transitions in self-reported musculoskeletal pain and interference with activities among newspaper...

P1: GDP/GCY

Journal of Occupational Rehabilitation [jor] pp551-joor-377505 July 13, 2002 19:45 Style file version June 4th, 2002

Journal of Occupational Rehabilitation, Vol. 12, No. 3, September 2002 (C© 2002)

Transitions in Self-Reported Musculoskeletal Painand Interference With Activities AmongNewspaper Workers

Donald C. Cole,1,2,5 Michael Manno,1,2 Dorcas Beaton,1,3,4

and Michael Swift1

Active surveillance of symptoms and disability due to musculoskeletal disorders (MSD)in working populations can map individual transitions in symptom intensity or disabilitylevel. Using repeat surveys, this study examined ifaetiologicalrisk factors for new symp-toms or disability, measured by interference with activities, were similar toprognosticriskfactors for subsequent outcomes of symptoms and disability. This paper reports on 379Toronto newspaper workers who completed questionnaires in 1996 and 1997. Questionson pain/discomfort during the last year, episode frequency and duration formed the basisfor constructing three mutually exclusive symptom levels: noncases (Level 1); mild cases(Level 2); and more severe or frequent pain cases (Level 3). A similar construction of overallinterference levels was based on the frequency with which musculoskeletal pain/discomfortinterfered with daily, recreational, social and family activities, or ability to do one’s jobduring the previous 12 months. The last was also examined as work interference alone.Levels of symptoms were cross-tabulated by overall and work interference at the two timepoints and Markov models of transitions between states were formulated. Results indicatethat period prevalence of symptoms and overall interference increased between phases,though only significantly for symptoms (Levels 2 and 3, 65.7–70.7%, p= 0.04), whilework interference was unchanged (17.9–17.0%). Equivalent proportions of workers im-proved as worsened in symptoms (21.1 and 22.4%, respectively), overall interference (16.7and 17.8%), and work interference (7.4 and 6.6%). The only significant predictor for thosewithout work interference at time one was job tenure, which was protective against workinterference (Odds Ratio (OR) 1.06, 95% Confidence Interval (CI) 1.01–1.12). Among thosewho had more severe/frequent symptoms at time one, lack of improvement over time waspredicted by longer job tenure (OR= 0.92per year, CI: 0.87–0.97), greater psychologicaldemands of work (OR= 0.65 per point, CI: 0.51–0.84) and marginally by greater upperextremity disability score. At time one, women with work interference were more likely to

1Institute for Work and Health, Toronto, Canada.2Department of Public Health Sciences, University of Toronto, Toronto, Canada.3St. Michael’s Hospital, Toronto, Canada.4Department of Occupational Therapy, Graduate Department of Rehabilitation Sciences, University of Toronto,Toronto, Canada.

5Correspondence should be directed to Donald C. Cole, MD, MSc, FRCP(C), Institute for Work and Health,481 University Avenue, Suite 800, Toronto, Ontario, Canada M5G 2E9; e-mail: [email protected].

163

1053-0487/02/0900-0163/0C© 2002 Plenum Publishing Corporation

P1: GDP/GCY


164 Cole, Manno, Beaton, and Swift

have persistent interference at time two (OR= 7.22, CI: 1.57–33.20). Suggestive findingsincluded opposite effects of social support at work: reduced risk of development of newsymptoms but increased risk of persistence at the highest symptom level.

KEY WORDS: longitudinal study; arm pain; repetitive strain injury/cumulative trauma disorders; disability;social support; psychological demands; questionnaire.

INTRODUCTION

Repeated health surveys are an important tool to examine change in risk factors, symp-toms, chronic conditions, or disability in populations. Both changes in overall prevalencee.g., physical activity in the U.S. Established Populations for Epidemiologic Studies ofthe Elderly (1) or arthritis in the Canadian National Population Health Survey (2), andexamination of individual changes, such as those in physical function (3), are important.Repeat cross-sectional data on musculoskeletal disorders (MSD) among working popula-tions has been generated during workplace surveillance programs or explicit longitudinalstudies (4).

Although whether the working population is doing “worse” or better as a whole(changes in prevalence) may be important, the argument can be made that determiningwhat happens to individuals between surveys (transitions) is also crucial. Considerablefluctuations in severity over time are likely, as mapped out by survey participants with lowback pain (5). Fluctuations in severity can be thought of as transitions between states ofsymptoms or disability. Shifting from a state with low intensity or frequency of symptoms(not meeting a surveillance case definition) into a state with enough symptoms to be desig-nated as a case is one type of transition. Similarly movement between states with differentintensities of disability could be considered transitions. Explicit description of the differentkinds of stability or movement between states of musculoskeletal function among workershas been most clearly delineated by Kaergaard and Andersen (6). To explore the factorsassociated with particular transitions (or lack thereof ), the population has usually beencategorized into groups (e.g., case, noncase) and logistic regression modeling carried outto assess the role of baseline predictors for membership in each group at subsequent timeperiods (7–9). Although methods for approaching transitions with multicategorical datahave been known for some time (10), they have only occasionally been applied to workforcesurveillance data (11,12).

