The Healthy Worker Project: AWork-Site Intervention for WeightControl and Smoking Cessation

Robert W. Jeffery, PhD, Jean L. Forster, PhD, MPH, Simone A. French,PhD, Steven H. Kelder, MPH, Hariy A. Lanido, PhD, Paul G. McGovem,PhD, David R. Jacobs, Jr., PhiD, and Judith E. Baxter, PhD

-o4^f, > 8 s .s .z, .:s. ..v.§v: ... ..................._...

_~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....

IntroductionWork-site health programs have been

discussed frequently in recent years as ameans for promoting behavior change inthe general population. Attractive fea-tures of these interventions include con-venient access to populations in need, theopportunity to capitalize on the social sup-port resources of work sites as communi-ties, and the potential for cost recoverythrough reduced absenteeism and healthcare expenditures. 12 Despite these appar-ent strengths, empirical data on the effec-tiveness of work-site health programs arerelatively sparse.

Evaluation of work-site health pro-motion programs is complicated by thefact that such programs vary widely in ob-jectives, content, organization, cost, andsetting. Some are simple and inexpensive(e.g., distribution of health informationpamphlets),3 while others are intensive(e.g., systematic risk factor screening andintensive follow-up counseling).4'5 Someconcentrate on a single risk factor such asobesity,6'7 smoking,8'9 or blood pres-sure, "' while others target multiple behav-ioral objectivcs.'A""-'3 In addition, somefocus on individual health behaviorchange (i.e., counseling of individuals athigh risk),8 while others address the socialaspects of the work environment (e.g.,work-site competitions)6"14"15 or institu-tional health policies (e.g., smoke-freeenvironments).1'-21

Several methodological factors alsomake interpretation of work-site interven-

tion data difficult.3'8'22 Methodologicalproblems include the following: studyingfew companies, which limits generaliz-ability; not including control groups or not

randomizing to intervention and controlconditions; limiting outcome evaluation to

program participants rather than evaluat-ing intervention effects on the entire workforce; evaluating effects over short timeperiods; and analyzing results using indi-viduals as the unit of analysis, even thoughthe work site is the unit of "assignment"and it is widely recognized that character-istics of work sites may have an importantinfluence on results.2

Results of work-site interventions todate have been mixed. Recruitment towork-site smoking programs, for exam-ple, has ranged from 0% to 88c.3 Dropoutrates from work-site weight loss programshave ranged from 0.5%1C3 to 80%..3_23 Mostimportant, the reported success of work-site interventions in changing health be-havior has ranged from minimal24 to verypromising.3.6 8'4'5 Clearly, additionalstudies involving strong research method-ologies would be helpful.

The present paper reports the resultsof such a study, a large-scale randomizedtrial called the Healthy Worker Project thatfocused on two common health risk fac-tors: obesity and cigarctte smoking. Thesetwo risk factors were selected because oftheir high prevalence in the US popula-tion,26-27 their important causal role in thedevelopment of disease,28299 the fact thatmost obese people and smokers want to

The authors are with the Division of Epidemi-ology, University of Minnesota School of Pub-lic Health, Minneapolis.

Requests for reprints should be sent toRobert W. Jeffery, PhD, Division of Epidemi-ology, University of Minnesota School of Pub-lic Health, 1300 S Second St (Suite 300), Min-neapolis, MN 55454-1015.

This paper was submitted to the JournalOctober 24, 1991, and accepted with revisionsOctober 27, 1992.

Editor's Note. See related editorial by Al-derman (p 313) in this issue.

American Journal of Public Health 395

Jeffery et al.

change,30,31 and the fact that traditionalmedical approaches have so far proved in-adequate to the task of ameliorating thesebehaviors at the population level.32,33

Intervention methods for the trialwere chosen on the basis of a careful re-view of existing work-site programs andwere pilot tested extensively to establishtheir attractiveness to employees and theirshort-term efficacy.2534-36 An importantstrength of the trial was its research meth-odology. Work sites were recruited from acomprehensive enumeration of sites in amajor metropolitan area. Sites were ran-domized to treatment or control condi-tions, and the effects of the interventionwere evaluated over an extended time pe-riod at several levels of employee aggre-gation.

This report focuses on the primaryoutcomes of the Healthy Worker Project,namely changes in obesity and smoking.The primary hypotheses ofthe studywerethat the prevalence of smoking and meanbodyweight would be reduced among em-ployees in treated work sites comparedwith those in control work sites. Resultsare reported for both cross-sectional andcohort designs.

