Rates of claims for cumulative trauma disorder of the upper extremity in Ontario workers during 1997

Vol. 25 No. 1, 2004

Dianne Zakaria, James Robertson, John Koval, Joy MacDermid and Kathleen Hartford

Abstract

Surveillance of work-related cumulative trauma disorder of the upper extremity (CTDUE) requires valid and reliable claim extraction strategies and should examine for confounding and interaction. This research estimated crude and specific rates of CTDUE claims in Ontario workers during 1997 while acknowledging misclassification and testing for confounding and interaction. Lower and upper limit event estimates were obtained by means of an algorithm applied to the Ontario Workplace Safety and Insurance Board (OWSIB) database and were combined with "at-risk" estimates obtained from the Canadian Labour Force Survey (LFS). Poisson regression was used to evaluate confounding and interaction. The method used to identify CTDUE claims had a substantial impact on the magnitude of rates, female to male rate ratios, the most commonly affected part of the upper extremity and the highest risk occupational categories. Poisson regression identified sex interactions. It allowed rigorous evaluation of the data and indicated that rates should be examined separately for men and women. Researchers should clearly define extraction strategies and examine the impact of misclassification.

Key words: algorithm misclassification; cumulative trauma disorder; Poisson regression; rates; sex interactions; surveillance; upper extremity; workers' compensation

Statement of problem

Cumulative trauma disorder of the upper extremity (CTDUE) is an umbrella term used to describe injuries that result from repeated use of the upper extremity over time rather than from a specific incident.1 Common examples of CTDUE include carpal tunnel syndrome, tendinitis and epicondylitis. Although the proportion of work-related claims attributable to cumulative trauma appears minimal, ranging from less than 1% to 8.7%,2,3 CTDUE claims are more costly and work disabling than acute upper extremity claims4-6 or workers' compensation claims in general.3,7 Hence, accurate identification of high-risk groups is important to ascertain risk factors, initiate appropriate control activities and monitor their effectiveness.

However, an extensive review of the literature has revealed that the range in rates and rate ratios is substantial when work-related cumulative trauma disorders are considered.8 A significant contributor to this variation may be the method of defining and extracting claims. Consequently, to provide more meaningful and comparable surveillance information, analyses should attempt to use well defined, valid and reliable extraction strategies. Furthermore, general conclusions on differences in the rate of CTDUE claims across gender, age groups, part of upper extremity or occupation should acknowledge confounding and interaction. For example, a statement regarding increased risk of CTDUE claims among women relative to men based solely on gender-specific rates may be inappropriate if male and female populations differ with respect to composition factors associated with the risk of CTDUE, such as age or occupation.

Rationale for present research

This research had three important objectives. First was the estimation of crude and specific rates of first-allowed, lost-time CTDUE claims among workers covered by the Ontario Workplace Safety and Insurance Board (OWSIB) during the 1997 calendar year. A first-allowed claim is a newly registered, accepted claim for a previously unreported injury or disease, and lost-time refers to the loss of wages.9 The second objective was to provide insight into the cause of the substantial variation in published rates and rate ratios by examining the impact of two different methods of defining and extracting CTDUE claims. The last was to demonstrate how Poisson regression could be used to identify and address confounding and interaction.

Methods

Identifying CTDUE claims

An algorithm10 was used to identify CTDUE claims in the OWSIB database. This algorithm used coded information regarding "nature of injury or disease", "part of body" and "event or exposure" to classify claims into one of three mutually exclusive categories: "definite", "possible" and "non-CTDUE". The definite category was developed to capture those claims that occurred gradually over time through voluntary actions of the worker but did not produce visible trauma. The possible category was used to capture those claims that could be related to a specific incident involving voluntary actions or free bodily motion but did not produce visible trauma. Finally, the non-CTDUE category captured those claims related to a specific, untoward event producing visible trauma.

Examination of agreement between the algorithm and claim review revealed that 96.3% of claims in the algorithm definite category, 29.1% of claims in the algorithm possible category and 2.8% of claims in the algorithm non-CTDUE category were actually defined as definite CTDUE by claim review. To acknowledge algorithm misclassification, two methods of identifying CTDUE claims were used. The lower limit estimate used algorithm definite CTDUE claims. According to claim review, this category contained a homogeneous group of definite CTDUE claims. The upper limit estimate was obtained by combining algorithm definite claims and algorithm possible upper extremity claims. The upper limit estimate resulted from the following reasoning: although the claimant may be able to attribute his or her injury to a particular incident, such as voluntary lifting, pulling or pushing, this incident may have been the proverbial "straw that broke the camel's back". That is, the identified incident may have been sufficient insult to an already compromised site rather than the only insult to a healthy site.

Estimating the population at risk

The Canadian Labour Force Survey (LFS) was used to obtain estimates of the population at risk of a CTDUE injury. This involved extracting the class of worker most likely to be insured by the OWSIB and using actual hours worked to estimate full-time equivalents at risk.11

Rate estimation

All first-allowed, lost-time claims occurring in those aged 15 years or greater with a date of injury or disease in the 1997 calendar year (105,556) were sorted by the algorithm into definite (3,279), possible (9,520), or non-CTDUE claims (92,757). Since the OWSIB and 1997 LFS collected information regarding sex, age, and occupation and the OWSIB collected additional information regarding part of body, specific rates were calculated by combining information from the two data sources. The following body part categories were used: "upper extremity", "neck & shoulder/shoulder & upper arm", "elbow & forearm", and "wrist & hand". Previous research examining these categories has indicated almost perfect agreement (kappa $ 0.81) between the OWSIB coders and claim review.10 The definite rates used algorithm definite claims while the definite plus possible rates combined algorithm definite and possible upper extremity claims. Rate standard errors were calculated according to Armitage and Berry12 and were used for standard 95% confidence intervals (CI).

