Objective: To evaluate the incremental effect of a second clientreminder postcard or an influenza tool kit targeted toward employerson increasing influenza vaccination rates among adults age <65years at high risk for complications from influenza illness.
Methods: In this demonstration study, enrollees of 3 managedcare organizations (n = 8881) were randomized at the employerlevel into 4 arms: 1 postcard, 2 postcards, 1 postcard + tool kit, and2 postcards + tool kit. The postcards and tool kits were mailed duringthe fall of 2001, and their effect on influenza vaccination rateswas assessed through a survey.
Results: Compared with a single postcard, 2 postcardsincreased vaccination rates by 4 percentage points (adjusted relativerisk = 1.05; < .05) among persons aged 50 to 64 years butdid not have any effect among younger adults. Older adults had agreater burden of disease and reported more favorable knowledgeand attitudes toward the influenza vaccine. The influenza tool kitdid not appear to have any incremental effect on vaccination rates.
Conclusions: Our findings underscore the necessity of evaluatingthe effectiveness of interventions in different population subgroupsand of identifying factors that modify the effectiveness ofinterventions. Rigorous assessment of intervention effectiveness inmanaged care settings will enable decision makers to optimize useof scarce healthcare dollars for improving the health and well-beingof enrollees.
(Am J Manag Care. 2004;10:698-702)
There is insufficient evidence of efficacy to recommendthat client reminders be used alone toincrease influenza vaccination among personswith high-risk conditions.1 Interventions to increaseclient demand in combination with provider-basedinterventions or interventions to increase access to vaccinationservices have been reported to increase vaccinationamong such persons.1 However, studies areneeded to resolve several issues, including the relationshipbetween frequency of reminders and effectiveness,and identification of the least and most effective combinationsof services in multicomponent interventions.2In this paper, we report results from a demonstrationproject where the incremental effectiveness of a secondclient reminder postcard or an influenza tool kit wasassessed.
InternationalClassification of Diseases, 9th Revision, ClinicalModification (ICD-9-CM)
The target population comprised Colorado residentsaged 18 to 64 years with high-risk conditions who wereenrolled in any of the 3 participating managed careorganizations (MCOs) as of September 1, 2001. Theproduct type was HMO for 2 MCOs, and both HMO andpreferred provider organization (PPO) for the third.High-risk conditions among commercially insuredadults were identified by using codes in insurance claimsdata, reviewed in September 2001 for the period fromMay 1, 2000, to April 30, 2001.3 The claims algorithmrequired 1 encounter in an acute-care hospital settingor 2 encounters in other settings for any of the followingconditions: diabetes, chronic cardiovascular disease,chronic pulmonary disease, immunosuppression,chronic renal failure, or hemoglobinopathy. Pharmacydata also were reviewed to identify diabetic patients(insulin or oral hypoglycemics/antihyperglycemics dispensedat least once during the 1-year period).
The postcards, mailed under each MCO's logo,included statements that flu shots were strongly recommendedfor persons with certain high-risk conditionsand that "it's time for your flu shot, it could save yourlife." Enrollees were instructed to call their primarycare physician (or influenza vaccination clinic sponsoredby 1 of the MCOs). The first postcard was mailedduring the last week of October 2001, and the secondpostcard was mailed a month later.
The influenza tool kit, mailed to employers onOctober 12, 2001, included (1) educational messagesfor employers, including the business case for influenzavaccination; (2) flyers, posters, a newsletter article, ande-mail and payroll stuffer communications to encouragevaccination; and (3) tips and a check list for implementingwork site, employer-sponsored influenza vaccinationclinics.
Because a large number of employers contract withthe MCOs and because there was concern about thecost effectiveness of sending tool kits to employers withfew eligible persons, we excluded employers with 3 orfewer high-risk persons (subscribers and dependents)enrolled in the 3 MCOs. This reduced the number ofemployers under consideration by 80% but resulted in aloss of only 14% of the persons with high-risk conditions.Out of 611 employers with 4 or more eligible personslocated in Colorado, 505 were selected usingsimple random sampling.
The 505 employers were randomly allocated to 1 ofthe following arms: 1 postcard, 1 postcard + tool kit, 2postcards, or 2 postcards + tool kit. For the 16 employerswith 126 or more persons at high risk, 125 personswere randomly selected so that a few extremely largeemployers did not overshadow the findings. For theremaining 489 employers, all persons with high-riskconditions were selected. Persons in the 4 arms totaled8881.
