An in-depth research study of work-related musculoskeletal disorders in Washington state from 1990-1998 based on worker’s compensation claims.
|Workers’ Compensation System|
|Claims Management Database|
|Definition of Outcome|
|Validity of Codes|
|Washington State Employment by Industry and Region|
2.1. Workers’ Compensation System
In Washington State, employers (except the self-employed) are required to obtain workers’ compensation insurance through the Department of Labor & Industries’ (L&I) industrial insurance system unless they are able to self-insure. L&I's State Fund covers approximately two-thirds of the workers in Washington State (the remainder works chiefly for the approximately 400 largest employers and is covered by their self-insured employers). Washington is the only state in which workers contribute a portion of their income to the medical aid portion of the State Fund.
2.1.1. Claims Management Data Base
Workers’ compensation claims data and employment data for the years 1990-1998 were obtained from L&I ‘s files. The L&I claims management data base consists of two major data processing systems. The Medical Information and Payment System (MIPS) receives all billing information generated by provider medical bills. This system records such relevant items as dates of service, all associated procedure and treatment (CPT) codes, and physician diagnosis by International Classification of Disease (ICD) version 9 code for each provider visit. The Labor and Industries' Industrial Insurance System (LINIIS) contains all data necessary for the administration of State Fund claims (e.g. claim type and nature, occupation, employer information. lost time, status, progress). Only those Self-Insured claims resulting in more than 3 days of lost time are coded in the LINIIS system. Rarely are there ICD9 codes or medical billing information in the MIPS database for the Self-Insured claims. Thus, the self-insured data in this report is not comparable to the State Fund data in terms of magnitude or cost.
Definition of Outcome
We used accepted State-Fund claims (for the 1990-1998 period, approximately 10% of the State Fund claims were rejected, Figure 1). Further, we only included claims with authorized or allowed bills for specific diagnoses codes or appropriate CPT procedure codes respectively. The final number of claims extracted from the MIPS database was comprised of claims that had either authorized or allowed CPT codes, or both. In addition, we extracted any claim that had a wrist or hand condition (or both) from the LINIIS claim history dataset by using the ANSI z16.2 code for body area. Similar methods were used to extract claims for general back, elbow and shoulder disorders. The specific disorders were defined as accepted claims based on claims with codes and/or CPT procedure codes (see Table 1 for codes).
a workers compensation claim in Washington State may include disorders
in more than one body part, only the primary site is assigned a z16.2
code. When specific disorders (like CTS for hand/wrist disorders) were
examined in detail to determine type of onset, disorders were required
to match the appropriate body area code (since type and nature of disorder
are only specified for the primary site of disorder).
Information collected for each claim included: claim status (“compensable” lost time claim of 4 or more days or medical treatment claim only codes); z16.2 codes for body area; nature; and type of disorder; 1 and 3 digit Standard Industrial Code (SIC); 4- digit Washington Industrial Code (WIC); claim identification number; social security number; date of injury; birth date; gender; total cost of claim; days of time loss; dollar amount of time loss payments; and dollar amount of medical aid payments. Using first date of injury allows us to estimate claims incidence. For example, if in LINIIS, a first date of injury year is 1990 and recorded for body area of hand/ wrist, but the first MIPS allowed bill with a CTS code is in 1991, for purposes of this analysis, this is a 1990 CTS claim.
We categorized non-traumatic and traumatic onset to differentiate “cumulative” trauma exposures from acute exposure, such as falls. A combination of BODY PART and NATURE and TYPE was required. Non-traumatic onset, TYPE codes (z16.2 code) included: rubbed or abraded (080), further restricted to disorders caused by leaning, kneeling, or sitting on objects (not vibrating) (081), those caused by objects being handled (not vibrating) (082), those caused by vibrating objects (083), those caused by repetition of pressure (085) and those caused by repetitive motion (086); overexertion (120); bodily reaction (100); and unknown (primarily strain, muscle soreness, pain with lifting etc.) (899-999). These TYPE codes were combined with the following NATURE codes: dislocation or herniation (190 for neck and back only); inflammation or irritation of the joints, tendons or muscles (260), including bursitis, tendinitis, synovitis and tenosynovitis; sprains and strains (310); multiple injuries (400 for upper extremity only); diseases of the nervous system (560), nerves and peripheral ganglia (562); symptoms and ill-defined conditions (580); and unclassified (999). Disorders not fulfilling the criteria for non-traumatic onset were considered traumatic (e.g., type was slips, trips, falls, struck by).
were extracted from L&I databases as of October, 1999. Claim costs
and time loss days reported here reflect actual totals for closed claims.
For claims that were not closed, costs and time loss days reflect actual
totals to this date plus total future estimated costs and time loss days,
as calculated by agency actuarial staff. Cost and lost time data are expected
to develop further for the most recent years. For example, as of October
1999, approximately 11% of all 1996 carpal tunnel syndrome claims and
19% of 1997 and 40% of 1998 carpal tunnel syndrome claims were still open.
For non-traumatic soft tissue disorders, approximately 10% of 1998 claims
were still open (Figure 1).
For Self-Insured compensable claims, we abstracted body part, nature, and type. We are less confident about the distinction between non-traumatic and traumatic onset status with the self-insured data because of more incomplete information in these data. We estimated the costs as a fraction of the State Fund costs and assumed a 25% lower compensable cost, because we considered the larger employers to be able to return workers to work more quickly than the small employers in the State Fund.
Validity of Codes
Numerous medical records abstraction exercises were conducted to evaluate the coding schemes used for both onset type (traumatic or non-traumatic) and specific diagnosis of upper extremity claims. In the first exercise, we took a random sample of 96 Washington State Fund compensable claims coded as CTS (N=56), epicondylitis (N=15) and rotator cuff disorders (N=25). One of the three diagnoses was recorded in each of the medical records. This exercise demonstrated that the physician’s statement on the medical records is fairly accurately translated into the coding system of the L&I claims management data base. We also observed that CTS and epicondylitis are often mentioned together in a single claim, and that CTS is usually filed as the main disorder. Thus, the incidence of “elbow” disorders would be underestimated because the epicondylitis case would be identified under "hand/wrist." Additionally, the cost and lost time information for that epicondylitis case would be lost because we required body part be elbow and diagnosis be epicondylitis in order to avoid overestimation of costs for specific conditions.
studies were used to confirm the diagnosis of CTS in 77% of cases.
We also checked whether our definition for a traumatic or non-traumatic onset disorder, based on our selected codes, agreed with information from the medical records, in which it was very clear whether the onset was traumatic or non-traumatic. There was 76% agreement for the hand/wrist 77% for elbow disorders and 64% for the shoulder disorders.
The second exercise involved abstracting medical records from 100 random claims from 1995 that were coded “traumatic carpal tunnel syndrome” and 98 “non-traumatic onset hand/wrist disorders”. The case definition for carpal tunnel syndrome included symptoms in the median nerve distribution, and one of the following: positive electrodiagnostic study, carpal tunnel release surgery or positive physical examination. Eighty-one percent of the first group met the case definition for carpal tunnel syndrome and 43% of the second group met the case definition for CTS. All of those coded as non-traumatic onset met the definition of non-traumatic onset. Of those that were coded traumatic onset, 64% were actually non-traumatic onset. This suggests that the incidence rate for non-traumatic onset hand/wrist disorders may be underestimated. For rotator cuff disorders, about 30% of the claims coded as traumatic were non-traumatic.
traumatic onset upper extremity claims, 50.8% were due to being struck
against or struck by, whereas 42% of the traumatic onset back disorders
were coded as fall related.
In addition to low back and upper extremity disorders, we reviewed neck and lower extremity disorders. For this purpose, the validity of the codes for nature and type was scrutinized. The purpose was to see whether a distinction could be made between non-traumatic soft tissue disorders and other musculoskeletal disorders.
For the neck disorders, the nature coding was in agreement with the information in the medical files in 86% of the cases. Of the types, 88% of the codes were in agreement.
most common neck disorders were sprain and strain (in 43% of bills), cervicalgia
(6%), dislocation (7%), displacement of intervertebral disc (3%), and
radicular syndrome or radiculitis (3%). Of these diagnostic groups, radicular
syndrome or radiculitis differed from the others in that about half of
the cases were non-traumatic, whereas the proportion of non-traumatic
cases for the other neck disorders was about 30%. The too small number
of cases consistently diagnosed with radiculitis precluded, however, the
consideration of this diagnostic group separately in the analysis.
2.2. Washington State Employment by Industry and Region
Employment Information is reported to L&I by State Fund employers as the number of hours worked by employees. However, hours by age and gender are not available. Numbers of employees working per year were calculated assuming that each full-time employee works 2,000 hours per year (40 hours per week for 50 weeks per year). Hours were converted to full time equivalent workers (FTEs) as FTEs = total hours reported/2,000. In those industries where there is a high proportion of part-time workers, the denominator may be an underestimate, making the incidence rate higher than it would be if there were all full-time workers. An industrial classification is a grouping of industries that share similar workplace exposures. Washington Industrial classifications are more specific than Standard Industrial Classifications (SIC) because employers must subclassify their employees based on type of work. Claim rates are presented also for SICs, because they allow a national comparison. To eliminate unstable rates, only those SIC codes with a minimum of 100 full time employees per year and those WIC codes with a minimum of 50 employees per year over the 9-year period were included in the industry analyses. The claims data includes the county, but employment information by county is not available. To obtain claims incidence rates for different parts of the State of Washington, we used Washington State Population Survey data from 1998 v , in which numbers of full time or part time workers are given for eight areas of the State. Self-insured compensable claims are not received and coded by L&I until they have been closed, thus this long lag time underestimates the number of claims in more recent years. Because claims data is incomplete for the self-insured, due to a high proportion of open claims, we used 1997 claims for the analysis.
Tunnel Syndrome in Clerical Office, Not Otherwise Classified
Descriptive analyses included a summary of claims by year, direct workers’ compensation costs, lost time, age and gender. Claim incidence rates were calculated by year and industry class, and are expressed as number of claims / 10,000 FTEs. Each industry code specific rate was compared to the industry-wide rate and a crude incident rate ratio or relative risk was calculated.Severity rates were calculated as lost days / 10,000 FTEs for claims matching the appropriate body area code. Test for trend of incidence rates over time was performed using a Poisson regression analysis in SAS Software (SAS Proprietary Software Version 7, SAS Institute Inc. Cary, NC, USA 1998). Differences between rates were tested by Poisson regression with an interaction term for the compared rates. In this comparison, musculoskeletal disorder claims were compared to non-musculoskeletal disorder claims and non-traumatic soft tissue disorder claims were compared to all claims excluding non-traumatic disorders. The different categories within non-traumatic soft tissue disorders were compared to all claims excluding non-traumatic soft tissue disorders.
industries for intervention purposes, frequencies of claims within an
industry as well as the relative risk compared to all industries are important
considerations. We combined the rank orders of both frequency and relative
risk to create a “Prevention Index” (PI)vi
PI=[Frequency Rank + Incidence Rank].