https://orcid.org/0000-0002-3027-6919
Abstract
Data on influenza vaccine effectiveness (IVE) against mortality are limited, with no Australian data to guide vaccine uptake. We aimed to assess IVE against influenza-related mortality in Australian hospitalized patients, assess residual confounding in the association between influenza vaccination and mortality, and assess whether influenza vaccination reduces the severity of influenza illness.
Data were collected between 2010 and 2017 from a national Australian hospital-based sentinel surveillance system using a case-control design. Adults and children admitted to the 17 study hospitals with acute respiratory symptoms were tested for influenza using nucleic acid testing; all eligible test-positive cases, and a subset of test-negative controls, were included. Propensity score analysis and multivariable logistic regression were used to determine the adjusted odds ratio (aOR) of vaccination, with IVE = 1 – aOR × 100%. Residual confounding was assessed by examining mortality in controls.
Over 8 seasons, 14038 patients were admitted with laboratory-confirmed influenza. The primary analysis included 9298 cases and 6451 controls, with 194 cases and 136 controls dying during hospitalization. Vaccination was associated with a 31% (95% confidence interval [CI], 3%–51%; P = .033) reduction in influenza-related mortality, with similar estimates in the National Immunisation Program target group. Residual confounding was identified in patients ≥65 years old (aOR, 1.92 [95% CI, 1.06–3.46]; P = .031). There was no evidence that vaccination reduced the severity of influenza illness (aOR, 1.07 [95% CI, .76–1.50]; P = .713).
Influenza vaccination is associated with a moderate reduction in influenza-related mortality. This finding reinforces the utility of the Australian vaccination program in protecting those most at risk of influenza-related deaths.
Influenza virus causes substantial morbidity and mortality and is implicated in up to 650 000 deaths globally each year [1]. The simplest preventive strategy against influenza infection is vaccination, with influenza vaccine effectiveness (IVE) primarily determined by the degree of match between the seasonal vaccine formulation and circulating virus strains [2]. While large clinical trials have demonstrated IVE against infection [3], it is not feasible to conduct clinical trials for uncommon influenza-related complications such as mortality. Therefore, monitoring of IVE against mortality relies on observational data, which are susceptible to confounding and bias [4–6].
Some IVE studies in the elderly have reported implausibly large reductions in all-cause mortality of 30%–60%, despite only 5%–10% of winter deaths being attributed to influenza [4, 7, 8]. Studies using influenza-specific mortality have estimated IVE between 35% and 89% [9–11]. This suggests that studies of influenza vaccination and all-cause mortality have overestimated the protective effect of vaccination, particularly when conducted in the elderly or when indirect outcome measures are used [4, 5, 12]. To supplement the usual statistical models used to adjust for confounding, “negative controls” are a useful tool to estimate residual confounding in IVE against mortality studies [13].
Using data collected from a national Australian hospital-based sentinel surveillance system, we aimed to address key questions regarding influenza vaccination and influenza-related mortality. First, we aimed to determine if influenza vaccination protects against influenza-related mortality by comparing the vaccination status of patients with confirmed influenza who died with a control group. Second, we aimed to determine whether there was residual confounding in this analysis by comparing the vaccination status of patients who died with noninfluenza respiratory illness with the same control group. Third, we explored whether influenza vaccination attenuates the severity of illness in hospitalized patients with influenza by comparing the vaccination status of influenza survivors with nonsurvivors.
METHODS
Study Design and Participants
The Influenza Complications Alert Network (FluCAN) is a national Australian hospital-based sentinel surveillance system with a case-control study design [14]. Data included in this analysis were collected from April to November, 2010–2017, at participating hospitals [15]. During the study period, both A/H1N1pdm and A/H3N2 subtypes and both B/Yamagata and B/Victoria lineages circulated in different seasons. Nonadjuvanted inactivated influenza vaccines were available during this time (Supplementary Table 1). Ethical approval was obtained for all participating sites and at Monash University (Monash approval ID 11056).
Study Procedures
Patients admitted to participating hospitals with acute respiratory symptoms suggestive of influenza were tested for influenza virus at the discretion of the treating clinician. Testing was conducted in certified laboratories using standard diagnostic procedures for influenza nucleic acid testing (NAT; reverse-transcription polymerase chain reaction). Potentially eligible patients were identified from influenza NAT lists; eligible test-positive patients were included as a case, and a subset of test-negative patients as controls (next available test-negative patient, 1 for each case where available). Clinical and demographic data were collected from hospital records during hospital admission, and stored in a REDCap database [16]. Vaccination history was obtained from the hospital record and, if not available, patient self-report was used. If neither were available, the patient’s general practitioner was contacted. Mortality was determined by assessment of the patient’s hospital record for all-cause, in-hospital mortality.
Exposure, Outcome, and Covariate Definitions
Patients were defined as vaccinated if they had received the seasonal influenza vaccine > 14 days before presentation [17]. Patients were defined as unvaccinated if no influenza vaccine was received, or if vaccine was received within 14 days. Patients were defined as having died of influenza-related mortality if they were influenza NAT positive and died during hospitalization. Four outcome groups were defined based upon mortality during hospitalization: case survivors (group 1), case deaths (group 2), control survivors (group 3), and control deaths (group 4) (Figure 1).
Patient flow in the Influenza Complications Alert Network (FluCAN) study, by case and control status. Abbreviations: ILI, influenza-like illness; NAT, nucleic acid testing.
Other covariates were defined as follows; smoking was defined as past or current smoking (or household for children); obesity was defined as a body mass index > 30 kg/m2 (or body weight > 120 kg for adults, or record of “obese” for children); age group was categorized as 0.5–15, 16–49, 50–64, 65–79, or ≥80 years, with adults ≥ 65 years defined as elderly; comorbidities were diabetes, chronic cardiac disease, chronic respiratory disease, chronic renal disease, chronic neurological disease, liver disease, malignancy, and immunosuppression, for which influenza vaccination is indicated [18]; number of comorbidities was a sum of comorbidities (none, 1 comorbidity, 2 or more comorbidities); functional status data was collected from 2013–2017, with full functional status (not restricted) defined as patients who were “fully active, able to carry on all predisease performance without restriction” with increasing disablement to mild restriction, self-caring, limited, and bed-bound [19].
Statistical Analysis
Analyses were conducted using Stata version 14 software (StataCorp, College Station, Texas). Demographic characteristics were compared using the χ 2 test.
A propensity score representing the probability of receiving influenza vaccination was calculated to adjust for potential confounding. Propensity scores were calculated by multivariable logistic regression of influenza vaccination onto the clinical covariates of control survivors (group 3). Covariates are shown in Table 1. Group 3 was chosen as the control group to provide an estimate of influenza vaccination status in the underlying source population. Patients were weighted using inverse probability weighting (IPW) to create exchangeable groups [20]. Propensity scores were asymmetrically trimmed to ensure vaccinated patients had corresponding unvaccinated patients at similar values of the propensity score (and vice versa). Covariate balance was assessed using standardized prevalence difference (SPD), where a SPD > 0.1 or < −0.1 was indicative of imbalance. Further details on model construction are available in the Supplementary Materials.
