The Burden of Pain on Employee Health and Productivity at...

FAST TRACK ARTICLE

The Burden of Pain on Employee Health andProductivity at a Major Provider ofBusiness Services

Harris Allen, PhDDavid Hubbard, MDSean Sullivan, JD

Objective: The objective of this study was to examine the burden of painon employee health and productivity at a Fortune 100 company headquar-tered in the northeastern United States to prioritize target areas for reducingthis burden. Methods: An electronic survey was conducted in late 2004,which produced a reasonably representative national sample of 1039 activeemployee respondents. Results: A total of 28.6% of respondents met thestudy definition for pain. Pain was linked to: 1) drops of more than 45%and 23%, respectively, in Overall Physical and Mental Health; 2) afivefold increase in health-induced limitations in work performance; and 3)nearly three and two thirds workdays lost to presenteeism and absenteeismover a 4-week period. Afflicted workers displayed considerable room forimprovement in their capacity for pain control and management. Conclu-sions: The prevalence of pain and its impact on those with the conditioncombine to make it an area of much opportunity for improving workforcehealth and productivity. Musculoskeletal diseases offer a promising initialtarget for corporate intervention. (J Occup Environ Med. 2005;47:658–670)

P ain is either the primary feature or aprominent comorbidity of a numberof medical diagnoses. It is a majordriver of utilization of group health,workers’ compensation, and disabil-ity services and often challengesefforts at medical and patient man-agement in ways that utterly defybeing ignored. As a medical condi-tion, it is also a significant contribu-tor to productivity loss— bothimpaired performance while at work(presenteeism) and time away fromwork (absenteeism)—striking manyemployees with a per-person impactthat is substantial. Four commonpain conditions (headaches, arthritis,back pain, and other musculoskeletalproblems) alone have recently beenestimated to cause productivity lossamong nearly 13% of the U.S. work-force at a total cost of $62.1 billionper year.1

Such characteristics are leadingpain to become a growing priorityfor employers as they look for newways to enhance employee healthand productivity. By way of context,for much of the private sector, theimprovement of existing workerproductivity has emerged as key tosurviving and thriving in the market-place. This development has beenreinforced by such trends as the re-lentless march toward greater glob-alization, the aging of babyboomerworkers, and the changing characterof work itself, which is making agrowing proportion of the workforceno longer interchangeable.2

Employers, in turn, have increas-ingly been drawn to a more proactive

From The Harris Allen Group, Brookline, Massachusetts (Dr Allen); CorSolutions, Poway,California (Dr Hubbard); and the Institute for Health & Productivity Management, Scottsdale, Arizona(Mr. Sullivan).

The authors gratefully acknowledge the sponsorship of the Institute for Health & ProductivityManagement (IHPM) and the financial support of a nonrestricted educational grant to IHPM by PurduePharma, Stamford, Connecticut.

Address correspondence to: Harris Allen, PhD, The Harris Allen Group, 41 Naples Road, Brookline,MA 02446; E-mail: [email protected].

Copyright © by American College of Occupational and Environmental Medicine

DOI: 10.1097/01.jom.0000171054.57677.4c

658 Burden of Pain on Employee Health and Productivity • Allen et al

stance toward worker health. Invest-ments in conventional routes towardproductivity enhancement like edu-cation and training have by and largebeen optimized. By comparison, be-yond paying benefits, health has forthe most part been ignored, and theprospects associated with taking onsuch stewardship functions as healthpromotion, disease prevention, therestoration of function, and return towork have gained an aura of muchpromise.3

The field has responded with em-ployer-sponsored studies on the im-pact of health on work performancethat date back to the late 1990s.4,5

Until recently, the focus has been ona variety of medical conditions, butrarely on pain per se. Examples in-clude studies of the impact of mus-culoskeletal disorders on the work ofbanktellers, of irritable bowel syn-drome at a hotel resort chain, ofallergies on the performance of aheavy manufacturing workforce, andof 19 diseases on productivity at nineTampa employers.6–9

Two developments are now spear-heading the increasing private sectorfocus on pain and productivity: 1)the recent resurgence in per-capitaspending for services by privatehealth insurance, which has out-stripped increases in average hourlyearnings of U.S. workers by 39% to14% during the 1999–2003 period10;2) analyses of claims data that haveamply documented pain’s role as aleading contributor to this trend.11,12

This article reports on a first-of-itskind company survey undertaken tobroaden this focus by examining theburden of pain on the health andproductivity of employees. Althoughclaims analyses suggest that this im-pact is likely to be substantial, theself-report method mounted by thesurvey offers unique data for de-scribing the burden of pain and forcharacterizing its magnitude. Thesurvey medium in effect providesemployees with a systematic way toconvey how their quality of life andwork performance are affected inways that speak to the corporate

bottom line. In so doing, it can gen-erate information to guide corporatedecision-making. Should a concertedeffort be made to mount a diseasemanagement initiative to reduce theburden of this condition? The resultshere offer a case study that addressesthis issue.

Materials and Methods

SampleTo participate, respondents (Rs)

needed to be active employees of thecompany residing in the UnitedStates with electronic access to theInternet either at work or at home.Recruitment took place through sev-eral modalities within the company’scommunications framework, whichwere tapped to introduce the projectand convey invitations to participate.The principal modalities includedelectronic messages transmitted inmass (ie, no personal addresses) for-mat through the company’s WebMDweb site and hard copy invitationsissued at the company’s “Huddle”calendar meetings (ie, meetings thatoccur the first of the month at sitesfor which electronic access is notavailable).

A total of 1039 employees re-sponded with completed surveys.The response rate could be atbest only approximated given the in-formation available. Company man-agement estimates that 40% of theworkforce had Internet accessthrough the worksite at the time ofthe survey (the percent with Internetaccess through home or some othermeans is unknown and still an issueof vigorous debate within the com-pany). Of those with Internet access,a further estimate is that roughly25% had signed up for the compa-ny’s WebMD program and was re-ceiving electronic messages throughthe program’s server at the time ofthe survey (the percent of employeesreached through other means whoalso had Internet access was notknown). These figures yield a re-sponse rate estimate of 43.3% [1039/(24,000*.40)*.25)].