In this study there was the opportunity to examine transitions in both symptoms andinterference with activities using data from two cross-sectional workforce surveys, con-ducted 1 year apart. The study’s research objectives were to model the transitions usinga multicategorical approach and to compare the predictors of transitions across differentlevels of symptoms and disability.

METHODS

Setting and Population

The research took place as part of a collaborative project with the management andunion of a large metropolitan newspaper. Of the 1007 (84% response rate) participants

P1: GDP/GCY


Transitions in Pain and Disability 165

in Phase 1 a cross-sectional survey (13), 548 gave consent to be recontacted. Of these,379 (69%) answered a follow-up questionnaire during Phase 2, approximately 1 year later.Participants in both phases were more commonly women (55% of this sample vs. 38% ofthose not participating), better educated (56% vs. 45% had completed university), and full-time workers (81% vs. 73%) (14). Identical percentages were married or in common lawrelationships and were the main wage earner in their household (59% for both across bothphases). The vast majority had similar office jobs with similar equipment for both phases. Asreported previously (14), 75% of those reporting pain or discomfort in the neck and upperextremity in Phase 2 perceived that their pain/discomfort was substantially or partiallycaused by work, while 85% reported that work aggravated their problem. Participantsreported a wide array of interventions for MSD between Phase 1 and Phase 2; these arereported in detail elsewhere (14).

Description of Transitions

All measures were based on questionnaires completed by participants at work, as setout previously (13–15). Here the focus is on the key health outcomes for categorization oflevels and the selection of potential predictors included in the transitions analysis. In bothPhase 1 and Phase 2 surveys, participants were asked if they had any pain or discomfortin the neck and upper extremity during the last year and if so, how frequently they hadpain/discomfort episodes and the duration of their symptoms for each of these episodes.These questions were asked using a modification of a previously published format (16).Participants were also asked to rate their pain/discomfort at any time in the past 7 dayson a 5-point scale ranging from no discomfort to unbearable pain. Their answers to thesequestions permitted construction of two case definitions: definition A required that theworker had pain episodes more frequently than three times in the past year or a pain episodelasting more than 1 week; and definition B required that the worker had pain episodes 12 ormore times in the past year or a pain episode lasting more than 1 week, and that they hadpain over the last 7 days which was of moderate or worse severity (15). Based on thesecase definitions, applied to the most severely affected body region, participants were placedin one of three mutually exclusive symptom levels: Level 1, noncases; Level 2, mild caseswhich met definition A but not B; and Level 3, cases with more severe or frequent painwhich met definition B.

Using questions from the Nordic questionnaire (17), participants also reported on thefrequency with which musculoskeletal pain interfered with their daily activities or withtheir recreational, social, and family activities during the last 12 months (from 1= never,2–5= sometimesto always). Further, participants reported whether their musculoskeletalproblem had interfered with their ability to do their job (1= no, 2= yes). Based on thesethree questions, overall interference levels were constructed as follows: none or Level 1, 1on all three questions; some or Level 2, 2 on one or two questions; and more or Level 3, 2or more on all three questions.

Cross-tabulations were carried out for each of the symptom levels and the interfer-ence levels resulting in two three-by-three tables (symptom, overall interference) and onetwo-by-two table (work interference). Phase 1 categories defined the rows and Phase 2categories defined the columns. Differences in row and column distributions were testedusing McNemar’s test.

P1: GDP/GCY



Prediction of Transitions

Variable Selection

Although a broad range of potential predictive variables were available, due to powerconsiderations, the selection of predictor variables was limited, due to the relatively smallnumbers in most Phase 1 levels. Even with consolidation of Phase 1 overall interferenceLevels 2 and 3, the minimum was 119 (symptom Level 2 as per Table II (a)), likely to bereduced further due to missing values for variables. To avoid reduced power and overfitting(18,19), 10 variables were selected as potential predictors. As there was the desire for somecommon potential predictors across the different transitions to be modelled, the followingapproach was used to arrive at a reduced list of potential predictors variables: consultingthe relevant literature, examining distributions of available variables in the current dataset, conducting preliminary univariate logistic regressions and applying categorical dataanalysis to groups of potential predictors within domains.

Both the etiologic and prognostic literatures on MSD were consulted, since the tran-sitions of interest encompassed the development of new cases, as well as the worsen-ing or improvement of existing cases. The identified potential predictors were broadlygrouped into three domains: demographic, work, and MSD. Building on the extensive re-view done by U.S. National Institute of Occupational Safety and Health colleagues (20)and other research conducted in office settings, the aim was to include both physical andpsychosocial/work organization factors relevant to etiology. Factors identified in earlierreviews of prognostic factors (21) and more recent prognostic cohorts (22–25) were alsoincluded.