Characteristics of Patients Included in the Primary Analysis
| Covariate | Case | Control | P Valuea | ||
|---|---|---|---|---|---|
| Survivors (n = 9104) | Deaths (n = 194) | Survivors (n = 6315) | Deaths (n = 136) | ||
| Influenza vaccinated | 4143 (45.5) | 125 (64.4) | 3579 (56.7) | 105 (77.2) | < .001 |
| Male sex | 4511 (49.5) | 114 (58.8) | 3174 (50.3) | 81 (59.6) | .008 |
| Age group, y | < .001 | ||||
| 0.5–15 y | 1745 (19.2) | 1 (0.5) | 740 (11.7) | 0 (0) | |
| 16–49 y | 1964 (21.6) | 19 (9.8) | 1506 (23.8) | 5 (3.7) | |
| 50–64 y | 1479 (16.2) | 25 (12.9) | 1236 (19.6) | 21 (15.4) | |
| 65–79 y | 2086 (22.9) | 61 (31.4) | 1602 (25.4) | 43 (31.6) | |
| ≥ 80 y | 1830 (20.1) | 88 (45.4) | 1231 (19.5) | 67 (49.3) | |
| Indigenous Australian | 583 (6.4) | 8 (4.1) | 515 (8.2) | 5 (3.7) | < .001 |
| Smoking | 1161 (12.8) | 18 (9.3) | 995 (15.8) | 12 (8.8) | < .001 |
| Obesityb | 2151/8210 (26.2) | 20/191 (10.5) | 1287/5743 (22.4) | 14/133 (10.5) | < .001 |
| Pregnant | 223 (2.4) | 0 (0) | 41 (0.6) | 0 (0) | < .001 |
| Homeless | 31 (0.3) | 2 (1.0) | 11 (0.2) | 0 (0) | .045 |
| Nursing home resident | 486 (5.3) | 29 (14.9) | 375 (5.9) | 26 (19.1) | < .001 |
| No. of comorbidities | < .001 | ||||
| 0 | 2487 (27.3) | 5 (2.6) | 1240 (19.6) | 4 (2.9) | |
| 1 | 2720 (29.9) | 59 (30.4) | 1878 (29.7) | 23 (16.9) | |
| ≥ 2 | 3897 (42.8) | 130 (67.0) | 3197 (50.6) | 109 (80.1) | |
| Diabetes | 1961 (21.5) | 55 (28.4) | 1392 (22.0) | 43 (31.6) | .005 |
| Chronic cardiac disease | 2828 (31.1) | 103 (53.1) | 2114 (33.5) | 82 (60.3) | < .001 |
| Chronic respiratory disease | 3216 (35.3) | 90 (46.4) | 2978 (47.2) | 83 (61.0) | < .001 |
| Chronic neurological disease | 1517 (16.7) | 53 (27.3) | 1045 (16.5) | 38 (27.9) | < .001 |
| Chronic renal disease | 316 (3.5) | 7 (3.6) | 364 (5.8) | 8 (5.9) | < .001 |
| Liver disease | 391 (4.3) | 13 (6.7) | 326 (5.2) | 11 (8.1) | .009 |
| Malignancy | 743 (8.2) | 30 (15.5) | 495 (7.8) | 25 (18.4) | < .001 |
| Immunosuppression | 1640 (18.0) | 54 (27.8) | 1436 (22.7) | 47 (34.6) | < .001 |
| Functional statusb | < .001 | ||||
| Not restricted | 3560/6274 (56.7) | 49/164 (29.9) | 2147/4108 (52.3) | 27/105 (25.7) | |
| Mild restriction | 1088/6274 (17.3) | 36/164 (22.0) | 744/4108 (18.1) | 28/105 (26.7) | |
| Self caring | 908/6274 (14.5) | 26/164 (15.9) | 691/4108 (16.8) | 19/105 (18.1) | |
| Limited | 624/6274 (9.9) | 37/164 (22.6) | 447/4108 (10.9) | 24/105 (22.9) | |
| Bed bound | 94/6274 (1.5) | 16/164 (9.8) | 79/4108 (1.9) | 7/105 (6.7) | |
| Study site | < .001 | ||||
| Hospital 1 | 699 (7.7) | 27 (13.9) | 654 (10.4) | 12 (8.8) | |
| Hospital 2 | 403 (4.4) | 4 (2.1) | 406 (6.4) | 3 (2.2) | |
| Hospital 3 | 615 (6.8) | 5 (2.6) | 359 (5.7) | 8 (5.9) | |
| Hospital 4 | 346 (3.8) | 5 (2.6) | 186 (2.9) | 3 (2.2) | |
| Hospital 5 | 739 (8.1) | 0 (0) | 214 (3.4) | 0 (0) | |
| Hospital 6 | 141 (1.5) | 3 (1.5) | 95 (1.5) | 0 (0) | |
| Hospital 7 | 480 (5.3) | 7 (3.6) | 218 (3.5) | 7 (5.1) | |
| Hospital 8 | 313 (3.4) | 21 (10.8) | 145 (2.3) | 5 (3.7) | |
| Hospital 9 | 195 (2.1) | 7 (3.6) | 144 (2.3) | 1 (0.7) | |
| Hospital 10 | 591 (6.5) | 15 (7.7) | 313 (5.0) | 14 (10.3) | |
| Hospital 11 | 198 (2.2) | 5 (2.6) | 176 (2.8) | 3 (2.2) | |
| Hospital 12 | 548 (6.0) | 0 (0) | 485 (7.7) | 0 (0) | |
| Hospital 13 | 1298 (14.3) | 49 (25.3) | 1111 (17.6) | 56 (41.2) | |
| Hospital 14 | 472 (5.2) | 13 (6.7) | 245 (3.9) | 3 (2.2) | |
| Hospital 15 | 705 (7.7) | 11 (5.7) | 723 (11.4) | 9 (6.6) | |
| Hospital 16 | 345 (3.8) | 4 (2.1) | 69 (1.1) | 2 (1.5) | |
| Hospital 17 | 1016 (11.2) | 18 (9.3) | 772 (12.2) | 10 (7.4) | |
| Year | < .001 | ||||
| 2010 | 206 (2.3) | 2 (1.0) | 430 (6.8) | 14 (10.3) | |
| 2011 | 209 (2.3) | 4 (2.1) | 46 (0.7) | 1 (0.7) | |
| 2012 | 988 (10.9) | 19 (9.8) | 1069 (16.9) | 14 (10.3) | |
| 2013 | 602 (6.6) | 6 (3.1) | 433 (6.9) | 8 (5.9) | |
| 2014 | 1330 (14.6) | 20 (10.3) | 1159 (18.4) | 20 (14.7) | |
| 2015 | 1533 (16.8) | 27 (13.9) | 1146 (18.1) | 31 (22.8) | |
| 2016 | 1400 (15.4) | 42 (21.6) | 1176 (18.6) | 26 (19.1) | |
| 2017 | 2836 (31.2) | 74 (38.1) | 856 (13.6) | 22 (16.2) | |
| Month of admission | < .001 | ||||
| April | 148 (1.6) | 5 (2.6) | 188 (3.0) | 1 (0.7) | |
| May | 217 (2.4) | 5 (2.6) | 281 (4.4) | 8 (5.9) | |
| June | 553 (6.1) | 9 (4.6) | 567 (9.0) | 11 (8.1) | |
| July | 1791 (19.7) | 35 (18.0) | 1330 (21.1) | 29 (21.3) | |
| August | 3207 (35.2) | 77 (39.7) | 2038 (32.3) | 51 (37.5) | |
| September | 2481 (27.3) | 49 (25.3) | 1411 (22.3) | 30 (22.1) | |
| October | 691 (7.6) | 14 (7.2) | 486 (7.7) | 6 (4.4) | |
| November | 16 (0.2) | 0 (0) | 14 (0.2) | 0 (0) |
| Covariate | Case | Control | P Valuea | ||
|---|---|---|---|---|---|
| Survivors (n = 9104) | Deaths (n = 194) | Survivors (n = 6315) | Deaths (n = 136) | ||
| Influenza vaccinated | 4143 (45.5) | 125 (64.4) | 3579 (56.7) | 105 (77.2) | < .001 |
| Male sex | 4511 (49.5) | 114 (58.8) | 3174 (50.3) | 81 (59.6) | .008 |
| Age group, y | < .001 | ||||
| 0.5–15 y | 1745 (19.2) | 1 (0.5) | 740 (11.7) | 0 (0) | |
| 16–49 y | 1964 (21.6) | 19 (9.8) | 1506 (23.8) | 5 (3.7) | |
| 50–64 y | 1479 (16.2) | 25 (12.9) | 1236 (19.6) | 21 (15.4) | |
| 65–79 y | 2086 (22.9) | 61 (31.4) | 1602 (25.4) | 43 (31.6) | |
| ≥ 80 y | 1830 (20.1) | 88 (45.4) | 1231 (19.5) | 67 (49.3) | |
| Indigenous Australian | 583 (6.4) | 8 (4.1) | 515 (8.2) | 5 (3.7) | < .001 |
| Smoking | 1161 (12.8) | 18 (9.3) | 995 (15.8) | 12 (8.8) | < .001 |
| Obesityb | 2151/8210 (26.2) | 20/191 (10.5) | 1287/5743 (22.4) | 14/133 (10.5) | < .001 |
| Pregnant | 223 (2.4) | 0 (0) | 41 (0.6) | 0 (0) | < .001 |
| Homeless | 31 (0.3) | 2 (1.0) | 11 (0.2) | 0 (0) | .045 |
| Nursing home resident | 486 (5.3) | 29 (14.9) | 375 (5.9) | 26 (19.1) | < .001 |
| No. of comorbidities | < .001 | ||||
| 0 | 2487 (27.3) | 5 (2.6) | 1240 (19.6) | 4 (2.9) | |
| 1 | 2720 (29.9) | 59 (30.4) | 1878 (29.7) | 23 (16.9) | |
| ≥ 2 | 3897 (42.8) | 130 (67.0) | 3197 (50.6) | 109 (80.1) | |
| Diabetes | 1961 (21.5) | 55 (28.4) | 1392 (22.0) | 43 (31.6) | .005 |
| Chronic cardiac disease | 2828 (31.1) | 103 (53.1) | 2114 (33.5) | 82 (60.3) | < .001 |
