness (0.778), frailty index before onset of illness (0.004), number of children aged ≤5 years living in the dwelling (0.0.12), medications before onset of illness (0.139).
The covariates included in the final multivariate conditional logistic regression were influenza vaccination, age (65–75 vs >75 years), number of children aged ≤5 years living in the dwelling, antiviral use prior to admission, baseline FI, and number of medications (0–4 or >4 prescription medications). The adjusted VE against any strain of influenza was 65.3% (95% CI, 23.5%–84.2%) for patients aged 65–75 years, and 54.4% (95% CI, 22.7%–73.1%) for those aged >75 years.
The VE of TIV against influenza hospitalization due to any strain without including baseline FI in the model was 43.4% (95% CI, 22.1%–58.9%). Besides frailty, the only covariate that contributed significantly to the final model was the number of children aged ≤5 years living in the dwelling (P = .05); age was not a significant factor (P = .24). Adjusting only for baseline FI, VE was 58.7% (95% CI, 36.2%–73.2%). Because frailty data were missing in 44% of cases, we performed sensitivity analyses of VE restricted to those with known frailty, so that the unadjusted and adjusted VE models would include the same number and subgroup of patients. Among patients with known FI, the unadjusted VE was 55.9% (95% CI, 32.6%–71.1%) and increased slightly when adjusting for frailty (VE 58.0 [95% CI, 34.2%–73.2%]). VE among those with missing frailty data was lower, at 35.4% (95% CI, –2.3% to 59.2%). Adjusted VE by level of frailty was 77.6% for nonfrail, 51.0% for prefrail, 59.6% for frail, and –24.8% for the most frail patients (Table 4).
Table 4.
Demographics and Vaccine Effectiveness Against Influenza Hospitalization by Level of Frailtya
| Characteristic | Nonfrail (n = 92) |
Prefrail (n = 229) |
Frail (n = 165) |
Most Frail (n = 19) |
Matched-Tests P Value |
|---|---|---|---|---|---|
| Mean age (SD), y | 76.0 (7.9) | 79.0 (7.7) | 81.9 (8.1) | 84.5 (7.7) | <.001 |
| Female | 51 (55.4) | 130 (56.8) | 104 (63.0) | 12 (63.2) | .59 |
| Influenza vaccination | 54 (58.7) | 156 (68.1) | 126 (76.4) | 15 (78.9) | .663 |
| Influenza case | 35 (38.0) | 64 (27.9) | 67 (40.6) | 10 (52.6) | .018 |
| 0–4 prescribed medications before admission | 43 (46.7) | 180 (78.9) | 152 (92.1) | 19 (100) | <.001 |
| Admitted from an LTCF | 0 (0.0) | 3 (1.3) | 16 (9.7) | 12 (63.2) | <.001 |
| Admitted to ICU | 15 (16.3) | 25 (10.9) | 19 (11.5) | 1 (5.3) | .36 |
| Died | 5 (5.4) | 9 (3.9) | 25 (15.2) | 5 (26.3) | .023 |
| VE against influenza hospitalization, % (95% CI) | 77.6 (39.3–91.7) | 51.0 (5.2–74.7) | 59.6 (8.0–82.3) | −24.8 (−1040.4 to 86.3) |
Data are presented as No. (%) unless otherwise indicated.
Abbreviations: CI, confidence interval; ICU, intensive care unit; LTFC, long-term-care facility; SD, standard deviation; VE, vaccine effectiveness.
aFrailty index score 0–0.1, nonfrail; >0.1–0.21, prefrail; >0.21–0.45, frail; >0.45, most frail.
The adjusted VE of TIV against influenza-related ICU admission or mechanical ventilation was 70.9% (95% CI, −93.2% to 95.6%) (Table 5). Among cases and controls, there were 37/320 (11.6%) and 67/564 (11.9%) deaths, respectively. Two patients missing 30-day survival status were assumed to have survived. The unadjusted VE estimate for the prevention of death due to any strain was 53.3% (95% CI, −8.9% to 80.0%); adjusted VE was 75.5% (95% CI, −74.5% to 96.6%).
Table 5.
Unadjusted and Adjusted Vaccine Effectiveness Estimates for Trivalent Influenza Vaccine Against Admission to the Intensive Care Unit or Mechanical Ventilation
| Strain | Cases | Controls | Unadjusted | Adjusted | ||
|---|---|---|---|---|---|---|
| No. | No. | VE, % | (95% CI) | VE, % | (95% CI) | |
| All strains | 37 | 64 | 32.3 | (−64.2 to 72.0) | 70.9a | (−93.2 to 95.6) |
| Influenza A | 14 | 27 | 22.9 | (−219.1 to 81.4) | … | … |
| A/H1N1 | 5 | 10 | 74.4 | (−180.2 to 97.7) | … | … |
| A/H3N2 | … | … | … | … | … | … |
| Influenza B | 23 | 37 | 37.5 | (−93.1 to 79.8) | 46.0b | (−286.6 to 92.5) |
| B/Victoria | … | … | … | … | … | … |
| B/Yamagata | 10 | 16 | −28.1 | (−614.4 to 77.0) | −34.2c | (−2466.8 to 93.0) |
Blank values represent categories for which logistic regression could not be performed due to a low number of subjects represented.
Abbreviations: CI, confidence interval; ICU, intensive care unit; TIV, trivalent influenza vaccine; VE, vaccine effectiveness.
Covariate (P value in model):
aInfluenza vaccination (0.201), age (0.656), antiviral usage before ICU/mechanical ventilation (0.995), frailty index before onset of illness (0.224).
bInfluenza vaccination (0.539), age (0.267), antiviral usage before ICU/mechanical ventilation (0.997), frailty index before onset of illness (0.413).
cInfluenza vaccination (0.845), age (0.332), antiviral usage before ICU/mechanical ventilation (0.998), frailty index before onset of illness (0.314).
DISCUSSION
We observed moderate yet statistically significant and clinically important protection against influenza hospitalization in people aged ≥65 years. Our fully adjusted estimate of VE against influenza hospitalization due to any strain was 58.0%. After removing frailty from the model, the VE estimate was 43.4%, and when adjusting for frailty only, the estimate was 58.7%. Some of this difference is attributable to very low VE among those who were missing frailty data (35.4%). We explored this further in a sensitivity analysis restricted to the subgroup of patients with known frailty, where unadjusted VE was 55.9% and adjusted VE was 58.7%. Based on previous frailty literature, in which the most frail individuals are more likely to have missing data, we would expect that those patients for whom frailty was missing would have in fact been among the most frail, further supporting the point that consideration of frailty is important [30].
Our findings highlight an important “frailty bias” in observational studies of VE. This likely arises because frailty is associated with vaccination status (vaccinated subjects had a higher baseline FI and were more likely to have been admitted from a long-term-care facility) and may also be associated with differential immunological responses to vaccination. VE was strikingly different across levels of frailty, with very good VE (77.6%) among the nonfrail older adults and lower VE among the prefrail and frail, and no apparent VE in the most frail (though this estimate is limited by small sample size and wide confidence limits). Taken together, these findings suggest that frailty is a strong confounder of TIV VE, highlighting the importance of accounting for frailty when assessing the impact of influenza vaccines in older populations.
