the assessment of chronic health conditions on work ... · the assessment of chronic health...

FAST TRACK ARTICLE

The Assessment of Chronic Health Conditionson Work Performance, Absence, and TotalEconomic Impact for Employers

James J. Collins, PhDCatherine M. Baase, MDClaire E. Sharda, RN, MBARonald J. Ozminkowski, PhDSean Nicholson, PhDGary M. Billotti, MSRobin S. Turpin, PhDMichael Olson, PhDMarc L. Berger, MD

Objective: The objective of this study was to determine the prevalence and estimatetotal costs for chronic health conditions in the U.S. workforce for the Dow ChemicalCompany (Dow). Methods: Using the Stanford Presenteeism Scale, information wascollected from workers at five locations on work impairment and absenteeism based onself-reported “primary” chronic health conditions. Survey data were merged withemployee demographics, medical and pharmaceutical claims, smoking status, biometrichealth risk factors, payroll records, and job type. Results: Almost 65% of respondentsreported having one or more of the surveyed chronic conditions. The most common wereallergies, arthritis/joint pain or stiffness, and back or neck disorders. The associatedabsenteeism by chronic condition ranged from 0.9 to 5.9 hours in a 4-week period, andon-the-job work impairment ranged from a 17.8% to 36.4% decrement in ability tofunction at work. The presence of a chronic condition was the most importantdeterminant of the reported levels of work impairment and absence after adjusting forother factors (P � 0.000). The total cost of chronic conditions was estimated to be10.7% of the total labor costs for Dow in the United States; 6.8% was attributable towork impairment alone. Conclusion: For all chronic conditions studied, the costassociated with performance based work loss or “presenteeism” greatly exceeded thecombined costs of absenteeism and medical treatment combined. (J Occup EnvironMed. 2005;47:547–557)

C linicians, policymakers, employers,and insurers are intently focused onthe cost of health care. As the currenthealthcare situation is analyzed, it isimportant to have a full appreciationof the total economic impact ofhealth conditions on business andour society. To enable the best in-vestment of limited societal re-sources, it would be helpful to have amore comprehensive understandingof the link between illness and over-all economic impact.

Health conditions increase work-related absences and reduce work-place productivity, creating asubstantial economic burden for in-dustry.1–3 Although this may not bedisputed in the abstract, it is muchmore difficult to assess the impact ofpoor employee health on a specificcompany’s productivity. First, mostemployers do not systematicallytrack absenteeism or do so for only aportion of their workforce. Second,methods to assess the impact of ahealth condition on work perfor-mance, known as presenteeism, haveonly been developed in the last fewyears.

From the perspective of an indi-vidual company, demonstrating thefinancial impact of health conditionsis critical for developing budgetarypriorities, including the amounts al-located for employee benefits suchas health insurance, health promotionprograms, and disease managementinterventions. Most studies have fo-cused on a particular disease or con-dition and have not characterized theimpact on all employees. Various

From the The Dow Chemical Company, Midland, Michigan (Dr Collins, Dr Baase, Mr Billotti);Outcomes Research and Management, Merck & Co., Inc. (Ms Sharda, Dr Turpin, Dr Berger); TheMedstat Group, Inc. (Dr Ozminkowski); The Wharton School (Dr Nicholson); and Personnel ResearchAssociates, Inc. (Dr Olson).

Address correspondence to: Catherine Baase, MD, Global Director, Health Services, The DowChemical Company, EDC Building, Midland, MI; E-mail: [email protected].

Copyright © by American College of Occupational and Environmental Medicine

DOI: 10.1097/01.jom.0000166864.58664.29

JOEM • Volume 47, Number 6, June 2005 547

chronic health conditions, includingasthma, allergic rhinitis, migraineheadaches, depression, respiratorydifficulties, back pain, and diabetes,have been shown to reduce produc-tivity.4–18 However, little is knownabout the relative contributions ofeach to a company’s overall workloss or how work loss varies by jobtype, demographics, or other factors.Limited data support the contentionthat interventions into disease statesand health risk can increase workeroutput,19 but more is needed.6,15,19

Some studies have suggested thathealth-related absences are lessimportant than health-related impair-ment on the job or “work impair-ment.” Work impairment, however,had not been assessed until veryrecently.4,17,20–22 Two studies esti-mated that work impairment mayreduce total work hours by onefifth.22,23 Thus, focusing only on ab-senteeism may significantly underes-timate lost productivity.5–18

Since 1997, the Dow ChemicalCompany (Dow) has been imple-menting an integrated health man-agement approach to help minimizehealth-related costs and maximizehealth-related productivity. Dow de-fines the total economic impact re-lated to health as the sum of directhealthcare costs and the indirectcosts of absenteeism and presentee-ism performance losses. Until re-cently, Dow had no estimate of thecost of presenteeism or health-related work performance. We reportthe results of a comprehensive char-acterization and economic analysisof chronic health conditions onDow’s U.S. workforce. To ourknowledge, this study is the mostcomprehensive survey of its kind todate and provides the most completepicture of the impact of chronichealth conditions on a diverse work-force. Given that the majority of thenonelderly in the United States ob-tain their health insurance throughtheir employers, employer decisionshave profound influence in thehealthcare system generally. Thiscase study can inform policymakers,

payers, and providers regarding thetotal economic impact of chronicconditions on business and our soci-ety and provides an important frame-work for considering investments ininterventions that improve function-ing for individuals with chronichealth conditions. For Dow, it estab-lishes a baseline against which futureassessments can be compared.