This figure could well be at thehigh end of the plausible range ofpossible response rates given that theunknown variables described heremay have had the net effect of in-creasing the denominator. Yet, irre-spective of the level of response, theobtained sample proved reasonablyrepresentative of the target popula-tion. The obtained sample’s averageage was virtually equivalent to thatof the target population, and its per-centage male was a modest sevenpoints below that of the latter (Table1). Accordingly, this sample offereda viable basis for generalizing to thepopulation targeted by the survey.

Data CollectionEmployees opting to participate

had to hyperlink to a URL listed inthese invitations, which routed themto a web site on which they, in turn,were able to take and complete thesurvey. Rs had the option of access-ing this URL while on breaks atwork or at home (or some othernonwork-related venue). The surveytook some 7 to 15 minutes to com-plete. Rs could do it all in one sitting,or alternatively take breaks and re-turn to their (the same) computerwithin the same 24-hour period tofinish it. The data collection periodextended from early Novemberthrough mid-December 2004.

Incentives and ConfidentialityTo incentivize participation, the

first 1000 Rs were awarded $10Amazon.com gift certificates. To re-ceive these certificates, Rs had toinclude their names and e-mail ad-dresses both obtained in a mannerthat was decoupled from the survey

TABLE 1Study Target Population andObtained Sample: A Comparison ofDemographic Characteristics

Target Obtained

Age (average inyears)

40.2 40.4

Percent male 52% 45%

JOEM • Volume 47, Number 7, July 2005 659

responses provided. To meet thecompany’s confidentiality require-ments, no other personal identifierswere elicited at any point in theprocess. As a result, all survey re-sponse data are and will remain dei-dentified to all parties, including thesurvey administrator.

SurveySurvey design served several ob-

jectives: 1) mount a general popula-tion approach that is relevant to allactive employees regardless ofhealth or clinical status and can iden-tify employees meeting study criteriafor pain; 2) enable the identificationof a healthy benchmark sample fornormative purposes; 3) focus on bothdescription of the pain experienceand quantification of the burden ofpain on health and productivity; and4) ask about ways affected employ-ees are coping with the condition.

To achieve these objectives, thesurvey examined a well-used set ofconcepts for measuring the burdenand management of disease, all tai-lored, when appropriate, for the as-sessment of pain: health status,chronic disease, presenteeism, ab-senteeism, medication use, conditionmanagement/coping, health risk be-haviors, job characteristics, anddemographics. As a set, these con-cepts—and their sequence of presen-tation in the context of a singleinstrument— have seen substantialuse in recent work.8,9 In most cases,the items used to assess them have aproven (ie, validated) track record byvirtue of prior studies. Appendix 1lists these concepts and the sourcesof the items used to measure each,and comments on the rationale forinclusion.

Pain DefinitionThe definition used to identify the

pain sample relied on a blending oftwo approaches whose capacity fordetecting patients with pain havebeen well documented in previouswork. One drew from the SF-36Health Status survey to field the itemassessing the extent of bodily pain

over the previous 4 weeks13; theother was adapted from the BriefPain Inventory to measure the pres-ence or absence of pain (other thaneveryday kinds of pain) the day ofthe survey.14

Rs had to report the experience ofpain on both items to meet the studydefinition. In so doing, they exhib-ited being impacted by pain in amanner that could be consideredboth sensitive and specific from anepidemiologic perspective. By re-porting the presence of out-of-the-ordinary pain the day of the survey,employees conveyed that their painwas immediate and compelling innature, thereby augmenting the sen-sitivity of the definition. By report-ing that pain had been experiencedover the prior 4 weeks, they alsorelated that their pain was also one ofduration, not simply a passing phe-nomenon, thereby augmenting thedefinition’s specificity. Together, thetwo indications made for a conserva-tive yet robust combination for em-piric study.

The two approaches were furthercombined to stratify employees withpain by severity. The item used forthe “bodily pain over 4 weeks” ap-proach had six response options, fiveof which rated the intensity of painranging from mild to very severe.The “non-everyday pain today” ap-proach followed the “presence/absence today” item with a series offour items, asking Rs to rate theirpain at its worst, its least, on average,and right now on 0- to 10-pointscales.

This second set of items was firstaveraged, then transposed into a 0- to100-point scale. Rs with scores rang-ing from 0 to less than 15.0 wereclassified as lowest-severity-pain pa-tients; Rs with scores from 15.0 to lessthan 42.5 were classified as medium-severity-pain patients; whereas Rswith scores ranging from 42.5 to 100were scored with highest severity.

Next, the 4-week item was trans-posed to a 0- to 100-point scale.Those scoring from 0 to 20 wereclassified as lowest severity; those

scoring from 21 to 60 were classifiedas medium severity; whereas thosescoring from 61 to 100 were classi-fied as highest severity.

In the analyses reported next, wetreated severity in some cases as asingle overall measure and in othersas a set of three modified dummyvariables for lowest, medium, andhighest severity. In each case, thesescores were calculated by taking theaverage of the “bodily pain over 4weeks” severity measure and thecorresponding “non-everyday paintoday” severity measure. For themodified dummy variable for eachseverity level, Rs received a score of0 if they reported no experience ofpain on either the corresponding“pain over 4 weeks” or the “non-everyday pain today” measure forthat severity level. They received ascore of 0.5 if they reported theseverity level in question on oneseverity measure but not the other(eg, they received a 0.5 score on thelowest-severity dummy variable ifthey reported lowest severity on the“bodily pain over 4 weeks” severitymeasure but not on the “non-everyday pain today” severity mea-sure). They were scored 1.0 if theyreported the severity level on bothmeasures.

Study GroupsAs shown in Table 2 and Figure 1,

297 of the 1039 Rs, or 28.6%, metboth pain criteria. Fully almost onein three employee respondents metthis study’s conservative definitionfor pain.

The severity stratification of thisgroup yielded a bell-shaped distribu-tion, with the majority of affectedrespondents falling in the mediumgroup. The modified variable scoringstrategy meant that each pain respon-dent could be designated a certainseverity level on one severity mea-sure and another severity level on theother severity measure. The rightpanel of Figure 1 conveys the resultof this scoring by differentiatingthose who scored a certain severitylevel on one but not both measures


from those who scored the severitylevel on both severity measures. Itcan be noted that this procedure re-sulted in the total number of “hits”across each of the three severity lev-els exceeding the total sample size ofthe pain group (ie, 297).