| Chronic respiratory disease | 3216 (35.3) | 90 (46.4) | 2978 (47.2) | 83 (61.0) | < .001 |
| Chronic neurological disease | 1517 (16.7) | 53 (27.3) | 1045 (16.5) | 38 (27.9) | < .001 |
| Chronic renal disease | 316 (3.5) | 7 (3.6) | 364 (5.8) | 8 (5.9) | < .001 |
| Liver disease | 391 (4.3) | 13 (6.7) | 326 (5.2) | 11 (8.1) | .009 |
| Malignancy | 743 (8.2) | 30 (15.5) | 495 (7.8) | 25 (18.4) | < .001 |
| Immunosuppression | 1640 (18.0) | 54 (27.8) | 1436 (22.7) | 47 (34.6) | < .001 |
| Functional statusb | < .001 | ||||
| Not restricted | 3560/6274 (56.7) | 49/164 (29.9) | 2147/4108 (52.3) | 27/105 (25.7) | |
| Mild restriction | 1088/6274 (17.3) | 36/164 (22.0) | 744/4108 (18.1) | 28/105 (26.7) | |
| Self caring | 908/6274 (14.5) | 26/164 (15.9) | 691/4108 (16.8) | 19/105 (18.1) | |
| Limited | 624/6274 (9.9) | 37/164 (22.6) | 447/4108 (10.9) | 24/105 (22.9) | |
| Bed bound | 94/6274 (1.5) | 16/164 (9.8) | 79/4108 (1.9) | 7/105 (6.7) | |
| Study site | < .001 | ||||
| Hospital 1 | 699 (7.7) | 27 (13.9) | 654 (10.4) | 12 (8.8) | |
| Hospital 2 | 403 (4.4) | 4 (2.1) | 406 (6.4) | 3 (2.2) | |
| Hospital 3 | 615 (6.8) | 5 (2.6) | 359 (5.7) | 8 (5.9) | |
| Hospital 4 | 346 (3.8) | 5 (2.6) | 186 (2.9) | 3 (2.2) | |
| Hospital 5 | 739 (8.1) | 0 (0) | 214 (3.4) | 0 (0) | |
| Hospital 6 | 141 (1.5) | 3 (1.5) | 95 (1.5) | 0 (0) | |
| Hospital 7 | 480 (5.3) | 7 (3.6) | 218 (3.5) | 7 (5.1) | |
| Hospital 8 | 313 (3.4) | 21 (10.8) | 145 (2.3) | 5 (3.7) | |
| Hospital 9 | 195 (2.1) | 7 (3.6) | 144 (2.3) | 1 (0.7) | |
| Hospital 10 | 591 (6.5) | 15 (7.7) | 313 (5.0) | 14 (10.3) | |
| Hospital 11 | 198 (2.2) | 5 (2.6) | 176 (2.8) | 3 (2.2) | |
| Hospital 12 | 548 (6.0) | 0 (0) | 485 (7.7) | 0 (0) | |
| Hospital 13 | 1298 (14.3) | 49 (25.3) | 1111 (17.6) | 56 (41.2) | |
| Hospital 14 | 472 (5.2) | 13 (6.7) | 245 (3.9) | 3 (2.2) | |
| Hospital 15 | 705 (7.7) | 11 (5.7) | 723 (11.4) | 9 (6.6) | |
| Hospital 16 | 345 (3.8) | 4 (2.1) | 69 (1.1) | 2 (1.5) | |
| Hospital 17 | 1016 (11.2) | 18 (9.3) | 772 (12.2) | 10 (7.4) | |
| Year | < .001 | ||||
| 2010 | 206 (2.3) | 2 (1.0) | 430 (6.8) | 14 (10.3) | |
| 2011 | 209 (2.3) | 4 (2.1) | 46 (0.7) | 1 (0.7) | |
| 2012 | 988 (10.9) | 19 (9.8) | 1069 (16.9) | 14 (10.3) | |
| 2013 | 602 (6.6) | 6 (3.1) | 433 (6.9) | 8 (5.9) | |
| 2014 | 1330 (14.6) | 20 (10.3) | 1159 (18.4) | 20 (14.7) | |
| 2015 | 1533 (16.8) | 27 (13.9) | 1146 (18.1) | 31 (22.8) | |
| 2016 | 1400 (15.4) | 42 (21.6) | 1176 (18.6) | 26 (19.1) | |
| 2017 | 2836 (31.2) | 74 (38.1) | 856 (13.6) | 22 (16.2) | |
| Month of admission | < .001 | ||||
| April | 148 (1.6) | 5 (2.6) | 188 (3.0) | 1 (0.7) | |
| May | 217 (2.4) | 5 (2.6) | 281 (4.4) | 8 (5.9) | |
| June | 553 (6.1) | 9 (4.6) | 567 (9.0) | 11 (8.1) | |
| July | 1791 (19.7) | 35 (18.0) | 1330 (21.1) | 29 (21.3) | |
| August | 3207 (35.2) | 77 (39.7) | 2038 (32.3) | 51 (37.5) | |
| September | 2481 (27.3) | 49 (25.3) | 1411 (22.3) | 30 (22.1) | |
| October | 691 (7.6) | 14 (7.2) | 486 (7.7) | 6 (4.4) | |
| November | 16 (0.2) | 0 (0) | 14 (0.2) | 0 (0) |
Data are presented as no. (%) unless otherwise indicated.
aχ 2 test, conducted across nonmissing groups.
bCovariate included in a sensitivity analysis only.
Characteristics of Patients Included in the Primary Analysis
| Covariate | Case | Control | P Valuea | ||
|---|---|---|---|---|---|
| Survivors (n = 9104) | Deaths (n = 194) | Survivors (n = 6315) | Deaths (n = 136) | ||
| Influenza vaccinated | 4143 (45.5) | 125 (64.4) | 3579 (56.7) | 105 (77.2) | < .001 |
| Male sex | 4511 (49.5) | 114 (58.8) | 3174 (50.3) | 81 (59.6) | .008 |
| Age group, y | < .001 | ||||
| 0.5–15 y | 1745 (19.2) | 1 (0.5) | 740 (11.7) | 0 (0) | |
| 16–49 y | 1964 (21.6) | 19 (9.8) | 1506 (23.8) | 5 (3.7) | |
| 50–64 y | 1479 (16.2) | 25 (12.9) | 1236 (19.6) | 21 (15.4) | |
| 65–79 y | 2086 (22.9) | 61 (31.4) | 1602 (25.4) | 43 (31.6) | |
| ≥ 80 y | 1830 (20.1) | 88 (45.4) | 1231 (19.5) | 67 (49.3) | |
| Indigenous Australian | 583 (6.4) | 8 (4.1) | 515 (8.2) | 5 (3.7) | < .001 |
| Smoking | 1161 (12.8) | 18 (9.3) | 995 (15.8) | 12 (8.8) | < .001 |
| Obesityb | 2151/8210 (26.2) | 20/191 (10.5) | 1287/5743 (22.4) | 14/133 (10.5) | < .001 |
| Pregnant | 223 (2.4) | 0 (0) | 41 (0.6) | 0 (0) | < .001 |
| Homeless | 31 (0.3) | 2 (1.0) | 11 (0.2) | 0 (0) | .045 |
| Nursing home resident | 486 (5.3) | 29 (14.9) | 375 (5.9) | 26 (19.1) | < .001 |
| No. of comorbidities | < .001 | ||||
| 0 | 2487 (27.3) | 5 (2.6) | 1240 (19.6) | 4 (2.9) | |
| 1 | 2720 (29.9) | 59 (30.4) | 1878 (29.7) | 23 (16.9) | |
| ≥ 2 | 3897 (42.8) | 130 (67.0) | 3197 (50.6) | 109 (80.1) | |
| Diabetes | 1961 (21.5) | 55 (28.4) | 1392 (22.0) | 43 (31.6) | .005 |
| Chronic cardiac disease | 2828 (31.1) | 103 (53.1) | 2114 (33.5) | 82 (60.3) | < .001 |
| Chronic respiratory disease | 3216 (35.3) | 90 (46.4) | 2978 (47.2) | 83 (61.0) | < .001 |
| Chronic neurological disease | 1517 (16.7) | 53 (27.3) | 1045 (16.5) | 38 (27.9) | < .001 |
| Chronic renal disease | 316 (3.5) | 7 (3.6) | 364 (5.8) | 8 (5.9) | < .001 |
| Liver disease | 391 (4.3) | 13 (6.7) | 326 (5.2) | 11 (8.1) | .009 |
| Malignancy | 743 (8.2) | 30 (15.5) | 495 (7.8) | 25 (18.4) | < .001 |
| Immunosuppression | 1640 (18.0) | 54 (27.8) | 1436 (22.7) | 47 (34.6) | < .001 |
| Functional statusb | < .001 | ||||
| Not restricted | 3560/6274 (56.7) | 49/164 (29.9) | 2147/4108 (52.3) | 27/105 (25.7) | |
| Mild restriction | 1088/6274 (17.3) | 36/164 (22.0) | 744/4108 (18.1) | 28/105 (26.7) | |
| Self caring | 908/6274 (14.5) | 26/164 (15.9) | 691/4108 (16.8) | 19/105 (18.1) | |
| Limited | 624/6274 (9.9) | 37/164 (22.6) | 447/4108 (10.9) | 24/105 (22.9) | |
| Bed bound | 94/6274 (1.5) | 16/164 (9.8) | 79/4108 (1.9) | 7/105 (6.7) | |
| Study site | < .001 | ||||
| Hospital 1 | 699 (7.7) | 27 (13.9) | 654 (10.4) | 12 (8.8) | |
| Hospital 2 | 403 (4.4) | 4 (2.1) | 406 (6.4) | 3 (2.2) | |
| Hospital 3 | 615 (6.8) | 5 (2.6) | 359 (5.7) | 8 (5.9) | |
| Hospital 4 | 346 (3.8) | 5 (2.6) | 186 (2.9) | 3 (2.2) | |
| Hospital 5 | 739 (8.1) | 0 (0) | 214 (3.4) | 0 (0) | |
