Methods

Top Methods Results Discussion Author & Article Info References

General Methods

Setting

The study was done at Kaiser Permanente, Northwest Region, a prepaid group practice health maintenance organization (HMO) that includes 20% of the Portland, Oregon-Vancouver, Washington, Standard Metropolitan Statistic Area population. Members receive almost all of their medical care from the health plan, and each contact they make with the HMO is recorded in a single, centralized chart.

Epidemic Periods

Epidemic periods were defined as periods during which influenza viruses circulated persistently in Oregon according to influenza virus isolation dates reported by the Oregon State Virology Laboratory. The eight epidemic periods defined using this approach are shown in Table 1. The study period for the nonepidemic 1981-1982 reference period was defined by averaging the months and days of the epidemic periods. The number of specimens tested ranged from 118 in 1981-1982 to 652 in 1988-1989.

Influenza viruses were isolated from only 9 of 118 specimens (8%) tested in the 1981-1982 nonepidemic reference period; B viruses were isolated from two specimens obtained in late April. During the epidemic study periods, influenza viruses were typically isolated from 25% of specimens, isolations were sustained, and isolation frequency distributions described epidemic curves.

Participants

Participants included in a study period were Kaiser Permanente members 65 years of age or older, each with at least 1 month of eligibility during that period and during the September-December vaccination period. These members represent almost all elderly persons enrolled in the HMO.

High-risk participants were defined as those who had had medical care contacts for chronic pulmonary (ICDA [International Classification of Diseases, Adapted] codes 9:491-493.9, 496, 500-505, 506.4, 508.1, 510-516, 714.8), cardiovascular (112.81, 130.3, 393-398, 401.0, 402-405, 410-414, 416, 420-429, 430-438, 440.1, 785.2-.3), metabolic (250-250.9, 571-571.9), renal (274.10, 580-583.9, 585, 587-588.0), or malignant (140-208.9) conditions. These risk factors for influenza-related complications are defined by the CDC's Advisory Committee for Immunization Practices.

Data Sources

An automated inpatient database was used to identify all hospitalizations for pneumonia and influenza during the nine study periods. This database contains discharge diagnoses and records of inpatient procedures and is the source of the discharge summaries.

The outpatient database contains outpatient diagnoses and records of procedures for a continuously updated research sample composed of 5% of elderly health plan members. Each month, a random 5% sample of all new elderly Kaiser Permanente subscriber units is added to the cumulative pool of enrollees whose records are abstracted. Automated outpatient diagnoses were available through 1987. Medical record abstractions were the source of outpatient diagnoses for the first quarter of 1988.

Inpatient high-risk discharge diagnoses made during the 3 years before the study period were obtained from the inpatient database. Outpatient high-risk diagnoses made during the year before the study period were obtained from the outpatient database and by chart abstraction.

Influenza and pneumococcal vaccination status was abstracted from the medical records for pneumonia and influenza cases that were not in the 5% research sample. Automated vaccination data through 1987 were available for cases and controls in the research sample. Medical record abstractions were the source of 1988-1989 vaccination data for all study participants. Participants receiving influenza vaccine after the start of epidemics were classified as unvaccinated.

Case-Control Assessment of Vaccine Effectiveness

Population-based, case–control designs were used to estimate influenza vaccine effectiveness in preventing medically attended episodes of pneumonia and influenza. To exclude pneumonias that were unlikely to be influenza related, we included only persons with pneumonia and influenza who were admitted to medical services. Because we based participant selection on discharge diagnoses, participants with both hospital- and community-acquired pneumonia and influenza were included.

Pneumonia and Influenza Cases and Controls

Participants who were hospitalized with pneumonia and influenza during the study period, who were eligible Kaiser Permanente members, and who had a pneumonia and influenza ICDA-9 code^{[480,481,482,483,485,486,487]} listed as a discharge diagnosis were included as cases of hospitalized pneumonia and influenza. Controls were members of the 5% research sample who met the study eligibility criteria and who were not hospitalized with pneumonia or influenza during the given period.

Pneumonia and influenza deaths during a study period were defined as hospitalized patients with pneumonia and influenza who died in the hospital. Controls were members of the 5% research sample who were not pneumonia and influenza hospital deaths during that period.