Prevention index

Since focussing intervention efforts on the highest risk occupations will have little impact on claim numbers if the at-risk populations are small, a prevention index was used to prioritize occupations for intervention purposes.3 All occupations were ranked according to their frequency of CTDUE claims and their CTDUE claim rate. The index is the mean of these two ranks. For example, an occupation that ranks first with respect to frequency of claims and claim rate will have a prevention index equal to one, making it worthy of increased attention and resources from a population, public health perspective.

Poisson regression modelling: the effect of sex, age, part of upper extremity and occupation on the rate of CTDUE claims

For each estimation method, claim counts of cumulative trauma disorder providing the most detail on sex, age, part of upper extremity and occupation as well as at-risk estimates were used in Poisson regression. Age was coded as a categorical variable because a curvilinear relation has been suggested.13 On the basis of previous research, the following interactions were considered:

  1. sex*age, as the highest risk age category may not be consistent across sex;13
  2. sex*part of upper extremity, as the female to male rate ratio for CTDUE seems to vary depending on whether the whole upper extremity or just carpal tunnel syndrome is considered;6,14-17
  3. sex*occupation, as the effect of occupation may not be consistent across sex;18,19 and
  4. part of upper extremity*occupation, as different jobs may be at risk for different subgroups of CTDUE.20-28

The model fitting process was executed as per Hosmer and Lemeshow.29 Briefly, all four explanatory variables were included in the initial model. An overall likelihood ratio test was conducted on the full main effects model to determine whether at least one of the explanatory variables was an important predictor of the log of the CTDUE rate. If the overall likelihood ratio test was statistically significant, a stepwise backward elimination procedure was applied using the partial likelihood ratio test at an alpha level of 0.10.30 A variable was removed if the likelihood ratio test p value was greater than 0.10 and its removal did not change the magnitude of any of the remaining regression coefficients by 10% or more. The latter requirement prevented the removal of important confounders29 with the 10% standard recommended by Koval.31 Once the main effects model had been established, interactions were added, one at a time, and their significance (alpha = 0.05) was examined using a partial likelihood ratio test. The significant interaction with the smallest p value decided how the initial model was split. The modelling process was then repeated on the sub-models. Model fit was examined by means of the Goodness of Fit test, regression diagnostics and pseudo-coefficients of determination.

Results

Crude and part of upper extremity-specific rates

The definite plus possible crude CTDUE claim rate was 3.12 times greater than the definite rate, but the increase in the rate was inconsistent across part of upper extremity (Table 1). Hence, the part of upper extremity rate ranking varied across estimation method. CTDUE claims accounted for 3.11% to 9.69% of all first-allowed, lost-time claims in those aged 15 years or greater.

TABLE 1
Crude and sex-specific CTDUE (cumulative trauma disorder of the upper extremity)
claim rates by part of upper extremity in Ontario workers, 1997

CTDUE Rate Estimation Method*

 


Part of upper extremity

Algorithm definite
(confidence intervals)

Algorithm definite + possible
(confidence intervals)


Inflation factor

All

Upper extremity**

81.68 (78.46, 84.91)

254.82 (247.80, 261.84)

3.12

Neck/shoulder/upper arm

12.18 (11.08, 13.29)

117.76 (113.68, 121.83)

9.67

Elbow/forearm

20.68 (19.21, 22.14)

37.59 (35.56, 39.63)

1.82

Wrist/hand

45.81 (43.53, 48.09)

89.38 (85.97, 92.79)

1.95

Men

Upper extremity

67.38 (63.79, 70.97)

254.99 (246.79, 263.18)

3.78

Neck/shoulder/upper arm

10.24 (8.92, 11.56)

125.64 (120.46, 130.82)

12.27

Elbow/forearm

19.02 (17.21, 20.83)

38.38 (35.75, 41.01)

2.02

Wrist/hand

36.32 (33.76, 38.87)

81.58 (77.58, 85.59)

2.25

Women

Upper extremity

101.35 (96.16, 106.55)

254.54 (245.44, 263.63)

2.51

Neck/shoulder/upper arm

14.85 (12.99, 16.71)

106.91 (101.56, 112.27)

7.2

Elbow/forearm

22.96 (20.63, 25.28)

36.51 (33.54, 39.47)

1.59

Wrist/hand

58.87 (55.04, 62.71)

100.05 (94.90, 105.21)

1.70


Sex and part of upper extremity-specific rates

Female to male rate ratios differed across estimation method (Table 1). Using the definite estimation method, female to male rate ratios ranged from a low of 1.21 for the elbow & forearm to a high of 1.62 for the wrist & hand. Using the definite plus possible estimation method, female to male rate ratios ranged from a low of 0.85 for the neck & shoulder/shoulder & upper arm to a high of 1.23 for the wrist & hand.