| Hospital 6 | 141 (1.5) | 3 (1.5) | 95 (1.5) | 0 (0) | |
| Hospital 7 | 480 (5.3) | 7 (3.6) | 218 (3.5) | 7 (5.1) | |
| Hospital 8 | 313 (3.4) | 21 (10.8) | 145 (2.3) | 5 (3.7) | |
| Hospital 9 | 195 (2.1) | 7 (3.6) | 144 (2.3) | 1 (0.7) | |
| Hospital 10 | 591 (6.5) | 15 (7.7) | 313 (5.0) | 14 (10.3) | |
| Hospital 11 | 198 (2.2) | 5 (2.6) | 176 (2.8) | 3 (2.2) | |
| Hospital 12 | 548 (6.0) | 0 (0) | 485 (7.7) | 0 (0) | |
| Hospital 13 | 1298 (14.3) | 49 (25.3) | 1111 (17.6) | 56 (41.2) | |
| Hospital 14 | 472 (5.2) | 13 (6.7) | 245 (3.9) | 3 (2.2) | |
| Hospital 15 | 705 (7.7) | 11 (5.7) | 723 (11.4) | 9 (6.6) | |
| Hospital 16 | 345 (3.8) | 4 (2.1) | 69 (1.1) | 2 (1.5) | |
| Hospital 17 | 1016 (11.2) | 18 (9.3) | 772 (12.2) | 10 (7.4) | |
| Year | < .001 | ||||
| 2010 | 206 (2.3) | 2 (1.0) | 430 (6.8) | 14 (10.3) | |
| 2011 | 209 (2.3) | 4 (2.1) | 46 (0.7) | 1 (0.7) | |
| 2012 | 988 (10.9) | 19 (9.8) | 1069 (16.9) | 14 (10.3) | |
| 2013 | 602 (6.6) | 6 (3.1) | 433 (6.9) | 8 (5.9) | |
| 2014 | 1330 (14.6) | 20 (10.3) | 1159 (18.4) | 20 (14.7) | |
| 2015 | 1533 (16.8) | 27 (13.9) | 1146 (18.1) | 31 (22.8) | |
| 2016 | 1400 (15.4) | 42 (21.6) | 1176 (18.6) | 26 (19.1) | |
| 2017 | 2836 (31.2) | 74 (38.1) | 856 (13.6) | 22 (16.2) | |
| Month of admission | < .001 | ||||
| April | 148 (1.6) | 5 (2.6) | 188 (3.0) | 1 (0.7) | |
| May | 217 (2.4) | 5 (2.6) | 281 (4.4) | 8 (5.9) | |
| June | 553 (6.1) | 9 (4.6) | 567 (9.0) | 11 (8.1) | |
| July | 1791 (19.7) | 35 (18.0) | 1330 (21.1) | 29 (21.3) | |
| August | 3207 (35.2) | 77 (39.7) | 2038 (32.3) | 51 (37.5) | |
| September | 2481 (27.3) | 49 (25.3) | 1411 (22.3) | 30 (22.1) | |
| October | 691 (7.6) | 14 (7.2) | 486 (7.7) | 6 (4.4) | |
| November | 16 (0.2) | 0 (0) | 14 (0.2) | 0 (0) |
| Covariate | Case | Control | P Valuea | ||
|---|---|---|---|---|---|
| Survivors (n = 9104) | Deaths (n = 194) | Survivors (n = 6315) | Deaths (n = 136) | ||
| Influenza vaccinated | 4143 (45.5) | 125 (64.4) | 3579 (56.7) | 105 (77.2) | < .001 |
| Male sex | 4511 (49.5) | 114 (58.8) | 3174 (50.3) | 81 (59.6) | .008 |
| Age group, y | < .001 | ||||
| 0.5–15 y | 1745 (19.2) | 1 (0.5) | 740 (11.7) | 0 (0) | |
| 16–49 y | 1964 (21.6) | 19 (9.8) | 1506 (23.8) | 5 (3.7) | |
| 50–64 y | 1479 (16.2) | 25 (12.9) | 1236 (19.6) | 21 (15.4) | |
| 65–79 y | 2086 (22.9) | 61 (31.4) | 1602 (25.4) | 43 (31.6) | |
| ≥ 80 y | 1830 (20.1) | 88 (45.4) | 1231 (19.5) | 67 (49.3) | |
| Indigenous Australian | 583 (6.4) | 8 (4.1) | 515 (8.2) | 5 (3.7) | < .001 |
| Smoking | 1161 (12.8) | 18 (9.3) | 995 (15.8) | 12 (8.8) | < .001 |
| Obesityb | 2151/8210 (26.2) | 20/191 (10.5) | 1287/5743 (22.4) | 14/133 (10.5) | < .001 |
| Pregnant | 223 (2.4) | 0 (0) | 41 (0.6) | 0 (0) | < .001 |
| Homeless | 31 (0.3) | 2 (1.0) | 11 (0.2) | 0 (0) | .045 |
| Nursing home resident | 486 (5.3) | 29 (14.9) | 375 (5.9) | 26 (19.1) | < .001 |
| No. of comorbidities | < .001 | ||||
| 0 | 2487 (27.3) | 5 (2.6) | 1240 (19.6) | 4 (2.9) | |
| 1 | 2720 (29.9) | 59 (30.4) | 1878 (29.7) | 23 (16.9) | |
| ≥ 2 | 3897 (42.8) | 130 (67.0) | 3197 (50.6) | 109 (80.1) | |
| Diabetes | 1961 (21.5) | 55 (28.4) | 1392 (22.0) | 43 (31.6) | .005 |
| Chronic cardiac disease | 2828 (31.1) | 103 (53.1) | 2114 (33.5) | 82 (60.3) | < .001 |
| Chronic respiratory disease | 3216 (35.3) | 90 (46.4) | 2978 (47.2) | 83 (61.0) | < .001 |
| Chronic neurological disease | 1517 (16.7) | 53 (27.3) | 1045 (16.5) | 38 (27.9) | < .001 |
| Chronic renal disease | 316 (3.5) | 7 (3.6) | 364 (5.8) | 8 (5.9) | < .001 |
| Liver disease | 391 (4.3) | 13 (6.7) | 326 (5.2) | 11 (8.1) | .009 |
| Malignancy | 743 (8.2) | 30 (15.5) | 495 (7.8) | 25 (18.4) | < .001 |
| Immunosuppression | 1640 (18.0) | 54 (27.8) | 1436 (22.7) | 47 (34.6) | < .001 |
| Functional statusb | < .001 | ||||
| Not restricted | 3560/6274 (56.7) | 49/164 (29.9) | 2147/4108 (52.3) | 27/105 (25.7) | |
| Mild restriction | 1088/6274 (17.3) | 36/164 (22.0) | 744/4108 (18.1) | 28/105 (26.7) | |
| Self caring | 908/6274 (14.5) | 26/164 (15.9) | 691/4108 (16.8) | 19/105 (18.1) | |
| Limited | 624/6274 (9.9) | 37/164 (22.6) | 447/4108 (10.9) | 24/105 (22.9) | |
| Bed bound | 94/6274 (1.5) | 16/164 (9.8) | 79/4108 (1.9) | 7/105 (6.7) | |
| Study site | < .001 | ||||
| Hospital 1 | 699 (7.7) | 27 (13.9) | 654 (10.4) | 12 (8.8) | |
| Hospital 2 | 403 (4.4) | 4 (2.1) | 406 (6.4) | 3 (2.2) | |
| Hospital 3 | 615 (6.8) | 5 (2.6) | 359 (5.7) | 8 (5.9) | |
| Hospital 4 | 346 (3.8) | 5 (2.6) | 186 (2.9) | 3 (2.2) | |
| Hospital 5 | 739 (8.1) | 0 (0) | 214 (3.4) | 0 (0) | |
| Hospital 6 | 141 (1.5) | 3 (1.5) | 95 (1.5) | 0 (0) | |
| Hospital 7 | 480 (5.3) | 7 (3.6) | 218 (3.5) | 7 (5.1) | |
| Hospital 8 | 313 (3.4) | 21 (10.8) | 145 (2.3) | 5 (3.7) | |
| Hospital 9 | 195 (2.1) | 7 (3.6) | 144 (2.3) | 1 (0.7) | |
| Hospital 10 | 591 (6.5) | 15 (7.7) | 313 (5.0) | 14 (10.3) | |
| Hospital 11 | 198 (2.2) | 5 (2.6) | 176 (2.8) | 3 (2.2) | |
| Hospital 12 | 548 (6.0) | 0 (0) | 485 (7.7) | 0 (0) | |
| Hospital 13 | 1298 (14.3) | 49 (25.3) | 1111 (17.6) | 56 (41.2) | |
| Hospital 14 | 472 (5.2) | 13 (6.7) | 245 (3.9) | 3 (2.2) | |
| Hospital 15 | 705 (7.7) | 11 (5.7) | 723 (11.4) | 9 (6.6) | |
| Hospital 16 | 345 (3.8) | 4 (2.1) | 69 (1.1) | 2 (1.5) | |
| Hospital 17 | 1016 (11.2) | 18 (9.3) | 772 (12.2) | 10 (7.4) | |
| Year | < .001 | ||||
| 2010 | 206 (2.3) | 2 (1.0) | 430 (6.8) | 14 (10.3) | |
| 2011 | 209 (2.3) | 4 (2.1) | 46 (0.7) | 1 (0.7) | |
| 2012 | 988 (10.9) | 19 (9.8) | 1069 (16.9) | 14 (10.3) | |
| 2013 | 602 (6.6) | 6 (3.1) | 433 (6.9) | 8 (5.9) | |
| 2014 | 1330 (14.6) | 20 (10.3) | 1159 (18.4) | 20 (14.7) | |
| 2015 | 1533 (16.8) | 27 (13.9) | 1146 (18.1) | 31 (22.8) | |
| 2016 | 1400 (15.4) | 42 (21.6) | 1176 (18.6) | 26 (19.1) | |
| 2017 | 2836 (31.2) | 74 (38.1) | 856 (13.6) | 22 (16.2) | |
| Month of admission | < .001 | ||||
| April | 148 (1.6) | 5 (2.6) | 188 (3.0) | 1 (0.7) | |
| May | 217 (2.4) | 5 (2.6) | 281 (4.4) | 8 (5.9) | |
| June | 553 (6.1) | 9 (4.6) | 567 (9.0) | 11 (8.1) | |
| July | 1791 (19.7) | 35 (18.0) | 1330 (21.1) | 29 (21.3) | |