Materials and MethodsBetween July and September

2002, 12,397 Dow full-time activeemployees at five locations in Mich-igan and Texas (representing 56% ofDow’s U.S. workforce) received aninvitation to participate in an onlinehealth survey. Part-time, temporary,work-study workers, interns, and

employees on extended leaves of ab-sence were excluded. Workers in abroad cross-section of job types par-ticipated, including top leadership,technical, professional, office profes-sional, and chemical production jobs.All participants provided informedconsent and confidentiality wasmaintained by employing a thirdparty for analysis of data. The proto-col was reviewed and approved byan Institutional Review Board, andthe study was designed and con-ducted using the principles of GoodEpidemiology Practices.24

All participants were asked tocomplete the Short-Form Health Sur-vey (SF-36)25 and the Stanford Pre-senteeism Scale (“SPS”) (Fig. 1).Validation of the SPS is reported

Fig. 1. Stanford Presenteeism Scale.

548 Assessment of Chronic Health Conditions • Collins et al

elsewhere.4,21 A random sample ofparticipants (10%) also completedthe Work Limitation Questionnaire(WLQ)26 as an additional measure ofhealth-related work loss. The SPS(Fig. 1) assesses self-reported ab-sence, work impairment, and workloss attributable to the identified“primary health condition.” The firstquestion asks the respondent to se-lect their chronic health conditionsfrom a list and to identify his or her“primary health condition,” definedas the condition that has affected theindividual most in the past 4 weeks.The chronic conditions listed on theSPS for this study are noted in Table1. Work impairment for each surveyrespondent was calculated from theresponses to 10 SPS items queryingthe frequency or intensity of specificmanifestations of the primary healthcondition, and how these manifesta-tions affected work (questions 2–11,Fig. 1). A five-point scale was usedfor scoring categorical responses: al-ways, frequently, about half the time,occasionally, or never. Each individ-ual’s responses were summed andnormalized to a percentage referredto as the Work Impairment Score(WIS). For example, an employeewho responded “about half the time”to all 10 questions would have a WISof 50% (50% impairment in abilityto work), whereas employees whoresponded “occasionally” to all 10questions would report a 25% im-pairment. As reported elsewhere,21

the reliability of the WIS was high inthis application (ie, Cronbach’s al-pha � 0.83), and evidence of contentand construct validity through com-parison to the SF-36 and the WLQwas included in that discussion.

The SPS also includes a single-item global assessment question thatasks the respondent to estimate thepercentage of “usual” productivitythey were able to achieve in the4-week period given the primaryhealth condition (question 12, Fig.1). We refer to this estimate as thework output score (WOS). We inter-pret the WIS as measuring the extentto which a health condition reduces a

person’s inputs into their job (eg,energy, ability to focus, ability towork with colleagues) and the WOSas measuring how much less outputis produced as a result of the dimin-ished inputs. Therefore, we use theWOS to quantify the monetary costof productivity losses because eco-nomic theory indicates that peopleare paid according to the value oftheir output, but we examine howhealth and other demographic char-acteristics affect the WIS because weare interested in identifying factorsthat affect a person’s underlyingability to work effectively.

The individual-level survey datawere merged with informationfrom many other sources, includingmedical and pharmaceutical claimsand company records that reporteddemographics, job category, pay-roll absence data, and plant loca-tion. Biometric data collected fromprevious health risk appraisalswere also merged, as was smokingstatus. The biometrics includedhigh-, medium-, and low-riskcategorical measures for diastolicblood pressure, systolic pressure,total cholesterol, the ratio of totalcholesterol to high-density lipopro-

TABLE 1Chronic Health Conditions on SPS and Corresponding ICD-9-CM Codes*

Primary Chronic HealthCondition on SPS ICD-9-CM Codes

Allergies 287.0, 372.14, 379.93, 446.20, 446.29,477.0–477.9, 478.8, 495.0–495.9, 558.3,693.1, 708.0–708.1, 708.5–708.9, 995.1,995.3, V14.0–V14.9, V15.01–V15.09

Arthritis or joint pain/stiffness 714.0–714.9, 715.00–715.98, 716.00–716.99,719.40–719.59

Asthma 493.xx

Back or neck disorder 720.0–720.9, 737.0–737.9, 738.2, 738.5,839.00–839.59, 846.0–846.9, 847.0–847.9

Breathing disorder (bronchitis,emphysema)

490, 491.0–491.9, 492.0–492.8

Depression, anxiety or emotionaldisorder

296.20–296.36, 296.4–296.7, 298.0, 300.4,301.12, 309.0–309.1, 311, 293.84,300.00–300.09, 300.20–300.30, 309.21,309.24, 301.0–301.9, 308.0–308.9, 309.22,309.23, 309.28–309.29, 309.3–309.9,312.0–312.9, 313.0–313.9

Diabetes 250.xx

Heart and circulatory problems(artery disease, high bloodpressure)

401.0–404.9, 405, 410.00–410.92, 411,411.1, 411.0, 411.81–411.89, 412, 413,414, 414.0, 414.1, 414.00–414.05, 414.10,414.8–414.9

Migraine/chronic headaches 346.0–346.9, 307.81, 784.0

Stomach or bowel disorder 530.81, 531.00–534.91, 535.00–535.91,536.0–536.9, 537.0–537.9, 556, 558.9,560.0–560.9, 569.81–569.9,562.00–562.13, 564.00–564.9, 782.7

* International Classification of Disease (ICD-9-CM) codes were used to find medicalclaims-based evidence of these conditions. Pharmacy codes were also used to find patientswith evidence of having these conditions.


tein (HDL) levels, and body massindex (BMI).