Sex, part of upper extremity and age-specific rates

Using the definite estimation method, both sexes demonstrated a parabolic relation between age and rate for each part of the upper extremity (Figure 1). The rates generally peaked in the 35 to 44 and 45 to 54 age group for men and women respectively, and the female to male rate ratio was usually the greatest in the 45 to 54 age group.

FIGURE 1 CTDUE claim rates by sex, part of upper extremity (NSU = Neck & Shoulder/Shoulder & Upper Arm; EF = Elbow & Forearm; WH = Wrist & Hand) and age using the definite estimation method

 

Using the definite plus possible estimation method, women continued to show a parabolic relation for all parts of the upper extremity whereas men demonstrated a parabolic relation for the elbow & forearm only (Figure 2). These parabolic relations were not as pronounced as in the definite method. Although male rates did not consistently peak in a particular age group, female rates again peaked in the 45 to 54 year age group and female to male rate ratios were greatest in the 45 to 54 year age group.

FIGURE 2 CTDUE claim rates by sex, part of upper extremity (NSU = Neck & Shoulder/Shoulder & Upper Arm; EF = Elbow & Forearm; WH = Wrist & Hand) and age using the definite + possible estimation method

Sex, part of upper extremity and occupation-specific rates

Occupational categories with the highest rates or prevention indexes were not always consistent across sex and part of upper extremity subgroups, or rate estimation method. However, regardless of estimation method, the occupational categories "textiles, furs & leather goods" and "other machining occupations" generally occurred in the top five highest rates and prevention indexes for both sexes across part of upper extremity, and "metal products, not elsewhere classified" generally occurred in the top five prevention indexes for both sexes across part of upper extremity (Figure 3).

FIGURE 3 CTDUE claim rates by sex, part of upper extremity (NSU = Neck & Shoulder/Shoulder & Upper Arm; EF = Elbow & Forearm; WH = Wrist & Hand) and occupation by estimation method

Poisson regression modelling

For both rate estimation methods, separate models were run for men and women because of significant interactions by sex (Tables 2 and 3). A part of upper extremity*occupation interaction could not be tested in the sex-specific models as a result of sparse data. For both sexes, each explanatory variable was a statistically significant ( = 0.05) predictor of the rate of CTDUE claims in the presence of the remaining variables.

TABLE 2
Poisson regression modelling of the definite CTDUE (cumulative trauma
disorder of the upper extremity) claim rate by sex in Ontario workers, 1997

Characteristic

Rate ratio (95% likelihood
ratio confidence interval)

Men (n = 585)

Women (n = 627)

*Age

15 to 24

0.355 (0.279, 0.446)

0.333 (0.263, 0.417)

25 to 34

0.760 (0.667, 0.866)

0.607 (0.530, 0.695)

35 to 44

1 (------------------)

1 (------------------)

45 to 54

0.954 (0.831, 1.094)

1.061 (0.937, 1.200)

55 plus

0.706 (0.569, 0.867)

0.717 (0.579, 0.880)

*Part of upper extremity

Wrist & hand

1 (------------------)

1 (------------------)

Elbow & forearm

0.525 (0.467, 0.591)

0.401 (0.355, 0.451)

Neck & shoulder/shoulder & upper arm

0.274 (0.235, 0.317)

0.261 (0.226, 0.300)

*Occupation

Metal products, NEC

1 (------------------)

1 (------------------)

Government officials & administrators

0.037 (0.006, 0.115)

0.434 (0.247, 0.712)

Other managers & administrators

0.004 (0.001, 0.011)

0.014 (.006, 0.027)

Management & administration related

-

0.053 (0.031, 0.086)

Physical & life sciences

0.067 (0.017, 0.176)

0.164 (0.050, 0.391)

Math, statistics, systems analysis & related

0.033 (0.012, 0.071)

0.050 (0.015, 0.119)

Architects & engineers

0.007 (0.000, 0.033)

0.072 (0.004, 0.324)

Architecture & engineering related

0.118 (0.050, 0.232)

0.082 (0.005, 0.368)

Social sciences & related

-

0.101 (0.057, 0.165)

University teaching & related

-

0.033 (0.002,0.149)

Elementary/secondary school teaching & related

-

0.007 (0.001, 0.022)

Other teaching & related

-

0.041 (.007, 0.130)

Nursing, therapy & related

0.035 (0.002, 0.155)

0.130 (0.091, 0.184)

Other medicine & health

-

0.149 (0.085, 0.245)

Artistic & recreational

0.034 (0.008, 0.090)

0.027 (0.004, 0.086)

Stenographic & typing

0.631 (0.036, 2.802)

0.270 (0.198, 0.366

Bookkeeping, account-recording & related

0.018 (0.001, 0.081)

0.240 (0.181, 0.316)

Office Machine & EDP operator

0.085 (0.014, 0.265)

0.363 (0.235, 0.542)

Material recording, scheduling & distribution

0.115 (0.065, 0.187)

0.171 (0.080, 0.319)

Reception, information, mail & message

0.554 (0.355, 0.824)

0.374 (0.264, 0.521)

Library, file, correspondence clerks & related

0.108 (0.046, 0.211)

0.265 (0.199, 0.353)

Sales, commodities

0.098 (0.064, 0.143)

0.293 (0.219, 0.391)

Sales, services & other sales

0.074 (0.032, 0.146)

0.045 (0.016, 0.100)

Protective services

0.037 (0.013, 0.080)