| August | 3207 (35.2) | 77 (39.7) | 2038 (32.3) | 51 (37.5) | |
| September | 2481 (27.3) | 49 (25.3) | 1411 (22.3) | 30 (22.1) | |
| October | 691 (7.6) | 14 (7.2) | 486 (7.7) | 6 (4.4) | |
| November | 16 (0.2) | 0 (0) | 14 (0.2) | 0 (0) |
Data are presented as no. (%) unless otherwise indicated.
aχ 2 test, conducted across nonmissing groups.
bCovariate included in a sensitivity analysis only.
An adjusted odds ratio (aOR) of vaccination was determined by logistic regression of the outcome (influenza-related mortality or noninfluenza mortality) onto influenza vaccination status, weighted by the IPWs from the propensity score. To determine if influenza vaccination protects against influenza-related mortality, the odds of vaccination in case deaths (group 2) was compared to control survivors (group 3) under the assumption that vaccination does not effect control admissions. IVE against influenza-related mortality was estimated from the aOR, with IVE = 1 – aOR × 100%. To assess residual confounding, the odds of vaccination in control deaths were compared to control survivors (group 4 vs group 3), as influenza vaccination should not affect death due to noninfluenza causes. To determine if influenza vaccination reduces the severity of influenza illness, the odds of vaccination in case deaths were compared to case survivors (group 2 vs group 1).
Subgroup analyses were conducted in the National Immunisation Program target group (the elderly, pregnant women, Indigenous Australians, and individuals with comorbidities), patients aged ≥ 50 years (the age group with the most mortality), and patients aged ≥ 65 years (an age group with high mortality and residual confounding). To assess whether IVE against influenza-related mortality varied by study year, an interaction parameter between year and influenza vaccination was included in the final outcome regression.
Several sensitivity analyses were conducted. First, we conducted a bias-corrected analysis using a ratio-of-ratios approach to estimate IVE against influenza-related mortality in the absence of residual confounding. To do this, we constructed a multinominal model with influenza vaccination as the exposure for both noninfluenza mortality and influenza-related mortality compared to noninfluenza survivors (control; group 3), with patients weighted using IPW. To obtain a bias-corrected coefficient, the difference between the log OR of vaccination for influenza-related mortality compared to control and the log OR of vaccination for the non-influenza mortality compared to control was calculated. IVE against influenza-related mortality was calculated as described for the primary analysis. Second, a multiple imputation procedure was performed using logistic regression models for missing vaccination status, and known vaccination status based on clinical covariates (including obesity and functional status) as well as influenza diagnosis. Fifty datasets were imputed and vaccine effectiveness was estimated as above [21]. Third, covariates with substantial missing data (obesity and functional status), excluded in the primary analysis, were included in a sensitivity analysis. Fourth, we included prior season influenza vaccination as a covariate. Fifth, nursing home residents were excluded as they may be protected from influenza infection by vaccination of nursing home staff, regardless of their own vaccination status [22]. Sixth, patients with influenza illness onset > 7 days were excluded to prevent misclassification of cases as test-negative controls. Seventh, we included all test-negative patients (control survivors and deaths; groups 3 and 4) as the control group. Last, we used an alternative propensity score methodology that stratified on the year of presentation and the quintile of propensity score using a conditional logistic regression. Sensitivity analyses were conducted as described for the primary analysis unless otherwise specified.
RESULTS
From 2010 to 2017, 14 038 patients with laboratory-confirmed influenza were admitted to the FluCAN study hospitals, and matched with a subset of test-negative controls (n = 10 223) (Supplementary Table 2). Of these patients, 4740 cases and 3772 controls were excluded from the primary analysis due to being collected outside of the April–November surveillance period (n = 47 cases, n = 37 controls), being aged < 6 months and therefore ineligible for influenza vaccination (n = 457 cases, n = 298 controls), having missing data (n = 4228 cases, n = 3431 controls), or having propensity scores for which there was no corresponding unvaccinated (or vaccinated) patient and were therefore trimmed from analysis (n = 8 cases, n = 6 controls) (Figure 1).
The primary analysis included 9298 patients with laboratory-confirmed influenza and 6451 test-negative controls, with 194 cases and 136 controls dying during hospitalization (Table 1). Of the cases who died, 64.4% were vaccinated, 76.8% were ≥ 65 years old, none were pregnant women, 4.1% were Indigenous Australians, and 97.4% had comorbidities.
Age group, smoking, number of comorbidities, chronic respiratory disease, chronic renal disease, study site, month of admission, and year were associated with propensity to be vaccinated (Supplementary Table 3). The model had good discrimination and calibration, and all covariates were well balanced except for month of admission (Supplementary Figure 1).
In the primary analysis, influenza vaccination reduced influenza-related mortality with an estimated IVE of 31% (95% confidence interval [CI], 3%–51%; P = .033; Table 2). Similar proportions of patients with noninfluenza mortality had influenza vaccination, compared to noninfluenza survivors; in this analysis the aOR of vaccination was 1.31 (95% CI, .84–2.02; P = .232). The proportion of patients with influenza-related mortality who received influenza vaccination compared to influenza survivors was similar, suggesting that the severity of influenza illness was not attenuated by vaccination, with an aOR of 1.07 (95% CI, .76–1.50; P = .713).
