who can help me?

n engl j med 357;14 www.nejm.org october 4, 2007 1373

The new england journal of medicine established in 1812 october 4, 2007 vol. 357 no. 14

Effectiveness of Influenza Vaccine in the Community-Dwelling Elderly

Kristin L. Nichol, M.D., M.P.H., M.B.A., James D. Nordin, M.D., M.P.H., David B. Nelson, Ph.D., John P. Mullooly, Ph.D., and Eelko Hak, Ph.D.

A B S T R A C T

From the Medicine Service and Center for Chronic Disease Outcomes Research, Min- neapolis Veterans Affairs Medical Center and University of Minnesota, Minneapo- lis (K.L.N., D.B.N.); the HealthPartners Research Foundation, Minneapolis (J.D.N.); Kaiser Permanente Northwest, Portland, OR ( J.P.M.); and the Julius Center for Health Services and Primary Care, Univer- sity Medical Center, Utrecht, the Nether- lands (E.H.). Address reprint requests to Dr. Nichol at Medicine Service (111), VA Medical Center, 1 Veterans Dr., Minne- apolis, MN 55417, or at nicho014@umn. edu.

N Engl J Med 2007;357:1373-81. Copyright © 2007 Massachusetts Medical Society.

Background

Reliable estimates of the effectiveness of influenza vaccine among persons 65 years of age and older are important for informed vaccination policies and programs. Short- term studies may provide misleading pictures of long-term benefits, and residual confounding may have biased past results. This study examined the effectiveness of influenza vaccine in seniors over the long term while addressing potential bias and residual confounding in the results.

Methods

Data were pooled from 18 cohorts of community-dwelling elderly members of one U.S. health maintenance organization (HMO) for 1990–1991 through 1999–2000 and of two other HMOs for 1996–1997 through 1999–2000. Logistic regression was used to estimate the effectiveness of the vaccine for the prevention of hospitalization for pneumonia or influenza and death after adjustment for important covariates. Ad- ditional analyses explored for evidence of bias and the potential effect of residual confounding.

Results

There were 713,872 person-seasons of observation. Most high-risk medical conditions that were measured were more prevalent among vaccinated than among unvaccinated persons. Vaccination was associated with a 27% reduction in the risk of hospitaliza- tion for pneumonia or influenza (adjusted odds ratio, 0.73; 95% confidence interval [CI], 0.68 to 0.77) and a 48% reduction in the risk of death (adjusted odds ratio, 0.52; 95% CI, 0.50 to 0.55). Estimates were generally stable across age and risk subgroups. In the sensitivity analyses, we modeled the effect of a hypothetical unmeasured con- founder that would have caused overestimation of vaccine effectiveness in the main analysis; vaccination was still associated with statistically significant — though lower — reductions in the risks of both hospitalization and death.

Conclusions

During 10 seasons, influenza vaccination was associated with significant reductions in the risk of hospitalization for pneumonia or influenza and in the risk of death among community-dwelling elderly persons. Vaccine delivery to this high-priority group should be improved.

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

n engl j med 357;14 www.nejm.org october 4, 20071374

Influenza continues to be a major cause of illness and death, especially among the elderly. Each year, influenza and its com- plications are responsible for about 186,000 ex- cess hospitalizations for respiratory and circula- tory illness1 and 44,000 excess deaths from all causes2 in this high-risk group. Influenza vaccines are safe and effective, and the elderly are included among the high-priority groups targeted for annual vaccination.3

Reliable estimates of the benefits of vaccination are important for establishing informed policies regarding resource allocation for the delivery of immunizations and identifying the need for new vaccines and strategies for the prevention and con- trol of inf luenza in this group.4 However, most studies assessing the effectiveness of inf luenza vaccination in the elderly have included one or only a few influenza seasons.5 Because of the variabil- ity of influenza from season to season, the results of these short-term studies — either favorable or unfavorable — might provide incomplete or mis- leading pictures about the benefits of vaccination over longer periods of time6; furthermore, hetero- geneity between studies may limit the ability to pool results appropriately across studies.7

Because of ethical constraints imposed by rec- ommendations that target the elderly for annual vaccination, most studies of the effectiveness of inf luenza vaccine in elderly persons have been observational studies and not randomized, con- trolled trials. Questions have been raised about the effects of potential bias and residual confounding on past estimates of vaccine effectiveness from these observational studies.8,9

In this study, we analyzed the effectiveness of influenza vaccination among 18 cohorts of com- munity-dwelling elderly members of health main- tenance organizations (HMOs) during 10 seasons. Our purpose was to provide a long-term view of the effectiveness of inf luenza vaccine while ad- dressing potential bias and residual confounding.