0.194 (0.076, 0.404)

Food & beverage preparations/lodging & accomodation

0.219 (0.144, 0.318)

0.258 (0.190, 0.349)

Personal, apparel & furnishings service

0.303 (0.120, 0.623)

0.266 (0.180, 0.385)

Other service

0.328 (0.241, 0.439)

0.944 (0.701, 1.265)

Farmers & farm management

-

1.432 (0.081, 6.416)

Other farming, horticultural & animal
husbandry

0.360 (0.226, 0.545)

0.760 (0.457, 1.199)

Fishing, hunting, trapping & related

3.284 (0.544, 10.246)

-

Forestry & logging

1.103 (0.547, 1.965)

-

Mining & quarrying

1.208 (0.815, 1.727)

-

Food, beverage & related processing

1.093 (0.833, 1.416)

2.051 (1.566, 2.687)

Other processing

0.373 (0.271, 0.502)

1.782 (1.263, 2.481)

Metal shaping & forming

0.795 (0.624, 1.005)

0.952 (0.466, 1.730)

Other machining occupations

1.120 (0.900, 1.388)

8.894 (6.625, 11.892)

Electrical, electronic & related equipment

0.185 (0.123, 0.269)

0.943 (0.679, 1.296)

Textiles, furs, and leather goods

4.249 (3.339, 5.363)

2.185 (1.718, 2.791)

Wood products, rubber, plastics & related & other

0.812 (0.652, 1.005)

1.773 (1.362, 2.311)

Mechanics & repairmen

0.381 (0.306, 0.473)

1.031 (0.435, 2.056)

Excavating, grading, paving & related

0.159 (0.075, 0.290)

-

Electrical power, lighting & wire
communication

0.352 (0.231, 0.514)

1.868 (0.660, 4.124)

Other construction

0.370 (0.288, 0.470)

1.289 (0.503, 2.696)

Motor transport operators

0.105 (0.071, 0.150)

0.030 (0.002, 0.133)

Other transportation operators

0.601 (0.359, 0.942)

1.970 (0.831, 3.927)

Material handling

0.221 (0.155, 0.307)

0.524 (0.366, 0.738)

Other crafts & equipment operators & NEC

0.218 (0.137, 0.330)

0.999 (0.646, 1.493)

Goodness of fit test

Deviance

516.1675, df = 539,
p = 0.7534

633.2577, df = 578,
p = 0.0553

Note: Reference categories are indicated by the estimate 1. Occupation was coded as per the Labour Force Survey 1997. The top five point estimates have been bolded for the occupation construct. Dashes indicate a lack of events in the occupation category or an at-risk estimate of zero.

NEC = not elsewhere classified; EDP = electronic data processor.

* Statistically significant (p < 0.0001) predictor of the rate of CTDUE claims in the presence of the remaining variables.

TABLE 3
Poisson regression modelling of the definite + possible CTDUE (cumulative trauma
disorder of the upper extremity) claim rate by sex in Ontario workers, 1997

Characteristic

Rate ratio (95% likelihood
ratio confidence interval)

Men (n = 660)

Women (n = 633)

*Age

15 to 24

0.826 (0.753, 0.906)

0.544 (0.480, 0.616)

25 to 34

0.965 (0.901, 1.034)

0.824 (0.758, 0.895)

35 to 44

1 (------------------)

1 (------------------)

45 to 54

0.855 (0.790, 0.924)

1.046 (0.963, 1.135)

55 plus

0.849 (0.762, 0.944)

0.776 (0.678, 0.884)

*Part of upper extremity

Wrist & hand

1 (------------------)

1 (------------------)

Elbow & forearm

0.469 (0.432, 0.509)

0.375 (0.341, 0.412)

Neck & shoulder/shoulder & upper arm

1.541 (1.452, 1.636)

1.098 (1.026, 1.176)

*Occupation

 

 

 

Metal products, NEC

1 (------------------)

1 (------------------)

Government officials & administrators

0.051 (0.020, 0.103)

0.323 (0.206, 0.482)

Other managers & administrators

0.006 (0.003, 0.010)

0.015 (0.009, 0.024)

Management & administration related

0.010 (0.004, 0.021)

0.039 (0.026, 0.057)

Physical & life sciences

0.088 (0.042, 0.160)

0.081 (0.025, 0.190)

Math, statistics, systems analysis & related

0.013 (0.005, 0.029)

0.024 (0.007, 0.056)

Architects & engineers

0.019 (0.008, 0.039)

0.033 (0.002, 0.146)

Architecture & engineering related

0.086 (0.046, 0.145)

0.079 (0.013, 0.245)

Social sciences & related

0.094 (0.051, 0.156)

0.201 (0.151, 0.263)

University teaching & related

-

0.065 (0.020, 0.154)

Elementary/secondary school teaching & related

0.019 (0.006, 0.044)

0.053 (0.036, 0.077)

Other teaching & related

0.014 (0.001, 0.061)

0.216 (0.134, 0.329)

Nursing, therapy & related

1.398 (1.115, 1.733)

0.859 (0.736, 1.006)

Other medicine & health

0.080 (0.028, 0.172)

0.180 (0.127, 0.248)

Artistic & recreational

0.061 (0.034, 0.102)

0.056 (0.026, 0.102)

Stenographic & typing

0.766 (0.190, 1.996)

0.167 (0.129, 0.214)