Examinations of distributions assisted in variable reduction. For example, although theliterature suggested that the number of body sites affected may be an important prognosticindicator for those with MSD, the overlap was too large with symptom level (i.e., more bodysites affected, higher symptom level) for this variable to be independently useful. OtherPhase 1 variables, such as Body Mass Index (weight in kg/height in m2) and a decisionauthority score on a reduced version of the Job Content Questionnaire (26), showed littlepredictive power (p > 0.25) on univariate logistic regressions of Phase 2 symptom-leveloutcome.

Selected Potential Predictors

The final set of potential predictor variables is set out in Table I. Gender was the onlydemographic predictor included. In the work domain, five variables were selected. Thesuboptimal positioning variable was based on diagrams indicating the participant’s usualplacement of workstation components and whether they fell outside a box representing bestpractice judgments of human factors’ professionals (26). The own workstation variablelikely represents a workstation more adapted to the participant’s dimensions and work styleat a time when adjustable components of workstations were uncommon at the newspaper.Psychological demands and social support variables were drawn from a reduced form ofthe Job Content Questionnaire (JCQ) used in the Canadian National Population HealthSurvey (27). Although the larger JCQ has undergone extensive validity testing (28), inthe NPHS version the assessment of “psychological demands” was limited to two items,

P1: GDP/GCY



Table I. Descriptors of Predictors of Transitions (n = 379)

Variable n missing Category n (%)

(a)CategoricalDemographic

Gender 0 Female 208 (54.9)Work factors

Suboptimal keyboard/monitor position 14 Yes 172 (47.1)Use own workstation 4 Yes 287 (76.5)

Mean (SD)

(b) ContinuousWork factors

Tenure in newspaper industry (years) 3 16.87 (9.03)Psychological demands on a 0 (Least)–10 (Most) Scale 3 6.88 (1.90)Social support on a 0 (Least)–10 (Most) Scale 7 6.40 (1.81)

WMSD factorsModified Pain Intensity Scale Scorea 3 34.12 (27.13)QuickDASH Score on a 0 (No Difficulty)–100 (Unable) 9 13.07 (14.70)

Scale

a0= No Discomfort–100= Unbearable.

asking about a hectic job and freedom from conflicting demands of others (reverse scored).Social support was tapped by three items, exposure to hostility or conflict from coworkers(negatively scored), supervisor help in getting the job done and coworkers’ help in gettingthe job done. As respondents answers were based on a 5-point scale (0= strongly agree;4= strongly disagree), scores were created to convert each into 10-point scales, with highernumbers indicating more of each construct.

Among participants who already had some or more symptoms/interference at Phase 1,two MSD-related factors were selected as potential predictors. Intensity of symptoms (pain/discomfort in the neck and upper extremities) was based on three questions (on averageover the past year, on average over the past week, and at its worst ever) adapted fromthe pain intensity scale of the Von Korff chronic pain grade (29,30). The modified painintensity scale score was obtained by summing the responses to the three items (5-pointscale) and converting the sum to a 0–100 scale (31). Difficulty carrying out daily tasks wasmeasured using theQuickDASH, a shortened (11-item) version of the DASH (Disabilitiesof the Arm Shoulder and Hand) outcome measure. Evidence is available on the validityand reliability of the full DASH (32,33). The 11 items selected had the highest item totalcorrelation in each of the domains covered by the DASH. Item responses (5-point scale)were summed and converted to a 0–100 scale with higher values reflecting greater disability.

Transition Modeling

Because the aim was to identify factors associated with getting better or getting worse,it was felt that a transition model would be most appropriate. To formulate a transitionmodel, assume the case level for subjecti at time j is given byYi j for i = 1, . . . ,n andj = 1, 2. The model describing transitions from initial case levelYi 1 = a toYi 2 is carried outby modeling the cumulative probabilityP(Yi 2 ≤ b | Yi 1 = a) for cutoffsb = 1, 2, through

P1: GDP/GCY



the logistic model

logP(Yi 2 ≤ b | Yi 1 = a, xi )

P(Yi 2 > b | Yi 1 = a, xi )= θab+ βa1xi 1+ · · · + βapxip, a = 1, 2, 3, b = 1, 2,

whereθab is the intercept term,xi = (xi 1, . . . , xip) are covariates for subjecti , andβa1, . . . ,

βap are regression coefficients. Thus, a separate ordered logistic model was fit for each ofthe three possible values ofYi 1. Notice that the intercept is allowed to vary for each value ofthe cutoffb, but that the covariate effects are constrained to be the same across all cutoffs. Theabove formulation is often called the proportional odds model. The proportional odds modelis a logical choice as there is a natural ordering to the outcome (1= noncase, 2= mild case,3= severe case). Suppressing subscripti to simplify the notation, the regression parametersstill have log odds ratio interpretations since for any covariatexj

βaj = logP(Y2 ≤ b | Y1 = a, xj + 1)/P(Y2 > b | Y1 = a, xj + 1)

P(Y2 ≤ b | Y1 = a, xj )/P(Y2 > b | Y1 = a, xj ), j = 1, . . . , p.