Figure 1 also shows the prevalenceof a second group, the healthy bench-mark, which functioned as a keynorm in the analyses discussed sub-sequently. To be eligible, Rs neededto report none of 24 diseases on thesurvey’s chronic condition checklistand above average health status onboth of the study’s two summaryhealth measures: overall physicalhealth and overall mental health (seesubsequently for a description). Thisdefinition allowed employees withcolds, the flu, and so on, but other-wise good health to be classified as

healthy, thereby providing a realisticbenchmark for comparison.

A total of 102 or 9.8% of thesample met these criteria. This per-centage is a little lower than the 15%to 20% levels we have typically seenin previous work with other employ-ers. A major reason for this was thehigh number of employees reportingone condition in particular: allergies.At 52%, this figure nearly doubledthe levels observed in our pastwork.8

Study MeasuresThe criteria used to compare these

groups spanned three categories:health, productivity, and pain man-agement. Table 3 lists these mea-sures and their descriptive statistics.

Health. The survey’s nine itemstaken from the Short-form 36 Health

Status Survey—each assessing ei-ther physical functioning, vitality,mental health, general health, orbodily pain—were combined to formtwo orthogonal measures of health:overall physical health and overallmental health, also known as thePhysical Component Summary(PCS) and Mental Component Sum-mary (MCS).

The PCS and MCS used here derivefrom the version of the PCS and MCSoriginally developed from the SF-12,itself a subset of the SF-36 HealthStatus Survey.16 These scales wereimputed through a modified regressionestimate technique—the same method-ology used to derive the SF-12 PCSand MCS with an additional errors-in-variables correction. 17 These imputa-tions used as norms for the calcula-tions a series of measures taken froman in-house active employee databasedeveloped in previous survey workwith a group of similar employers (ie,the Xerox, Digital Equipment, andGTE Corporations18).

In addition, a disease count wascomputed from all “yes” reportsgiven in response to a chronic con-dition checklist adapted from theMedical Outcomes Study19 and adepression measure developed fromthe three items taken from a newscreener for major depression anddysthymia.20 Rs reporting no chronicconditions and scores greater than 50on both the PCS and MCS measurescomprised the healthy benchmarkgroup.

Productivity loss. The study’smeasures of productivity loss cov-ered three conceptual distinctions:workdays lost resulting from health,limitation in performance on speci-fied job dimensions resulting fromphysical or mental health problems,and limitation in overall effective-ness at work resulting from physicalor mental health problems.

The workdays lost distinction wascomprised of: 1) the absenteeism or“time away from work resultingfrom health” developed by Kessler etal21 and 2) the days less than 100%resulting from health measure firstFig. 1. Key study groups.

TABLE 2Study Definition for PainBlending Two Validated ApproachesBrief Pain Inventory (Cleeland & Ryan, 1994)

Throughout our lives, most of us have had pain from time to time (such asminor headaches, sprains and toothaches). Have you had pain other thanthese everyday kinds of pain today?

SF-36 Health Status Survey (Ware, 1993)How much bodily pain have you had during the past 4 weeks?

SF-36: Amt of pain last 4 wks

BPI: Has Pain Now

No Yes Total

None 229 18 247Very Mild to Very Severe 495 297 792

Total 724 315 1039


used by Allen and Bunn in the Inter-national Truck and Engine Allergyproject.15

The “limitation on job dimen-sions” distinction was operational-ized by 12 items taken from theWork Limitations Questionnaire.22

These items assessed capacity tomeet performance requirements onone of four dimensions as a func-tion of health: time demands, phys-ical demands, mental/interpersonaldemands, and output demands.Eight of these items (two for eachdimension), comprising the WLQshort form, were combined to forma summary WLQ scale.23

The third distinction, overall effec-tiveness at work, was captured by thesingle-item index first deployed byAllen and colleagues in the TampaHealthy People Survey.9 Both thisitem and the WLQ items were eachtransposed to the 0- to 100-pointcontinuum, with high limitationscored high.

Pain management. Three mea-sures comprised this final category.Two were three-item scales focusingon satisfaction with treatment forpain and on coping with life in gen-eral. The other measure was a single-item scale asking Rs to rate thedegree of completeness of their pain

control. All three measures weredrawn from the recently validatedTreatment Outcomes of Pain Surveydeveloped by Rogers et al.24

AnalysesThe Stata software program was

used to implement an analytic strat-egy that consisted primarily of theestimation of direct effect modelsusing ordinary least-squares regres-sion.25 All multivariate analyseswere controlled for age and gender.

ResultsTable 3 also reports certain findings

that help to provide context. With re-

TABLE 3Study Measures

HealthNo. ofitems Range Mean S.D. Meaning of high score

Health StatusGeneral health 1 0–100 69.1 19.5 Excellent healthPhysical functioning scale 4 0–100 90.9 16.6 No physical limitationVitality—energy level 1 0–100 60.6 24.1 A lot of energy all the timeMental health—down in dumps 1 0–100 81.6 21.3 None of the timePhysical component summary 9 0–100 52.0 8.2 Excellent physical healthMental component summary 9 0–100 47.8 11.3 Excellent mental healthDepression screener 3 0–100 25.2 43.4 At risk for major depression

and/or dysthymiaDiseases: count (out of 24) 24 0–24 2.7 2.5 Many diseases

PainPain Severity 2 0–100 37.2 16.8 High severity

Low 2 0–100 23.9 35.8 High proportionModerate 2 0–100 60.1 38.1 High proportionHigh 2 0–100 16.0 29.7 High proportion

Health Risk behaviors% Smoking 1 0–100 17.1 37.7 % smoking �1 cig/day% Overweight 1 0–100 29.8 45.8 % BMI � 30.0% Drinking 1 0–100 10.2 30.3 % �1 drink/day% Sedentary lifestyle 1 0–100 30.3 46.0 % exercising � once/wk

ProductivityDays Lost—Last 4 wks

Days absent 1 0–28 .47 1.78 More daysDays at work �100% due to health 1 0–28 1.91 4.32 More days

Presenteeism: —Last 4 wksTime demands 2 0–100 27.9 34.6 Health severely limitsPhysical demands 2 0–100 16.2 23.4 Health severely limitsMental/interpersonal demands 2 0–100 12.2 17.4 Health severely limitsOutput demands 2 0–100 11.1 24.2 Health severely limitsWLQ scale 8 0–100 16.9 16.6 Health severely limitsOverall work effectiveness 1 0–100 9.7 17.1 Health severely limits

Self-reported Accidents—Last yr% reporting 1� accidents 1 0–100 4.9 .22 More accidents at work

Pain ManagementSatisfaction w/Treatment 3 0–100 60.7 24.7 Highly satisfiedPain control 1 0–100 66.3 27.7 Complete controlCoping: past week 3 0–100 74.1 22.5 Completely able to cope


spect to health, the sample exhibitedbetter-than-average physical health butlower-than-average mental health (ie,its mean scores of 52.0 for PCS and47.8 for MCS were, respectively, 2points above and 2.2 points below thecorresponding norms in our in-housedatabase). Roughly one fourth of thesample indicated through the depres-sion screener that they were at riskfor major depression and/or dysthy-mia. Some 30% gave responses indi-cating that they were at risk for beingoverweight and having a sedentarylifestyle.