Bookkeeping, account-recording & related

0.054 (0.025, 0.102)

0.209 (0.171, 0.255)

Office machine & EDP operator

0.034 (0.006, 0.104)

0.200 (0.134, 0.286)

Material recording, scheduling
& distribution

0.379 (0.311, 0.457)

0.337 (0.233, 0.473)

Reception, information, mail & message

1.007 (0.811, 1.237)

0.351 (0.275, 0.445)

Bookkeeping, account-recording & related

0.054 (0.025, 0.102)

0.209 (0.171, 0.255)

Office machine & EDP operator

0.034 (0.006, 0.104)

0.200(0.134, 0.286)

Material recording, scheduling & distribution

0.379 (0.311, 0.457)

0.337 (0.233, 0.473)

Other service

0.624 (0.537, 0.723)

1.451 (1.207, 1.746)

Farmers & farm management

-

0.769 (0.044, 3.413)

Other farming, horticultural &
animal husbandry

0.451 (0.354, 0.567)

0.757 (0.542, 1.034)

Fishing, hunting, trapping & related

1.094 (0.182, 3.392)

-

Forestry & logging

1.020 (0.653, 1.510)

1.337 (0.076, 5.961)

Mining & quarrying

0.742 (0.538, 0.995)

-

Food, beverage & related processing

1.131 (0.954, 1.335)

1.844 (1.519, 2.238)

Other processing

0.612 (0.518, 0.721)

1.748 (1.375, 2.208)

Metal shaping & forming

0.853 (0.733, 0.990)

1.350, 0.891, 1.965)

Other machining occupations

1.629 (1.436, 1.847)

7.254 (5.846, 8.972)

Electrical, electronic & related
equipment

0.195 (0.150, 0.248)

0.708 (0.549, 0.906)

Textiles, furs, and leather goods

6.903 (6.044, 7.873)

2.259 (1.908, 2.682)

Wood products, rubber, plastics & related & other

0.887 (0.774, 1.015)

1.555 (1.285, 1.881)

Mechanics & repairmen

0.502 (0.440, 0.572)

1.510 (0.936, 2.305)

Excavating, grading, paving & related

0.181 (0.117, 0.266)

-

Electrical power, lighting & wire communication

0.491 (0.390, 0.611)

1.252 (0.533, 2.458)

Other construction

0.460 (0.397, 0.531)

1.522 (0.879, 2.449)

Motor transport operators

0.463 (0.403, 0.530)

0.164 (0.084, 0.286)

Other transportation operators

2.242 (1.868, 2.674)

4.578 (3.118, 6.504)

Material handling

0.741 (0.647, 0.847)

0.676 (0.539, 0.844)

Other crafts & equipment operators & NEC

0.550 (0.453, 0.664)

0.885 (0.646, 1.191)

Goodness of fit test

Deviance

1134.1943, df = 609,
p < 0.0001

1197.1515, df = 583,
p < 0.0001


Note: Reference categories are indicated by the estimate 1. Occupation was coded as per the Labour Force Survey 1997. The top five point estimates have been bolded for the occupation construct. Dashes indicate a lack of events in the occupation category or an at-risk estimate of zero.

NEC = not elsewhere classified; EDP = electronic data processor.

* Statistically significant (p < 0.0001) predictor of the rate of CTDUE claims in the presence of the remaining variables.


Men and women demonstrated a parabolic relation between the rate of CTDUE claims and age, peaking in the 35 to 44 and 45 to 54 year age categories respectively. The parabolic relation was less pronounced in the definite plus possible estimation method, particularly for the men.

Using the definite rate estimation method, the rates of elbow & forearm and neck & shoulder/shoulder & upper arm claims were significantly less than was the rate of wrist & hand claims, with rate ratios of approximately 0.5 and 0.25 respectively. When using the definite plus possible rate estimation method, the rate ratios for part of upper extremity were not consistent across sex. Although men and women demonstrated a significantly lower rate of elbow & forearm claims relative to the wrist & hand, the rate of neck & shoulder/shoulder & upper arm claims was significantly greater than the wrist & hand in men, with no practically important difference noted in women.

The five occupational categories with the highest rate ratios, according to the point estimates, are bolded in Tables 2 and 3. When the effect of occupation across sex was compared, there were indications of both qualitative and quantitative interactions. With a qualitative interaction, an exposure's effect is opposite across sub-groups, whereas with a quantitative interaction, an exposure's effect varies in magnitude across subgroups.31 As an example of a qualitative interaction in the definite Poisson regression models, the rate of CTDUE claims in the occupation "other processing" was significantly less than that of "metal products, nec" among men (rate ratio = 0.373, 95% CI 0.271- 0.502), whereas among women it was significantly greater (rate ratio = 1.782, 95% CI 1.263-2.481). As an example of quantitative interaction, the rate of CTDUE claims in "textiles, furs, and leather goods" was 4.249 times greater (95% CI 3.339-5.363) than that of "metal products, nec" among men, whereas among women it was 2.185 times greater (95% CI 1.718-2.791).