The model used here is an example of a first order Markov model. The key feature forsuch a model is that the probability of being in a future state depends on the current stateonly. Higher order models assume that transitions can depend on states farther in the past.Because data was only available for two time points, by default, the model is first-orderMarkov. Implementation is straightforward using any statistical software that can fit orderedpolytomous logistic models such as SAS. Models were fit using the LOGISTIC procedurein SAS (34).

The choice was made to model the outcome so that positive regression coefficientsare associated with staying at a lower level or improvement by moving to a lower level.Negative regression coefficients are associated with staying at a higher level or gettingworse by moving to a higher level. For overall interference transitions, small cell sizes inthe 1, 3 and 3, 1 cross-tabulations created difficulties in achieving stable estimates. Hence,overall interference: Levels 2 and 3 were consolidated for modeling purposes. For workinterference transitions, only two levels were available.

RESULTS

Description of Transitions

Changes in period prevalence varied by outcome: symptom Levels 2 and 3, 65.7–70.7% (p = 0.04); overall interference Levels 2 and 3, 50.8–53.1% (p = 0.38); and workinterference Level 2, 17.9–17.0% (p = 0.67). Individually, the majority of participantsremained at the same symptom (56.5%), overall interference (65.4%), and work interference(86.0%) levels across the two phases (diagonals in Table II (a)–(c)). As highlighted by theshading in the tables, similar proportions improved (light grey) as worsened (dark grey) inboth symptoms (21.1 and 22.4%, respectively), overall interference (16.7 and 17.8%), andwork interference (7.4 and 6.6%) between phases. For symptoms and overall interference,one-level transitions were more common than two-level transitions both in development of“new” cases e.g., incident mild symptoms (Level 2 from Level 1) was 9.8% versus incident

P1: GDP/GCY



Table II. Transitions Between Phase 1 and Phase 2 (n, cell %) (n = 379)

To: Phase 2 symptom level

Level 1 Level 2 Level 3 Total

From: Phase 1 symptom level(a)Symptom level transitions(n missing= 0)

Noncase (Level 1) Stable Incident Incident77 (20.3%) 37 (9.8%) 16 (4.2%) 130 (34.3%)

Mild (Level 2) Resolved Ongoing Worsened23 (6.1%) 64 (16.9%) 32 (8.4%) 119 (31.4% )

More severe or frequent (Level 3) Resolved Improved Chronic11 (2.9%) 46 (12.1%) 73 (19.3%) 130 (34.3%)

Total 111 (29.3%) 147 (38.8%) 121 (31.9%) 379 (100%)

To: Phase 2 overall interference level

From: Phase 1 overall interference level(b) Overall interference level transitions(n missing= 27)

None (Level 1) Stable Incident Incident128 (36.4%) 41 (11.6%) 4 (1.1%) 173 (49.2%)

Some (Level 2) Resolved Ongoing Worsened30 (8.5%) 71 (20.2%) 18 (5.1%) 119 (33.8%)

More (Level 3) Resolved Improved Chronic7 (2.0%) 22 (6.2%) 31 (8.8%) 60 (17.0%)

Total 165 (46.9%) 134 (38.1%) 53 (15.1%) 352 (100%)

To: Phase 2 work interference level

From: Phase 1 work interference level(c) Work interference level transitions(n missing= 15)

None (Level 1) Stable Incident275 (75.6%) 24 (6.6%) 299 (82.1%)

Some (Level 2) Resolved Chronic27 (7.4%) 38 (10.4%) 65 (17.9%)

Total 302 (83.0%) 62 (17.0%) 364 (100%)

severe/frequent symptoms (Level 3 from Level 1), 4.2%, and improvement of “old” casese.g., interference improvement (Level 3 to Level 2), 6.2% versus interference resolution(Level 3 to Level 1), 2.0%.

Prediction of Transitions

Symptoms

The results for the three symptom level transition models are shown in separate columnsof Table III. The first model (first column), describes transitions for individuals who werenoncases (a = 1) at Phase 1. Because it identifies factors associated with onset of symptoms,it can be thought of as an “etiology” model. Tenure at work was the only predictor thatachieved even marginal significance (Odds Ratio, OR= 1.04 per year, 95% ConfidenceInterval, CI: 1.00–1.09). Nevertheless, it suggested a survivor effect associated with reducedrisk of becoming a case.