As for productivity loss, less than5% of the sample recorded at leastone accident or injury on the job overthe last year—a rate that comparesfavorably with rates observed forother employers in our in-house da-tabase. In contrast, of the perfor-mance dimensions, time demandsstood out with an average score thatwas strikingly high, both with re-spect to the other work dimensionsobserved for this company and withrespect to the time demand scores wehave observed for other employers inpast studies. Time pressures and re-lated job stresses would appear to bea cardinal feature of this company’swork environment for many employ-ees irrespective of the burden ofdisease.

Pain BurdenMarked by a linear pattern that

increased with severity on all nineproductivity measures examined, theburden exacted by pain on produc-tivity loss was substantial (Table 4).The healthy benchmark posted neg-ligible loss scores on virtually all ofthese measures. In each case, theadded loss resulting from the pres-ence of pain—whether expressed ineither average number of days, per-cent reporting one or more accidents,or the average extent of limitation—was statistically dramatic. This loss,indexed by comparing the pain groupoverall with the healthy benchmark,registered highly significant (P �0.0001) increases on eight of themeasures and a significant increase

(0.02) on the ninth (accidents). Fig-ure 2 shows these relationships fordays absent and days less than 100%during the previous 4 weeks. As canbe seen, the two combined totaled toan average of just over one third of aday for healthy employees; the cor-responding figure for those with painwas approximately 4 days, leavingthe difference (three and two thirdsdays) directly attributable to the im-pact of pain.

The stepwise nature of this in-crease was evident with tests com-paring the severity and healthygroups (Table 4). On five of thenine measures, the lowest severitygroup posted a significant jump inloss relative to the healthy group.On eight measures, the burden ofpain for the medium-severity groupwas a significant jump relative to

the lowest severity group, whereasfor all nine measures, the highestseverity group’s increment relativeto the medium group was highlysignificant.

Figure 3 summarizes this patternby showing the results for the fourWLQ demand scales.

Particularly revealing was the re-sults for time demands. Reflectingan evidently time-pressured envi-ronment, the healthy group’s scoreregistered a not inconsequentialburden at 16.8 on the 0 –100 limi-tation scale. Yet, even here, theburden associated with highest se-verity pain almost tripled this lim-itation score to 46.1.

This pattern replicated on thehealth measures (Table 5). Thehealth status measures—each scoredin the opposite direction to link high

Fig. 2. Pain and workdays lost to health problems (last 4 weeks).

Fig. 3. Burden of pain on performance at work (difficulty/ability to meet job demands).


scores with better health—yieldedthe mirror reverse, with the healthygroup scoring high and each severitygroup registering successively lowerscores. Figure 4 shows this patternby displaying the results for the sum-mary PCS and MCS measures. Over-all physical health fell sharply withpain severity, progressively worsen-

ing with each level of severity. Over-all mental health also fell sharply,with the full effect of pain achievedat medium pain severity with aver-age scores equivalent to a clinicaldepression diagnosis (ie, a score of43 on the MCS scale).

The disease count measure like-wise posted stepwise increments,

from zero diseases for the healthybenchmark (reflecting the group’sdefinition) to an average of morethan six diseases for the highestseverity pain group. With respect tothe health risk behavior measures,in contrast, these groups did notdisplay quite the linear pattern. Forexample, for employees in the me-

TABLE 4Productivity by Pain Status

Healthy

Pain Severity LevelHealthy vs. All Severity

Levels

Lowest Medium Highest Test df P

Days Lost—last 4 wksDays absent 0.03 0.33 0.86*† 1.67*† F � 15.5 1,1032 �0.0001Days at work �100% due to health 0.12 1.73*† 3.2*† 5.8*† F � 40.1 1,1032 �0.0001

Self-reported accidents—Last yr% reporting 1� accidents 2.6 5.1 7.5 26.6*† X2 � 5.2 1 0.02

Limitations while at work: performance dimensions—last 4 wksTime Demands2 16.8 30.2* 32.6* 46.1*† F � 19.0 1,997 �0.0001Physical Demands2 6.2 12.8* 25.4*† 51.5*† F � 70.1 1,1011 �0.0001Mental/Interpersonal Demands2 2.1 11.3* 18.2*† 32.2*† F � 75.8 1,1022 �0.0001Output Demands 3.1 13.2* 14.3* 22.9* F � 19.3 1,1021 �0.0001WLQ Scale2 7.1 17.0 22.2*† 38.5*† F � 87.0 1,1030 �0.0001

Limitations while at work: Performance Overall—Last 4 wksOverall Effectiveness2 0.49 6.6* 16.2*† 32.4*† F � 74.8 1,1032 �0.0001

1 Row entries are least squares (or logistic) regression estimates adjusted for age and gender.2 Range: 0 � no limitation; 100 � severely limited.* Score significantly different (P � .05) from score for Healthy group. No asterisk indicates that this difference is not significant.† Score significantly different (P � .05) from cell to immediate left. No † indicates that this difference is not significant.