Total annual health costs were de-fined as the sum of the medicaltreatment costs, absenteeism costs,and presenteeism costs. Treatmentcosts included the company’s expen-ditures for employees’ medical, men-tal health, and pharmacy costs for the12-month period before the survey.Absenteeism costs were based onhours away from work resultingfrom the primary condition reportedon the SPS (question 13, Fig. 1).These hours were annualized andmultiplied by the average annualwage for the worker’s job type to putabsenteeism in dollar terms. Presen-teeism or work performance losseswere similarly estimated using theWOS. For example, if a person indi-cated she was 10% less productivethan usual over the past 4 weeks as aresult of her primary health condi-tion, we assumed the value of the lostoutput over the entire year was equalto 10% of the average annual wageof that job type. For some people, theproductivity impact over the past 4weeks is likely to be larger or smallerthan the impact for the year as awhole, but the reported impactshould be an accurate measure onaverage. The prevalence of chronichealth conditions by job type wasused to project the costs from oursample to the entire U.S. Dowworkforce.

AnalysisTwo multiple regression analyses

were conducted to estimate the im-pact of various factors on work im-pairment and absenteeism. Thesefactors included the “primary” healthcondition reported on the survey, anindicator denoting whether morethan one chronic condition was re-ported, the employee’s job type,plant location, age, ethnicity, sex,and biometrics, and his or her aver-age number of hours worked perweek during the recall period. Eth-nicity was reported by the study par-ticipants. The WIS was used as thedependent variable measure of work

loss in the first regression analysis.Because almost 85% of the respon-dents reported no absence-associatedlost hours in the 4-week reportingperiod, we used a binary (yes or no)indicator for absences as the depen-dent variable in a logistic regressionanalysis. We evaluated residual plotsto assess the fit of the regressionmodels, determine the influence ofoutliers, and assure regression as-sumptions were not violated. Condi-tion indices were used to evaluatecollinearity between independentvariables.7,27

Self-reported survey prevalence ofcertain chronic conditions was com-pared with those same conditionsgenerated from medical claims, us-ing both International Classificationof Diseases, 9th Revision, ClinicalModification codes for health condi-tions (Table 1) and evidence of theuse of pharmacotherapy for theseconditions. Medical and pharmacyclaims files were then used to esti-mate the direct medical costs oftreatment for the chronic conditionsof interest based on inpatient, outpa-tient, emergency room, and phar-macy costs. Finally, the economicimpact of the chronic conditions onthe entire Dow U.S. workforce wasestimated. This was done by multi-plying the average total cost of workloss (based on medical and drugexpenditures, absenteeism dollars,and work impairment dollars) ineach job category by the number ofworkers in each job categorythroughout the Dow U.S. workforceand dividing the result by total salaryand benefits paid to all Dow U.S.employees. This resulted in an esti-mate of the economic burden ofchronic illness at Dow as a percent-age of labor costs for the chronicconditions reported in the survey.

ResultsOf the 12,397 workers contacted,

7797 responded (63%). Comparedwith nonrespondents, the respon-dents were similar in age, morelikely to be female, less likely to becurrent smokers, and less likely to be

union members. In addition, a major-ity of both responders and nonre-sponders were likely to have a BMIless than 30 and to have normaldiastolic pressure (see Table 2). Be-cause of missing values on healthrisk factors and incomplete surveys,the number of survey participantsused in the regression analyses was5108 for work impairment and 5240for the absenteeism; the number ofparticipants completing the entiresurvey was 5369. There were nomajor differences among the casesleft out as a result of missing healthrisk factors when compared withthose who were included in the re-gression analyses. Medical claimsdata were available for 95% (7410 of7797) of respondents reporting achronic condition, and these surveyparticipants were used for the costanalysis.

PrevalenceOverall health status, as assessed by

the single global question in the SF-36for respondents, were reported as fol-lows: excellent (12%), very good(45%), good (36%), fair (6%), andpoor (0.5%). Almost two thirds (65%)of the survey participants reported oneor more chronic health conditions.Those most frequently reported wereallergies (37.1%), arthritis/joint pain orstiffness (21.8%), and back and neckdisorders (16.3%). These results aresimilar to those reported in anothersurvey.28 If we examine only theprimary health conditions, or thecondition that affected the partici-pant the most in the last 4 weeks (seeFig. 1, question 1), 58% of thosesurveyed reported a primary healthcondition. The most common pri-mary health conditions reported wereallergies (18.9%), arthritis/joint painor stiffness (9.0%), heart or circula-tory problems (7.1%), and back orneck disorders (7.0%). For all condi-tions, the general self-reportedprevalences, whether a condition wasdesignated as primary, were higherfor all conditions than the prevalenceof the same condition noted by ex-amination of medical claims (Fig. 2).