The second model (second column) describes transitions among those who had mildsymptoms (a = 2) at Phase 1 and the third model (third column) describes predictors of

P1: GDP/GCY



Table III. Multivariable Predictors of Symptom Level Transitionsa

Baseline symptom level

Noncase (Level 1) Mild (Level 2) Severe (Level 3)(n = 119)b (n = 115)b (n = 122)b

Potential predictors OR (95% CI) OR (95% CI) OR (95% CI)

DemographicFemale gender 1.69 (0.79–3.62) 1.49 (0.65–3.40) 0.90 (0.38–2.10)

Work factorsTenure (per year) 1.04 (1.00–1.09) 1.02 (0.98–1.06) 0.92 (0.88–0.97)Suboptimal workstation 0.95 (0.45–2.02) — —Own workstation 1.20 (0.51–2.84) 1.73 (0.60–5.00) 1.70 (0.69–4.22)Psychological demands (per point) 1.01 (0.82–1.25) 1.00 (0.82–1.23) 0.68 (0.53–0.86)Social support (per point) 1.21 (0.94–1.55) 1.06 (0.84–1.33) 0.87 (0.71–1.07)

WMSD factorsPain (per point) NA 0.99 (0.96–1.02) 0.96 (0.92–1.00)QuickDASH (per point) NA 0.98 (0.94–1.03) 0.97 (0.93–1.00)

Model fit statistics% concordant 64.6 60.6 76.6% discordant 35.0 38.3 22.1% tied 0.4 1.1 1.3

aDependent variables are transitions from each column or level to another column or level. Odds ratios greaterthan one signify a greater probability of staying in the same column for those in column 1, i.e., staying the same,or moving into columns to the left for those in the second and third columns, i.e., getting better. Conversely oddsratios less than one signify moving right for those in the first two columns, i.e., getting worse, or staying in thethird column, i.e., staying the same.

bBecause of missing values for the outcome and explanatory variables,ns for each model were less than all ateach level in Phase 1.

lack of improvement between years. In the latter, job tenure (OR= 0.92 per year, 95%CI: 0.88–0.97) and psychological demands of work (OR= 0.68 per point, 95% CI: 0.53–0.86) were both significant. Participants’ pain intensity score (OR= 0.96 per point, 95%CI: 0.92–1.00) andQuickDASH score were marginal (OR= 0.97 per point, 95% CI: 0.93–1.00 ) indicating that increased pain and disability were associated with marginally poorerprognosis.

Interference

For overall interference level transitions there were no statistically significant asso-ciations. However, pain intensity was marginal (OR= 0.97, 95% CI: 0.94–1.00) withinthe Level 2 and 3 group. For work interference level transitions, job tenure was protectiveagainst work interference for those without it at Phase 1 (OR= 1.06, 95% CI: 1.01–1.12)and female gender was associated with dramatic increases in likelihood of persistence ofsymptoms (OR= 7.22, 95% CI: 1.57–33.20) among those with interference at Phase 1 (seeTable IV).

Trends in Both

Tracking particular variables across all models, one can notice some trends thatmight be significant in larger samples. Odds ratios for having one’s own workstation,

P1: GDP/GCY



Table IV. Multivariable Predictors of Interference Level Transitions

Baseline overall interference levela Baseline work interference levela

None Some/more None Some(Level 1) (Levels 2 and 3) (Level 1) (Level 2)

(n = 159)b (n = 172)b (n = 278)b (n = 63)b

Potential predictors OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)

DemographicFemale gender 1.08 (0.51–2.28) 0.71 (0.31–1.65) 1.27 (0.54–3.00) 7.22 (1.57–33.20)

Work factorsTenure (per year) 1.02 (0.98–1.07) 1.00 (0.96–1.05) 1.06 (1.01–1.12) 1.00 (0.92–1.08)Suboptimal workstation 0.76 (0.36–1.60) — 0.92 (0.39–2.17) —Own workstation 1.10 (0.44–2.76) 1.24 (0.48–3.23) 1.45 (0.55–3.87) 0.29 (0.07–1.16)Psychological demands 0.98 (0.79–1.22) 0.96 (0.76–1.21) 0.90 (0.69–1.16) 0.87 (0.61–1.24)Social support 1.20 (0.95–1.51) 0.89 (0.71–1.11) 0.91 (0.71–1.17) 0.88 (0.60–1.30)

WMSD factorsPain NA 0.97 (0.94–1.00) NA 0.97 (0.93–1.02)QuickDASH NA 0.98 (0.94–1.02) NA 0.96 (0.92–1.01)

Model fit statistics% concordant 61.9 69.5 65.6 79.1% discordant 37.8 30.0 33.0 20.6% tied 0.3 0.5 1.4 0.3

aDependent variables are transitions from each column or level to another column or level. Odds ratios greaterthan one signify a greater probability of staying in the same column for those in column 1, i.e., staying the same,or moving into columns to the left for those in the second and third columns, i.e., getting better. Conversely oddsratios less than one signify moving right for those in the first two columns, i.e., getting worse, or staying in thethird column, i.e., staying the same.

bBecause of missing values for the outcome and explanatory variables,n’s for each model were less than all ateach level in Phase 1.

perhaps permitting greater adjustment to one’s needs, were consistently greater than one.In contrast, social support at work seemed to be protective in symptom models 1 and 2and interference model 1 (odds ratios all>1) but associated with worse prognosis insymptom model 3 and interference model 2 and 3 (odds ratio<1), indicating two dif-ferent functions of social support. Similarly, female gender had odds ratios>1 in themore etiologic models (#’s 1), declining some for the mixed model (symptoms 2) andbecoming negative in two of the prognostic models (symptoms 3 and overall interfer-ence 2 and 3). Interestingly it became very positive in the work interference prognosticmodel (i.e., women were less likely to remain with work interference from Phase 1 toPhase 2).