M e t h o d s

We pooled subject-level data from 18 cohorts of elderly members of one U.S. HMO for the 1990– 1991 through the 1999–2000 seasons and of two other U.S. HMOs for the 1996–1997 through the 1999–2000 seasons. Each cohort provided data for more than 20,000 person-seasons, for a total of 713,872 person-seasons during 10 seasons. The

study authors were solely and independently re- sponsible for the study design, the data analysis, the writing and preparation of all drafts of the manuscript, and the submission of the manuscript. The basic study design and methods have been de- scribed previously.10-13 The institutional research committees of the HMOs approved the project. In- formed consent was not required.

Study Sites and Participants

The participating HMOs were HealthPartners in Minnesota and Wisconsin (1990–1991 through 1999–2000); Kaiser Permanente Northwest in the Portland, Oregon, and Vancouver, Washington, area (1996–1997 through 1999–2000); and Oxford Health Plans in New York City and surrounding counties (1996–1997 through 1999–2000). All non- institutionalized members of the plans were in- cluded in that season’s cohort if they were 65 years of age or older as of October 1, had been con- tinuously enrolled in the plan for the preceding 12 months, were alive on the first day of the in- fluenza season, and were either continuously en- rolled or died during the outcome period. We se- lected these criteria to ensure adequate baseline and follow-up data and to reduce the potential for survivor bias.

Study Data

Data were extracted retrospectively from the ad- ministrative and clinical databases of the HMOs by the HMO research teams using standardized definitions that were consistent across the 10 sea- sons. Data elements included age, sex, baseline co- existing medical conditions (defined according to outpatient or inpatient International Classification of Diseases, Ninth Revision, Clinical Modif ication [ICD- 9-CM] codes denoting heart disease, lung disease, diabetes, renal disease, hematologic or nonhema- tologic cancer, vasculitis or rheumatologic disease, dementia or stroke, and immune deficiency or or- gan transplantation), health care use during the previous 12 months (the number of hospitaliza- tions and outpatient visits), year, site, and influ- enza-vaccination status. The study outcomes in- cluded hospitalization for pneumonia or influenza (ICD-9-CM codes 480 through 487) and death from any cause.

Influenza Seasons

Region-specific dates for the first and last influ- enza isolates reported to the Centers for Disease

Influenz a Vaccinations in the Elder ly

n engl j med 357;14 www.nejm.org october 4, 2007 1375

Control and Prevention defined the influenza sea- sons. Outcomes were included if they occurred during this period or, to capture delayed compli- cations, within 2 weeks after the end of the in- fluenza season.

Statistical Analysis

The baseline characteristics of vaccinated and un- vaccinated subjects were compared with the use of the chi-square test and Student’s t-test. Logistic regression (SPSS for Windows, version 13.0) was used to compare study outcomes between vacci- nated and unvaccinated subjects after adjustment for covariates. The propensity score (the proba- bility of being vaccinated given the observed co- variates) was calculated for each subject and was included in the models with the use of strata based on quintiles of propensity score.14 Demographic characteristics, coexisting medical conditions, pre- vious use of health care, site, and year were also included. Age (65 to 69 years, 70 to 79 years, 80 to 89 years, and 90 years and older) and number of outpatient visits (<4, 4 to 8, 9 to 15, and >15) were included as categorical variables. A similar model derived from and validated among vacci- nated and unvaccinated cohorts from the three HMOs for the 1996–1997 and 1997–1998 seasons discriminated well between persons who did or did not enter the hospital or die during the influenza season.15 Vaccine effectiveness was estimated as a percentage: (1 − adjusted odds ratio) × 100.

We conducted subgroup analyses to explore for heterogeneity in levels of vaccine effectiveness and to divide the study population into more homo- geneous strata that might reduce the effect of re- sidual confounding or bias.16 The subgroups were defined by sex, age, risk status (high risk was de- fined as the presence of one or more major coex- isting conditions at baseline, and low risk as the presence of no major coexisting conditions at baseline), propensity-score quintile, previous hos- pitalization, and previous use of outpatient care. To test for a healthy-vaccinee bias, we compared the risk of hospitalization for vaccinated and for unvaccinated persons during noninfluenza peri- ods; data about hospitalization but not about mor- tality were available for the summer months of June through September after the 1998–1999 and 1999–2000 influenza seasons.

For our sensitivity analysis, we modeled how a hypothetical unmeasured confounder might have influenced our estimates of vaccine effectiveness.

In our main analyses, vaccinated subjects appeared sicker than unvaccinated subjects, and we con- trolled for those measured differences. If an un- measured confounder were also present so that persons with the confounder were less likely to be vaccinated but more likely to be hospitalized or die, then our main analyses would have overes- timated vaccine effectiveness.17 Some examples of possible unmeasured confounders are race, in- come, and functional status. Limited data for two of the HMOs, however, suggest that the study populations were racially fairly homogeneous, with 85 to 90% of persons indicating that they were white (unpublished observations). Other investiga- tors have reported that, after age, sex, and coexist- ing conditions have been controlled for, functional status is a stronger predictor of hospitalization or death among the elderly than is income.18 We therefore modeled our hypothetical confounder on the basis of published data about impaired func- tional status in the elderly.