The differences were larger for themore symptomatic conditions suchas allergies, arthritis, and migraineheadaches, and smaller for heart dis-ease and diabetes.

Across all job types except un-skilled laborers, a majority of work-ers reported at least one chroniccondition as follows: clerical and of-fice workers (70.6%), craft workers–

skilled (69.2%), officials and manag-ers (68.5%), technicians (68.5%),professionals (66.6%), operatives–semiskilled (64.4%), service workers(60.5%), sales workers (58.2%), andlaborers–unskilled (25.0%).

Work LossFor those individuals reporting a

primary chronic condition, the as-

sociated absenteeism during the4-week recall period varied bychronic condition, ranging from 0.9to 5.9 hours, whereas work impair-ment varied from a 17.8% to 36.4%decrement in ability to function (Ta-ble 3). Employees reporting depres-sion, anxiety, or emotional disorders(36.4% decrement) or breathing dis-orders (bronchitis, emphysema;23.8% decrement) had the highestwork impairment. Employees withthese conditions also reported thehighest absences. Overall, greaterwork impairment was associatedwith lower health status according tothe SF-36 and the correlation wasstrongly negative (�0.65).21

Factors associated with work im-pairment and absences are shown inTable 4. Important predictors forwork impairment included sex (P �0.012), age (P � 0.000), location(P � 0.000), job (P � 0.000), thepresence of a chronic condition (P �0.000), number of chronic conditions(P � 0.000), and hours worked (P �0.000). Work impairment decreasedwith increasing age and was highestamong the employees whose jobswere classified as service workersand operatives (semiskilled). Depres-sion, anxiety, or emotional disorderresponses were associated with themost work impairment, but migraine/chronic headaches, “other” condi-tions, breathing disorders, and backor neck disorders were also impor-tant predictors of work impairment.The degree of work impairment in-creased with the number of chronicconditions reported. Work impair-ment was highest for personsworking less than 40 hours. This re-gression model explained 18% of thevariance of work impairment.

Important predictors of absence,defined as one or more absencehours in a 4-week period, includedsex (P � 0.002), age (P � 0.003),location (P � 0.000), job (P �0.000), chronic condition (P �0.000), number of chronic conditions(P � 0.000), and hours worked (P �0.000). Females were 40% morelikely to report absence than males.

TABLE 2Characteristics of Survey Population

SurveyRespondents

SurveyNon-Respondents

Number 7797 4600Demographics

Average Age 43.2 yrs 43.9 yrsFemale 72% 28%Male 60% 40%Union jobs 12.5% 33.3%

Site location#1 72% 28%#2 63% 37%#3 56% 42%#4 72% 28%#5 61% 39%

Outpatient visits 2.47 visits 2.18 visitsHealth Indicators

Current smoking statusYes 13% 19%No 87% 81%

BMI�25 23% 18%25–29.9 40% 39%30� 36% 42%

Systolic blood pressureNormal (�140) 88% 83%Low risk (140–159) 11% 15%High risk (160�) 1% 2%

Diastolic blood pressureNormal (�90) 87% 85%Low risk (90–99) 10% 13%High risk (100�) 1.9% 2.2%

HDLLow risk (4.0–4.9) 54% 49%Mild risk (5.0–5.9) 35% 39%High risk (6.0�) 10% 12%

LDLLow risk 62% 59%Mild risk 25.7% 26.6%High risk 12% 15%

Total CholesterolLow risk (�200) 54% 49%Mild risk (200–239) 33% 35%High risk (240�) 13% 16%

All characteristics were significantly different statistically (P � 0.001). Other characteristicsthat were checked that did not significantly differ between responders and non-responderswere: Expenditures (total, inpatient, outpatient, emergency room, mental health and drug),inpatient admissions, inpatient days, ER visits, mental health utilization (inpatient admissionsand days, outpatient visits, ER visits and drug scripts).


The risk of being absent decreasedwith age, with workers 56 and olderbeing 50% (odds ratio � 0.5) lesslikely to be absent than workers lessthan age 25. The risk of being absentwas lowest at the location 3 (70%)and highest at location 5 (130%).Sales workers (180%) and office and

clerical workers (140%) had thehighest risk of being absent, whereascraft workers (90%) and managers(100%) had the lowest risk. Thelargest risk for absence from chronicconditions occurred among workerswith breathing disorder (bronchitis,emphysema; 440%), followed by de-

pression, anxiety, or emotional disor-der (220%) and then by migraine/chronic headaches (170%). The riskof absence also increased with thenumber of health condition reported.Finally, the risk of being absent de-creased with the more hours worked:the risk being 50% lower amongemployees working 60 of more hourscompared with employees working40 to 44 hours. This regressionmodel explained 11% of the varianceof absence.

CostsBased on the experiences of those

employees who reported a primarycondition, we estimated the total costof each condition individually (seeFig. 3). Among employees who re-ported at least one primary condition,the highest total cost per worker peryear was for those reporting depres-sion, anxiety, or emotional disorderas their “primary health condition”($18,864 in year 2002 dollars), and

Fig. 2. *Prevalence rates from the SPS of any chronic conditions in the last 4 weeks andchronic conditions reported from medical and pharmacy claims in the last year.**Breathingdisorders (bronchitis and emphysema).