Meth4dsWork Sites

The trial was conducted between fall1987 and fall 1990 in the seven-county met-ropolitan area surroundingMinneapolis/St.Paul, Minn. The region has a population ofapproximately 2.5 million. Participatingwork siteswere recruited from a listingpur-chased from Dun's Marketing Service. Asite was defined as a physically contiguouslocation with awork force ofprimarily full-time employees. For design and logisticalreasons, sites with between 400 and 900employees were recruited. Prospectivework siteswere approached first with a let-ter describing the study, and then were fol-lowed up by phone. For sites with contin-ued interest, a personal interview wasscheduled after which the work sites eitherdecided to participate or declined. Of the154 sites approached, 36 were found to beineligible, usually because site-size infor-mation was incorrect; 83 declined partici-pation; 2 could not be reached by phone;and 1 was excluded by the investigatorsaftercompletion ofthe baseline survey, butbefore randomization, as a result of poorresponse rate. Thus, 32 siteswere includedin the study (a 27% response rate). Prelim-inary analyses compared sites agreeing to

participate with those refusingon a number

of characteristics. Sites electing to partici-pate were slightly larger (610 vs 545 em-ployees,P < .05), had a lower average in-crease in number of employees over theprevious 5 years (17% vs 67%, P < .04),and were more likely to be in the publicsector (39% of participating vs 5% of re-fusing sites, P < .0005). The most com-mon reasons given for refusing to partici-pate were time, inconvenience, ordisinterest (54%); specific conditions of theproject, such as randomization (29%); con-flicts with existing programs (18%); andgeneral business conditions such asmerger, expansion, or economic downturn(19%). Primary functions of sites includedin the study included insurance, primaryhealth care, financial services, manufac-turing, education, electronic assembly,bulk mail distribution, research and devel-opment, and city, county, state, and fed-eral government operations.

Research Design and EvaluationSurveys

Sites agreeing to participate in thestudy completed a baseline survey, wererandomized to either intervention or notreatment for 2 years, and then were resur-veyed. The baseline survey was given to200 employees selected at random fromeach site. The same individuals were alsocontacted 2 years later at follow-up as acohort sample. In addition, a new sampleof200 employeeswas selected at follow-upfor a cross-sectional comparison. Samplingwith replacement was used so that thecross-sectional surveyswould be represen-tative of all employees at each time point.Thus, an average of about one third of theemployees in the second random samplewere also included in the cohort. This de-sign allowed evaluation of intervention ef-fects both in those continuously employed,and thus exposed to intervention for 2 fullyears, and on the employee population asa whole, including attrition and replace-ment.

Evaluation surveys included assess-ment by questionnaire of demographiccharacteristics, smoking and weight losshistories, diet and exercise habits, job char-acteristics, health history, and job satisfac-tion. Direct measures were taken of heightand weight. Expired air carbon monoxidewas also assessed for all employees as avalidation of reported smoldng status.37 In-dividuals who did not attend on-site surveysessions were contacted by telephone in 29ofthe 32 sites (employee telephone numberscould not be obtained for 3 sites). Individ-uals in the cohortwho had left employmentwiththe site atfollow-upwere also surveyed

by phone. The phone surveys included as-sessment by self-report of height, weight,and smoling status. Survey response ratesat baseline averaged 75% (range = 61% to90%) for the on-site surveys and 77%(range = 36% to 90%o) forthe telephone sur-veys, an average net total response rate of92% (range = 61% to 99%). At follow-up,average completion rates for on-site surveysin the cross section and cohort were 77%(range = 50% to 93%) and 76%(range = 48% to 90%o), respectively. Tele-phone follow-up completion rates averaged87% (range = 62% to 100%) and 83%(range = 71% to 100%) and total responserates were 94% (range = 87% to 99%) and93% (range = 49% to 100%) for the cross-section and cohort, reswtively. Analysisofdifferences between individuals surveyedat work vs over the phone showed that thelatter were older (38.5 vs 37.8 years,P < .05), had higherjob status (46%vs41%managerial, P < .007), were more likely tosmoke (30.8% vs 23.0%, P < .0001),weighed less (relative weight: 1.19 vs 1.23;P < .001), and were more likely to reporthaving had diabetes, hypertension, heartdisease, or chronic obstructive lung disease(19% vs 16%,P < .02). Follow-up surveyswere completed in all 32 sites. Net em-ployee attrition between baseline and fol-low-up, defined as the percentage of em-ployees in the baseline surveywho were nolonger employed at that site at follow-up,averaged 19% (range = 6% to 43%).

InterventionIntervention was based on proce-

dures developed in successful pilotstudies.25,3436 These procedures con-sisted of a combination of on-site classesand an incentive system organized throughpayroll deduction. Four rounds were of-fered at each site during the 2 years of in-tervention, each round consisting of 11 bi-weekly sessions. Classes were held on siteand on employee time, and multiple classtimes were offered to accommodate shiftworkers. Class content consisted of state-of-the-art interventions for smoking cessa-tion35 and weight loss34 involving behaviormodification principles. The interventionswere led by professional health educatorswho were trained, supervised, and paid bythe study. All employees were eligible forthe weight program. All smokers were in-vited to participate in the smoking cessa-tion program, as were ex-smokers con-

cerned about remaining abstinent. Weightloss goals were selected by participantsthemselves, with a minimum of 0 lb and a

maximum of 1% of body weight loss eachweek. Permanent cessation was the goal

March 1993, Vol. 83, No. 3396 American Journal of Public Health

Weight Control and Smoking Intervention

for all smokers in the smoking program(i.e., reduction in smoking rate was not ac-cepted as a personal goal). An incentivecomponent was selected for use in thisstudybecause many ofthe most successfulprior work-site studies had used incentivestrategies.38-41 Participants selected anamount ofmoney tobe deducted from eachpaycheck (a minimum of $5 biweekly).Employees in the weight program receiveda refund at each session if they had madeprogress toward their weight loss goals.Expired carbon monoxide level was mea-sured for participants in the smoking ces-sation program. They received a refund ofall money in their account at any session inwhich they had carbon monoxide valuesless than 8 ppm (no-smoking level). At theend of the program, all incentive funds thathad not been returned to participants wereused in a way negotiated at each site (theywere usually given to charity).

Employees were recruited to interven-tion programs systematically. Before eachround of intervention classes, all employeeswere contacted through company mail andinvited to participate. Recruitment also in-cluded posters, word of mouth, cafeteriabooths, and a variety ofspecial promotionalactivities specific to sites.

Analysis StrategyAnalyses reported in this paper are

basedon all employees surveyed, whetherin person or by phone and whether or notthey were employed throughout the 2years of the study. Body mass index (kg/mm2) was used as the measure of obesity.The body mass indices of employees re-porting their heights and weights by tele-phone were adjusted, on the basis of theobserved relationship between self-re-ported and measured body mass index inthe on-site sample, by means of gender-specific regression equations. The out-come variable related to smoking was siteprevalence. When carbon monoxide mea-sureswere available, smokingwas definedeither as a self-report of smoking or as aself-report ofnonsmoking with an expiredair carbon monoxide value above 5 ppmafter subtracting the work-site ambientcarbon monoxide level.42 Reports ofsmoking among those surveyed by phonewere accepted at face value.

It is believed that these methods fordealingwith self-report data are without sig-nificant bias. Several investigators whohave explored the self-reportbias forweighthave noted that people tend to underreporttheir weight.4344 However, underreportingis no greater among those in treatment forweight loss than it is in the general popula-

tion.4546A recent review ofthe literature onthe assessment of outcome in smoldng ces-sation studies has similarly concluded thatfalse reports of nonsmoking are generallylow.47 Others, however, have reported sub-stantial discrepancies between self-reportedquitting and biochemically validated cessa-tion.48 In this study, direct assessments offalse reports of nonsmoking were availableamong those for whom carbon monoxidemeasures were taken. The false report rate,as defined above, averaged about 2%.These rates did not differ significantly atbaseline and follow-up or between treat-ment and control work sites at either timepoint.

In addition to the analyses reportedhere, separate analyses restricted to sub-sets of the population (e.g., those sur-veyed in person or by phone only) werealso performed. These analyses yieldedsimilar estimates ofboth the direction andmagnitude of treatment effects. Statisticalpower, however, tended to be reduced be-cause the sample sizes per site are smallerand thus produce less stable estimates ofcompany means (e.g., restricting analysesto those employees who had direct mea-surements only reduces the sample sizeper site in the cohort analysesby over50%[from 175 to 80].

The analysis strategy used for this re-search is based on a conceptual frameworkthat explicitly recognizes that both differ-ences between individuals within worksites and differences in organizational char-

acteristics between sites contnbute to vari-ance in health behavior. As descnibed morefully elsewhere,49 the selection of 32 sitesand of 200 individuals per site in surveysrepresents an effort to balance these twofactors. Analyses ofstudyhypotheses usedwork site as the unit of analysis and wereperformed in two steps. First, adjustedmeanbodymass indices and adjustedprev-alences of smoking were generated foreachwork site; age, gender, education, oc-cupation, and marital status were used ascovariates. Body mass index analyseswere also adjusted for smoking. The sec-ond step in the analysiswasto compare thecompany means by repeated-measuresanalysis of variance.

ResultsSelected characteristics of employees

and work sites involved in this trial areshown in Table 1. Sites averaged about 600employees each; 12 were public sector and20 private sector. Employees averaged 38years of age, and slightly over half werewomen. About 40o identified themselvesas having professional or managerial job ti-tles, a simflar proportion identified them-selves as having clerical or sales positions,and slightly less than 20%o were blue-collarworkers (i.e., laborers, skilled productionworkers, caftspeople, etc.). Both men andwomen had mean body mass indices thatwere about 10%o above actuarial ideals.5Between 24% and 25% reported being cur-

American Journal of Public Health 397March 1993, Vol. 83, No. 3

Jeffeiy et al.

rent smokers. Treatment and control sitesdid not differ significantly.

Table 2 summarizes participationrates and results of the smoking cessationand weight loss classes held in the inter-vention work sites. In the 2 years of in-tervention, 270 and 2041 employees par-ticipated in the smoking and weightprograms, respectively. Fifteen peopleenrolled in the smoking program morethan once, as did 465 people who enrolledin the weight program. Over the 2 years,about 12% of all smokers participated inthe smoking program. Sixteen percent ofall employees, and 36% of obese employ-ees (obesity was defined as .120% of theactuarially defined "ideal" weight5) par-ticipated in the weight program over 2years. Participation was highest the firsttime the programs were offered. Subse-quent enrollment fell quickly and stabi-lized at about a third ofthat achieved in thefirst round. Analyses of participation pat-terns showed that women were morelikely to participate than men, that profes-sional and clerical/sales personnel wereabout 1.6 times as likely as blue-collarworkers to participate in the weight pro-gram, and that professionals were 2.6times as likely as either clerical/sales orblue-collar workers to participate in thesmoking program."8 Across the entire2-year intervention period, 43% of smok-

ers quit (defined as having a carbon mon-oxide level of 8 ppm or less at the lastsession attended), and the average per-person weight losswas 4.8 lbs. Short-termsmoking cessation rates and mean weightlosses were best in round 1, suggesting,perhaps, that employees participating inlater rounds tended tobemore recalcitrantcases.

Table 3 shows the adjusted smokingprevalence at baseline and follow-up foreach of the control and treatment worksites in the trial. Table 4 shows comparabledata with respect to body mass index. Ta-ble means are adjusted for age, sex, edu-cation, occupation, and marital status.Body mass index means are also adjustedfor smoking status. In the cross-sectionalsamples, smoking prevalence increased byan average of one percentage point be-tween baseline and follow-up in the controlcompanies (range = -4.88% to 12.25%).Smoking prevalence in the treated sites de-creasedby approximately three percentagepoints (range = -10.84% to 4.28%), F (1,30) 3.73,P = .0581. In the cohort, smokingprevalence decreased by approximatelyone percentage point in the control sites(range = -5.53% to 2.59%) and decreasedby three percentage points in the treatmentsites (range = -10.05% to 0.44%), a netdifference of approximately two percent-age points, F (1, 30) 5.19,P = .030. These

results supported the hypothesis that theprogram would have a beneficial effect onemployee smoking rates.

The likelihood that the smoking re-sults in the studywere caused by the pres-ence of treatment is supported by the ob-servation that the percentage of smokersparticipating in the smoking program ineach of the treatment companies and thecompany-specific decrease in smokingprevalence were significantly correlated(r = .45, P < .08). Higher participationwas associated with greater reductions insmoking prevalence. Additional subgroupanalyses also revealed that the treatmenteffect for smoking was present in long-term as well as short-term employees andin those surveyed in person or by phone.(Analyses restricted only to individuals forwhom biochemical validation of smokingwas obtained yielded similar estimates ofthe direction and magnitude of treatmenteffects for smoking, although power wasreduced as a result of smaller samplesizes. In cross-sectional analysis, the netdifference between treatment and controlwork sites was 4.06 percentage points,P = .08. In cohort analysis, the differencewas 1.38 percentage points, P = .29.)

Unfortunately, little change was ob-served in either treatment or control sitesin body mass index over the 2years of thisstudy. In the cross-sectional survey, em-ployees in both the treatment and controlsites had slightly lower body mass indice'sat follow-up than at baseline. In the co-hort, the control sites showed avery slightweight gain and the treatment sites a veryslight weight loss. In neither case were thetreatment and control differences suffi-cient to approach conventional levels ofstatistical significance (all Ps > .50).Thus, these results provide little supportfor the hypothesis that this work-sitehealth promotion program was beneficialin promoting companywide weight con-trol. An interesting additional observa-tion, however, was that even though therewas no main effect of treatment on bodymass index, rates of participation inweight control programs in treatment siteswere positively correlated with change inbody mass index (r = .55, P < .03).Higher participation was associated withgreater weight loss. (Analyses restrictedonly to individuals whose weights weredirectly assessed produced similar results.Change in body mass index in treatmentsites averaged -0.15 and 0.05 in the cross-sectional and cohort samples, respec-tively. Corresponding changes observedin control sites were -0.11 and 0.15, re-

spectively. Differences between treat-

March 1993, Vol. 83, No. 3398 American Journal of Public Health

Weight Control and Sminolg Intervention

ment and control sites did not approachconventional levels of statistical signifi-cance in either case.)

iswussionThis project evaluated the effects of

work-site health promotion interventionsfor weight loss and smoking cessationover a period of 2 years. In comparisonwith previous studies in this area, it had anumber of methodologic strengths. Worksites participating in the project were re-cruited from a population enumeration ofsites, thus avoiding, to some extent, biasdue to the use of convenience samples.Also, work sites were randomized totreatment and control conditions. Inter-ventions were applied systematically overan extended time period, and the studydesign allowed analysis at several levels ofemployee aggregation. Finally, sitesrather than individuals were used as theunit of analysis. Thus, it is believed thatthis trial provides a robust statement aboutthe efficacy of one approach to work-sitehealth promotion.

The experience of this study in re-cruiting sites provides information aboutthe enthusiasm ofemployers forwork-sitehealth promotion. At the outset, we be-lieved that this project would be attractiveto employers, since it offered a 50-50chance of receiving free weight loss andsmoking cessation programs in exchangefor the modest cost ofcompleting two em-ployee surveys. Nevertheless, a majorityof sites that were eligible to participatedeclined (72%). These results suggest thatthe overall receptiveness of employers towork-site health promotion may be moremodest than some advocates have sug-gested.

Findings regarding the participationand short-term results of the HealthyWorker Project intervention activities arealso instructive. In the case of smokingcessation, participation rates were rela-tively low, totaling only 12% ofsmokers ina 2-year period. The low rates contraststrikingly with our own pilot work25 andwith some previous studies in which up to88% participation has been reported.3 A12% recruitment rate, however, shouldnot be viewed as entirely negative. Othercommunity programs for smoking cessa-tion (e.g., smoking contests42 and coffe-spondence courses5l) have fared less well.The short-term smoking cessation rates of43% in this study also compare favorablywith other programs available in the com-munity.52

Participation rates for the weight lossprogram were better than for the smokingprogram, confirming previous findingsthat people are less likely to seek profes-sional help for smoking cessation thanthey are for weight loss.51 Mean weightlosses achieved byparticipantswere mod-est, however, and lower than in manystudies. This may, in part, have been dueto the fact that employees did not have tobe overweight to participate in the pro-gram and chose their own weight lossgoals.

Despite low participation, the preva-lence of smoking in treatment work sitesdecreased significantly in relation to con-trol sites, a finding that supports the valueofwork-site health promotion for reducingsmoking rates. Results in the cohort sam-

ple suggest nearly a tripling of spontane-ous quit rates in the intervention worksites, and the cross-sectional results werequantatively larger. Two points regardingthe findings deserve emphasis. The first isthat the intervention effect cannot be ac-counted for solely by the people quittingas a direct result of being in classes. The4% net reduction in cross-sectional smok-ing prevalence and 2% net reduction incohort prevalence translates into 255 and127 quitters, respectively. Since only 116smokers quit while participating in thetreatment program and some of themsurely relapsed, it seems likely that thepresence of the intervention programstimulated some employees to quit ontheir own, even though they did not enrollin classes. The second point is that the

American Joumal of Public Health 399March 1993, Vol. 83, No. 3

Jeffery et al.

S......,,,,,,f::S5~~~~~~~~~~~~7~~~~:~~~~z:~~~x~~~~:~~~~xi:x~~~~~x:S:S:fSS:f:.S-ffffi:SS:SfffSS.fy~~~.................S:.a.Ohu'-,f.-,f,ff.--.s.-.''''S°'-''-'...........'-.S'-'..-

0:=:,'p'. ,:$ 0.3$*5.03S6.1~2 S - 'o .;o S 'sy-- S S ' '-: 'S Sf ' ' 'S

o.,n°5, 25'4"50 02s5~ *5A 0412''""'''''':

.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.. S

14 279 212 S 3 25..5 S.. } S14,,,s,,,,f,S,,,,,S-''''' '0.342S"4'''''""'0'"R 2"'50"9''8''ss''S- - f '''S'f'' ---.....

.....9-........-e.- .72 7 8.-0

31o.'''.° 27404"'''"'"'' 2 4-01''''-1 28 ''0"'"22.

3D *L 0F.47 0.4.7g. 017S .0.71i tttttSgS |rWEgtt0432f0

2 25.07 2843§|0 lll043 --0 0|g21924 -0- 26*1e 2x* 0*8 257 2f51 0404

I 26.12SSSlyffiSlrg00SgBg.........5... ..0410. 2.1..2543 -04210 25*7 25.46 -040.............. 04

it26.085.70-04*9 25.662122 025~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..............................129847 5.10 -0407 2W26*7 -0409~~~~~~~~~~~~~~~~~~~~~~~~~~~................

......01.....17 25432657 0489~~~~~~~~~~~~~~~~~~~~~~~~~~.... .28..72.25....-04121625*72417 -0~~~~~~~~~~~~~~~~~~~~~~~~~.......1..........4.......4.4*7..02*217025.16-0.53 2574 25.0-0.09~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....................................*126*120*4 0*326.76 27.10~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...0....................

2226*4 26*10* 28~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....7...2..041272134 26* -0412 28432111 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"0*~.....510.26..2.02......2021..425*42546~~~~~~~~~~~~~~~~~~~~~~~~~4*?*5*8 *543~~~~~~~~~~~~~~~~~~~~~......Mew25*2188-0.08 2188 2588 -0402~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~......8904404680* 0446 0.50 04t~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....

1~~~~~UP1bfUp~~~~~~~~~~~~~~~~~~ ~~~~~C@OLffii~~~~~~~~~~~~~~~~~~ u~~~~~~deu~~~~~~~~~..........

popularity ofa program should not be con-fused with its health effects. Our weightloss programs were far more popular thanour smoking programs, but the smokingprogram would clearly be a better invest-ment for an employer interested in healthoutcomes. It has been estimated that asmoking employee costs an employer sev-eral hundred dollars per year more than anonsmoking employee.53 We estimatethat the cost ofprovidingour smoking pro-gram for 2 yearswas approximately $1500($30 per session in instructor time pluscost of materials) per intervention site andthat the number of quitters produced persite was between 8 and 16. Thus, this pro-gram appears to be justified on purely eco-nomic grounds.

400 American Journal of Public Health

The failure of the weight loss pro-gram to produce positive results in the em-ployee population as a whole, despite highparticipation rates and a dose-responserelationship with mean company weightloss, invites further discussion. One pos-sible explanation is that the HealthyWorker Project weight loss program didwork, but only in people who would havelost weight anyway. In other words, ourwork-site program may have given em-ployees a new option for weight control,but it did not enhance interest in weightcontrol above that already existing. Theplausibility of this explanation is strength-ened somewhat by previously publisheddata on dieting behaviors in this samestudy population.54 Lifetime prevalence

of participation in formal weight loss pro-grams was 13% in men and 26% inwomenat baseline, and rates ofinformal dieting tolose weight were much higher. It is thuslikely that many of these individualswould have spontaneously tried to loseweight during the 2 years of this study,whether or not a work-site program wasavailable.

In summary, we conclude that work-site intervention programs for smokingcessation of the type evaluated here areeffective in reducing the prevalence ofsmoking in employee populations. Al-though we observed considerable site-to-site variation in employee participationand in overall treatment effectiveness, webelieve that, on average, the effects persite are sufficient to justify an investmentin such programs in work sites of this sizeor larger, evenwhen absolute rates of par-ticipation are relatively low. The longerterm potential of such programs for accel-erating downward trends in smoking ratesmerits additional research attention. An-other important research issue in this areafrom a public health perspective is how toinduce more work sites to invest in suchprograms and how to increase employeeparticipation to maxinize their potential.In contrast to smoking, this study pro-vides no support for the beneficial effectsof work-site weight control programs.Their popularity suggests that they maymeet a perceived need of employees and,thus, may be good for employee morale.Their effect on health risks, however, maybe limited. []

AcknowledgmentsThis research was supported by grant 3R01-HI34740 to Robert W. Jeffery, PhD, from theNational Heart, Lung, and Blood Institute.

References1. Fielding JE. Health promotion and disease

prevention at the worksite.Annu Rev Pub-lic Health. 1984;5:237-265.

2. Glasgow RE, Terborg JR. Occupationalhealth promotion programs to reduce car-diovascular risk. J Consult Clin PsychoL1988;56:365-373.

3. Klesges RC, Cigrang J, Glasgow RE.Worksite smoking modification programs:a state-of-the-art review and directions forfuture research. Curr Psychological ResRev. 1987;6:26-56.

4. Wilbur CS. Live for Life: The Johnson &Johnson program. Prev Med. 1983;12:672-681.

5. Erfurt JC, Foote A, Heirich MA. Worksiteweilness programs: incremental compari-son of screening and referral alone, healtheducation, follow-up counseling, and plantorganization. Am J Health Promotion.1991;5:438-448.

March 1993, Vol. 83, No. 3

6. Brownell KD, Cohen RY, Stunkard AJ,Felix MR, Cooley NB. Weight loss com-petitions at theworksite: impact onweight,morale and cost-effectiveness.AmJPublicHealth 1984;74:1283-1285.

7. Scignar CB. Mandatory weight controlprogram for 550 police officers choosing ei-ther behavior modification or "willpow-er": impact on weight, morale and cost-effectiveness. Obes/BanaticMed 1980;9:88-92.

8. The Health Consequences of Smoking-Cancer and Chronic Lung Disease in theWorkplace:A Report ofthe Surgeon Gen-eraL Washington, DC: US Department ofHealth and Human Services; 1985. DHHSpublication PHS 85-50207.

9. Klesges RC, Glasgow RE, KIesges LM,Morray K, Quayle R. Competition and re-lapse prevention training in worksite smok-ing modification. Health Educ Res. 1987;2:5-14.

10. Alderman M, Madhavan S. Davis T. Re-duction ofcardiovascular disease eventsbyworksite hypertension treatment. Hyper-tension. 1983;5(suppl V):V138-V143.

11. Shipley RH, Orleans CT, Wilbur CS, Pis-cheria PV, McFadden DW. Effect of theJohnson & Johnson Live for Life programon employee smoking. Prev Med. 1988;17:15-34.

12. Fielding JE, Piseichia PV. Frequency ofworksite health promotion activities.AmJPublic Health 1989;79:16-20.

13. Kronenfeld JJ, Jackson K, Blair SN, et al.Evaluating health promotion: a longitudi-nal quasiexperimental design. Health EducQ. 1987;14:123-139.

14. Stunkard AJ, Cohen RY, Felix MRJ.Weight loss competitions at the worksite:how they work and how well. Prev Med1989;18:460-474.

15. Klesges RC, Vasey MM, Glasgow RE. Aworksite smoking modification competi-tion: potential for public health impact.AmJPublic Heakh 1986;76:198-200.

16. Orleans CS, Shipley RH. Worksite smok-ing cessation initiatives: review and recom-mendations. Addict Behav. 1982;70:1-16.

17. Bennett D, Levy BS. Smoking policies andsmoking cessation programs of large em-ployers in Massachusetts. Am J PublicHealt 1980;70:629-631.

18. Sorensen G, Rigotti N, Rosen A, Pinney J,Pieble R. Effects of a worksite nonsmokingpolicy: evidence for increased cessation.Am JPublicHealth. 1991;81:202-204.

19. BeckerDM, Conner HF, Waranch R, et al.The impact of a total ban on smoking in theJohns Hopkins Children's Center. JAMA.1989;262:799-802.

20. Rosenstock U, Stergachis A, Heaney C.Evaluation ofsmoking prohibition policy ina health maintenance organization. Am JPublic Health. 1986;76:1014-1015.

21. Petersen LR, Helgerson SD, Gibbons CM,Calhoun CR, Ciacco KH, Pitchford KC.Employee smoking behavior changes andattitudes following a restrictive policy onworksite smoking at a large company. Publie Health Rep. 1988;103:115-120.

22. Cook TD, Campbell DT. Quasi-expen-

March 1993, Vol. 83, No. 3

mentation. Desgn andAnalysis IssuesforField Settings. Chicago, Ill: Rand Mc-Nally; 1979.

23. Foshee V, McLeroy KR, Summer SK, Bi-beau DL. Evaluation of worksite weightloss programs: a review of data and issues.JNutrEduc. 1986;18:S38-S43.

24. Fielding JE. Effectiveness of employeehealth improvement programs. J OccupMed. 1982;24:907-916.

25. Jeffery RW, Forster JL, Schmid TL.Worksite health promotion: feasibility test-ingofrepeated weight control and smokingcessation classes.AmJHealth Pomotion.1989;3:11-16.

26. Abraham S, Johnson CL. Overweightadults in the United States. Advancedata.1979;51:1-9.

27. Pechacek TF. An overview ofsmoking be-havior and its modification. In: KrasnegorN, ed. BehavioralApproaches toAnalysisand TreatmentofSubstanceAbuse. Wash-ington, DC: National Institute on DrugAbuse; 1979:92-113. NIDA ResearchMonograph No. 25.

28. Health Implications ofObesity. Bethesda,Md: National Institutes of Health; 1985;5(9).

29. Koop CE. The Health Consequences ofSmoking--Cancer:A reportofthe SurgeonGeneraL Washington DC: US Departmentof Health and Human Services; 1982.DHHS publication 82-5-179.

30. Kent DC, Cenci L. Smoking and thework-place: tobacco smoke health hazards to theinvoluntary smoker. J Occup Med. 1982;24:469-472.

31. Jeffeiy RW, Folsom AR, Luepker RV, etal. Prevalence of overweight and weightloss behavior in a metropolitan adult pop-ulation: The Minnesota Heart Surveyexperience. Am JPublc Heakh. 1984;74:349-352.

32. JefferyRW, ForsterJL. Obesityasapublichealth problem. In: Johnson WG, ed. Ad-vances in Eating Disorders. Greenwich,Conn: JAI Press; 1987:253-271.

33. Trends in Smoking Intervention. A Na-tional StatusReport. Washigton, DC: USDepartment of Health and Human Serv-ices; 1986:2:27-37. DHHS publicationHHS/PHS/CDC-87-8396.

34. Jeffery RW, Forster JL, Snell MK. Pro-moting weight control at the worksite: apilot program of self-motivation using pay-roll-based incentives. Prev Med. 1985;14:187-194.

35. Jeffery RW, Pheley AM, Forster JL, Kra-mer FM, Snell MKL Payroll contracting forsmoking cessation: a worksite pilot study.Am JPrev Med. 1988;4:83-86.

36. Forster JL, Jeffeiy RW, Sullivan S, SnellMK. A worksite weight control programusing financial incentives collected throughpayroll deduction. J Occup Med. 1985;27:804-0.

37. Glasgow RE, Klesges RC, Godding PR,Vasey MW, O'Neill HK. Evaluation of aworksite-controlled smoking program. JConsult Chin PsychoL 1984;52:137-138.

38. Follick MJ, Fowler JL, Brown RA. Attri-tion in worksite weight-loss interventions:the effects of an incentive procedure. I

Weight Control and Smolkng Intervention

Consult Clin PsychoL 1984;52:139-140.39. Browneli KD, Cohen RY, Stunkard AJ,

Felix MRJ, Cooley NB. Weight loss com-petitions at theworksite: impact onweight,morale, and cost-effectiveness.Am JPub-lic Health. 1984;74:1283-1285.

40. Dow Chemical Company. TheDow Chem-ical Company I Quit Pmgrawm Freeport,Tex: Dow Texas Division; 1976.

41. Rosen GM, Lichtenstein E. An employeeincentive program to reduce cigarettesmoking. J Consult Clin PsychoL 1977;45:957.

42. Lando HA, McGovern PG, Kelder SH,Jeffery RW, Forster JL. Use of carbonmonoxide in assessing smoke exposure in aworksite population. HealthPsychoL 1991;10:296-301.

43. Stunkard AJ, Albaum JM. The accuracy ofself-reported weights. Am J Clin Nutr.1981;34:1593-1599.

44. Pirie PL, Jacobs DR, Jeffery RW, HannanP. Distortion in self-reported height andweight data. JAm Diet Assoc. 1981;78:601-606.

45. Tell GS, Jeffery RW, Kramer MF. Canself-reported body weight be used to eval-uate long-tenn follow-up of a weight-lossprogram?JAm DietAssoc. 1987;87:1198-1201.

46. MurphyJK, Bruce BK, Williamson DA. Acomparison of measured and self-reportedweights in a four year follow-up of spouseinvolvement in obesity treatment. BehavTher. 1985;16:524-530.

47. Velicer WF, Prochaska JO, Rossi JS,Snow MG. Assessing outcome in smokingcessation studies. Psychol BulL 1992;3:23-41.

48. WeissfeldJL, HollowayJJ, KirschtJP. Ef-fects ofdeceptive self-reports of quitting onthe results of treatment trials for smoking:a quantitative assessment. J Clin Epide-mioL 1989;42:231-243.

49. Jacobs DR Jr, Jeffery RW, Forster JL,Perry CL. Methodological issues in work-site health intervention research: general-izability, evaluation framework, and unit ofanalysis. Presented at the National Heart,Lung, and Blood Institute Conference onMethodological Issues in Worksite Re-search; April 10, 1988; Airlie House, Va.

50. Metropolitan Life Insurance Co. Newweight standards for men and women. StatBulL 1959;40.

51. Schmid TL, Jeffery RW, Hellerstedt WL.Direct mail recruitment to home-basedsmoking and weight control programs: acomparison of strategies. P?ev Med. 1989;18:503-517.

52. Lando HA, McGovern PG, Barrios FX,Etringer BD. Comparative evaluation ofAmerican Cancer Society and AmericanLung Association smoking cessation clin-ics.Am JPublic Health. 1990;80:554-559.

53. Kristein MM. How much can business ex-pect to profit from smoking cessation?PrevMed 1983;12:358-381.

54. Jeffery RW, Adlis SA, Forster JL. Preva-lence of dieting among working men andwomen: The Healthy Worker Project.HealthPsychL 1991;104:274-281.

American Jousrnal of Public Health 401