On the basis of previous studies of functional status and the observed levels of association be- tween important covariates with outcomes in our multivariable models, we estimated that plausible associations with vaccination and outcomes for a strong confounder would mean that persons with the confounder would be half as likely to be vac-

Table 1. Baseline Characteristics of the Study Subjects.*

Characteristic Unvaccinated (N = 298,623)

Vaccinated (N = 415,249)

Age (yr) 73.6±6.9 73.9±6.3

Male sex (%) 41.7 44.4

Presence of one or more high-risk medical conditions (%)

45.6 55.6

Diabetes 11.0 14.4

Heart disease 22.7 26.8

Lung disease 15.2 19.2

Renal disease 2.0 2.3

Vasculitis or rheumatologic disease 1.4 1.9

Immune deficiency 1.0 1.2

Cancer 13.3 14.5

Dementia or stroke 4.7 3.4

No. of outpatient visits during baseline period

10.3±15.6 12.8±13.3

Hospitalization during baseline period (%) 13.3 14.5

* Plus–minus values are means ±SD. All P values are less than 0.001.

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

n engl j med 357;14 www.nejm.org october 4, 20071376

cinated19 and two to three times as likely to be hospitalized or die18,20-22 as would persons with- out the confounder. We varied the prevalence of the confounder from 20 to 60%.23 The method of Lin et al.24 was used to quantify the effect of the hypothetical confounder on our estimates of vac- cine effectiveness. All reported P values are two- sided and were not adjusted for multiple testing.

R e s u l t s

There were 713,872 person-seasons of observation. Vaccinated subjects were slightly older and had higher prevalence rates of all the baseline medi- cal conditions except dementia or stroke (Table 1).

The predominant circulating virus strains for the 10 influenza seasons were type A/H3N2 for 6 seasons (1991–1992, 1993–1994, 1996–1997, 1997–1998, 1998–1999, and 1999-2000), type B for 1 season (1990–1991), mixed A/H3N2 and B for 2 seasons (1992–1993 and 1994–1995), and mixed A/H3N2 and A/H1N1 for 1 season (1995– 1996).1,2 The vaccine–virus antigenic match was good to excellent in all seasons except 1992–1993 and 1997–1998, when the circulating A/H3N2 strains represented drifted variants.10,12

During the 10 influenza seasons, there were 4599 hospitalizations for pneumonia or influenza and 8796 deaths. The observed hospitalization rates for unvaccinated and for vaccinated partici- pants were, on average, 0.7% and 0.6% per season, respectively, with corresponding death rates of 1.6% and 1.0% per season. Increasing age and the presence of one or more high-risk medical condi- tions at baseline were the strongest predictors of the risk of hospitalization or death in our models, with patterns of increasing risk being virtually identical for vaccinated and for unvaccinated per- sons (Fig. 1).

Influenza vaccination was associated on aver- age with substantial reductions in hospitalizations for pneumonia and for influenza (vaccine effec- tiveness, 27%; adjusted odds ratio, 0.73; 95% con- fidence interval [CI], 0.68 to 0.77) and in death (vaccine effectiveness, 48%; adjusted odds ratio, 0.52; 95% CI, 0.50 to 0.55). Estimates varied from season to season and across the 18 cohorts (Fig. 2). In the two seasons with a poor match between the vaccine and the virus strain, vaccine effectiveness was lower for reducing death (in seasons with a poor match, vaccine effectiveness was 37% [ad- justed odds ratio, 0.63; 95% CI, 0.57 to 0.69]; in seasons with a good match, vaccine effectiveness was 52% [adjusted odds ratio, 0.48; 95% CI, 0.46 to 0.51]) but not for reducing hospitalization. Our multivariable regression models showed good to excellent discrimination, with C statistics of 0.76 (95% CI, 0.75 to 0.77) for hospitalization and 0.81 (95% CI, 0.80 to 0.81) for death.

Estimates of vaccine effectiveness suggested clinically significant benefits across the subgroups

Adjusted Odds Ratio

22p3

AUTHOR:

FIGURE:

JOB: ISSUE:

4-C H/T

RETAKE

SIZE

ICM

CASE

EMail Line H/T Combo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st 2nd 3rd

Nichol

1 of 3

10-04-07

ARTIST: ts

35714

1 10 100

<70 yr 70–79 yr 80–89 yr ≥90 yr

Low risk Vaccinated

<70 yr 70–79 yr 80–89 yr ≥90 yr

High risk

<70 yr 70–79 yr 80–89 yr ≥90 yr

Low risk Unvaccinated

<70 yr 70–79 yr 80–89 yr ≥90 yr

High risk

<70 yr 70–79 yr 80–89 yr ≥90 yr

Low risk Vaccinated

<70 yr 70–79 yr 80–89 yr ≥90 yr

High risk

<70 yr 70–79 yr 80–89 yr ≥90 yr

Low risk Unvaccinated Hospitalization

Death

<70 yr 70–79 yr 80–89 yr ≥90 yr

High risk

Figure 1. Risk of Outcomes According to Age and Risk Status for Unvacci- nated and Vaccinated Groups.