Adjusted Odds Ratios of Vaccination and Influenza Vaccine Effectiveness in the Primary Analysis and Subgroups
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Primary analysis model | ||||
| Influenza-related mortalitya | 6495 | 0.69 (.49–.97) | 31% (3%–51%) | .033 |
| Noninfluenza mortalityb | 6437 | 1.31 (.84–2.02) | … | .232 |
| Severity of influenza illnessc | 9282 | 1.07 (.76–1.50) | … | .713 |
| Subgroups | ||||
| National immunisation program target groupd | ||||
| Influenza-related mortality | 5648 | 0.68 (.48–.97) | 32% (3%–52%) | .031 |
| Noninfluenza mortality | 5591 | 1.28 (.83–1.99) | … | .265 |
| Severity of influenza illness | 7625 | 0.97 (.69–1.37) | … | .868 |
| Patients aged ≥ 50 y | ||||
| Influenza-related mortality | 4240 | 0.76 (.52–1.11) | 24% (−11% to 48%) | .162 |
| Noninfluenza mortality | 4197 | 1.45 (.91–2.31) | … | .121 |
| Severity of influenza illness | 5564 | 1.15 (.79–1.67) | … | .463 |
| Patients aged ≥ 65 y (elderly) | ||||
| Influenza-related mortality | 2982 | 0.83 (.54–1.28) | 17% (−28% to 46%) | .405 |
| Noninfluenza mortality | 2943 | 1.92 (1.06–3.46) | … | .031 |
| Severity of influenza illness | 4063 | 1.15 (.75–1.74) | … | .525 |
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Primary analysis model | ||||
| Influenza-related mortalitya | 6495 | 0.69 (.49–.97) | 31% (3%–51%) | .033 |
| Noninfluenza mortalityb | 6437 | 1.31 (.84–2.02) | … | .232 |
| Severity of influenza illnessc | 9282 | 1.07 (.76–1.50) | … | .713 |
| Subgroups | ||||
| National immunisation program target groupd | ||||
| Influenza-related mortality | 5648 | 0.68 (.48–.97) | 32% (3%–52%) | .031 |
| Noninfluenza mortality | 5591 | 1.28 (.83–1.99) | … | .265 |
| Severity of influenza illness | 7625 | 0.97 (.69–1.37) | … | .868 |
| Patients aged ≥ 50 y | ||||
| Influenza-related mortality | 4240 | 0.76 (.52–1.11) | 24% (−11% to 48%) | .162 |
| Noninfluenza mortality | 4197 | 1.45 (.91–2.31) | … | .121 |
| Severity of influenza illness | 5564 | 1.15 (.79–1.67) | … | .463 |
| Patients aged ≥ 65 y (elderly) | ||||
| Influenza-related mortality | 2982 | 0.83 (.54–1.28) | 17% (−28% to 46%) | .405 |
| Noninfluenza mortality | 2943 | 1.92 (1.06–3.46) | … | .031 |
| Severity of influenza illness | 4063 | 1.15 (.75–1.74) | … | .525 |
Abbreviations: CI, confidence interval; IVE, influenza vaccine effectiveness.
aCase deaths vs control survivors.
bControl deaths vs control survivors to detect residual confounding.
cCase deaths vs case survivors to assess attenuation of illness severity.
dTarget group includes patients ≥65 years old, pregnant women, Indigenous Australians, and individuals with comorbidities.
Adjusted Odds Ratios of Vaccination and Influenza Vaccine Effectiveness in the Primary Analysis and Subgroups
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Primary analysis model | ||||
| Influenza-related mortalitya | 6495 | 0.69 (.49–.97) | 31% (3%–51%) | .033 |
| Noninfluenza mortalityb | 6437 | 1.31 (.84–2.02) | … | .232 |
| Severity of influenza illnessc | 9282 | 1.07 (.76–1.50) | … | .713 |
| Subgroups | ||||
| National immunisation program target groupd | ||||
| Influenza-related mortality | 5648 | 0.68 (.48–.97) | 32% (3%–52%) | .031 |
| Noninfluenza mortality | 5591 | 1.28 (.83–1.99) | … | .265 |
| Severity of influenza illness | 7625 | 0.97 (.69–1.37) | … | .868 |
| Patients aged ≥ 50 y | ||||
| Influenza-related mortality | 4240 | 0.76 (.52–1.11) | 24% (−11% to 48%) | .162 |
| Noninfluenza mortality | 4197 | 1.45 (.91–2.31) | … | .121 |
| Severity of influenza illness | 5564 | 1.15 (.79–1.67) | … | .463 |
| Patients aged ≥ 65 y (elderly) | ||||
| Influenza-related mortality | 2982 | 0.83 (.54–1.28) | 17% (−28% to 46%) | .405 |
| Noninfluenza mortality | 2943 | 1.92 (1.06–3.46) | … | .031 |
| Severity of influenza illness | 4063 | 1.15 (.75–1.74) | … | .525 |
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Primary analysis model | ||||
| Influenza-related mortalitya | 6495 | 0.69 (.49–.97) | 31% (3%–51%) | .033 |
| Noninfluenza mortalityb | 6437 | 1.31 (.84–2.02) | … | .232 |
| Severity of influenza illnessc | 9282 | 1.07 (.76–1.50) | … | .713 |
| Subgroups | ||||
| National immunisation program target groupd | ||||
| Influenza-related mortality | 5648 | 0.68 (.48–.97) | 32% (3%–52%) | .031 |
| Noninfluenza mortality | 5591 | 1.28 (.83–1.99) | … | .265 |
| Severity of influenza illness | 7625 | 0.97 (.69–1.37) | … | .868 |
| Patients aged ≥ 50 y | ||||
| Influenza-related mortality | 4240 | 0.76 (.52–1.11) | 24% (−11% to 48%) | .162 |
| Noninfluenza mortality | 4197 | 1.45 (.91–2.31) | … | .121 |
| Severity of influenza illness | 5564 | 1.15 (.79–1.67) | … | .463 |
| Patients aged ≥ 65 y (elderly) | ||||
| Influenza-related mortality | 2982 | 0.83 (.54–1.28) | 17% (−28% to 46%) | .405 |
| Noninfluenza mortality | 2943 | 1.92 (1.06–3.46) | … | .031 |
| Severity of influenza illness | 4063 | 1.15 (.75–1.74) | … | .525 |
Abbreviations: CI, confidence interval; IVE, influenza vaccine effectiveness.
aCase deaths vs control survivors.
bControl deaths vs control survivors to detect residual confounding.
cCase deaths vs case survivors to assess attenuation of illness severity.
dTarget group includes patients ≥65 years old, pregnant women, Indigenous Australians, and individuals with comorbidities.
Subgroup Analyses
The vaccine target group comprised 84.1% (13239/15749) of the patients analyzed. In this group, the estimates of vaccine effectiveness against influenza-related mortality, residual confounding, and attenuation of illness severity were similar to the primary analysis (Table 2). Patients aged ≥ 50 years old comprised 62.0% (9769/15749) of all patients analyzed and accounted for the majority of all mortality (305/330 [92.4%]). In patients aged ≥ 50 years, the point estimate for IVE against influenza-related mortality was lower than in the primary analysis (24% vs 31%; Table 2). Conversely, the aOR for influenza vaccination was further from the null than in the primary analysis (1.45 vs 1.31), suggesting increased residual confounding. Patients aged ≥ 65 years comprised 44.5% (7008/15749) of all patients analyzed and accounted for 78.5% (259/330) of all mortality. In patients aged ≥65 years, the point estimate for IVE against influenza-related mortality was lower than in the subgroup analysis in patients aged ≥ 50 years, and in the primary analysis (17% vs 24% vs 31%, respectively; Table 2). In patients aged ≥ 65 years, there was no evidence of IVE against influenza-related mortality or severity of influenza illness (P > .05). However, there was evidence of residual confounding, with an aOR of 1.92 (95% CI, 1.06–3.46; P = .031).
Impact of Vaccination by Year
The aORs of vaccination by year are shown in Figure 2. As 2010–2013 had < 20 influenza-related deaths, they were omitted from the analysis.
Adjusted odds ratios of vaccination (with 95% confidence intervals) in cases with influenza-related mortality compared to controls who survived, by study year. Years 2010–2013 are not displayed due to imprecision in the estimates (< 20 deaths per year).