Managed care organization administrative databaseswere used to obtain information on age, sex, subscriber/dependent status, and MCO product type. Dataon receipt of influenza vaccination in September 2001through March 2002 and other variables were collectedthrough a bimodal mail-telephone survey conductedduring summer 2002. The survey was administered to6657 persons who were still enrolled in the MCOs as ofthe time of the survey. The survey administration rateswere similar across the 4 arms, ranging from 74% to76%. The survey response rate was 60% (3996/6657).
Because the interventions were assigned at theemployer level, adjustment for potential clustering ofresponses within employers was performed by fittinglogistic regression models with generalized estimatingequations using STATA, Version 8.2.4,5 Adjusted oddsratios were transformed to adjusted relative risks.6
The distributions of age, sex, ethnicity, self-report ofhigh-risk condition, knowledge and attitudes towardinfluenza vaccination, and receipt of prior influenzavaccination were similar across the 4 arms (Table 1).However, differences were noted in the distributions ofrace, marital status, education, subscriber status, MCO,PPO product, and availability of influenza vaccinationsat the work site.
Among adults aged 18 to 64 years, 69% (2753/3996)received an influenza vaccination during the 2001-2002season. Because regression modeling indicated that agemodified the effectiveness of the intervention, we presentresults separately for the 18 to 49 and 50 to 64 agegroups.
Adjusted analysis shows that a second postcard orthe influenza tool kit did not significantly increaseinfluenza vaccination rates among persons aged 18 to49 years (Table 2). For persons aged 50 to 64 years,those in the "2 postcard" or "2 postcard + tool kit" armshad a 6 percentage point higher vaccination rate thanthose receiving a single postcard (adjusted relative risk= 1.08; < .05).
Because the tool kit did not appear to increase therate of vaccination (Table 2), we repeated the analysisby collapsing the 4 arms into 2 groups (Table 3). Amongpersons aged 50 to 64 years, the 2-postcard group had a4 percentage point higher vaccination rate than the 1-postcard group (adjusted relative risk = 1.05; < .05).
To assess why age modified the effect of the intervention,we compared characteristics of persons in the2 age groups. Older persons had claims evidence of moredisease: 16% of 50- to 64-year-olds had 2 or more high-riskconditions, compared with 7% of 18- to 49-year-olds( <.001). Older adults also reported more favorableknowledge and attitudes toward influenza vaccine(Table 4).
Our study showed that a second postcard reminderincreased influenza vaccination by 4 percentagepoints among 50- to 64-year-old persons with high-riskconditions, but did not have any effect among youngeradults. Influenza tool kits mailed to employers did nothave any incremental effect among persons who weremailed single or multiple postcard reminders.
Our study had certain potential limitations. First, weused self-report to ascertain vaccination status.However, self-report of influenza vaccination has beenshown to be a reliable and valid measure.7,8 Any misclassificationin vaccination status is likely to be nondifferentialacross the study arms, which would resultin underestimation of the intervention effect.9 Second,the effect of 2 postcards could have been distortedbecause of attrition due to disenrollment, nonresponse,or incomplete information. However, because attritionrates were similar in the study groups, the effect estimatorwill be unbiased.10 Third, it could be assertedthat, although the findings were internally valid, theresults may not apply to persons who were excludedbecause of attrition. Excluded persons were similar toincluded persons with regard to distributions of plan,product type, and sex, but the age distribution differed(persons aged 50-64 years comprised 55% of the includedvs 46% of the excluded adults; <.001).Nonetheless, because we stratified our analysis by age,bias because of differing age distributions was minimized.Fourth, the 2001-2002 influenza vaccinationseason was characterized by vaccine supply problems.11 Although it was recommended that priority begiven to persons with high-risk conditions,11 it is possiblethat the staggered or delayed supply of vaccine mayhave attenuated the effect of the intervention. Finally,our regression model adjusted for differences in theavailability of work-site influenza vaccinations acrossthe study arms, which could have masked the effect oftool kit—motivated work-site vaccination clinics.However, excluding the work-site vaccination variablefrom the model did not alter our finding that the toolkit was ineffective. If an employer received a tool kit inOctober, it would likely be too late to implement a vaccinationclinic that year.
The observation that 2 postcard reminders wereeffective among older persons may be because of fewerbarriers in the form of negative attitudes toward andbeliefs about influenza vaccination.12 Two postcardreminders are likely to be cost effective, largely due toavoidance of influenza-associated productivity losses.The finding that the influenza tool kit did not have anincremental effect might be because employers did notuse the tool kit or because the tool kit may be adding littleto the postcard messages. However, we cautionagainst extrapolating our findings to interventionsinvolving tool kits with more intensive outreach efforts.Also, our findings may not be generalizable to populationswith lower vaccination rates.
Our findings have several implications. First, mailingtool kits to employers did not have any incrementaleffect on vaccination rates. Developing methods tomore effectively encourage employers to use such toolkits merits attention. Second, 2 postcards produced anincrease in vaccination among persons aged 50 to 64years but not among younger adults. This observationunderscores the fact that interventions need to beevaluated in different population subgroups. Furtherresearch is needed to identify factors that modifyeffectiveness. With 170 million enrollees in MCOs,efforts to further improve preventive service deliveryby MCOs are laudatory, although limited resourcesshould not be committed to interventions before rigorousevaluation proves themeffective.
The collaborating sponsors of thedemonstration project were America'sHealth Insurance Plans, Alliance ofCommunity Health Plans, Blue Crossand Blue Shield Association, Centers forDisease Control and Prevention,Employers Managed Health CareAssociation, National Business Group onHealth, National Business Coalition onHealth, and National Institute for HealthCare Management Foundation. TheColorado participants included projectmanagement by the Colorado BusinessGroup on Health, and participation ofThe Alliance, the Colorado Foundationfor Medical Care, the ColoradoDepartment of Public Health andEnvironment Immunization Program,the Colorado Influenza andPneumococcal Alert Coalition, AnthemBlue Cross Blue Shield of Colorado,Kaiser Permanente Colorado, andPacifiCare of Colorado.
From the Division of Prevention Research and Analytic Methods, Epidemiology Program Office, Centers for Disease Control and Prevention, Atlanta, Ga (FA, CF, AF, PS); Anthem Blue Cross and Blue Shield, Denver, Colo, and the Department of Obstetrics and Gynecology, University of Colorado Health Science Center, Denver (LML); the Clinical Research Unit, Kaiser Permanente Colorado, Denver (EWN, EKF); and the National Immunization Program, Centers for Disease Control and Prevention, Atlanta, Ga (SN).
This research was supported by the Centers for Disease Control and Prevention.
Address correspondence to: Faruque Ahmed, PhD, Centers for Disease Control and Prevention, Mail Stop K53, 4770 Buford Highway NE, Atlanta, GA 30341-3717. E-mail: firstname.lastname@example.org.
The Guide to Community PreventiveServices: What Works to Promote Health?
1. Briss P, Ndiaye SM, Rodewald L, et al. Vaccine-preventable diseases. In: TaskForce on Community Preventive Services, ed. New York, NY: Oxford University Press.In press.
Am J Prev Med.
2. Briss PA, Rodewald LE, Hinman AR, et al. Reviews of evidence regarding interventionsto improve vaccination coverage in children, adolescents, and adults. The TaskForce on Community Preventive Services. 2000; 18(1 suppl):97-140.
International Classification ofDiseases, Ninth Revision, Clinical Modification
3. US Department of Health and Human Services. . Washington, DC: US Dept ofHealth and Human Services; 1980. Publication PHS 80-1260.
4. Stata Corporation. Stata statistical software, version 8.2. College Station, Tex:Stata Corporation; 2003.
5. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. 1986; 42:121-130.
J Chronic Dis.
6. Flanders WD, Rhodes PH. Large sample confidence intervals for regression standardizedrisks, risk ratios, and risk differences. 1987;40:697-704.
Am JPrev Med.
7. Nichol KL, Korn JE, Baum P. Estimation of outpatient risk characteristics andinfluenza vaccination status: validation of a self-administered questionnaire. 1991;7(4):199-203.
Am J Prev Med.
8. Mac DR, Baken L, Nelson A, Nichol KL. Validation of self-report of influenzaand pneumococcal vaccination status in elderly outpatients. 1999;16(3):173-177.
9. Gordis L. More on causal inferences: bias, confounding, and interaction. In: Philadelphia, Pa: WB Saunders Company;2000:204-208.
Am J Epidemiol
10. Stang A, Jockel KH. Studies with low response proportions may be less biasedthan studies with high response proportions. . 2004;159:204-210.
MMWR Morb Mortal Wkly Rep
11. Delayed influenza vaccine availability for 2001-02 season and supplementalrecommendations of the Advisory Committee on Immunization Practices.. 2001;50(27):582-585.
12. Chapman GB, Coups EJ. Predictors of influenza vaccine acceptance amonghealthy adults. . 1999;29(4):249-262.