DISCUSSION

The finding that period prevalence of symptoms significantly increased while periodprevalence of work interference remained the same is new. Most workplace longitudinalstudies focus primary case definitions on symptoms with or without physical examinationand have paid less attention to shifts in disability. This finding may indicate either greaterwillingness among those affected to work with their MSD or greater efforts towards im-proved workplace accommodations between phases. If the latter, then this is good newsfor occupational rehabilitation providers who aim to maintain ability to carry out one’s jobeven in the face of persistent symptoms.

P1: GDP/GCY



Transitions between levels of symptoms within the workforce were considerable, evenover a 1-year period. The extent of improvement or worsening overall (about 20%) wassimilar to that observed in symptom-based case versus noncase transitions between aninitial survey and more in-depth examination about 6 months apart in another newspaperstudy (35). It was greater than that reported for transitions between symptom duration-based levels among postal workers over 3-month intervals (December–March—worsening9.8%, 11.4% improvement; March–May—worsening 4.5%, improvement 17.6%) (4) andbetween yearly surveys of symptom-based cases among nursing personnel of between10 and 14% shifting to cases or to noncases (36). It was also greater than that amongfemale sewing machine operators (about 5% each way between year 1 and year 2), wherephysical examination findings were required in case definition (6). Although the workingpopulations, definitions of levels and time periods between repeat surveys vary across thesestudies, each indicates the fluctuating character of severity and the importance of trackingindividual level changes over time. The demonstration of interference transitions indicatesthat disability measures should also be added to such surveys and similarly tracked.

Categorizing the population into different levels permitted examination of predictorsof different kinds of transitions over time. Despite creating integrated transition models inwhich improvement and worsening could be simultaneously integrated into odds ratios foreach predictor, the power to detect predictors was limited. Investigators of transitions inpain among a cohort of Finnish forestry workers (8) experienced similar challenges. In fact,for etiological predictors, there was barely the ability to detect a survivor effect on symp-tom levels (i.e., those that had been at the newspaper longest were least likely to developnew symptoms), although one was detected in the larger sample of those without workinterference. Suggestive positive effects of social support reducing the likelihood of be-coming a symptom case, similar to conclusions in recent systematic reviews on the role ofpsychosocial factors in musculoskeletal pain, were also identified (37).

Greater success was achieved with the use of the prognostic predictors. Althoughgreater work demands has been associated with worse musculoskeletal pain (38) and greaterlikelihood of switching jobs (21), this study is one of the first demonstrations of such effectson symptom level in a segment of the workforce already with significant symptoms. Thefinding of the marginal negative effect of greater existing disability (QuickDash scores) isconsistent with the finding among forestry workers that the best predictor of future painwas current disability (8). The finding for the strong effect of female gender on resolvingwork restrictions is different from some work on prognosis for those with compensablesoft tissue injuries, where women take longer to return to work, perhaps indicating greaterchronicity of disability (39).

Some caveats on findings are in order. As in many follow-up studies, those who par-ticipated in both phases differed significantly from the overall population, more commonlyreporting neck and upper extremity pain/discomfort (71% on Phase 1 among Phase 2 par-ticipants versus 53% among nonparticipants in Phase 2) (14). Such differences precludeeasy extrapolation of the proportions changing case level to the workforce as a whole. Thelimited sample size restricted the range of predictors that could be used, because of thelimited proportions of the workforce that changed levels of symptoms or interference be-tween surveys. Having more years of data through repeat cross-sectional surveys on largerpopulations would permit more precise confidence limits around existing predictors as inthe larger forestry workers cohort (11). Multiple time points would also permit use of higher

P1: GDP/GCY



order Markov transition models that incorporate states farther in the past, more in keepingwith the fluctuating nature of MSD (40).

Future researchers are advised to repeat measurement of both symptoms andinterference/disability over time along with their potential predictors in work cohorts amonglarger samples. Such data would permit more nuanced understanding of transitions in levelsof symptoms and disability and their predictors. Further, it would gradually move the focusaway from traumatic or acute framings of MSD at work and inform the understanding of theprocesses of movement between states of ability and disability as part of enabling–disablingprocesses, so essential to policy-making (41).

ACKNOWLEDGMENTS

The authors acknowledge other members of the Worksite Upper Extremity Group at theInstitute for Work and Health, members of the RSI (Repetitive Strain Injury) Watch SteeringCommittee, and participants in the study. This project was sponsored by the Institute forWork and Health. The Institute, an independent not-for-profit research organization, receivessupport from the Ontario Workplace Safety and Insurance Board.