Sensitivity Analyses
The sensitivity analyses are shown in Table 3. The bias-corrected analysis gave higher IVE against influenza-related mortality estimates of 47% (95% CI, 9%–70%; P = .022). Restriction to patients with an onset of illness ≤ 7 days increased the residual confounding, but had little effect on the IVE against influenza-related mortality or severity analyses. The other sensitivity analyses were broadly similar to the primary analysis.
Adjusted Odds Ratios of Vaccination and Influenza Vaccine Effectiveness in the Sensitivity Analyses
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Bias-corrected analysis | ||||
| Influenza-related mortalitya | 6645 | 0.53 (.30–.91) | 47% (9%–70%) | .022 |
| Multiple imputation of influenza vaccination status | ||||
| Influenza-related mortality | 7442 | 0.66 (.48–.90) | 34% (10%–52%) | .008 |
| Inclusion of obesity and functional status | ||||
| Influenza-related mortality | 4227 | 0.66 (.45–.96) | 34% (4%–55%) | .029 |
| Noninfluenza mortalityb | 4174 | 1.49 (.86–2.56) | … | .151 |
| Severity of influenza illnessc | 6084 | 0.92 (.63–1.34) | … | .647 |
| Adjustment for prior season vaccination | ||||
| Influenza-related mortality | 5583 | 0.70 (.45–1.11) | 30% (−11% to 55%) | .128 |
| Noninfluenza mortality | 5442 | 1.19 (.71–2.00) | … | .514 |
| Severity of influenza illness | 7584 | 0.94 (.59–1.51) | … | .807 |
| Exclusion of nursing home residents | ||||
| Influenza-related mortality | 6092 | 0.62 (.42–.90) | 38% (10%–58%) | .012 |
| Noninfluenza mortality | 6037 | 1.18 (.74–1.89) | … | .489 |
| Severity of influenza illness | 8767 | 1.01 (.70–1.47) | … | .945 |
| Exclusion of patients swabbed > 7 d after onset of illness | ||||
| Influenza-related mortality | 4864 | 0.73 (.49–1.10) | 27% (−10% to 51%) | .136 |
| Noninfluenza mortality | 4820 | 1.67 (.94–2.97) | … | .078 |
| Severity of influenza illness | 7076 | 1.16 (.77–1.73) | … | .476 |
| All test-negative patients included in PS | ||||
| Influenza-related mortality | 6633 | 0.68 (.48–.96) | 32% (4%–52%) | .028 |
| Stratified by year and quintile of PS | ||||
| Influenza-related mortality | 5405 | 0.72 (.52–1.00) | 28% (0%–48%) | .047 |
| Noninfluenza mortality | 4615 | 1.07 (.70–1.65) | … | .754 |
| Severity of influenza illness | 7413 | 0.96 (.69–1.33) | … | .799 |
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Bias-corrected analysis | ||||
| Influenza-related mortalitya | 6645 | 0.53 (.30–.91) | 47% (9%–70%) | .022 |
| Multiple imputation of influenza vaccination status | ||||
| Influenza-related mortality | 7442 | 0.66 (.48–.90) | 34% (10%–52%) | .008 |
| Inclusion of obesity and functional status | ||||
| Influenza-related mortality | 4227 | 0.66 (.45–.96) | 34% (4%–55%) | .029 |
| Noninfluenza mortalityb | 4174 | 1.49 (.86–2.56) | … | .151 |
| Severity of influenza illnessc | 6084 | 0.92 (.63–1.34) | … | .647 |
| Adjustment for prior season vaccination | ||||
| Influenza-related mortality | 5583 | 0.70 (.45–1.11) | 30% (−11% to 55%) | .128 |
| Noninfluenza mortality | 5442 | 1.19 (.71–2.00) | … | .514 |
| Severity of influenza illness | 7584 | 0.94 (.59–1.51) | … | .807 |
| Exclusion of nursing home residents | ||||
| Influenza-related mortality | 6092 | 0.62 (.42–.90) | 38% (10%–58%) | .012 |
| Noninfluenza mortality | 6037 | 1.18 (.74–1.89) | … | .489 |
| Severity of influenza illness | 8767 | 1.01 (.70–1.47) | … | .945 |
| Exclusion of patients swabbed > 7 d after onset of illness | ||||
| Influenza-related mortality | 4864 | 0.73 (.49–1.10) | 27% (−10% to 51%) | .136 |
| Noninfluenza mortality | 4820 | 1.67 (.94–2.97) | … | .078 |
| Severity of influenza illness | 7076 | 1.16 (.77–1.73) | … | .476 |
| All test-negative patients included in PS | ||||
| Influenza-related mortality | 6633 | 0.68 (.48–.96) | 32% (4%–52%) | .028 |
| Stratified by year and quintile of PS | ||||
| Influenza-related mortality | 5405 | 0.72 (.52–1.00) | 28% (0%–48%) | .047 |
| Noninfluenza mortality | 4615 | 1.07 (.70–1.65) | … | .754 |
| Severity of influenza illness | 7413 | 0.96 (.69–1.33) | … | .799 |
Abbreviations: CI, confidence interval; IVE, influenza vaccine effectiveness; PS, propensity score.
aCase deaths vs control survivors.
bControl deaths vs control survivors to detect residual confounding.
cCase deaths vs case survivors to assess attenuation of illness severity.
Adjusted Odds Ratios of Vaccination and Influenza Vaccine Effectiveness in the Sensitivity Analyses
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Bias-corrected analysis | ||||
| Influenza-related mortalitya | 6645 | 0.53 (.30–.91) | 47% (9%–70%) | .022 |
| Multiple imputation of influenza vaccination status | ||||
| Influenza-related mortality | 7442 | 0.66 (.48–.90) | 34% (10%–52%) | .008 |
| Inclusion of obesity and functional status | ||||
| Influenza-related mortality | 4227 | 0.66 (.45–.96) | 34% (4%–55%) | .029 |
| Noninfluenza mortalityb | 4174 | 1.49 (.86–2.56) | … | .151 |
| Severity of influenza illnessc | 6084 | 0.92 (.63–1.34) | … | .647 |
| Adjustment for prior season vaccination | ||||
| Influenza-related mortality | 5583 | 0.70 (.45–1.11) | 30% (−11% to 55%) | .128 |
| Noninfluenza mortality | 5442 | 1.19 (.71–2.00) | … | .514 |
| Severity of influenza illness | 7584 | 0.94 (.59–1.51) | … | .807 |
| Exclusion of nursing home residents | ||||
| Influenza-related mortality | 6092 | 0.62 (.42–.90) | 38% (10%–58%) | .012 |
| Noninfluenza mortality | 6037 | 1.18 (.74–1.89) | … | .489 |
| Severity of influenza illness | 8767 | 1.01 (.70–1.47) | … | .945 |
| Exclusion of patients swabbed > 7 d after onset of illness | ||||
| Influenza-related mortality | 4864 | 0.73 (.49–1.10) | 27% (−10% to 51%) | .136 |
| Noninfluenza mortality | 4820 | 1.67 (.94–2.97) | … | .078 |
| Severity of influenza illness | 7076 | 1.16 (.77–1.73) | … | .476 |
| All test-negative patients included in PS | ||||
| Influenza-related mortality | 6633 | 0.68 (.48–.96) | 32% (4%–52%) | .028 |
| Stratified by year and quintile of PS | ||||
| Influenza-related mortality | 5405 | 0.72 (.52–1.00) | 28% (0%–48%) | .047 |
| Noninfluenza mortality | 4615 | 1.07 (.70–1.65) | … | .754 |
| Severity of influenza illness | 7413 | 0.96 (.69–1.33) | … | .799 |
| Analysis and Outcome | No. | Adjusted Odds Ratio (95% CI) | Estimated IVE (95% CI) | P Value |
|---|---|---|---|---|
| Bias-corrected analysis | ||||
| Influenza-related mortalitya | 6645 | 0.53 (.30–.91) | 47% (9%–70%) | .022 |
| Multiple imputation of influenza vaccination status | ||||
| Influenza-related mortality | 7442 | 0.66 (.48–.90) | 34% (10%–52%) | .008 |
| Inclusion of obesity and functional status | ||||
| Influenza-related mortality | 4227 | 0.66 (.45–.96) | 34% (4%–55%) | .029 |
| Noninfluenza mortalityb | 4174 | 1.49 (.86–2.56) | … | .151 |
| Severity of influenza illnessc | 6084 | 0.92 (.63–1.34) | … | .647 |
| Adjustment for prior season vaccination | ||||
| Influenza-related mortality | 5583 | 0.70 (.45–1.11) | 30% (−11% to 55%) | .128 |
| Noninfluenza mortality | 5442 | 1.19 (.71–2.00) | … | .514 |
| Severity of influenza illness | 7584 | 0.94 (.59–1.51) | … | .807 |
| Exclusion of nursing home residents | ||||
| Influenza-related mortality | 6092 | 0.62 (.42–.90) | 38% (10%–58%) | .012 |
| Noninfluenza mortality | 6037 | 1.18 (.74–1.89) | … | .489 |
| Severity of influenza illness | 8767 | 1.01 (.70–1.47) | … | .945 |
| Exclusion of patients swabbed > 7 d after onset of illness | ||||
| Influenza-related mortality | 4864 | 0.73 (.49–1.10) | 27% (−10% to 51%) | .136 |
| Noninfluenza mortality | 4820 | 1.67 (.94–2.97) | … | .078 |
| Severity of influenza illness | 7076 | 1.16 (.77–1.73) | … | .476 |
| All test-negative patients included in PS | ||||
| Influenza-related mortality | 6633 | 0.68 (.48–.96) | 32% (4%–52%) | .028 |
| Stratified by year and quintile of PS | ||||
| Influenza-related mortality | 5405 | 0.72 (.52–1.00) | 28% (0%–48%) | .047 |
| Noninfluenza mortality | 4615 | 1.07 (.70–1.65) | … | .754 |
| Severity of influenza illness | 7413 | 0.96 (.69–1.33) | … | .799 |
Abbreviations: CI, confidence interval; IVE, influenza vaccine effectiveness; PS, propensity score.