TABLE 3Number of Respondents, Prevalence Rates, and Mean Impairment Scores (WIS) and Absence Hours by Self-ReportedHealth Conditions

Reported ChronicHealth Condition

Number (%) ofRespondentswho Reported

Having thisCondition,

Regardless ofWhether it was

Selected as TheirPrimary Health

Condition

Number (%)of

RespondentsWho ChoseCondition as

PrimaryHealth

Condition andAnswered

ImpairmentScore

Questions

MeanImpairmentScore For

Those WhoChose

Condition asPrimaryHealth

Condition

95%Confidence

IntervalAround MeanImpairment

Score

Number (%)of

RespondentsWho ChoseCondition as

PrimaryHealth

Condition andAnswered

Question onAbsenteeism

Mean HoursAbsent forThose who

ChoseCondition as

PrimaryHealth

Condition

95%Confidence

IntervalAround MeanHours Absent

Allergies 2890 (37.1) 1472 (18.9) 18.2 17.5,18.8 1482 (19.0) 0.9 0.7,1.1Arthritis/joint pain . . . 1697 (21.8) 704 (9.0) 19.7 18.6,20.7 707 (9.1) 1.1 0.8,1.4

Asthma 348 (4.5) 99 (1.3) 17.9 15.2,20.5 98 (1.3) 0.9 0.5,1.4Back/neck disorder 1274 (16.3) 545 (7.0) 21.7 20.5,22.8 549 (7.0) 2.1 1.5,2.6Breathing disorder . . . 104 (1.3) 20 (0.3) 23.8 17.8,29.7 20 (0.3) 5.9 1.6,10.2

Depression, 721 (9.2) 337 (4.3) 36.4 34.4,38.3 339 (4.3) 3.7 2.8,4.6anxiety . . .

Diabetes 253 (3.2) 189 (2.4) 17.8 15.9,19.6 189 (2.4) 1.3 0.6,1.9Heart/circulatory 929 (11.9) 554 (7.1) 19.9 18.7,21.1 556 (7.1) 1.4 0.9,1.9

problem . . .Migraines/chronic

headaches635 (8.1) 243 (3.1) 23.4 21.7,25.0 243 (3.1) 2.4 1.7,3.2

Stomach/boweldisorder

677 (8.7) 262 (3.4) 21.7 20.0,23.3 266 (3.4) 1.9 1,2.7

Musculoskeletal 49 (0.6) 46 (0.6) 21.5 17.6,25.4 49 (0.6) 2.6 0.5,4.7Other 906 (11.6) 434 (5.6) 23.0 21.5,24.4 442 (5.7) 4.0 3.1,4.9


TABLE 4Predictors of Work Impairment and Absence Due to a Chronic Health Condition

Variable

Work Impairment Absence

Linear RegressionCoefficient

VariableP-value

GroupP-value

Logistic RegressionOdds Ratio

VariableP-value

GroupP-value

Constant 12.5 0.000 NA 0.000Sex

Male 0.0 Reference 0.012 1.0 Reference 0.002Female 0.7 0.159 1.4 0.000Unknown �3.4 0.022 0.9 0.640

RaceWhite 0.0 Reference 0.228 1.0 Reference 0.433Black �1.1 0.263 1.2 0.312Hispanic �1.1 0.158 1.2 0.077Asian 0.6 0.631 1.1 0.613Amer. Indian 5.7 0.070 1.2 0.768Unknown �1.3 0.672 0.6 0.595

Age�25 2.8 0.091 0.000 1.0 0.932 0.00326–35 0.0 Reference 1.0 Reference36–45 �1.4 0.017 0.7 0.00046–55 �2.8 0.000 0.6 0.00056� �4.5 0.000 0.5 0.000

Location1 0.2 0.688 0.000 1.1 0.322 0.0002 1.2 0.016 1.1 0.4413 3.5 0.000 0.7 0.0304 0.0 Reference 1.0 Reference5 0.8 0.283 1.3 0.054Unknown 1.4 0.431

JobManagers 0.0 Reference 0.000 1.0 Reference 0.000Craft workers (skilled) 1.9 0.047 0.9 0.908Office & clerical 1.8 0.053 1.4 0.020Operatives (semi-skilled) 3.1 0.000 1.0 0.997Professionals 0.3 0.701 1.1 0.298Sales workers 2.5 0.237 1.8 0.087Service Workers 3.3 0.145 0.6 0.221Technicians 1.7 0.033 1.2 0.126Unknown 0.1 0.922 1.6 0.108

Chronic ConditionArthritis 0.0 Reference 0.000 1.0 Reference 0.000Allergies �0.5 0.397 1.0 0.989Asthma �3.5 0.014 1.0 0.856Back 1.6 0.034 1.2 0.086Breathing 1.7 0.574 4.4 0.007Depression 14.4 0.000 2.2 0.000Diabetes �2.4 0.028 1.2 0.346Heart 0.4 0.592 0.9 0.396Migraines 2.3 0.022 1.7 0.001Stomach 1.0 0.297 1.3 0.037Other 2.3 0.004 1.0 0.910

Number of ConditionsNumber 3.0 0.000 0.000 1.4 0.000 0.000

Body Mass Index�30 0.0 Reference 1.0 Reference30� 0.9 0.092 1.1 0.403Missing 0.1 0.922 1.2 0.124

SmokingNon-smoker 0.0 Reference 1.0 ReferenceSmoker 0.6 0.402 1.0 0.987Unknown 0.4 0.592 1.0 0.985

(Continued)


the lowest was for those reportingallergies ($6947).