TABLE 5Health by Pain Status

Healthy

Pain Severity LevelHealthy vs. All Severity

Levels

Lowest Medium Highest Test df P

Health StatusGeneral health2 83.0 67.5* 61.5*† 50.4*† F � 94.3 1,1032 �0.0001Physical functioning scale2 98.9 93.4* 87.1*† 72.0*† F � 50.2 1,1032 �0.0001Vitality—energy level2 80.5 63.0* 51.6*† 45.3* F � 86.1 1,1032 �0.0001Mental health—down in dumps2 98.3 83.0* 73.3*† 67.8*† F � 83.1 1,1032 �0.0001Physical component summary 57.9 53.4* 46.0*† 31.7*† F � 334.8 1,1032 �0.0001Mental component summary 57.0 48.6* 43.2*† 43.7* F � 76.1 1,1032 �0.0001

Health Risk behaviors% Smoking 8.9 15.5 18.5* 35.6*† X2 � 6.7 1 0.01% Overweight 12.5 31.0*† 40.7*† 50.4 X2 � 21.5 1 0.00% Drinking 5.5 17.5* 9.9 16.8*† X2 � 5.3 1 0.02% Sedentary lifestyle 13.1 32.3* 41.2* 48.8 X2 � 20.8 1 0.00

Diseases# of conditions (out of 24) 0 2.6* 4.3*† 6.3*† F � 330.3 1,1032 �0.0001

1 Row entries are least squares (or logistic) regression estimates adjusted for age and gender.2 Range: 0 � poor health; 100 � excellent health.* Score significantly different (P � .05) from score for Healthy group. No asterisk indicates that this difference is not significant.† Score significantly different (P � .05) from cell to immediate left. No † indicates that this difference is not significant.


dium-severity group, pain had nogreater risk prevalence than thelowest severity pain on smoking,drinking, or sedentary lifestylerisks. Yet, even here, the paingroup as a whole was more at riskthan the healthy benchmark oneach of the four health risk behav-ior measures assessed. Those in thepain group were significantly morelikely than healthy employees to beat risk for being overweight, aswell as for smoking, drinking, anda sedentary lifestyle.

Pain ManagementThe survey uncovered a number of

ways that employees with pain re-ported they are explicitly using tomanage the condition. These methodsfall into four categories: medicationuse, complementary methods, health-care utilization, and health-relatedbehaviors.

Of the 297 employees in the studypain sample, 81% indicated that theyhad taken at least one medication(from the survey’s list of 30� med-ications) for pain over the previousyear (Table 6). The majority (ie,more than three in five) reported thatthey had taken Advil, Motrin, aspi-rin, and/or Tylenol either for head-ache pain and/or others kind of pain.Slightly more than 10% reported tak-ing one or more of the followingprescriptions: Celebrex, Vicodin, orNaprosyn/Naproxen. (All other med-ications recorded less than 10% use.)

The second most frequently usedcategory, at 59%, consisted of variousforms of healthcare utilization. Some55% reported at least one outpatientvisit to a medical clinician for painover the previous year, whereas 15%reported at least one emergency roomvisit and 6% at least one hospital inpa-tient night explicitly for pain. In addi-

tion, 5% of the sample reported thatthey had had at least one psychiatric/psychologic visit for pain.

The third most often reported cat-egory was health-related behaviors(28%). Some 17% of employees inthe pain sample reported that theywere exercising to alleviate pain,whereas 12% said they had changedtheir diet. Some 10% reported drink-ing at least one alcoholic beverageper day for pain relief. At 25%,various options for use of comple-mentary medicine comprised the fi-nal category. Some 16% reported useof herbal medicine, whereas just over10% reported use of acupressure anda chiropractor.

Use of these methods was notmutually exclusive, either within oracross categories. However, regard-less of the combination, the painsample gave much reason to inferthat—at least from the employee per-spective—there was considerableroom for improvement in how theirpain was being managed. As shownin Table 7 and Figure 5, those in themedium-severe category indicatedthat their pain in the last week wason average less than two thirds (ie,65.1%) completely controlled. Thosein the highest severity group reportedthat their pain on average was lessthan 40% completely controlled.

These pain control scores werereflected in the extent of satisfactionwith their treatment program forpain. The average employee withmedium severe pain was less thansomewhat satisfied with his or hertreatment; the average employee

Fig. 4. Burden of pain on health status (overall physical and mental health).

TABLE 6Methods Used to Manage Pain

Pain Sample (n � 297)

Medication 81% Complementary 25% Utilization 59% Health Behaviors 28%

Advil/Motrin 62% Herbal 16% Inpatient admission 6% Exercise 17%Aspirin/Tylenol 61% Acupressure 13% ER visit 15% Changes in diet 12%Celebrex 12% Chiropractic 11% Outpatient Psych visit 5% Drinking alcohol 10%Vicodin 12% 3 others: each �10% Outpatient Medical visit 55%Naprosyn/Naproxen 11%26 other meds: each �10%


with highest severe pain was in factsomewhat dissatisfied with his or herprogram. Those in the lowest sever-ity were more positive, giving anaverage score of very satisfied. Yet,this average still fell short of com-pletely satisfied; even this group ex-pressed room for improvement intheir program.

The extent of success realized bythese treatment programs in control-ling pain evidently had broader ram-ifications for the capacity of affectedemployees to cope with life chal-lenges. The study’s coping measureaddressed the extent of perceivedcontrol employees felt they had overtheir life more generally, their capac-ity to handle problems, and theircapacity to cope with stressful situa-tions. Pain severity showed a strongrelationship on this measure as well.The lowest severity group reportedan average coping score, suggestingthat their capacity in this regard wasroughly four fifths what it could be,

whereas the highest severity group’sscore suggested an average capacitylevel that was not even three fourthsof that for the lowest severity group.

Reducing Pain BurdenWith the evidence for pain bur-

den—and room for improvement inreducing this burden—so substantial,what did the data have to say aboutwhere to start in this regard? Thesurvey offered one way for identifyingand prioritizing the opportunities froma clinical/epidemiologic perspective:ranking pain burden by disease.Given that the resources available forsuch an initiative would likely in anycase to be limited, this analysis waspredicated on the assumption thattackling those conditions exertingthe greatest burden first would pro-vide the best chances for getting the“biggest bang for the buck.”

To operationalize this notion ofpain burden, survey data were usedto calculate two factors for each dis-

ease in the pain sample: its preva-lence and its average pain severity.Table 8 provides these calculationsand reports rankings by individualdisease and by disease groupings. Asshown, allergies—fueled primarilyby its high prevalence—ranked firstamong individual diseases. Whendiseases were grouped, however, thehighest ranking belonged to the mus-culoskeletal group. All things con-sidered, if the decision is made to actand resources are limited, the com-bined prevalence of arthritis, backpain, and neck and shoulder prob-lems, together with their respectiveper-patient impacts, would appear tomake this disease grouping the mostfertile starting point for intervention.