In a recent meta-analysis of test-negative case-control studies of TIVs in older adults, of 13 datasets that reported influenza hospitalization, 9 reported both adjusted and unadjusted data, but only 2 studies (7 datasets) included consideration of functional status [5]. In the first large-scale prospective controlled study to assess PCR-confirmed influenza in older (>50 years) hospitalized patients during 3 seasons (2006–2009) in the United States, the unadjusted VE estimate was 71.3% [14]. After adjusting for propensity score based on 16 covariates not including a specific measure of frailty, VE was 61.2% [14]. In a study of seasonal influenza vaccination in Europe in the 2012–2013 influenza season, VE estimates against influenza hospitalization in patients aged ≥65 years were adjusted for various covariates including functional status (Barthel index) [18]. Their unadjusted VE estimate against influenza hospitalization was 43% whereas the adjusted estimate was 58% [18]. Our results support previous studies which show that function is relevant for influenza VE. While the consideration of underlying functional status may mitigate the frailty bias to some degree, we argue that assessment of frailty is optimal as it considers function in the context of overall health status. Unlike the Barthel index, which measures only functional disability, the FI is a measure of an older individual’s overall vulnerability to adverse health outcomes (including consideration of comorbidities, sensory impairments, function, mobility, and symptoms) and is thus the optimal assessment to include in observational studies of VE to address frailty bias.
As shown in Figure 2, and in relation to published literature, 2011–2012 was a relatively late influenza season in Canada. Our finding of moderate VE among older adults is thus of particular interest, given evidence that VE may tend to wane with time after vaccination, even within a single influenza season [31, 32].
The main limitations of our study were its observational design and limited generalizability across seasons (eg, inability to predict viral circulation and vaccine match, and differences in viral epidemiology across regions). However, our study has numerous strengths. The sentinel approach provides consistency within a large-scale observational setting and has been used in Canada to develop a robust test-negative case-control design over multiple seasons [33–37]. The test-negative case-control design has been shown to provide accurate and precise estimates of VE, and is an accepted alternative to randomized controlled trials [38]. Methodological strengths of our study include the rigor by which cases were matched to controls and our use of multiplex PCR methods to confirm the diagnosis of influenza and characterize influenza types, subtypes, and lineages, enabling us to estimate strain-specific VE. Furthermore, the assessment of outcomes and health status using measures particularly relevant to older adults (frailty, function, and mobility) were unique features of our study.
In summary, we demonstrated moderate VE of TIV against influenza hospitalization in older adults. We found that frailty is the most important confounder to take into account when estimating VE for older adults. We identified an important frailty bias that impacted VE estimates; not accounting for frailty would have led to underestimating VE in this vulnerable population. Frailty should therefore be considered in future studies of influenza VE. Despite commonly held views that VE is poor in older adults as a whole, our findings clarify that VE appears to be good in older adults who are not frail but does diminish as frailty increases. In community-dwelling older adults, the prevalence of frailty is approximately 24% and the majority of older adults are not frail [27, 39]. Systematically evaluating the impact of frailty on VE and outcomes of influenza illness is critically important for understanding the important health benefits of influenza vaccination in older adults. The potential of influenza vaccination to contribute to healthy aging should not be underestimated.
Supplementary Data
Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Supplementary Material
Notes
Author contributions. All authors participated in the design, implementation, analysis, and/or interpretation of the study and the development of this manuscript. All authors had full access to the data and gave final approval before submission.
Acknowledgments. The authors thank the many dedicated SOS Network surveillance monitors for their hard work and diligence. We also thank hospital staff in participating hospitals for their collaboration and assistance and the many patients and their families whose participation made this study possible. Finally, we thank Annick Moon (Moon Medical Communications Ltd, UK) for providing medical writing services and Bruno Dumont (Business and Decision Life Sciences, on behalf of GSK, Wavre, Belgium) for editorial assistance and manuscript coordination.
Disclaimer. The authors received no financial support or other form of compensation related to the development of the manuscript. The authors are solely responsible for final content and interpretation.
Financial support. This work was supported by GlaxoSmithKline Biologicals SA and the Public Health Agency of Canada/Canadian Institutes of Health Research Influenza Research Network (PCIRN).