Shown are the adjusted odds ratios, with bars indicating the 95% confi- dence intervals. The reference category (with an odds ratio of 1) for both outcomes within each group was persons at low risk who were under 70 years of age. High risk is defined as having one or more major coexisting conditions at baseline, and low risk as having no major coexisting condi- tions at baseline.

Influenz a Vaccinations in the Elder ly

n engl j med 357;14 www.nejm.org october 4, 2007 1377

(Fig. 3). There was, however, evidence for interac- tion between vaccination and high-risk status for hospitalization (P = 0.004) and between vaccination and sex (P = 0.03) and outpatient visits (P = 0.03) for death. All other P values for interaction be- tween vaccination and subgroup variables were greater than 0.05. Our analysis for evidence of a healthy-vaccinee bias was negative; during the two noninfluenza periods for which we had informa- tion, vaccinated and unvaccinated persons had similar risks of hospitalization (for June through September 1999, the adjusted odds ratio was 1.0 [95% CI, 0.78 to 1.28]; for June through September 2000, the adjusted odds ratio was 0.94 [95% CI, 0.74 to 1.19]).

The sensitivity analysis shows how our esti- mates of vaccine effectiveness might have been influenced by residual confounding (Table 2). With increasing prevalence of the confounder and with increasing risk of an outcome because of the con- founder, estimates of vaccine effectiveness were incrementally lower though still significant. In the most extreme scenario that we evaluated, at a prevalence of 60% and an increased risk by a fac- tor of three of an outcome, the estimates of vac- cine effectiveness were reduced to 7% for hospi- talization and 33% for death.

D i s c u s s i o n

In this study, influenza vaccination of community- dwelling elderly persons during 10 seasons was associated with substantial reductions in hospital- izations for pneumonia or influenza and in death. By pooling patient-level data from 18 cohorts spanning the decade of the 1990s and including 713,872 person-seasons of observation, we have provided an important perspective on the benefits of vaccination among the elderly. We documented both the year-to-year variability that can be seen and the average long-term level of benefit from vaccination. Previous case–control and cohort stud- ies from other populations and countries have also demonstrated benefits of vaccination in elderly populations.5,6 However, the limitations of stud- ies based on a single or just a few influenza sea- sons, because of the substantial variability from year to year in circulating viruses and in antigenic match between circulating viruses and vaccine strains, and the challenges associated with at- tempts to pool results from heterogeneous studies in the absence of patient-level data6,7 have been ac-

knowledged. Our study during 10 consecutive sea- sons attempts to overcome these limitations.

The large number of subjects in our data set permitted considerable precision in our estimates, which showed substantial benefits across multiple subgroups, a result suggesting that vaccination benefits probably extend to a broad spectrum of elderly persons. We also included subjects from

33p9

B

A

AUTHOR:

FIGURE:

JOB: ISSUE:

4-C H/T

RETAKE

SIZE

ICM

CASE

EMail Line H/T Combo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st 2nd

3rd

Nichols

2 of 3

10-04-07

ARTIST: ts

35714

10060 804020−20−40 0−60

1990–1991 HP 1991–1992 HP 1992–1993 HP 1993–1994 HP 1994–1995 HP 1995–1996 HP 1996–1997 OX 1996–1997 HP

1997–1998 OX 1996–1997 KP

1997–1998 HP

1998–1999 OX 1997–1998 KP

1998–1999 HP 1998–1999 KP 1999–2000 OX 1999–2000 HP 1999–2000 KP

1997–1998 OX 1996–1997 KP

1997–1998 HP

1998–1999 OX 1997–1998 KP

1998–1999 HP 1998–1999 KP 1999–2000 OX 1999–2000 HP 1999–2000 KP

Overall Season

1990–1991 HP 1991–1992 HP 1992–1993 HP 1993–1994 HP 1994–1995 HP 1995–1996 HP 1996–1997 OX 1996–1997 HP

Overall Season

Effectiveness of Vaccine in Reducing the Risk of Hospitalization (%)

10060 804020−20−40 0−60

Effectiveness of Vaccine in Reducing the Risk of Death (%)

Figure 2. Effectiveness of the Inf luenza Vaccine in Reducing the Risk of Hospitalization for Pneumonia or Influenza (Panel A) and Death (Panel B).

The estimates of vaccine effectiveness were calculated as (1 − adjusted odds ratio) × 100. The bars indicate the 95% confidence intervals. HP de- notes HealthPartners, OX Oxford Health Plans, and KP Kaiser Permanente.