DiscussionThis study makes a compelling

case for steps to reduce the burden ofpain on employee health and produc-tivity at this company. It also pro-vides pointers as to where to start inthis regard. First, the company’sworkforce and its work environmentsport a profile that is ripe for suscep-tibility to pain. Although employeephysical health is better than average(relative to employee populationselsewhere), employee overall mentalhealth would appear to be worse thanaverage. Just over one fourth of em-ployees exhibit risk for depression, afrequent comorbidity of pain.26 Thiscompany’s employees also averagemore than two and a half medicalconditions per employee, and almostone third are at risk for overweightand another third are at risk for asedentary lifestyle—all characteris-tics heavily implicated in the painexperience. Moreover, the companyworkplace is one whose time de-mands many employees evidentlyfind to be considerable and condu-cive to acerbating the condition.

Second are the epidemiologicfindings. Nearly one in three em-ployees is affected by pain—a lessconservative definition than thatused here would have found an evengreater proportion. Moreover, theburden of pain on employee healthFig. 5. Management of pain (outcomes).

TABLE 7Outcomes of Pain Management

Pain Severity

Lowest Medium Highest

Satisfaction w/Treatment* 80.3 58.6 39.3Pain Control† 89.4 65.1 36.0Coping† 83.1 74.2 60.0

* Range: 0 � completely dissatisfied; 100 � completely satisfied.† Range: 0 � none; 100 � complete.


and work productivity is consider-able. For employees in the highestseverity group (relative to thehealthy benchmark), pain is associ-ated with: 1) drops of more than 45%and 23%, respectively, in overallphysical and mental health (as exhib-ited by the summary WLQ scale); 2)a fivefold increase in health-inducedlimitations in performance while onthe job; and 3) the addition of nearlythree and two thirds workdays lost topresenteeism and absenteeism over a4-week period.

Third, although a variety of meth-ods are being used to manage pain,many employees with the conditionrate their current treatment approach

as far short of optimal. Those withmedium- and highest severity pain,in particular, indicate that there ismuch room for improvement, both intheir capacity for pain control as wellas in their treatment(s) for managingthe condition. Nowhere is the oppor-tunity for improvement more evidentthan among those employees withpain attributable to musculoskeletaldiseases. Comparisons of pain bur-den by disease—generated by exam-ining the combined prevalence andper person impact of 24 individualconditions—found that musculoskel-etal diseases offered the greatest“bang for the buck” potential of allgroupings studied.

Empiric studies are rarely with-out their possible limitations, andthis undertaking is no exception. Afirst question concerns sample rep-resentativeness. The age/sex con-trols in the analyses adjusted forthe fact that female employees,who are somewhat more likely toreport pain (and a major comorbid-ity, depression), were slightly over-represented in the data. Yet, theextent to which our sample wasrepresentative of all company em-ployees with electronic access, orfor that matter the extent to whichemployees with electronic accessare representative of all of the com-pany’s active employees, remains

TABLE 8Ranking Pain Burden by Disease: Pain Sample

Disease

Individual Disease

Prevalence(Pain Sample)

Pain Severity(0–100)

Prevalence �Pain Severity Disease Rank

Allergy 58.2% 38.0 22.1 1Neck/spine 42.8% 43.2 18.4 2Low back 38.1% 45.3 17.2 3Depression 38.1% 41.2 15.7 4Arthritis/rheumatism 33.3% 43.4 14.5 5Sleeping problems 27.6% 42.9 11.9 6Migraine 27.9% 41.9 11.7 7Heartburn 25.6% 42.8 10.9 8Hypertension 21.5% 39.3 8.5 9Irritable bowel syndrome 13.5% 44.5 6.0 10Deafness 12.5% 43.1 5.4 11Limitations 9.1% 52.5 4.8 12Dermatitis 10.8% 42.7 4.6 13Chronic lung disease 8.8% 44.0 3.9 14Urinary 8.4% 43.4 3.7 15Blindness 7.4% 41.4 3.1 16Anemia 6.7% 41.6 2.8 17Diabetes 5.7% 42.5 2.4 18Ulcer 3.4% 53 1.8 19Cancer 3.4% 52.5 1.8 20Heart disease 4.7% 34.3 1.6 21Skin cancer 2.4% 36.1 .85 22Seizures 1.0% 38.8 .39 23Kidney disease 1.0% 29.6 .30 24

Disease Grouping

Disease Grouping

Disease GroupRanking

Diseases(indexed by above ranking)

Combined Prevalence� Pain Severity

Musculoskeletal 2,3,5 50.1 1Mental & nervous 4,6,7 39.3 2Respiratory 1,14 26.0 3Gastrointestinal 8,10,14,15,19,24 22.7 4Cardiovascular 9,18,21 12.5 5Cancer 20,22 2.65 6All other measured conditions 11,12,13,16,17,23 21.09 —


unknown. Further work clarifyingthis issue is needed before esti-mates of pain burden can be final-ized in this employee population.

Second, pain detection relied ex-clusively on survey self-report. Thevalidity of self-report has been thesubject of considerable past work bythe first author and others.27 Al-though pain is inherently a subjectivephenomenon that arguably is system-atically best examined by thismethod, it remains the case that con-firming diagnostic data from admin-istrative sources (eg, group health)would have bolstered the credibilityof the pain designations.

Third, the pain severity classifica-tion used here differed in two mainrespects from what has been reportedin the literature and could be ques-tioned in light of previous research.First was the classification’s combina-tion of the pain severity item drawnfrom the SF-36 survey with the sever-ity measure developed from the fourBPI severity items—a synthesis that toour knowledge was used for the firsttime in this context. Second were thecut points used for the BPI componentof the severity measure. These cutpoints were an approximation of thecut points first advocated by Serlin etal.28 and were a departure from anumber of other studies that have sys-tematically explored other cut pointcombinations.29–31

For the SF-36 component of theoverall severity measure, it can benoted the severity classification wasbased on a straightforward designationof the labels employees could choosefrom when answering the SF-36 item.For example, those respondents whoanswered by saying their pain wassevere or very severe (the two optionsat the most severe end of the six-pointresponse scheme) were here assignedto the highest severity group. For theBPI component, the severity classifica-tion was based largely on distributionalcharacteristics. Here we endeavored tobalance the need to have at least areasonable minimum number of re-spondents in each group with the de-sire to have an overall distribution of

group percentages that was a reason-able approximation of the correspond-ing group percentages for the SF-36component.