Potential conflicts of interest. M. K. A. reports grant funding from GSK, Pfizer, and Sanofi Pasteur. V. S. was employed by GSK Vaccines at the time of the study and is now employed by Novavax Vaccines, and holds shares in the GSK group of companies. T. H. reports payments from the GSK group of companies, Pfizer, and AbbVie. F. H. is employed by the GSK group of companies. G. D. S. was external consultant at Business and Decision Life Sciences (on behalf of GSK) at the time of the study, and is currently employed by the GSK group of companies and holds shares in the GSK group of companies. J. E. M. reports payments to her institution from the GSK group of companies and Sanofi Pasteur. A. C. reports payments from Sanofi. M. E. reports payments from the GSK group of companies, Public Health Agency of Canada, and Canadian Institutes of Health Research (CIHR). S. H. reports payments from the GSK group of companies. B. I. was employed by the GSK group of companies at the time of the study. J. La. reports payments from the GSK group of companies and CIHR. A. M. reports payments to her institution from the GSK group of companies, and payments from Hoffman La Roche and Sanofi Pasteur. J. P. reports payments from the GSK group of companies, Merck, Roche, and Synthetic Biologics. L. V. reports payments from the GSK group of companies. S. A. M. reports payments from the GSK group of companies, Pfizer, Merck, Novartis, and Sanofi. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1. Castilla J, Guevara M, Martínez-Baz I et al. Enhanced estimates of the influenza vaccination effect in preventing mortality: a prospective cohort study. Medicine (Baltimore) 2015; 94:e1240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Chan TC, Fan-Ngai Hung I, Ka-Hay Luk J, Chu LW, Hon-Wai Chan F. Effectiveness of influenza vaccination in institutionalized older adults: a systematic review. J Am Med Dir Assoc 2014; 15:226 e1–6. [DOI] [PubMed] [Google Scholar]
- 3. Nichol KL, Nordin JD, Nelson DB, Mullooly JP, Hak E. Effectiveness of influenza vaccine in the community-dwelling elderly. N Engl J Med 2007; 357:1373–81. [DOI] [PubMed] [Google Scholar]
- 4. Rivetti D, Jefferson T, Thomas R et al. Vaccines for preventing influenza in the elderly. Cochrane Database Syst Rev 2006; CD004876. doi:10.1002/14651858.CD004876.pub2. [DOI] [PubMed] [Google Scholar]
- 5. Darvishian M, Bijlsma MJ, Hak E, van den Heuvel ER. Effectiveness of seasonal influenza vaccine in community-dwelling elderly people: a meta-analysis of test-negative design case-control studies. Lancet Infect Dis 2015; 14:1228–39. [DOI] [PubMed] [Google Scholar]
- 6. Chen Q, Griffin MR, Nian H et al. Influenza vaccine prevents medically attended influenza-associated acute respiratory illness in adults aged ≥50 years. J Infect Dis 2015; 211:1045–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Cheng AC, Kotsimbos T, Kelly PM. Influenza vaccine effectiveness against hospitalisation with influenza in adults in Australia in 2014. Vaccine 2015; 33:7352–6. [DOI] [PubMed] [Google Scholar]
- 8. Casado I, Dominguez A, Toledo D et al. Effect of influenza vaccination on the prognosis of hospitalized influenza patients. Expert Rev Vaccines 2016; 15:425–32. [DOI] [PubMed] [Google Scholar]
- 9. Castilla J, Martínez-Artola V, Salcedo E et al. ; Network for Influenza Surveillance in Hospitals of Navarre Vaccine effectiveness in preventing influenza hospitalizations in Navarre, Spain, 2010-2011: cohort and case-control study. Vaccine 2012; 30:195–200. [DOI] [PubMed] [Google Scholar]