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

n engl j med 357;14 www.nejm.org october 4, 20071378

three geographically diverse HMOs to enhance the likelihood that our study population would be representative of other HMO populations across the country. However, elderly enrollees in HMOs may differ from elderly persons without HMO coverage in important ways, including race, in- come, functional status, and urban versus non- urban residence,25 and caution should be used in generalizing our results to other groups.

In addition, our study was not designed to evaluate levels of vaccine effectiveness among the frailest elderly, such as those living in nursing homes. The elderly may have impaired humoral26 and cell-mediated27,28 immune responses to influ- enza vaccines. Institutionalized and other frail, elderly persons may be especially likely to exhibit such impaired immune responses,29,30 and there- fore the levels of vaccine effectiveness in this population may be lower than those we have re- ported for the community-dwelling elderly.

Inactivated influenza vaccines are most effec- tive when there is a good match between circulat- ing viruses and vaccine strains; protection may also be substantial, though sometimes lower, dur- ing years with a poor match among healthy young adults,31,32 healthy adults and adults at high risk aged 50 through 64 years,33 and institution- alized elderly persons.34 Our findings are similar. During the two seasons with a poor match in our study, vaccination was associated with significant reductions in hospitalization and death, although the reduction in the risk of death was smaller than it was during the seasons with a good match.

Our results provide useful information on the

Figure 3. Results of the Subgroup Analyses Assessing Vaccine Effectiveness in Reducing the Risk of Hospitalization (Panel A) and Death (Panel B).

The subgroups included quintile of propensity score, sex, age, risk group (high risk was defined as the pres- ence of one or more major coexisting conditions at baseline, and low risk as the presence of no major co- existing conditions at baseline), quartile of the number of outpatient visits during the baseline period, and pre- vious hospitalization during the baseline period. Esti- mates of vaccine effectiveness were calculated as (1 − adjusted odds ratio) × 100, with bars indicating the 95% confidence intervals. There was interaction be- tween vaccination and high-risk status for hospitaliza- tion (P = 0.004), between vaccination and sex for death (P = 0.03), and between vaccination and outpatient vis- its at baseline for death (P = 0.03). All other P values for interaction between vaccination status and subgroups were greater than 0.05.

Overall Quintile of propensity score

Quartile of number of out- patient visits

Previous hospitalization

Sex

Low risk

High risk

Subgroup

1 2 3 4

1 2 3 4

5

Male Female

70–79 yr <70 yr

80–89 yr

<70 yr

≥90 yr

70–79 yr

≥90 yr 80–89

Yes No

33p9

100

Effectiveness of Vaccine in Reducing the Risk of Hospitalization (%)

Effectiveness of Vaccine in Reducing the Risk of Death (%)

50 7525−25 0−50

10050 7525−25 0−50

Overall Quintile of propensity score

Quartile of number of out- patient visits

Previous hospitalization

Sex

Low risk

High risk

Subgroup

1 2 3 4

1 2 3 4

5

Male Female

70–79 yr <70 yr

80–89 yr

<70 yr

≥90 yr

70–79 yr

≥90 yr 80–89

Yes No

B

A

AUTHOR:

FIGURE:

JOB: ISSUE:

4-C H/T

RETAKE

SIZE

ICM

CASE

EMail Line H/T Combo

Revised

AUTHOR, PLEASE NOTE: Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd 3rd

Nichols

3 of 3

10-04-07

ARTIST: ts

35714

Influenz a Vaccinations in the Elder ly

n engl j med 357;14 www.nejm.org october 4, 2007 1379

benefits that elderly persons may receive from vac- cination. How these results might relate to popu- lation-level trends is unclear. For example, influ- enza-attributable excess mortality rates in the United States have not declined to the degree that might be expected in light of increasing vaccina- tion rates during the 1980s and the early 1990s.35 However, nation-level data do not include the risk profile or vaccination status of those who have died. Critical information is therefore lacking, making it difficult to estimate what the expected excess mortality rates would be if vaccination rates were 0%, what benefits have already been realized given current patterns of vaccine use, and what additional benefits might be realized with more effective vaccine delivery. Because of the large geographic variations in vaccine delivery in the United States, as well as disparities in vaccination rates according to race and ethnic group,36 the aging of the population,2 and the increasing numbers of elderly persons with high-risk con- ditions37,38 for whom the risk of dying increases exponentially, attempts to correlate population- level trends with individual levels of protection due to vaccination or vice versa will be susceptible to many pitfalls, including the ecologic fallacy.39

This study has several limitations. Because this was an observational study, we adjusted for impor- tant covariates in our analytic models. These mod-

els discriminated well between persons who did and those who did not become hospitalized or die. Our inclusion criteria were designed to minimize the possibility of survivor bias, and we did not find evidence for a healthy-vaccinee effect in our analy- ses. Nevertheless, residual confounding may have influenced our results, and our sensitivity analyses indicate how our estimates of vaccine effectiveness would be lower, though still significant, after ad- justment for the effect of a strong hypothetical unmeasured confounder. Misclassification of vac- cination status may also have occurred; the most likely cause of misclassification would have been a failure to record receipt of vaccine. However, at one of the HMOs, more than 90% of members at high risk who were vaccinated received the vaccine at a health plan site.40 Furthermore, agreement between medical records and computerized data- bases has been excellent, with more than 95% agreement at two of the study sites.12 Even if sub- stantial misclassification occurred, it probably would have biased the results toward the null hy- pothesis.