REFERENCES

1. Ferucci L, Izmirlian G, Leveille S, Phillips CL, Corti MC, Brock DB, Guralnik JM. Smoking, physical activityand active life expectancy.Am J Epidemiol1999; 149: 645–653.

2. Federal, Provincial and Territorial Advisory Committee on Population Health.Toward a healthy future.SecondReport on the Health of Canadians. Cat H39-468/1999E. Ottawa, ON: Health Canada and Minister of PublicWorks and Government Services, 1999.

3. Beckett LA, Brock DB, Lemke JH, Mendes de Leon CF, Guralnik JM, Fillenbaum GG, Branch LG, WetleTT, Evans DA. Analysis of change in self-reported physical function among older persons in four populationstudies.Am J Epidemiol1996; 143: 766–778.

4. Takala E-P, Viikari-Juntura E, Moneta GB, Saarenmas K, Kaivanto K. Seasonal variation in neck and shouldersymptoms.Scand J Work Environ Health1992; 18: 257–261.

5. McGorry RW, Webster BS, Snook SH, Hsiang SM. The relations between pain intensity, disability and theepisodic nature of chronic and recurrent low back pain.Spine2000; 25: 834–841.

6. Kaergaard A, Andersen JH. Musculoskeletal disorders of the neck and shoulders in female sewing machineoperators: Prevalence, incidence and prognosis.Occup Environ Med2000; 57: 528–534.

7. Jonsson BG, Persson J, Kilbom A. Disorders of the cervicobrachial region among female workers in theelectronics industry. A two-year follow up.Int J Ind Ergon1988; 3: 1–12.

8. Estlander A-M, Takala E-P, Viikari-Juntura E. Do psychological factors predict changes in musculoskeletalpain? A prospective, two-year follow-up of a working population.J Occup Environ Med1998; 40: 440–453.

9. Tousignant M, Rossignol M, Goulet L, Dassa C. Occupational disability related to back pain: Application ofa theoretical model of work disability using prospective cohorts of manual workers.Am J Ind Med2000; 37:410–422.

10. Bishop YMM, Fienberg SE, Holland PW.Discrete multivariate analysis: Theory and practice.Cambridge,MA: MIT press, 1975.

11. Luukkonen R, Martikainen R, Mutanen P, Viikari-Juntura E, Riihim¨aki H, Takala E-P. Analysis of longitudinalquestionnaire data using three different statistical methods. In:Proceedings of the 13th Triennial Congressof the International Ergonomics Association, Vol. 4: Musculoskeletal disorders and rehabilitation, Tampere,Finland, 1997, pp. 78–80.

12. Lusa S, Luukkonen R, Viikari-Juntura E, Punakallio A. Factors predicting musculoskeletal pain in Finnishfire-fighters: A three-year follow up study. Abstract atPREMUS, 4th International Scientific Conference onprevention of work-related musculoskeletal disorders, Amsterdam, The Netherlands, 2001, p. 223.

13. Polanyi MFD, Cole DC, Beaton DE, Chung J, Wells R, Abdolell M, Beech-Hawley L, Ferrier SE, MondlochMV, Shields SA, Smith JM, Shannon HS. Upper limb work-related musculoskeletal disorders among news-paper employees: Cross-sectional survey results.Am J Ind Med1997; 32: 620–628.

P1: GDP/GCY



14. Swift M, Cole DC, Beaton DE, Manno M. Health care utilization and workplace interventions for neck andupper limb problems among newspaper workers.J Occup Environ Med2001; 43: 265–275.

15. Beaton DE, Cole DC, Manno M, Bombardier C, Hogg-Johnson S, Shannon H. Estimating the burden ofwork-related musculoskeletal disorders in newspaper workers: What difference do case definitions make?J Occup Rehabil2000; 10: 39–53.

16. Hagberg M, Silverstein B, Wells R, Smith MJ, Hendrick HW, Carayon P. Perusse Health and risk factor surveil-lance for work-related musculoskeletal disorders. In: Kuorinka I, Forcier L, eds.Work-related musculoskeletaldisorders (WMSDs): A reference book for prevention.London: Taylor & Francis, 1995, pp. 213–243.

17. Kuorinka I, Jonsson B, Kilbom A, Vinterberg H, Biering-Sorensen F, Andersson G, Jørgenson K. StandardisedNordic questionnaires for the analysis of musculoskeletal symptoms.Appl Ergo1987; 18: 233–237.

18. Norman GR, Streiner DL.PDQ statistics. Toronto, ON: B.C. Decker, 1986, p. 172.19. Copas JB. Regression, prediction and shrinkage (with discussion).J R Stat Soc Ser B1983; 45: 311–354.20. Bernard BP (ed.).Musculoskeletal disorders and workplace factors. A critical review of epidemiologic evi-

dence for work-related musculoskeletal disorders of the neck, upper extremity and low back.DHHS (NIOSH)Publication No. 97–141. Cincinnati, OH: National Institute for Occupational Safety and Health, 1997.