When model fit for the definite estimation method was examined, the goodness of fit test, regression diagnostics and pseudo-coefficient of determination suggested that the male model fit the data well. For the female model, the goodness of fit test was borderline significant. Examination of the standardized residuals identified one extreme outlier. When the observation producing this residual was removed from the data set and the model re-fitted, the goodness of fit test p value increased (Deviance = 615.210, degrees of freedom = 577, p = 0.1313), but the model parameters remained virtually the same, suggesting that the outlier was not influential. Hence, the original model was considered to fit the data well. When model fit for the definite plus possible estimation method was examined, the goodness of fit test and regression diagnostics suggested poor model fit for both sexes.

Discussion

Part of upper extremity-specific rates

The crude CTDUE claim rate derived using the definite method, 81.68 per 100,000 full-time equivalents (FTEs) (Table 1), was congruous with the rate for Ontario in 1991.14 However, acknowledging the algorithm misclassification inflated the crude rate by a factor of 3.12 to 254.82 per 100,000 FTEs. Similarly, the proportion of all first-allowed, lost-time claims attributable to CTDUE varied substantially, from 3.11% to 9.69%. These findings indicate that considerable variation in rates and proportions can be ascribed to the method used to define and extract claims. The variation was so great that the neck & shoulder/shoulder & upper arm, which was at lowest risk using the definite method, was at greatest risk using the definite plus possible method (Table 1). Thus, by acknowledging potential misclassification, attention is drawn to the vulnerability of the neck & shoulder/shoulder & upper arm and risk factors previously associated with this area.20

Sex and part of upper extremity-specific rates

The overall female to male rate ratio calculated using the definite method, 1.5 (Table 1), is comparable to that noted for Ontario in 1991,14 but acknowledging the potential misclassification reduced the ratio to 1.0. Despite the equality of the overall definite plus possible rates, female to male rate ratios continued to vary across part of upper extremity. Men had a higher neck & shoulder/shoulder & upper arm claim rate, and women had a higher wrist & hand claim rate. Several reasons may account for this differential. First, men and women had different occupational distributions and thus were exposed to different job-related risk factors in 1997 (2 = 1334310, df = 48, p < 0.0001). Second, there may be sex differences in tasks within the same job title.18,19 Finally, there are many sex differences not examined by this research.8

Sex, part of upper extremity and age-specific rates

When the definite method is used, a parabolic relation between the rate of CTDUE claims and age was demonstrated (Figure 1). This is counterintuitive, as one would expect the rate of CTDUE claims to increase with age as a result of the degenerative effect of aging and confounding with duration of exposure.8 The decreased rate after the peak may be the result of the healthy worker survivor effect;32-34 workers progressing to physically less stressful jobs with seniority; or OWSIB policy, which indicates that recurrences or associated disorders should be documented on the initially established claim.9

With algorithm misclassification taken into account, the male rate varied little with age for the neck & shoulder/shoulder & upper arm and demonstrated a statistically significant linear decline for the wrist & hand (F1,3 = 98.79, p = .0022, r2 = 0.97) (Figure 2). The varying effect of age may be the consequence of the type of claims falling into the algorithm possible category. These claims were primarily from males; generally diagnosed as sprains, strains, tears; mainly affected the neck & shoulder/shoulder & upper arm and wrist & hand; and predominantly resulted from some form of overexertion. Perhaps the high force component of these injuries negates the need for prolonged exposure that is associated with age.

Regardless of estimation method and part of upper extremity, the female to male age-specific rate ratios commonly peaked in the 45 to 54 age group, indicating that this age period is of particularly high risk for women. This vulnerability may be related to the hormonal changes or hormone replacement therapy associated with menopause.8

Sex, part of upper extremity and occupation-specific rates

Congruous with previous research, the effect of occupation on the CTDUE claim rate was not consistent across sex16,17 or part of upper extremity.20-28 Potential reasons for the first interaction have been discussed. The latter interaction suggests that different occupations are characterized by different typical duties that may stress different parts of the upper extremity. For both men and women, the occupational categories "textiles, furs & leather goods" and "other machining occupations" generally ranked in the top five highest rates and prevention indexes for each part of the upper extremity across estimation methods. These occupational categories had relatively stable rates and collectively accounted for 2.1% of the employee FTEs in 1997. The importance of the occupational category "metal products, nec" was identified through the prevention index. Although this occupational category did not consistently appear among the highest rates for each part of the upper extremity across estimation methods, it generally ranked in the top five prevention indexes because it accounted for a large proportion of employee FTEs in 1997 - i.e. 3.5%. All these occupational categories would be worthy of greater scrutiny to determine which specific occupations and associated duties or work organization factors are responsible for the increased risk.

Poisson regression modelling

Poisson regression modelling allowed a more rigorous evaluation of the data than did the calculation and comparison of specific rates. In fact, this is one of the primary advantages of Poisson regression: to identify and quantify systematic trends that are not easily appreciated in a large volume of data.35,36 Poisson regression usually identified statistically significant sex*part of upper extremity, sex*age and sex*occupational category interactions, which were reflected in the specific rates (Table 1, Figures 1, 2 and 3). Hence, using conventional standardization techniques to make comparisons across sex or occupation would not convey the complexity of the differences.37 Thus, Poisson regression indicates that male and female rates should be examined separately.