aCase deaths vs control survivors.
bControl deaths vs control survivors to detect residual confounding.
cCase deaths vs case survivors to assess attenuation of illness severity.
DISCUSSION
In this case-control study, we examined the effectiveness of influenza vaccination on influenza-related hospital mortality in Australian patients. We found that influenza vaccination reduced the risk of influenza-related mortality by 31% (95% CI, 3%–51%), including in the National Immunisation Program target group. This reduction demonstrates the utility of influenza vaccination in not only preventing medically attended influenza infection and influenza-related hospitalization [23, 24], but also in preventing death. These findings are in line with a study in a French elderly population that found IVE against influenza-attributable deaths was 35% (95% CI, 6%–55%) using routinely collected data [9]. Other studies using laboratory-confirmed outcomes have reported higher IVE against influenza-related mortality, ranging from 65% to 89% [10, 11, 25]. However, the authors of 2 of these studies found IVE reduced when restricting the analysis to children with high-risk conditions (51% vs 65%) [25], and when inpatient controls were used instead of outpatient controls (72% vs 89%) [10]. In our study, 84% of the patients were at increased risk of severe outcomes, and inpatient controls were used, which may partially explain the differences in observed IVE.
The subgroup analyses revealed that the magnitude of IVE against influenza-related mortality was attenuated slightly in patients aged ≥ 50 years and was further reduced in patients aged ≥ 65 years. The phenomenon of immunosenescence (a decline in the immune system with age) is generally well accepted, but the relationship between immunological decline and IVE remains unclear [26]. One meta-analysis of IVE studies against hospitalization found that IVE differed by age, with an IVE of 51% (95% CI, 44%–58%) in patients 18 to <64 years old, and 37% (95% CI, 30%–44%) among patients ≥ 65 years of age [24]. Another meta-analysis of IVE against infection in the community-dwelling elderly found lower IVE estimates in adults aged > 75 years, compared to adults ≤ 75 years, although these results were not statistically different (16% [95% CI, −8% to 55%) vs 33% [95% CI, 17%–45%]) [23]. In line with the latter meta-analysis, we found some evidence that IVE against influenza-related mortality decreased with age. This most likely reflects immunosenescence, but could also be due to increasing residual confounding in the older population.
The use of an internal control (the effect of vaccination on noninfluenza mortality) identified that there may be residual confounding in our analysis model, particularly in patients aged ≥ 65 years. This may be due to unmeasured factors, or differences in confounders between vaccination mortality type (influenza or noninfluenza mortality) [13]. Although our analyses showed that influenza vaccination protects against influenza-related mortality, the presence of residual confounding cautions against overinterpretation of other quantitative findings. Using a ratio-of-ratios approach, we estimated that IVE against influenza-related mortality may be as high as 47%, although there is debate in the literature about whether bias-corrected estimates of association are valid as they assume that bias and measurement error in the case and control groups are similar [1, 2]. The bias-corrected analysis may also be consistent with some degree of attenuation of severity, which was not observed in the primary analysis.
We found little evidence that influenza vaccination attenuates influenza illness, in line with 2 other studies [27, 28]. One of these studies found vaccination did not reduce disease severity for most outcomes assessed, with a marginal reduction in intensive care unit length of stay for patients aged 50−64 years only [28]. Research undertaken by the US Influenza Hospitalization Surveillance Network found vaccination attenuated influenza illness when circulating viruses and vaccines were antigenically similar [29], but not when there was vaccine mismatch [28], suggesting that attenuation is dependent on IVE. As 38% of the case deaths in this study were from 2017 when vaccine match with circulating strains was poorer against the A/H3N2 subtype [30], further data are required to allow investigation of the impact of vaccination on influenza illness severity in years with good vaccine match compared to years with poor vaccine match. Published studies of IVE in Australian primary care and hospital settings have reported comparable estimates over this period [15, 31,–38]. Although the finding of residual confounding introduces some uncertainty, it is likely that the benefit of influenza vaccination against influenza-related mortality is primarily due to prevention of influenza illness.
The point estimates for IVE against influenza-related mortality appeared to vary over influenza seasons, as expected with changes in the circulating virus strains, seasonal vaccine formulation, and the match between the strains and vaccine. It was difficult to perform a meaningful analysis to assess IVE against influenza-related mortality over influenza seasons as the number of deaths, and consequently statistical power, in any single year was small.
Using data collected over 8 influenza seasons, this study provides the first estimates of IVE against influenza-related mortality in Australia. However, this study has some limitations. We used all-cause mortality, which may include deaths due to other underlying disease, as attribution of deaths to influenza is challenging. Misascertainment of cases may occur due to delays between symptom onset and testing, although the sensitivity analysis restricting to ≤ 7 days between onset and testing had little effect on influenza-related mortality. Influenza vaccination status was missing in a higher proportion of patients who died, which was due to difficulties with ascertainment. Misclassification of vaccination status may have occurred, although previous studies have found self-report of influenza vaccination is highly sensitive (94%–98%), but overestimates vaccination coverage [39, 40]. Testing for influenza was performed at the discretion of the treating physician where indicated for acute respiratory illness, and surveillance was passive, rather than active. Although influenza testing in hospitalized patients is routine (for both treatment and infection control purposes), we cannot exclude the possibility of selection bias. This may result in a biased estimate of vaccine effectiveness if testing was linked to both severity of illness and vaccination status, or if vaccination was associated with a reduced risk of presentation to hospital. Controls were frequency matched, rather than explicitly matched, as in some periods there were more cases than controls. To account for this in the analysis, we included month of presentation in the propensity score. These findings predominately apply to patients in the target group of the National Immunisation Program, who comprised 84% of the patients analysed.
This study reinforces the public health benefit of the influenza immunisation program in protecting against mortality, as well as reducing the risk of hospitalization. Recent changes to the National Immunisation Program introducing “enhanced” vaccines (high-dose and adjuvanted vaccines) for the elderly, and a pediatric program for children < 5 years of age, are likely to protect those at highest risk of complications. Ongoing surveillance is required to monitor vaccine effectiveness and the overall impact of the program. However, there is clearly a need for a more effective vaccine to reduce the burden of influenza.
Supplementary Data
Supplementary materials are available at Clinical 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.
Notes
Acknowledgments. The authors thank the Influenza Complications Alert Network (FluCAN) study investigators (Allen Cheng, Dominic Dwyer, Mark Holmes, Louis Irving, Grant Waterer, Tony Korman, Louise Cooley, Anna Howell, Deb Friedman, Peter Wark, Graham Simpson, John Upham, Simon Bowler, Sanjaya Senenayake, Tom Kotsimbos, and Paul Kelly); the Paediatric Active Enhanced Disease Surveillance system (PAEDS) team (Kristine Macartney, Christopher Blyth, Helen Marshall, Julia Clark, Josh Francis, and Jim Buttery); and Jill Garlick, Janine Roney, and study staff at each site, who have all contributed to FluCAN and PAEDS.
Financial support. FluCAN and PAEDS are supported by the Australian Department of Health. A. C. C. was supported by a National Health and Medical Research Council (NHMRC) Career Development Fellowship (APP1068732). Several sites are also funded by an NHMRC Partnership Grant. M. J. S. is a recipient of an Australian Research Council Future Fellowship (project number FT180100075).
Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.
Comments