Outpatient pneumonia and influenza cases included members of the 5% research sample who had an outpatient episode of pneumonia and influenza with no inpatient component during the given study period. Controls were members of the 5% research sample who did not have an outpatient episode of pneumonia and influenza during the given period.

Cases and controls were group matched for high-risk status. Logistic regressions were used to adjust influenza vaccine effects on occurrence of pneumonia and influenza episodes for differences in age, sex, months of HMO eligibility, and pneumococcal vaccine status.

Vaccine Effectiveness

Relative risks for pneumonia and influenza episodes in unvaccinated and vaccinated populations were estimated by the adjusted exposure (unvaccinated) odds ratios. The effectiveness of influenza vaccine in preventing pneumonia and influenza episodes was computed from the estimated relative risk as effectiveness =1-(1/relative risk).

Pneumonia and Influenza Episodes Prevented

The population-based, case–control design allowed unbiased estimation of the total number of participants vaccinated and of the pneumonia and influenza episode rates among unvaccinated participants. The product of these estimates multiplied by the vaccine effectiveness is the estimated number of pneumonia and influenza episodes prevented by influenza vaccination: pneumonia and influenza episodes prevented = (number of participants vaccinated)(effectiveness of vaccination)(pneumonia and influenza episode rate among unvaccinated participants).

Statistical Methods

The units of analysis for period-specific vaccine effectiveness analyses were persons. A case for a given period had an episode of pneumonia or influenza during that period. A control for a given period did not have an episode of pneumonia or influenza during that period. The same person could appear as a case in more than one period.

Person-periods were the units of analysis for aggregate analyses. A person contributed a person-period of observation to each period when study criteria were met. Some members of the 5% research sample were controls in some periods and cases in other periods. Study participants not in the 5% research sample could only contribute person-periods as hospitalized cases.

Risks for pneumonia and influenza episodes during high-risk and non-high-risk person-periods are separately modeled as logistic functions of influenza vaccine status, pneumococcal vaccine status, age, sex, and length of observation time.

The SAS logistic procedure was used to estimate odds ratios and confidence intervals for model variables. The Hosmer-Lemeshow statistic was used to assess the goodness of fit of logistic models.

Cost Analysis Methods

Medical Care Costs of Pneumonia and Influenza Episodes

Costs of providing inpatient care for pneumonia and influenza episodes were obtained from Kaiser Permanente accounting systems for the various patient care cost centers. Costs of providing outpatient care were estimated from the 5% research sample, using relative value units included in the outpatient database and unit costs for accounting cost centers.

Cost of Vaccine Delivery

Costs of the vaccine delivery program were estimated by a micro-costing approach and included overhead costs. The purchase cost of vaccine was obtained from administrative records and included all vaccines used and wastage. Distribution costs were allocated based on the proportion of direct drug costs represented by influenza vaccine.

Two methods of promoting the vaccination program were identified: an article in the member newsletter and referrals by nurses, generally by the telephone advice nurse. Time estimates were obtained from interviews with providers, and labor costs per hour were obtained from Kaiser Permanente administrative records.

Cost-effectiveness Ratios

Three cost-effectiveness measures were computed: cost per prevented episode of outpatient pneumonia and influenza, cost per prevented episode of hospitalization, and cost per prevented hospital death.

Net Financial Benefit

Only direct monetary costs and benefits to the HMO were considered. The formula, treating costs as negative benefits, is as follows: net benefits = (number of hospitalizations prevented) (mean cost per hospitalization) + (outpatient contacts prevented) (mean cost per outpatient contact) –(number of vaccinations) (unit cost of vaccine and supplies + unit cost of vaccine wastage + unit cost of vaccine promotion + unit cost of vaccine delivery + unit cost of overhead) –(number of adverse reactions to vaccine)(mean cost of treating adverse reactions).