- 10. Cheng AC, Brown S, Waterer G et al. Influenza epidemiology, vaccine coverage and vaccine effectiveness in sentinel Australian hospitals in 2012: the Influenza Complications Alert Network (FluCAN). Commun Dis Intell Q Rep 2013; 37:E246–52. [DOI] [PubMed] [Google Scholar]
- 11. Cheng AC, Holmes M, Irving LB et al. Influenza vaccine effectiveness against hospitalisation with confirmed influenza in the 2010-11 seasons: a test-negative observational study. PLoS One 2013; 8:e68760. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Gefenaite G, Rahamat-Langendoen J, Ambrozaitis A et al. Seasonal influenza vaccine effectiveness against influenza in 2012–2013: a hospital-based case-control study in Lithuania. Vaccine 2014; 32:857–63. [DOI] [PubMed] [Google Scholar]
- 13. Kwong JC, Campitelli MA, Gubbay JB et al. Vaccine effectiveness against laboratory-confirmed influenza hospitalizations among elderly adults during the 2010–2011 season. Clin Infect Dis 2013; 57:820–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Talbot HK, Griffin MR, Chen Q, Zhu Y, Williams JV, Edwards KM. Effectiveness of seasonal vaccine in preventing confirmed influenza-associated hospitalizations in community dwelling older adults. J Infect Dis 2011; 203:500–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Talbot HK, Zhu Y, Chen Q, Williams JV, Thompson MG, Griffin MR. Effectiveness of influenza vaccine for preventing laboratory-confirmed influenza hospitalizations in adults, 2011–2012 influenza season. Clin Infect Dis 2013; 56:1774–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Turner N, Pierse N, Bissielo A et al. The effectiveness of seasonal trivalent inactivated influenza vaccine in preventing laboratory confirmed influenza hospitalisations in Auckland, New Zealand in 2012. Vaccine 2014; 32:3687–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Trucchi C, Paganino C, Orsi A, De Florentiis D, Ansaldi F. Influenza vaccination in the elderly: why are the overall benefits still hotly debated? J Prev Med Hyg 2015; 56:E37–43. [PMC free article] [PubMed] [Google Scholar]
- 18. Puig-Barbera J, Natividad-Sancho A, Launay O et al. 2012–2013 seasonal influenza vaccine effectiveness against influenza hospitalizations: results from the global influenza hospital surveillance network. PLoS One 2014; 9:e100497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet 2013; 381:752–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Johnstone J, Parsons R, Botelho F et al. Immune biomarkers predictive of respiratory viral infection in elderly nursing home residents. PLoS One 2014; 9:e108481. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. McElhaney JE. The unmet need in the elderly: designing new influenza vaccines for older adults. Vaccine 2005; 23(suppl 1):S10–25. [DOI] [PubMed] [Google Scholar]
- 22. McElhaney JE, Dutz JP. Better influenza vaccines for older people: what will it take? J Infect Dis2008; 198:632–4. [DOI] [PubMed] [Google Scholar]
- 23. Ridenhour BJ, Campitelli MA, Kwong JC et al. Effectiveness of inactivated influenza vaccines in preventing influenza-associated deaths and hospitalizations among Ontario residents aged ≥65 years: estimates with generalized linear models accounting for healthy vaccinee effects. PLoS One 2013; 8:e76318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Searle SD, Mitnitski A, Gahbauer EA, Gill TM, Rockwood K. A standard procedure for creating a frailty index. BMC Geriatr 2008; 8:24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. World Health Organization. Recommended composition of influenza virus vaccines for use in the 2011–2012 Northern Hemisphere influenza season http://wwwwhoint/influenza/vaccines/virus/2011_12north/en/. Accessed 15 October 2015.