Achieving optimal success in preventing and controlling influenza among the elderly may re- quire more immunogenic vaccines and new strat- egies that induce greater levels of herd immunity and thereby interrupt inf luenza transmission in communities. More effective vaccines for the el-

Table 2. Sensitivity Analysis to Quantify the Effects of a Hypothetical Unmeasured Confounder on the Study Results.*

Increase in the Risk of Outcome on Account

of the Confounder Prevalence

of Confounder Hospitalization for

Pneumonia or Influenza Death

Vaccine Effectiveness

Adjusted Odds Ratio (95% CI)

Vaccine Effectiveness

Adjusted Odds Ratio (95% CI)

% % %

— 0 27 0.73 (0.68–0.77) 48 0.52 (0.50–0.55)

Doubled 20 20 0.80 (0.75–0.85) 43 0.57 (0.55–0.60)

Doubled 40 15 0.85 (0.80–0.90) 40 0.60 (0.58–0.63)

Doubled 60 14 0.86 (0.81–0.92) 39 0.61 (0.59–0.65)

Tripled 20 14 0.86 (0.81–0.92) 38 0.62 (0.59–0.64)

Tripled 40 9 0.91(0.86–0.97) 35 0.65 (0.63–0.69)

Tripled 60 7 0.93 (0.87–0.99) 33 0.67 (0.64–0.70)

* We modeled our hypothetical confounder on impaired functional status in the elderly. The results of the main analysis without adjustment for the hypothetical unmeasured confounder are shown in the first row, which shows a prevalence of 0. Persons with the confounder were assumed to be half as likely to be vaccinated as persons without it. Vaccine ef- fectiveness is estimated as (1 − adjusted odds ratio) × 100. Adjusted odds ratios in the presence of the hypothetical con- founder were estimated by the method described by Lin et al.24

T h e n e w e n g l a n d j o u r n a l o f m e d i c i n e

n engl j med 357;14 www.nejm.org october 4, 20071380

derly are under development but have not yet been approved for use in the United States. Vaccination of children in the United States has been associ- ated with reductions in illness in households41 and in the community,42 and in Japan with lower mor- tality rates among the elderly.43 However, these studies are not conclusive,44,45 and additional re- search is needed to define the benefits among the elderly that might be realized from vaccinating children. In the meantime, vaccination rates of el- derly persons remain stagnant and well below the 2010 goal of 90%.3 Even as we wait for new vac- cines and new strategies, patients, their health care providers, and policymakers should renew efforts to improve the delivery of current influenza vac- cines to this high-priority group. Hospitalizations and deaths will be prevented if we can succeed.

Supported by the National Vaccine Program Office and the Centers for Disease Control and Prevention (CDC) through an agreement with the American Association of Health Plans, for the obtaining of the original data for 1996–1997 through 1999–2000 from HealthPartners in Minnesota and Wisconsin, Kaiser Permanente Northwest in the Portland, Oregon, and Vancouver, Washington, area, and Oxford Health Plans in New York City and surrounding counties; by the HealthPartners Re- search Foundation and Connaught Laboratories for the obtain- ing of the original data for 1990–1991 through 1995–1996 from HealthPartners; in part by the Center for Chronic Disease Out- comes Research at the Minneapolis VA Medical Center (to Dr. Nelson) for analysis of data for the data-pooling project; and by a grant (916.56.109) from the Netherlands Scientific Organiza- tion (to Dr. Hak).

Dr. Nichol reports serving as a consultant to or as a member of medical advisory boards of Sanofi Pasteur, MedImmune, GlaxoSmithKline, and Novartis and receiving grant support from Sanofi Pasteur and GlaxoSmithKline; and Dr. Nordin, re- ceiving grant support from Sanofi Pasteur and the CDC. No other potential conflict of interest relevant to this article was reported.

References

Thompson WW, Shay DK, Weintraub E, et al. Influenza-associated hospitaliza- tions in the United States. JAMA 2004;292: 1333-40.

Thompson WW, Shay DK, Weintraub E, et al. Mortality associated with influ- enza and respiratory syncytial virus in the United States. JAMA 2003;289:179-86.