Results

Top Methods Results Discussion Author & Article Info References

Over the nine study periods, 58% of person-periods were contributed by women, 37% were contributed by persons older than 74 years of age, and 40% were contributed by persons at high risk for complications of influenza. The influenza and pneumococcal vaccination rates for high-risk elderly persons were 32% and 16%, respectively. For non-high-risk person-periods, the vaccination rates were 22% for influenza and 7% for pneumococcal infection. Vaccinations done after the start of epidemics accounted for 6% of total vaccinations.

There were 765 hospitalizations due to pneumonia and influenza in 734 persons, and 106 in-hospital deaths due to pneumonia and influenza occurred among high-risk elderly persons during the nine study periods. Among non-high-risk elderly persons, there were 129 hospitalizations in 128 persons and 7 in-hospital deaths. Sixty-six outpatient episodes among 62 high-risk persons and 47 outpatient episodes among 47 non-high-risk persons were identified in the 5% research sample during the 1980-81 to 1987-88 study periods.

The aggregate analyses of hospitalizations, deaths, and outpatient episodes among high-risk elderly persons included 1474, 1476, and 1250 controls, respectively. The corresponding numbers of controls included in the aggregate analyses for non-high-risk elderly persons were 2291, 2292, and 1612, respectively. Controls in both risk strata contributed person-periods of observation to an average of 3.3 study periods.

Chronic pulmonary and cardiovascular conditions were the major high-risk conditions; these accounted for 87% and 82%, respectively, of high-risk conditions among high-risk cases and controls. The direction of this small difference would be expected to result in a slight underestimation of vaccine effectiveness. No direct medical care attributable to adverse reactions to influenza vaccine was detected in the 5% research sample during the nine periods. Table 2 shows excess pneumonia and influenza hospitalization rates for the influenza epidemic periods relative to the 1981-1982 nonepidemic reference period. The high-risk populations experienced significant excesses during all epidemic periods except 1986-1987. Significant excesses occurred in the non-high-risk populations only during 1982-1983, 1983-1984, 1984-1985, and 1988-1989.

Table 3 presents the fitted logistic regression models for pneumonia and influenza episodes when the data from all nine study periods were pooled. Nonuse of influenza vaccine was associated with increased risk for pneumonia and influenza hospitalization. Age was associated with increased risk for all types of pneumonia and influenza episodes. Male sex was associated with an increased risk for hospitalization due to episodes of pneumonia and influenza among high-risk elderly persons and with increased risk for outpatient episodes of pneumonia and influenza episodes among non-high-risk elderly persons. Among high-risk elderly persons, use of pneumococcal vaccine was associated with increased risk for episodes of pneumonia and influenza requiring hospitalization. High-risk elderly persons were more likely to have received pneumococcal vaccine. Controlling this proxy for level of risk slightly increased the estimated effectiveness of influenza vaccination. All the logistic models fit the data well, except for the outpatient pneumonia and influenza model for non-high-risk elderly persons, which was estimated from only 47 episodes.

Vaccine Effectiveness

Influenza vaccination was 30% (CI, 17% to 42%) effective in preventing hospitalizations due to pneumonia and influenza among high-risk elderly persons when the data from all nine study periods were pooled. The aggregate effectiveness of influenza vaccines in preventing in-hospital death due to pneumonia and influenza among high-risk elderly persons was 33% (CI, –7% to 58%), not significantly different from zero. Aggregated influenza vaccine effectiveness in preventing hospitalizations for pneumonia and influenza among non-high-risk elderly persons was 40% (CI, 1% to 64%). Influenza vaccines were 83% (CI, 54% to 94%), 70% (CI, 9% to 90%), and 51% (CI, 13% to 73%) effective in preventing hospitalizations for pneumonia and influenza among high-risk elderly persons during the 1980-1981, 1982-1983, and 1986-1987 epidemics, respectively. Effectiveness estimates for the other five epidemic periods ranged from 15% to 32% and were not statistically significant. The effectiveness estimates are generally consistent with the levels of cross-reaction between epidemic and vaccine strains during these epidemics (Table 2).

Cost Analysis

The total medical care costs for pneumonia and influenza were $45.50 and $5.14 per person-period for high-risk and non-high-risk elderly persons, respectively. Overall, the total medical care cost for pneumonia and influenza was $21.27 per person-period (Appendix Table).