Additional tests that reran analysesreported here using an adaptation ofthe cut points for the BPI severitymeasure advocated by one of the mostrecent studies in this area, Zelman etal32 (3.0/6.0 on the average of the“worst” and “average” BPI severityitems) found no substantive differ-ences. The means for each severitygroup stayed virtually the same, thepattern of significant differences be-tween groups stayed intact, and theoverall interpretation and conclusionsremained unchanged. Yet, to avoidany unintended connotations, it wasultimately deemed best to couch theseverity designations in terms that em-phasized the distributional relation-ships (lowest, medium, and highest)and to leave the substantiation of des-ignations rooted in clinical terms(mild, moderate, severe) for futureresearch.

Fourth, for many readers, convert-ing into dollars the productivitylosses attributable to pain wouldlikely strengthen the burden analy-ses. Those not familiar with surveytechniques tend to relate morereadily to estimates conveyed in“dollarized” metrics than to thoseused here (eg, percent-limited).

These possible limitations with-standing, it would appear that a sys-tematic effort to target and reduce theburden of pain offers this company asignificant opportunity to create a new“win–win”—nurturing a better qualityof life for many of employees while atthe same time promoting a more pro-ductive workforce. What steps couldbe taken to better equip this companyfor determining whether to pursue thisopportunity? We close with three sug-gestions that pertain, not only to thiscompany, but to other employers aswell, which may be contemplating tak-ing action to measure and/or reducethe burden of disease on workforcehealth and productivity.

First, an exhaustive effort should bemade to tap the available databases to

refine the information for shaping andguiding strategy development. In thiscase, a more expansive picture of thecombinations of methods that affectedemployees are using to manage paincould be generated from the existingdatabase that would help to interven-tion developers to better anticipatewhat treatment methods and ap-proaches need be attended to and per-haps changed. Explicit comparisons ofthe relationships that these methodcombinations have with the variousmanagement outcome measures couldbe used to better characterize the com-binations and to prioritize them interms of effectiveness. Teasing apartsome exploratory results (not reportedhere) that have provocatively linkedjob characteristics measured in the sur-vey (ie, job demands vs. job control33)with productivity loss could similarlyhelp inform strategy development.

Second, identifying the alternativeintervention programs available andconducting analyses that compare thereturn-on-investment (ROI) for theseoptions would provide company man-agement with the framework needed tomake what is essentially a businessdecision. One promising mechanismfor conducting this kind of exercise isthe Economic Valuation Tool, whichthe Institute for Health and Productiv-ity Management is distributing to tar-geted members free of charge.34 Theimplementation costs associated withthe program alternatives for interven-tion would need to be identified. Thesecosts would need to be comparedagainst projected savings in terms ofrecouped productivity loss, reducedclaims costs, and so on, to generate theROI ratios that would help to guide thedecision.

A final consideration concernsbalancing employee confidentialitysensitivities on one hand with theprerequisites for effective and effi-cient study design and interventionimplementation on the other. In gen-eral, the use of personal identifiersgreatly facilitates—indeed is criticalto—the accuracy of any analysis ofpre–post change as well as the capac-ity of any intervention to make the


kind of connections with employees/patients who have any hope ofachieving real behavioral change.Yet, in this engagement, the decisionwas made upfront to forego any pro-curement of personal identifiers dur-ing data collection in a manner thatpermitted linkage of these identifierswith survey responses.

Corporate concerns over employeesensitivities in this area drove this de-termination, and it could well havebeen the best decision to make at thetime given the company’s culture andcircumstances. Yet, if an interventionand its evaluation are ever elected inthe future, at some point, personalidentifiers will need to be obtained andanother round of “pre” data collected.In this regard, it is of note that themaintenance of participant confidenti-ality, while implementing and evaluat-ing interventions, is now an estab-lished precedent in the field (as oneexample, see Allen et al35). For thoseintent on taking the next step, a num-ber of “how to” Health InformationPortability and Accountability Act(HIPAA)-compliance guides are nowsurfacing that can be drawn on forreference purposes to help direct futurework in this area.37

References1. Stewart W, Ricci J, Chee E, et al. Lost

productive time and cost due to commonpain conditions in the US workforce.JAMA. 2003;290:2443–2254.

2. Sullivan S. Employee health: 21st cen-tury human capital. Health & Productiv-ity Management. 2003;2:4–5.

3. Reynolds C. Why health and productivitymanagement? Health & ProductivityManagement. 2003;2:6–8.

4. Burton W, Conti D, Chen C, et al. Therole of health risk factors and disease onworker productivity. J Occup EnvironMed. 1999;41:863–877.

5. Goetzel R, Guindon A, Turshen I,Ozminkowski R. Health and productivitymanagement: establishing key perfor-mance measures, benchmarks, and bestpractices. J Occup Environ Med. 2001;43:10–17.

6. Queroze B. The invisible costs of presen-teeism: a study of health & productivityin the workplace. Health & ProductivityManagement. 2003;2:10–12.

7. Groves D, Dubois R. Productivity and

work loss in irritable bowel syndrome onemployee performance. Presented at theSecond Annual Institute for Health andProductivity Management Conference,‘Putting a Bottom Line under HealthCare’; September 2002; Scottsdale, AZ.

8. Bunn W, Pikelny D, Paralkar S, et al. Theburden of allergies—and the capacity ofmedications to reduce this burden—in aheavy manufacturing environment. J Oc-cup Environ Med. 2003;45:941–955.

9. Employers Health Coalition. The HiddenCompetitive Edge: Employee Health andProductivity. Newton, MA: ManagedCare Communications; 2000.

10. Ginsburg P. Controlling health care costs.N Engl J Med. 2004;351:1591–1593.

11. Landis D, Hubbard D, Cua P, Walker D.Disease management of chronic pain im-proves employee productivity and lowershealth care costs. Health & ProductivityManagement. 2003;2:21–23.

12. CorSolutions. Prevalence, cost and pro-ductivity implications of pain. Availableat: www.CorSolutions.com. 2005: PainSolutions page.

13. Ware J. SF-36 Health Survey: Manualand Interpretation Guide. Boston: TheHealth Institute, New England MedicalCenter; 1993.

14. Cleeland C, Ryan K. Pain assessment.Global use of the brief pain inventory.Ann Acad Med. 1994;23:129–138.

15. Allen H, Bunn W. Validating self-reported measures of productivity atwork: a case for their credibility in aheavy manufacturing setting. J OccupEnviron Med. 2003;45:926–940.

16. Ware J, Kosinski M, Keller S. A 12-itemshort-form health survey. Med Care.1996;34:220–223.

17. Rogers W. A New Scale-Scoring Algo-rithm. Boston: New England MedicalCenter; 2004.

18. Allen H, Darling H, McNeill D, BastienF. The employee health care value sur-vey: Round one. Health Aff (Millwood).1994;13:26–41.

19. Stewart A, Ware J, eds. Measuring Func-tioning and Well-Being: The MedicalOutcomes Study Approach. Durham, NC:Duke University Press; 1992.

20. Rogers W, Wilson I, et al. Assessing theperformance of a new depressionscreener for primary care (PC-SAD).J Clin Epidemiol. 2002;55:164–175.

21. Kessler R, Barber C, et al. The worldhealth organization health and work per-formance questionnaire. J Occup EnvironMed. 2003;45:156–174.

22. Lerner D, Amick B, et al. Work limitationsquestionnaire. Med Care. 2001;39:72–85.

23. Lerner D, Rogers B, Chang H. TechnicalReport: Scoring the Work Limitations

Questionnaire (WLQ) Scales and the WLQIndex for Estimating Work ProductivityLoss. New England Medical Center; 2003.

24. Rogers W, Wittink H, et al. Using the‘TOPS,’ an outcomes instrument for mul-tidisciplinary outpatient pain treatment.Pain Medicine. 2000;1:55–67.

25. Stata Corp. Stata Statistical Software:Release 6.0. College Station, TX: StataCorp; 1999.

26. Lin E, Katon W, et al. Effect of improv-ing depression care on pain and func-tional outcomes among older adults witharthritis: a randomized control trial.JAMA. 2003;290:2428–2434.

27. Allen H. Gaining greater confidence inproductivity self-reports. Health & Pro-ductivity Management. 2004;3:9–12.

28. Serlin RC, Mendoza TR, Nakamura Y,Edwards KR, Cleeland CS. When is can-cer pain mild, moderate or severe? Grad-ing pain severity by its interference withfunction. Pain. 1995;61:277–284.

29. Jensen M, Smith D, Ehde D, Robinsin L.Pain site and the effects of amputationpain: further clarification of the meaningof mild, moderate, and severe pain. Pain.2001;91:317–322.

30. Mendoza T, Chen C, Brugger A, et al.Lessons learned from a multiple-dosepost-operative analgesic trial. Pain. 2004;109:103–109.

31. Zelman D, Hoffman D, Seifeldin R, DukesE. Development of a metric for a day ofmanageable pain control: derivation of painseverity cut-points for low back pain andosteoarthritis. Pain. 2003;106:35–42.

32. Zelman D, Dukes E, Brandenburg N,Bostrom A, Gore M. Identification of cutpoints for mild, moderate and severe paindue to diabetic peripheral neuropathy.Pain. 2005;115:29–36.

33. Karasek R, Gordon G, et al. Job ContentQuestionnaire: Questionnaire and User’sGuide, version 1.1. Los Angeles: Univer-sity of Southern California; 1986.

34. Economic Evaluation Tool. Scottsdale,AZ: Institute for Health & ProductivityManagement; 2004.

35. Allen H, Borden S, Pikelny D, et al. Anintervention to promote appropriate man-agement of allergies in a heavy manufac-turing setting: an evaluation of health andproductivity outcomes. J Occup EnvironMed. 2003;45:956–983.

36. Centers for Disease Control and Preven-tion. Behavioral Risk Factor SurveillanceSystem User’s Guide. Atlanta: US De-partment of Health and Human Services,Centers for Disease Control and Preven-tion; 1998.

37. AISHealth.com. HIPAA Security Compli-ance Guide. Washington, DC: AtlanticInformation Services, Inc; 2005.


APPENDIX 1The Survey: Content, Sources and Rationale

Concept Survey Source Notes/Comments

Health status: past 4 weeks SF-36 Health Status13 Enabled two health status scales—PhysicalHealth & Mental Health—to be formed andnormed via multiple other databases. Itemson bodily pain served as the “4 week”component of the pain screen and enabledstratification of EEs with pain by severity

Physical health items re: upper extremities Treatment Outcomes of Pain24 Added discrimination where SF-36 has beenshown to fall short re: pain patients

Pain status: today Brief Pain Inventory14 Served as the “today” component of the painscreen and will add to capacity to enablestratify by severity

Co-morbidity checklist Medical Outcomes Study19 Yielded count and itemized listing ofco-morbid medication conditions

Screener for major depression disorder anddysthymia

Primary Care Screener forAffective Disorders20

Screened for DSM-IV diagnosis of MDD anddysthymia

Absenteeism (4 wks) Health & Work PerformanceQuestionnaire21

Allowed calculation of EE health-induced timeaway from work in days and hours

Presenteeism (attribute-specific—4 wks) Work LimitationsQuestionnaire22

Allowed 4 job attribute-specific scales—timedemands, physical demands, etc.—to becomputed and normed via other databases.

Presenteeism (overall) ITEC Allergy Survey8 Allowed overall assessment of impairment atwork

Health care utilization(overall and pain specifically)

— Obtained conventional utilization info (neededas no administrative data were available)

Work place injuries ITEC Allergy Survey8 Gave a self-report of injuries and a read oninjury severity (by asking about doctor visits)

Disability support via workers compensation Treatment Outcomes of Pain24 Obtained disability support info (needed as noadministrative data were available)

Medications for pain Healthy People Survey9 Asked only of those who meet criteria for painsample

Alternative treatments/remedies for pain Healthy People Survey9 Asked only of those who meet criteria for painsample

Treatment: outcomes and control over pain Treatment Outcomes of Pain24 Assessed how well the overall managementprogram now in place is working

Job description: demand vs. control Job Content Questionnaire33 Enabled description of EE’s job position interms of EE’s views of demand vs. control

Demographics — Elicited demographic chars known to beassociated with health and productivity

Health risk behaviors Behavioral Risk FactorSurveillance System36

Elicited reads on health risks that areimplicated in the diagnosis of chronic pain


The Burden of Pain on Employee Health and Productivity at...

Documents

Transcript of The Burden of Pain on Employee Health and Productivity at...