Advisory Committee on Immuniza- tion Practices. Prevention and control of influenza: recommendations of the Advi- sory Committee on Immunization Practices (ACIP). MMWR Recomm Rep 2006;55 (RR-10):1-42. [Erratum, MMWR Morb Mor- tal Wkly Rep 2006;55:800.]

Schwartz B, Hinman A, Abramson J, et al. Universal influenza vaccination in the United States: are we ready? Report of a meeting. J Infect Dis 2006;194:Suppl 2: S147-S154.

Nichol KL. Influenza vaccination in the elderly: impact on hospitalisation and mortality. Drugs Aging 2005;22:495-515.

Jefferson T, Rivetti D, Rivetti A, Rudin M, Di Pietrantonj C, Demicheli V. Efficacy and effectiveness of influenza vaccines in elderly people: a systematic review. Lancet 2005;366:1165-74. [Erratum, Lancet 2006; 367:986.]

Jacobson RM, Targonski PV, Poland GA. Meta-analyses in vaccinology. Vaccine 2007;25:3153-9.

Glezen WP, Simonsen L. Benefits of influenza vaccine in US elderly — new studies raise questions. Int J Epidemiol 2006;35:352-3.

Jefferson T. Influenza vaccination: poli- cy versus evidence. BMJ 2006;333:912-5.

Nichol KL, Margolis KL, Wuorenma J, Von Sternberg T. The efficacy and cost ef- fectiveness of vaccination against influ- enza among elderly persons living in the

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

community. N Engl J Med 1994;331:778- 84.

Nichol KL, Wuorenma J, Von Stern- berg T. Benefits of influenza vaccination for low-, intermediate-, and high-risk se- nior citizens. Arch Intern Med 1998;158: 1769-76.

Nordin J, Mullooly J, Poblete S, et al. Influenza vaccine effectiveness in pre- venting hospitalizations and deaths in persons 65 years or older in Minnesota, New York, and Oregon: data from 3 health plans. J Infect Dis 2001;184:665-70.

Nichol KL, Nordin J, Mullooly J, Lask R, Fillbrandt K, Iwane M. Influenza vac- cination and reduction in hospitalizations for cardiac disease and stroke among the elderly. N Engl J Med 2003;348:1322-32.

Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med 1997;127:757-63.

Hak E, Wei F, Nordin J, Mullooly J, Po- blete S, Nichol KL. Development and vali- dation of a clinical prediction rule for hos- pitalization due to pneumonia or influenza or death during inf luenza epidemics among community-dwelling elderly per- sons. J Infect Dis 2004;189:450-8.

Hak E, Verheij TJ, Grobbee DE, Nichol KL, Hoes AW. Confounding by indication in non-experimental evaluation of vaccine effectiveness: the example of prevention of influenza complications. J Epidemiol Community Health 2002;56:951-5.

Hak E, Hoes AW, Nordin J, Nichol KL. Benefits of influenza vaccine in US elderly — appreciating issues of confounding bias and precision. Int J Epidemiol 2006;35: 800-2.

Mayo NE, Nadeau L, Levesque L, Mill- er S, Poissant L, Tamblyn R. Does the ad- dition of functional status indicators to

11.

12.

13.

14.

15.

16.

17.

18.

case-mix adjustment indices improve pre- diction of hospitalization, institutionaliza- tion, and death in the elderly? Med Care 2005;43:1194-202.

Jackson LA, Nelson JC, Benson P, et al. Functional status is a confounder of the association of influenza vaccine and risk of all cause mortality in seniors. Int J Epidemiol 2006;35:345-52.

Walter LC, Brand RJ, Counsell SR, et al. Development and validation of a prog- nostic index for 1-year mortality in older adults after hospitalization. JAMA 2001; 285:2987-94.

Carey EC, Walter LC, Lindquist K, Co- vinsky KE. Development and validation of a functional morbidity index to predict mortality in community-dwelling elders. J Gen Intern Med 2004;19:1027-33.

Lee SJ, Linkdquist K, Segal MR, Co- vinsky KE. Development and validation of a prognostic index for 4-year mortality in older adults. JAMA 2006;295:801-8.

Lethbridge-Cejku M, Rose D, Vickerie J. Summary health statistics for US adults: National Health Interview Survey, 2004. Vital and health statistics. Series 10. No. 228. Washington, DC: Government Print- ing Office, 2006:table 19. (Accessed Sep- tember 7, 2007, at http://www.cdc.gov/ nchs/data/series/sr_10/sr10_228.pdf.)

Lin DY, Psaty BM, Kronmal RA. As- sessing the sensitivity of regression re- sults to unmeasured confounders in ob- servational studies. Biometrics 1998;54: 948-63.

Banthin JS, Taylor AK. Medical Expen- diture Panel Survey — research findings #15: HMO enrollment in the United States: estimates based on household reports, 1996. Rockville, MD: Agency for Health- care Research and Policy. (Accessed Sep-

19.