- 26. Public Health Agency of Canada. An Advisory Committee Statement (ACS)—National Advisory Committee on Immunization (NACI): statement on seasonal influenza vaccine for 2011–2012. Can Commun Dis Rep 2011; 37 http://www.phac-aspc.gc.ca/publicat/ccdr-rmtc/11vol37/acs-dcc-5/index-eng.php. Accessed 11 October 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Hoover M, Rotermann R, Sanmartin C, Bernier J. Validation of an index to estimate the prevalence of frailty among community-dwelling seniors Statistics Canada Catalogue no 82-003-X health reports; 2013. http://www.statcan.gc.ca/pub/82-003-x/2013009/article/11864-eng.pdf. Accessed 11 October 2015. [Google Scholar]
- 28. Mahoney FI, Barthel DW. Functional evaluation: the Barthel index. Md State Med J 1965; 14:61–5. [PubMed] [Google Scholar]
- 29. MacKnight C, Rockwood K. A hierarchical assessment of balance and mobility. Age Ageing 1995; 24:126–30. [DOI] [PubMed] [Google Scholar]
- 30. Andrew MK. Social vulnerability in old age. In: Fillit HM, Rockwood K, Young J, eds. Brocklehurst’s Textbook of Geriatrics and Clinical Gerontology. 8th ed Philadelphia: Saunders Elsevier, 2016. [Google Scholar]
- 31. Ferdinands JM, Fry AM, Reynolds S et al. Intraseason waning of influenza vaccine protection: evidence from the US influenza vaccine effectiveness network, 2011–2012 through 2014–2015 [manuscript published online ahead of print 29 December 2016]. Clin Infect Dis 2017. doi:10.1093/cid/ciw816. [DOI] [PubMed] [Google Scholar]
- 32. Kissling E, Nunes B, Robertson C et al. I-MOVE multicentre case-control study 2010/11 to 2014/15: is there within-season waning of influenza type/subtype vaccine effectiveness with increasing time since vaccination? Euro Surveill 2016; 21. doi:10.2807/1560-7917.ES.2016.21.16.30201. [DOI] [PubMed] [Google Scholar]
- 33. McNeil S, Shinde V, Andrew M et al. Interim estimates of 2013/14 influenza clinical severity and vaccine effectiveness in the prevention of laboratory-confirmed influenza-related hospitalisation, Canada, February 2014. Euro Surveill 2014; 19. [DOI] [PubMed] [Google Scholar]
- 34. McNeil SA, Andrew MK, Ye L et al. Interim estimates of 2014/15 influenza vaccine effectiveness in preventing laboratory-confirmed influenza-related hospitalisation from the Serious Outcomes Surveillance Network of the Canadian Immunization Research Network, January 2015. Euro Surveill 2015; 20:21024. [DOI] [PubMed] [Google Scholar]
- 35. Skowronski D, Gilbert M, Tweed S, Petric M, Li Y, Mak A. Effectiveness of vaccine against medical consultation due to laboratory-confirmed influenza: results from a sentinel physician pilot project in British Columbia, 2004–2005. Can Commun Dis Rep 2005; 31:181–91. [PubMed] [Google Scholar]
- 36. Skowronski DM, De Serres G, Dickinson J et al. Component-specific effectiveness of trivalent influenza vaccine as monitored through a sentinel surveillance network in Canada, 2006–2007. J Infect Dis 2009; 199:168–79. [DOI] [PubMed] [Google Scholar]
- 37. Skowronski DM, Masaro C, Kwindt TL et al. Estimating vaccine effectiveness against laboratory-confirmed influenza using a sentinel physician network: results from the 2005–2006 season of dual A and B vaccine mismatch in Canada. Vaccine 2007; 25:2842–51. [DOI] [PubMed] [Google Scholar]
- 38. De Serres G, Skowronski DM, Wu XW, Ambrose CS. The test-negative design: validity, accuracy and precision of vaccine efficacy estimates compared to the gold standard of randomised placebo-controlled clinical trials. Euro Surveill 2013; 18. [DOI] [PubMed] [Google Scholar]
- 39. Hoover M, Rotermann M, Sanmartin C, Bernier J. Validation of an index to estimate the prevalence of frailty among community-dwelling seniors. Health Rep 2014; 24:10–7. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.