For all chronic conditions, with theexception of breathing disorders(bronchitis, emphysema), the direct

costs of treatment through medicalcare exceeded the condition-relatedabsenteeism costs. However, the costof presenteeism, or work perfor-mance loss, was the largest compo-

nent cost for every chronic condition.When weighted by the survey prev-alence of each condition across allDow U.S. workers, the average costsper employee in 2002 dollars were$2278 for medical care, $661 fromabsenteeism, and $6721 from workimpairment. Projecting these resultsto Dow’s entire U.S. workforce, thetotal cost estimate was 10.1% of totallabor costs in 2002: 6.8% from pre-senteeism, 2.3% from use of medicalcare, and 1.0% from absenteeism.

DiscussionThis survey represents the most

comprehensive attempt by a com-pany to assess the prevalence ofwork impairment from chronic healthconditions in its workforce and toestimate the relative contributions ofhealth-related absenteeism, work im-

TABLE 4(Continued)

Variable

Work Impairment Absence

Linear RegressionCoefficient

VariableP-value

GroupP-value

Logistic RegressionOdds Ratio

VariableP-value

GroupP-value

Diastolic Blood Pressure�90 0.0 Reference 1.0 Reference90–99 �1.7 0.038 1.0 1.000100� �0.3 0.882 1.0 1.000

Systolic Blood Pressure�140 0.0 Reference 0.424 1.0 Reference 0.654140–159 �0.2 0.782 0.8 0.051160� �1.2 0.548 0.5 0.063

Total Cholesterol�200 0.0 Reference 1.0 Reference200–239 �0.3 0.563 1.1 0.421240� 0.0 0.985 1.1 0.436

Total Cholesterol/HDL�4.0 0.0 Reference 1.0 Reference4.0–4.9 0.1 0.480 1.2 0.0155.0–5.9 0.0 0.902 1.1 0.6506.0� 0.3 0.059 1.3 0.041

Hours Worked�40 11.4 0.000 0.000 1.6 0.249 0.00040–44 0.0 Reference 1.0 Reference45–50 0.6 0.166 0.7 0.00050–55 1.0 0.084 0.6 0.00055–60 3.4 0.000 0.6 0.00060� 1.3 0.127 0.5 0.000Missing �5.0 0.107 0.4 0.439

RegressionNumber 5108 5240R-Squared 0.179 0.111*Sign 0.000 0.000

*Cox and Snell R-Squared.Analysis in this table was based on linear multiple regression of the Work Impairment Score and logistic regression on reported absence.

Fig. 3. Chronic Conditions.


pairment, and medical care. The re-sponse rate to the survey was highand although there were some differ-ences between survey respondentsand nonrespondents, the partic-ipants represented a broad cross-section of the Dow workforce. Thestudy clearly shows that chronichealth conditions affect a majority ofworkers across the broad range ofknowledge-based and production-based jobs, and these conditionssignificantly impacted work impair-ment, absenteeism, and medial costs.

There are some important limita-tions to our study, which makecausal assessments difficult. First,our survey was cross-sectional indesign. A longitudinal study mayprovide a better assessment of thecauses and impacts of work impair-ment and absenteeism. However,some aspects of our design are lon-gitudinal. For instance, the biometrichealth indicators were collected be-fore the survey. Second, it is likelythat the existence of a chronic con-dition began well before the surveywas conducted. The survey asks forinformation about productivity in ashort, 4-week recall period. Acuteflareups of some chronic conditionsmay be short-lived so we may not beaccurately characterizing the impactof chronic illness. The relationshipbetween chronic conditions and pro-ductivity may differ if a longer recallperiod had been used. Third, ourinformation on work impairment andabsences was taken from a surveyand not directly measured. Also, wegeneralized the costs to the DowU.S. locations based on the findingsfrom several locations in two states.This decision was based on the sim-ilarity of the population, the fact thatthe survey group represented morethan half of the U.S. workforce, andthe high response rate. Our estimatedcosts may be different if we surveyedall Dow U.S. locations.

This study is unique in severalways, however. The availability ofmedical claims, biometric, and otherdata for most of the workers sur-veyed enabled us to develop a com-

prehensive assessment of workforcehealth status, work impairment,health-related absenteeism, and totalcosts. Dow has a diverse workforceand all job types were included in thesurvey. Although some previousstudies have examined work loss in avariety of jobs, most have focusedprimarily on either production or ad-ministrative jobs.4,8,17,22

The prevalence of every chroniccondition was higher based on thesurvey results as compared with theestimates from medical claims (seeFig. 2). This is expected, because notall individuals seek treatment, not alleligible claims are filed, not everydiagnosis is coded, claims were notavailable for all respondents as somewere enrolled in HMOs (�5%), andrespondents may have had claimsthat occurred outside our 12-monthstudy period. However, the magni-tude of the differences in chroniccondition prevalence by data sourcehighlights the importance of lookingbeyond medical claims. Indeed, self-report of a chronic condition was themost important predictor of workimpairment and absences.

Several previous studies havefound a relationship between biomet-ric risk factors such as body massindex and work impairment or ab-senteeism.29–31 We did not find anysuch relationship, but we only exam-ined the biometric risk factors whentaking the chronic conditions intoaccount. Because biometric risk fac-tors often precede and are often as-sociated with chronic conditions, it isnot surprising we did find a relation-ship between these risk factors andwork impairment or absenteeismwhen the chronic conditions areconsidered.