The estimated cost of delivering a vaccination during the nine seasons was $7.11, of which $2.76 was for overhead, $2.35 for delivery of the injection, $1.45 for vaccine and supplies, $0.29 for promotion, and $0.26 for wastage.

Aggregated over the nine periods, the estimated cost per hospitalization prevented among high-risk elderly persons was $3234 (Table 4). The aggregate cost per hospitalization prevented in non-high-risk elderly persons was $18 817.

For the nine seasons, the high-risk vaccination program prevented 83 inpatient and 194 outpatient episodes, saving $496 154. When this savings is compared with the cost of the vaccination program, a net savings of $6.11 (CI, -$0.23 to $11.53) per vaccination was achieved. Cumulated over the 9 seasons, the non-high-risk vaccination program prevented 12 inpatient and 135 outpatient episodes, saving $72 759. This savings was less than the cost of the vaccination program and resulted in a net cost of $4.82 (CI, $3.34 to $7.06) per vaccination. For all elderly persons, the HMO saved $1.10 in direct medical care costs per vaccination (CI,-$3.36 to $4.71).

Although influenza vaccines were 40% (CI, 1% to 64%) effective in preventing inpatient episodes among non-high-risk elderly persons, the low hospitalization rate in this population did not result in enough prevented episodes to produce net savings. The finding that influenza vaccination programs were cost saving for the elderly is sensitive to the vaccine effectiveness estimates. For high-risk elderly persons, the upper confidence estimate of net cost per vaccination is only $0.23, and the HMO recovered 97% of the $7.11 cost per vaccination. For non-high-risk elderly persons, essentially none of the HMO's costs were recovered. Optimistically, for high-risk elderly persons, the HMO saves $11.53 per vaccination and incurs a net cost of $3.34 per vaccination for non-high-risk elderly persons. There is an overall savings of $4.71 per vaccination.

Discussion

Top Methods Results Discussion Author & Article Info References

We show that, in the long term, a program of annual influenza vaccination is cost-effective and cost-saving or cost-neutral for high-risk elderly persons. Indirect benefits such as prevention of suffering, incapacity, and lost wages are likely to compensate for the small net cost of vaccinating non-high-risk elderly persons. Because our cost-effectiveness and net financial benefit results are based on HMO costs rather than charges, they cannot be directly generalized to the fee-for-service sector in which most elderly persons receive their medical care. However, we would expect equally effective vaccination programs in the fee-for-service sector to have achieved even greater net financial benefits because fee-for-service hospital admission rates and lengths of stay were higher than in HMOs during the 1980s. Although vaccination program costs are expected to be lower in HMOs, we would speculate that the HMO-fee-for-service differential in vaccination program costs would not entirely compensate for the differential in medical care costs. On the basis of the 1988-1992 Medicare Influenza Vaccine Demonstration, the Health Care Financing Administration applied various cost, vaccine coverage, and vaccine effectiveness estimates and concluded that influenza vaccine was cost-effective for Medicare and was possibly cost-saving, depending on the effectiveness of the vaccine and the level of vaccine coverage [15].

Our estimate of $1.10 cost savings (in 1985 dollars) per vaccination over the nine study periods is considerably smaller than the U.S. Congressional Office of Technology Assessment estimate of $4.00 cost savings (in 1978 dollars) per vaccination [16, 17]. Their simulation study was based on Kaiser Permanente's records during the 1968-1969 and 1972-1973 epidemics, when both vaccine effectiveness and excess pneumonia and influenza hospitalization rates were higher [18, 19]. Our study results provide a basis formulating influenza vaccination policies and financing annual vaccination programs in HMOs and other managed care organizations. Substantial additional medical care cost-saving would be expected to accrue to the study HMO if the influenza vaccination rate among high-risk elderly were increased from the 32% rate achieved in the 1980s. The ultimate cost benefit to such organizations depends, of course, on their ability to overcome organizational, provider, and, most of all, member barriers to vaccination programs. Consequently, further research is needed to fully understand the nature of these barriers and the likely success of various educational programs in overcoming them [20-22]. Our study shows, however, that managed care organizations have a financial incentive to institute influenza vaccination programs for their elderly members.