Several factors may have contributed to poor model fit when misclassification was acknowledged (Table 3). First, the final model presented assumed no interactions, but Figures 2 and 3 suggested potential age*part of upper extremity and part of upper extremity*occupation interactions respectively. Adequate data did not exist to test the latter interaction, but the former interaction was statistically significant (p < 0.05) for both the men and women. Second, the models did not include any work organization or detailed ergonomic measures previously associated with CTDUE. Third, the definite plus possible method of estimation may have combined claims with different risk factors into one overall rate to be predicted by the same model. For example, algorithm definite claims tended to be related to repetitiveness whereas algorithm possible claims were primarily related to overexertion.10 Thus, although the overall Poisson regression models and each of their components were statistically significant when using the definite plus possible rate estimation method, the observed summary measures for the effect of age, part of upper extremity and occupational category may not be accurate across worker subgroups.

Choice of estimation method

Estimation method had a dramatic impact on the conclusions. If information on the cost and disability associated with claims falling into the algorithm definite and possible categories was available, attention could be focused on the estimation method that identified the most costly and disabling claims.

Limitations

Several limitations need to be acknowledged. First, the specificity of the occupational categories was limited by the level of detail used in the LFS. Consequently, some occupations at high risk may be masked by the aggregation, but an elevated risk despite the aggregation is certainly worthy of increased attention. Hence, this type of surveillance activity can be used to stimulate more detailed epidemiologic investigations, target resources for ergonomic evaluations and prevention, and evaluate control activities.38,39

Second, exposure was quantified using broad occupational categories rather than accurate measurements of risk factors. This crude measure of exposure probably contributed to the poor fitting models. Third, first-allowed, lost-time claims were used rather than all first-allowed claims, because only the former were adequately coded for algorithm application. Hence, the rates reflect those injuries significant enough to result in a loss of wages. It is possible that occupational categories identified as low risk may have a substantial occurrence of CTDUE claims that do not result in lost wages.

Finally, as the rates became more and more specific, stability was compromised by a decreasing number of events and smaller population at-risk estimates.40 One solution to this problem could be the combining of data from consecutive calendar years to increase the number of events and population at-risk estimates for the more specific rates. However, when choosing which years to combine, changes in OWSIB policy or claim coding and LFS methodology should be considered.

Conclusions

The method used to identify CTDUE claims had a substantial impact on the magnitude of rates, female to male rate ratios, the most commonly affected part of the upper extremity and the highest risk occupational categories. Adjusting for the potential misclassification of an extraction algorithm increased the crude rate of CTDUE claims in OWSIB-covered workers by a factor of 3.12, decreased the female to male rate ratio from 1.50 to 1.00 and identified the neck & shoulder/shoulder & upper arm as being just as vulnerable as the wrist & hand. The 45 to 54 year age category was noted to be a particularly high-risk period for women. The occupational categories "textiles, furs & leather goods", "other machining occupations" and "metal products, nec" were identified as being worthy of greater investigation. Consistent with previous research, Poisson regression identified sex interactions indicating that rates in men and women should be examined separately.

Acknowledgements

This work has been supported by an Ontario Graduate Studies Scholarship; Ontario Graduate Studies in Science and Technology Scholarship; and Physiotherapy Foundation of Canada Ann Collins Whitmore Memorial Award.