Consistent with other publishedfindings, work impairment repre-sented a far greater proportion of lostproductivity compared with absen-teeism.17 Indeed, our study foundthat almost two thirds of total healthand productivity management costswere attributable to work impair-ment. A similar magnitude (63%)was reported recently by Bank One,

using a different instrument to assesswork impairment.32,33 These find-ings suggest that interventions thatfocus on absenteeism and ignore pre-senteeism not only underestimate thetrue magnitude of the impact ofhealth on productivity, but alsomay not accurately characterize thefinancial return on various healthinterventions. These findings alsosuggest that the decision about whatinstrument should be used to assesspresenteeism is less important thanwhether work impairment is as-sessed.

Because almost two thirds ofworkers reported one or morechronic health conditions, the annualfinancial impact on Dow is quitelarge. The current survey allowedDow to quantify the magnitude interms of direct and indirect costs forits U.S. workforce. Chronic condi-tions alone are estimated to cost Dowmore than $100 million annually inlost productivity for its U.S. work-force. Interventions that prevent dis-ease or improve treatment outcomesmay reap significant returns for thecompany. Although some workimpairment and absenteeism forchronic conditions is unavoidableeven with the most successful healthinterventions, the high prevalenceand total costs of the chronic condi-tions represent a significant opportu-nity for management to increaseworkforce productivity through im-proved worker health. Policymakersand healthcare providers would bewell served to consider the total eco-nomic impact of health conditions,including direct and indirect costs asillustrated in this analysis. Forhealthcare providers, the importanceand value of health outcomes andfunctional health status are rein-forced.

Measurement of presenteeism isan evolving discipline. We note thatthe estimates reported here includethe mitigating effects that currenttreatment had on health status andwork loss; thus, the impact of thesechronic conditions may be greater inother populations if they do not have


comparable health benefits. In thisstudy, we employed the WIS to mea-sure presenteeism. In other research,we are comparing the WIS with thesingle question of WOS, which pro-vides a direct estimate of productiv-ity. Although we believe that theWIS may be more sensitive tochange over time and yield betterdetail for health program manage-ment and evaluation endeavors, werecognize that it may provide a dif-ferent estimate of the magnitude ofwork impairment compared with theWOS (eg, the WIS provides greatervalues than the WOS to the extentthat work impairment is not reflectedin actual worker productivity). Thesedifferences, however, would notchange our findings that work im-pairment represents a greater loss ofvalue in comparison to absenteeismand direct medical costs combined.

We also note that the frameworkfor assessing costs presented here isincomplete. The full impact of aworker’s illness on productivity of-ten reaches beyond the individual’swork loss or equivalent wage. Inrelated research, we have found thatthe costs associated with an employ-ee’s absenteeism will be largestwhen it is difficult to replace thatemployee, he or she operates as partof a team, and the employee’s workcannot be easily postponed.34 Thesethree factors formed the basis for aset of multipliers for 35 differentjobs, which we are applying toDow’s survey results in a separateanalysis.35 Applying these multipli-ers, the total costs of lost productiv-ity at Dow would be significantlygreater.

We have found that illness has asignificant impact on the productiv-ity of our worker in all types of jobs.This information can also helpful inguiding the development of cost-effective interventions for particularconditions, providing a baseline fromwhich to assess intervention effec-tiveness, and design further researchto investigate other illnesses.

Many CEOs have made statementsthat their employees are their com-

panies’ greatest asset. There is littledoubt that the creativity and produc-tivity of the workforce is the enginefor corporate success. Chronic healthconditions are common among alljob types and have the potential tosignificantly impact a company’s fi-nancial performance. Although mostmanagement attention to date hasfocused on direct medical costs andabsenteeism, our experience suggeststhere is far greater loss of productiv-ity resulting from decrements in pre-senteeism, representing a substantialmanagement opportunity as well as acompelling focus for healthcare pro-viders and policymakers.

AcknowledgmentsThe authors gratefully acknowledge the

support and guidance of Drs. Ron Z. Goetzeland Steven M. Teutsch. This study wasfunded by a grant from Merck & Co, Inc.

References1. Rizzo JA, Abbott TA, Berger ML. The

labor productivity effects of chronicbackache in the United States. Med Care.1998;36:1471–1488.

2. Clark CE, MacMillan L, Sondhi S, et al.Economic and social impact of migraine.QJM. 1996;89:77–84.

3. van Tulder MW, Koes BW, Bouter LM.A cost-of-illness study of back pain inThe Netherlands. Pain. 1995;62:233–240.

4. Koopman C, Pelletier KR, Murray JF, etal. Stanford presenteeism scale: healthstatus and employee productivity. J Oc-cup Environ Med. 2002;44:14–20.

5. Van Roijen L, Essink-Bot M, Koopman-schap M, et al. Societal perspective onthe burden of migraine in The Nether-lands. Pharmacoeconomics. 1995;7:170–179.

6. Simon GE, Barber C, Birnbaum HG, etal. Depression and work productivity: thecomparative costs of treatment versusnontreatment. J Occup Environ Med.2001;43:2–9.

7. Cady RD, Ryan R, Jhnigran P, et al.Sumatriptan injection reduces productiv-ity loss during a migraine attack; resultsof a double-blind, placebo-controlledtrial. Arch Intern Med. 1998;158:1013–1018.

8. Burton WN, Conti DJ, Chen CY, et al.The role of health risk factors and diseaseon worker productivity. J Occup EnvironMed. 1999;41:863–877.

9. Lerner DJ, Amick BC, Malspeis S, et al.The migraine work and productivity lossquestionnaire: concepts and design. QualLife Res. 1999;8:699–710.

10. Lipton RB, Stewart WF, Korff M V. Theburden of migraine. Pharmacoeconom-ics. 1994;6:215–221.

11. Mushet GR, Miller D, Clements B, et al.Impact on sumatriptan on workplace pro-ductivity nonwork activities, and health-related quality of life among hospitalemployees with migraine. Headache.1996;36:137–143.

12. Lofland JH, Locklear JC, Frick KD.Different approaches to valuing the lostproductivity of patients with migraine.Pharmacoeconomics. 2001;19:917–925.

13. Davies GM, Santanello N, Gerth W, et al.Validation of a migraine work and pro-ductivity loss questionnaire for use inmigraine studies. Cephalalgia. 1999;19:497–502.

14. Osterhaus JT, Gutterman DL, PlachetkaJR. Healthcare resource and lost labourcosts of migraine headaches in the US.Pharmacoeconomics. 1992;2:67–76.

15. Pelletier B, Boles M, Lynch W. Changein health risks and work productivity overtime. J Occup Environ Med. 2004;46:746–754.

16. Cortelli P, Dahlof C, Bouchard J, et al. Amultinational investigation of the impactof subcutaneous sumatriptan III: work-place productivity and non-workplace ac-tivity. Pharmacoeconomics. 1997;11:35–42.

17. Wahlqvist P, Carlsson J, StalhammarNO, et al. Validity of a work productivityand activity impairment questionnairefor patients with symptoms of gastro-esophageal reflux disease (WPAI-GERD)—results from a cross-sectionalstudy. Value in Health. 2002;5:106–113.

18. Blanc PD, Trupin L, Eisner M, et al. Thework impact of asthma and rhinitis: find-ings from a population based survey.J Clin Epidemiol. 2001;54:610–618.

19. Stang P, Cady R, Batenhorst A, et al.Workplace productivity a review of theimpact of migraine and its treatment.Pharmacoeconomics. 2001;19:231–244.

20. Burton WN, Pransky G, Conti DJ, et al.The association of medical conditionsand presenteeism. J Occup Environ Med.2004;46:S38–S45.

21. Turpin RS, Ozminkowski RJ, Sharda CE,et al. Reliability and validity of the Stan-ford Presenteeism Scale. J Occup Envi-ron Med. 2004;46:1123–1133.

22. Stewart WF, Ricci JA, Chee E, et al. Costof lost productive work time among USworkers with depression. JAMA. 2003;289:3135–3144.


23. Goetzel RZ, Guindon AM, Turshen IJ, etal. Health and productivity management:establishing key performance measures,benchmarks, and best practices. J OccupEnviron Med. 2001;43:10–17.

24. Chemical Manufacturers Association’sEpidemiology Task Group. Guidelinesfor good epidemiology practices for oc-cupational and environmental epidemio-logical research. J Occup Med. 1991;33:1221–1229.

25. Ware JE, Sherbourne CD. The MOS36-item Short Form Health Survey (SF-36), I: conceptual framework and itemselection. Med Care. 1992;30:247–283.

26. Lerner D, Reed J, Massarotti E. TheWork Limitations Questionnaire’s valid-ity and reliability among patients with

osteoarthritis. J Clin Epidemiol. 2002;55:197–208.

27. Norusis MJ. SPSS 11.0 Guide to DataAnaylsis. Chicago: Prentice Hall; 2002.

28. Kessler RC, Barber C, Beck A, et al. TheWorld Health Organization and WorkPerformance Questionnaire (HPQ). J Oc-cup Environ Med. 2003;45:156–174.

29. Reidel JE, Lynch W, Baase C, et al. Theeffect of disease prevention and healthpromotion on workplace productivity.Am J Health Promot. 2001;15:iii–v.

30. Burton WN, Conti DJ, Chen CY, et al.The role of health risk factors and diseaseon worker productivity. J Occup EnvironMed. 1999;41:863–877.

31. Tsai SP, Gilstrap EL, Colangelo TA, etal. Illness absence at an oil refinery and

petrochemical plant. J Occup EnvironMed. 1997;39:455–462.

32. Hemp P. Presenteeism: at work–but outof it. Harvard Bus Rev. 2004;49–58.

33. Goetzel RZ, Long SR, Ozminkowski RJ,et al. Health, absence, disability, andpresenteeism cost estimates of certainphysical and mental health conditionsaffecting US employers. J Occup Envi-ron Med. 2004;46:398–412.

34. Pauly M, Nicholson S, Xu J, et al. Ageneral model of the impact of absentee-ism on employers and employees. HealthEcon. 2002;11:221–231.

35. Nicholson S, Pauly M, Polsky D, et al.Measuring the Effects of Work Loss onProductivity with Team Production.NBER Working Paper 10632; 2004.


the assessment of chronic health conditions on work ... · the assessment of chronic health...

Documents