20.

21.

22.

23.

24.

25.

Influenz a Vaccinations in the Elder ly

n engl j med 357;14 www.nejm.org october 4, 2007 1381

tember 7, 2007, at http://www.meps.ahrq. gov/mepsweb/data_files/publications/rf15/ rf15.shtml.)

Goodwin K, Viboud C, Simonsen L. Antibody response to influenza vaccina- tion in the elderly: a quantitative review. Vaccine 2006;24:1159-69.

McElhaney JE. The unmet need in the elderly: designing new influenza vaccines for older adults. Vaccine 2005;23:Suppl 1: S10-S25.

Targonski PV, Jacobson RM, Poland GA. Immunosenescence: role and measure- ment in influenza vaccine response among the elderly. Vaccine 2007;25:3066-9.

Gross PA, Quinnan GV Jr, Weksler ME, Setia U, Douglas RG Jr. Relation of chronic disease and immune response to influenza vaccine in the elderly. Vaccine 1989;7:303-8.

Fulöp T Jr, Wagner JR, Khalil A, We- ber J, Trottier L, Payette H. Relationship between the response to influenza vacci- nation and the nutritional status in insti- tutionalized elderly subjects. J Gerontol A Biol Sci Med Sci 1999;54:M59-M64.

Jefferson T, Rivetti D, Di Pietrantonj C, Rivetti A, Demicheli V. Vaccines for prevent- ing influenza in healthy adults. Cochrane Database Syst Rev 2007;2:CD001269.

Ohmit SE, Victor JC, Rotthoff JR, et al. Prevention of antigenically drifted influ-

26.

27.

28.

29.

30.

31.

32.

enza by inactivated and live attenuated vaccines. N Engl J Med 2006;355:2513-22.

Herrera GA, Iwane MK, Cortese M, et al. Influenza vaccine effectiveness among 50-64-year-old persons during a season of poor antigenic match between vaccine and circulating influenza virus strains: Colo- rado, United States, 2003-2004. Vaccine 2007;25:154-60.

Gross PA, Hermogenes AW, Sacks HS, Lau J, Levandowski RA. The efficacy of in- fluenza vaccine in elderly persons: a meta- analysis and review of the literature. Ann Intern Med 1995;123:518-27.

Simonsen L, Reichert TA, Viboud C, Blackwelder WC, Taylor RJ, Miller MA. Impact of influenza vaccination on sea- sonal mortality in the US elderly popula- tion. Arch Intern Med 2005;165:265-72.

Racial/ethnic disparities in influenza and pneumococcal vaccination levels among persons aged > or =65 years — United States, 1989–2001. MMWR Morb Mortal Wkly Rep 2003;52:958-62.

Fry AM, Shay DK, Holman RC, Curns AT, Angerson LJ. Trends in hospitaliza- tions for pneumonia among persons aged 65 years or older in the United States, 1988-2002. JAMA 2005;294:2712-9.

Mokdad AH, Ford ES, Bowman BA, et al. Diabetes trends in the US: 1990-1998. Diabetes Care 2000;23:1278-83.

33.

34.

35.

36.

37.

38.

Morgenstern H. Ecologic studies in epi- demiology: concepts, principles, and meth- ods. Annu Rev Public Health 1995;16:61-81.

Mac Donald R, Baken L, Nelson A, Nichol KL. Validation of self-report of in- fluenza and pneumococcal vaccination status in elderly outpatients. Am J Prev Med 1999;16:173-7.

King JC, Stoddard JJ, Gaglani M, et al. Effectiveness of school-based inf luenza vaccination. N Engl J Med 2006;355:2523- 32.

Piedra PA, Gaglani MJ, Kozinetz CA, et al. Herd immunity in adults against in- fluenza-related illnesses with the use of the trivalent-live attenuated influenza vac- cine (CAIV-T) in children. Vaccine 2005; 23:1540-8.

Reichert TA, Sugaya N, Fedson DS, Glezen WP, Simonsen L, Tashiro M. The Japanese experience with vaccinating schoolchildren against influenza. N Engl J Med 2001;344:889-96.

Jordan R, Connock M, Albon E, et al. Universal vaccination of children against influenza: are there indirect benefits to the community? A systematic review of the evidence. Vaccine 2006;24:1047-62.

Fukuda K, Kieny MP. Different ap- proaches to influenza vaccination. N Engl J Med 2006;355:2586-7. Copyright © 2007 Massachusetts Medical Society.

39.

40.

41.

42.

43.

44.

45.

clinical trial registration The Journal requires investigators to register their clinical trials

in a public trials registry. The members of the International Committee of Medical Journal Editors (ICMJE) will consider most clinical trials for publication

only if they have been registered (see N Engl J Med 2004;351:1250-1). Current information on requirements and appropriate registries

is available at www.icmje.org/faq.pdf.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.