21. Cole DC, Hudak PL. Prognosis of non-specific work-related musculoskeletal disorders of the neck and upperextremity.Am J Ind Med1996; 29: 657–668.

22. Katz JN, Keller RB, Fossel AH, Punnett L, Bessette L, Simmons BP, Mooney N. Predictors of return to workfollowing carpal tunnel release.Am J Ind Med1997; 31: 85–91.

23. Katz JN, Lew RA, Bessette L, Punnett L, Fossel AH, Mooney N, Keller RB. Prevalence and predictors oflong-term work disability due to carpal tunnel syndrome.Am J Ind Med1998; 33: 543–550.

24. Brox JI, Staff PH, Ljunggren AE, Brevik JI. Arthroscopic surgery compared with supervised exercises inpatients with rotator cuff disease (stage II impingement syndrome).Br Med J1993; 307: 899–903.

25. Feuerstein M, Huang GD, Haufler AJ, Miller JK. Development of a screen for predicting clinical outcomesin patients with work-related upper extremity disorders.J Occup Environ Med2000; 42: 749–761.

26. Karlqvist L, Hagberg M, Wenemark M,Anell R, Koster M. Musculoskeletal symptoms among computer-assisted design (CAD) operators and evaluation of a self-assessment questionnaire.Int J Occup EnvironHealth1996; 2: 185–194.

27. Wilkins K, Beaudet MP. Work stress and health.Health Rep1998; 10: 47–62.28. Karasek R, Brisson C, Kawakami N, Houtman I, Bongers P, Amick B. The Job Content Questionnaire (JCQ):

An instrument for internationally comparative assessments of psychosocial job characteristics.J Occup HealthPsychol1998; 3: 322–355.

29. Von Korff M, Dworkin SF, Le Resche L. Graded chronic pain status: An epidemiologic evaluation.Pain1990;40: 279–291.

30. Von Korff M, Ormel J, Keefe FJ, Dworkin SF. Grading the severity of chronic pain.Pain1992; 50: 133–149.31. Smith JM, Cole DC, Beaton DE, Manno M, and the Worksite Upper Extremity Research Group.Compar-

ing severity indices for upper extremity work-related musculoskeletal disorders across multiple indicators.Institute for Work and Health Working Paper #110. Toronto, ON: Institute for Work and Health, 2001.

32. Hudak PE, Amadio PC, Bombardier C, and the Upper Extremity Collaborative Group. Development of anupper extremity outcome measure: The DASH (Disabilities of the Arm, Shoulder and Hand).Am J Ind Med1996; 29: 602–608.

33. McConnell S, Beaton DE, Bombardier C.The DASH outcome measure: A user’s manual, 1st edn., Toronto,ON: Institute for Work and Health, 1999.

34. SAS Institute Inc.SAS/STAT users’ guide,Version 6, Vol. 12, 4th edn. Cary, NC: SAS Institute Inc., 1989.35. National Institute for Occupational Safety and Health (NIOSH).NIOSH health hazard evaluation report.

NIOSH Report No. HETA 89-250-2046. Newsday, Inc., Melville, New York. Cincinnati, OH: U.S. Departmentof Health and Human Services, Public Health Service, Centers for Disease Control, National Institute forOccupational Safety and Health.

36. Josephson M, Lagerstr¨om M, Hagberg M, Hjelm EW. Musculoskeletal symptoms and job strain amongnursing personnel: A study over a three year period.Occup Environ Med1997; 54: 681–685.

37. Hoogendoorn WE, van Poppel MNM, Bongers PM, Koes BW, Bouter LM. Systematic review of psychosocialfactors at work and private life as risk factors for back pain.Spine2000; 25: 2114–2125.

38. Cole DC, Ibrahim SA, Shannon HS, Scott F, Eyles J. Work correlates of back problems and activity restrictiondue to musculoskeletal disorders in the Canadian National Population Health Survey (NPHS) 1994/95 Data.Occup Environ Med2001; 58: 728–734.

39. Clarke J, Chung J, Cole DC, Hogg-Johnson S, Haidar and the ECC Prognosis Working Group.Gender andbenefit duration in lost time work-related soft tissue disorders: Relationships with work and social factors.Institute for Work and Health Working Paper #85. Toronto, ON: Institute for Work and Health, 1999.

40. Hagberg M, Christiani D, Courtney TK, Halperin W, Leamon TB, Smith TJ. Conceptual and definitionalissues in occupational injury epidemiology.Am J Ind Med1999; 32: 106–115.

41. Brandt EJ, Pope AM (eds.).Enabling America: Assessing the role of rehabilitation science and engineering.Division of Health Sciences Policy, Institute of Medicine. Washington, DC: National Academy Press, 1997.

transitions in self-reported musculoskeletal pain and interference with activities among newspaper...

Documents