References

  1. Stobbe T. Occupational ergonomics and injury prevention. Occup Med 1996;11(3):531-43.
  2. Brogmus G, Sorock G, Webster B. Recent trends in work-related cumulative trauma disorders of the upper extremities in the United States: an evaluation of possible reasons. J Occup Environ Med 1996;38:401-11.
  3. Silverstein B, Viikari-Juntura E, Kalat J. Use of a prevention index to identify industries at high risk for work-related musculoskeletal disorders of the neck, back, and upper extremity in Washington State, 1990-1998. Am J Ind Med 2002;41:149-69.
  4. Silverstein B, Welp E, Nelson N, Kalat J. Claims incidence of work-related disorders of the upper extremities: Washington State, 1987 through 1995. Am J Public Health 1998;88:1827-33.
  5. Webster B, Snook S. The cost of compensable upper extremity cumulative trauma disorders. J Occup Med 1994;36(7):713-17.
  6. Yassi A, Sprout J, Tate R. Upper limb repetitive strain injuries in Manitoba. Am J Ind Med 1996;30:461-72.
  7. Hashemi L, Webster B, Clancy E, Courtney T. Length of disability and cost of work-related musculoskeletal disorders of the upper extremity. J Occup Environ Med 1998;40(3):261-69.
  8. Zakaria D, Robertson J, MacDermid J, Hartford K, Koval J. Work-related cumulative trauma disorders of the upper extremity: navigating the epidemiologic literature. Am J Ind Med 2002;42:258-69.
  9. Workplace Safety and Insurance Board of Ontario. Operational policy. Toronto: Workplace Safety and Insurance Board of Ontario, 1998.
  10. Zakaria D, Mustard C, Robertson J et al. Identifying cumulative trauma disorders of the upper extremity in workers' compensation databases. Am J Ind Med 2003;43(5):507-18.
  11. Zakaria D, Robertson J, MacDermid J, Hartford K, Koval J. Estimating the population at risk for Ontario Workplace Safety and Insurance Board-covered injuries or diseases. Chronic Dis Can 2002;23:17-21.
  12. Armitage P, Berry G. Statistical methods in medical research. 3rd ed. Oxford: Blackwell Scientific Publications, 1994:91.
  13. Tanaka S, Seligman P, Halperin W et al. Use of workers' compensation claims data for surveillance of cumulative trauma disorders. J Occup Med 1988;30:488-92.
  14. Ashbury F. Occupational repetitive strain injuries and gender in Ontario. J Occup Environ Med 1995;37:479-85.
  15. Feuerstein M, Miller V, Burrell L, Berger R. Occupational upper extremity disorders in the federal workforce. J Occup Environ Med 1998;40:546-55.
  16. Gun R. The incidence and distribution of RSI in South Australia 1980-81 to 1986-87. Med J Aust 1990;153:376-80.
  17. Sprout J. The gender differences in upper-extremity occupational repetitive strain injuries in Manitoba [dissertation]. University of Manitoba, 1997.
  18. Messing K, Dumais L, Courville J, Seifert A, Boucher M. Evaluation of exposure data from men and women with the same job title. J Occup Med 1994;36:913-17.
  19. Nordander C, Ohlsson K, Balogh I, Rylander L, Palsson B, Skerfving S. Fish processing work: the impact of two sex dependent exposure profiles on musculoskeletal health. Occup Environ Med 1999;56:256-64.
  20. Bernard B, Putz-Anderson V, Burt S et al. Musculoskeletal disorders and workplace factors: a critical review for work-related musculoskeletal disorders of the neck, upper extremity, and low back. Cincinnati: National Institute for Occupational Safety and Health, 1997.
  21. Bernard B, Sauter S, Fine L, Petersen M, Hales T. Job task and psychosocial risk factors for work-related musculoskeletal disorders among newspaper employees. Scand J Work Environ Health 1994;20:417-26.
  22. Cherniak M. Epidemiology of occupational disorders of the upper extremity. Occup Med 1996;11:513-30.
  23. English C, Maclaren W, Court-Brown C et al. Relations between upper limb soft-tissue disorders and repetitive movements at work. Am J Ind Med 1995;27:75-90.
  24. Hagberg M, Wegman D. Prevalence rates and odds ratios of shoulder-neck diseases in different occupational groups. Br J Ind Med 1987;44:602-10.
  25. Kurppa K, Viikari-Juntura E, Kuosma E, Huuskonen M, Kivi P. Incidence of tenosynovitis or peritendinitis and epicondylitis in a meat-processing factory. Scand J Work Environ Health 1991;17:32-7.
  26. Ranney D, Wells R, Moore A. Upper limb musculoskeletal disorders in highly repetitive industries: precise anatomical physical findings. Ergonomics 1995;38:1408-23.
  27. Stenlund B, Goldie I, Hagberg M, Hogstedt C. Shoulder tendinitis and its relation to heavy manual work and exposure to vibration. Scand J Work Environ Health 1993;19:43-9.
  28. Welch L, Hunting K, Kellogg J. Work-related musculoskeletal symptoms among sheet metal workers. Am J Ind Med 1995;27:783-91.
  29. Hosmer D, Lemeshow S. Applied logistic regression. 2nd ed. New York: John Wiley & Sons Inc., 2000:91-9.
  30. Kennedy W, Bancroft T. Model building for prediction in regression based upon repeated significance test. Ann Math Stat 1971;42:1273-84.
  31. Koval J. Epidemiology 512 course notes. In: Biostatistical methods. London, Ontario: University of Western Ontario, 1997:9,21.
  32. Hernberg S. Validity aspects of epidemiological studies. In: Karvonen M, Mikheev M, editors. Epidemiology of occupational health. Copenhagen, Europe: WHO Regional Office for Europe, 1986.
  33. Monson R. Observations on the healthy worker effect. J Occup Med 1986;28:425-33.
  34. Steenland K, Deddens J, Salvan A, Stayner L. Negative bias in exposure-response trends in occupational studies: modeling the healthy worker survivor effect. Am J Epidemiol 1996;143:202-10.
  35. Gill J. Generalized linear models: a unified approach. Thousand Oaks: Sage Publications, Inc., 2001.
  36. Little R. Generalized linear models for cross-classified data from the WFS. International Statistical Institute. World Fertility Survey. Technical Bulletins, 1978.
  37. Fleiss J. Statistical methods for rates and proportions. 2nd ed. New York: John Wiley & Sons, 1981.
  38. Canadian Standards Association. Z795-96 coding of work injury or disease information. Etobicoke, Ontario: Canadian Standards Association, 1996.
  39. Schwartz E. Use of workers' compensation claims for surveillance of work-related illness - New Hampshire, January 1986-March 1987. MMWR 1987;36:713-20.
  40. Pagano M, Gauvreau K. Principles of biostatistics. Belmont: Wadsworth Publishing Company, 1993:76.

Author References

Dianne Zakaria, Continuing Care Reporting System, Canadian Institute for Health Information, Ottawa, Ontario

James Robertson, John Koval, Department of Epidemiology & Biostatistics, University of Western Ontario, London, Ontario

Joy MacDermid, Hand and Upper Limb Centre, St. Joseph's Health Care, London, Ontario, and School of Rehabilitation Science, McMaster University, Hamilton, Ontario

Kathleen Hartford, Lawson Health Research Institute, London, Ontario

Correspondence: Dianne Zakaria, Canadian Institute for Health Information, 377 Dalhousie Street, Suite 200, Ottawa, ON Canada, K1N 9N8; Fax: (613) 241-8120; E-mail: dzakaria@cihi.ca


Report a problem or mistake on this page
Please select all that apply:

Thank you for your help!

You will not receive a reply. For enquiries, contact us.

Date modified: