Survey Structure and Sampling Strategy

As interest in the importance of work and musculoskeletal disorders has grown over the past two decades, the most common data reference has been the Annual Report on Injuries and Illnesses provided by the Bureau of Labor Statistics. Since 1982, the source for these data has been a national sample of private industry reports of injury and illness according to the requirements of the Occupational Safety and Health Administration (OSHA Form 200). This source is therefore limited to employment in the private sector and includes only cases that have been determined to be work-related by the employer, using definitions provided by OSHA. Annually, a stratified random sample of establishments is selected to provide data. (For the mining workforce, the Mine Safety and Health Administration requires that full details of all work-related injuries and illnesses be reported to BLS. Currently, the sample of nonmining establishments is approximately 165,000, significantly reduced from the sample size of 250,000 establishments selected for the annual survey a decade ago).

When an employer is selected to be part of an annual sample, two-part forms are provided for recording and submitting the necessary summary information about occupational injuries and illnesses. In Part I (Summary of Occupational Injury and Illnesses), the employer is instructed to provide information from personnel records and from the Log and Summary of Occupational Injury and Illnesses (OSHA 200) that reports: (a) the size of the establishment, (b) the number of hours worked during the year, (c) the total number of injuries separated into those with and without lost workdays, (d) the total types of illnesses (using OSHA's classification of skin diseases or disorders, dust diseases of the lungs, respiratory conditions due to toxic agents, poisoning, disorders due to physical agents, disorders associated with repeated trauma, and all other occupational illnesses), and (e) the number of illnesses separated into those with and without lost workdays. At the level of this survey, BLS is able to estimate total employment by industry or by occupation, thus permitting the calculation of injury and illness incidence rates. On the basis of reports in Part I, musculoskeletal disorders are identified only by category of illness or “disorders associated with repeated trauma.” For the guidance of those completing the OSHA log and summary, this category is described with the following examples: noise-induced hearing loss; synovitis, tenosynovitis, and bursitis; Raynaud's phenomena; and other conditions due to repeated motion, vibration, or pressure. Thus, while the category includes a case of hearing loss, it is unlikely to include a case of back strain or pain from overexertion.

In Part II (Reporting of Cases with Days Away from Work), each establishment covered by OSHA is required to maintain detailed information about each injury and illness on the form entitled Supplemental Record of Occupational Injury and Illnesses (OSHA 101). Since 1992, BLS has requested that each establishment included in the annual sample provide information from this supplemental record for cases reported to result in at least one day away from work. The Mine Safety and Health Administration requires all mines to report the full details of all cases. Information extracted from this supplemental form provides the following details about the lost-time cases: date of event, days away from work, days of restricted work activity, length of service, race, age, gender, and occupation. Three additional details are provided in a descriptive or narrative form: event (what the employee was doing and how the injury occurred); nature (what the injury or illness was), and source (what object or substance directly harmed the employee). These details are then coded according to a BLS-developed system that allows better focus on musculoskeletal disorders, although it is not based on the standard ICD categories. Participating establishments take information from workers' compensation reports, insurance forms, or other supplementary records to complete this second survey.

While all establishments are required to maintain the details used to complete Part II of the BLS survey—with the exception of requirements of the Mine Safety and Health Administration for mine operators to report the added details on all injuries and illnesses—BLS requests the additional details only on events that are associated with lost time. Therefore, events that result only in restricted work activity or no restrictions are not sampled for the available greater detail.

Lost-workday events included in the Part II sample account for approximately 25 percent of the reported conditions. The added detail available from these reports allows more refined specification of illnesses and injuries. For example, event includes information that allows classification into categories relevant to musculoskeletal disorders, such as overexertion or repetitive trauma, and nature includes relevant categories such as sprain/strain/tears, carpal tunnel syndrome, tendinitis, back pain, and other pain. If the data OSHA requires for all injuries and illnesses were reported to BLS, the data system would be much more robust and would allow a better understanding of these conditions.

Survey Results

BLS injury and illness data from Part I of the annual summary show an increase in all reported illnesses between 1982 and 1994, rising from an estimated 105,600 reports in 1982 to 514,700 in 1992. This trend was followed by a small decline over the past four years, with the total reports in 1998 estimated to be 392,000. No appropriate denominator for these numbers is available to calculate rates based on the current reporting scheme. The subset of conditions described as “disorders associated with repeated trauma” showed a similar but much more striking trend, with an estimated 22,600 reports in 1982 rising to 332,100 in 1992; the most recent estimate is 253,300. Events attributed to repeated trauma have therefore risen from 21 to 65 percent of all reported illnesses in the portion of the private sector sampled by BLS.

BLS examined the recent decline in overall occupational injury and illness rates in an effort to understand these trends (Conway and Svenson, 1998). Although the examination did not address musculoskeletal disorders specifically, the general observations probably apply to these disorders as well. The BLS study speculated that the downward trend was not due to a shift in employment from high- to low-hazard industries, nor an increase in underreporting, but rather from better recognition of the causes of occupational injuries and illnesses by all parties involved. This led to a nationwide adoption and implementation of safety and health programs that contain accident-related costs.

BLS injury and illness data from Part II are available only since 1992. During this period there has been a gradual decline in the total number of lost-time injuries and illnesses. Using the information on the nature of the injury associated with the different body parts allows a more refined way to examine trends in musculoskeletal disorders than is possible from Part I data. The relevant nature of injury designations that may be related to musculoskeletal disorders are sprains and strains, carpal tunnel syndrome, tendinitis, and soreness/pain.

In 1997, Part I data were used to estimate the total number of illnesses related to repeated trauma as 276,600. In the same year, the added information provided in the Part II dataset for conditions related to musculoskeletal disorders provided an estimate of 846,000 lost-workday cases. The large majority of these conditions occur in the neck, shoulder, upper extremity, and back. Note that this number counts only the events that occurred in 25 percent of the cases—those reporting lost work time. In 1997, only 1.8 million of the 7.1 million total injuries and illnesses were reported as lost-workday cases in Part II.

Some differences exist by industry when the results of Part I and Part II are compared. For the Part I subcategory of illnesses that are identified as due to repeated trauma, the largest proportion of reports come from manufacturing (72 percent), with service (10 percent) and retail trade (6 percent) the two next largest. When the same data are examined by incidence rates, manufacturing has three times the total industry rate (106/ 10,000 full-time workers versus 32/10,000). It is difficult to compare these results with those available from the Part II data because the categories are different. However, among the categories of conditions that could be related to musculoskeletal disorders, the proportional distribution of events changes substantially. Manufacturing is responsible for only 22 percent of sprains/strains, carpal tunnel syndrome, or tendinitis, while the service industry accounts for 26 percent. When the incidence rates for carpal tunnel syndrome alone are examined (the only specific musculoskeletal disorder), manufacturing still leads the other sectors showing a twofold excess over all private industry (6.5/10,000 versus 3.4/1,000), but both the transportation and finance sectors also exceed the overall average rates. Among the musculoskeletal disorder-related categories, the most common, but less specific, sprains and strains have the highest overall incidence rates (92.5/1,000); but now the transportation sector has the highest sector-specific rate (191/1,000) with construction, mining, agriculture, and wholesale trade all showing rates higher than overall industry. These data suggest that there is no single industrial sector in which musculoskeletal disorders are a unique problem and, furthermore, that the problem is not limited to the traditional environments of heavy labor represented by agriculture, mining, and manufacturing.

The Part II detailed information on each lost-time case is particularly important, as it permits much better examination of trends and the association of specific risk factors with occupation. Some examples of the possible use of the detailed data are provided to illustrate the potential of the added detail.

1.

Three conditions—carpal tunnel syndrome, tendinitis of the upper extremity, and hurt back/back pain—show age-specific time trends that deserve attention. For example, in 1996, median days away from work for female typists is 50 percent longer for younger (ages 35 to 44) than for older typists (ages 55 to 64). The same direction in age difference was seen for receptionists (median 5 times longer for younger versus older workers), but billing clerks showed the opposite trend. Median days away from work were 10 times longer for the older age group compared with the younger.

2.

Details about type of condition in 1997 provide further insights. Injuries or illnesses attributed to repetitive motion were examined. The proportion of repetitive motion events attributed to typing or key entry was determined for three conditions: carpal tunnel syndrome, tendinitis, and sprains or strains (the only conditions common enough not to be suppressed). According to BLS, repetitive motion accounts for 98 percent of all carpal tunnel syndrome, 63 percent of tendinitis, and only 2 percent of sprains and strains. Among the subset of these three conditions attributed to repetitive motion, almost one-quarter of all carpal tunnel syndrome cases were associated with typing or key entry, compared with only 13 percent of tendinitis and 5 percent of sprains and strains.

3.

Trends between 1992 and 1997 deserve further investigation. During this period, when all injury and illness reports decreased, sprains, strains, and tears of the wrist decreased by 30 percent, while carpal tunnel reports decreased by only 13 percent. When examined by industrial sector, carpal tunnel rates in the manufacturing sector decreased from 9.7 to 6.5 per 10,000 full-time employed workers, while rates in transportation and public utilities increased from 3.1 to 3.6. Differing trends were also observed for sprains, strains, and tears between 1992 and 1997. In the agriculture sector, incidence rates were significantly reduced (from 183.2 to 97.6 per 10,000 full-time employed workers), while the rates for transportation and public utilities remained fairly stable (from 211.2 to 190.8 per 10,000 full-time employed workers). Explanations for these different trends should be undertaken.

An example of a less obvious value derived from these data is provided in a BLS study examining the changing composition of lost-workday cases (Ruser, 1999). The Part I injury and illness reports between 1976 and 1997 show the annual incidence of lost-workday cases has not changed substantially. However, there is a trend toward decreasing days off the job and an increasing use of restricted work activity assignments. Even though the Part II data have been collected by BLS only since 1992, the 1992 to 1996 data provide important insights not available from the Part I data. The Part II data show a slight reduction in the number of days away from work coupled with a growing use of restricted activity days and a substantial increase in the proportion of cases with lost workdays accompanied by restricted workdays. Furthermore, the Part II data allow examination of the trends by the nature of the injury. They show the highest percentage increase in restricted work activity days among those with carpal tunnel syndrome. The increased fraction was from 31 to 47 percent of lost-workday cases between 1992 and 1996. Considerable value would come from being able to examine all injuries and illnesses using the added detail available from the Part II survey data already routinely collected by employers but not reported to the BLS.

Survey Structure and Sampling

In 1988, the National Institute for Occupational Safety and Health (NIOSH) provided support to the National Center for Health Statistics (NCHS) to undertake a supplement to its annual survey that collected detailed information on interviewees' work histories and on self-reported health conditions of the back and hand. The data available from this source have been examined in order to estimate the work-relatedness of selected musculoskeletal disorders of the back and hand (specifically, carpal tunnel syndrome). The National Health Interview Survey (NHIS) is a continuous personal interview survey conducted by NCHS on a probability sample of the households of noninstitutionalized civilians. In 1988, interviews were completed on 44,000 individuals, 30,000 of whom were working. The detail available from this survey provides a much richer resource than BLS data to understand the relationship of work to health conditions, but, at the same time, the sampled workforce is much smaller than that represented by the BLS data. As a result, the level of detail about industry or job-specific risk is not as great.

NIOSH scientists have taken primary responsibility for the analysis and publication of results from this special survey related to musculoskeletal disorders, primarily back pain and carpal tunnel syndrome. Recently, tendinitis and related conditions as well as arthritis have been examined for their relationship to work.

In order to examine the reports of back pain that are likely to reflect musculoskeletal disorders of the back, analysts examined responses of interviewees who reported back pain occurring for at least a week in the past 12 months. These analyses were carried out for all back pain as well as for a subset defined as work-related back pain (back pain brought on by repeated activities or resulting from a single accident or injury at work). For those with report of hand discomfort, subjects could volunteer their sense of the cause of discomfort.

Survey Results for Back Pain

The NHIS survey data for those who reported to have worked in the past 12 months were used to estimate that 22.4 million people (prevalence of 17.6 percent) suffered from back pain, without regard to whether it was related to work. A small majority of those with back pain (56 percent) were males (Guo et al., 1995). When cause of back pain was sought, 78 percent reported that the pain was associated with repeated activity, a single accident or injury, or both, and two-thirds of those reported the activity or event occurred while at work. Risk of back pain by occupation was examined by comparing specific occupations to all workers. Among males, the highest-risk occupations—prevalence ratio >2.0—were construction laborers, carpenters, and industrial truck and tractor equipment operators. Among females, the highest-risk occupations with a prevalence ratio of >2.0 were nursing aides/orderlies/attendants, licensed practical nurses, maids, and janitor/cleaners. A literature review indicated that many of the highest-risk occupations were under study, but it was noted that several that had relative risks among the top 15 for gender were previously unrecognized occupations, including: carpenters, hairdressers, automobile mechanics, janitors, and maids, often employed in small businesses or self-employed.

In a separate study utilizing the same data set, NIOSH examined the subset of reports of back pain that were specifically attributed by the respondent to work. Severity was measured by recording the number of days away from work (Guo et al., 1999). With this more stringent definition, 5.6 million work-related back pain cases were estimated. In this analysis, NIOSH reported cases by industry rather than by occupation. Among males, the industries with a prevalence ratio of >2.0 were lumber and building material retailing, crude petroleum and natural gas extraction, and sawmills/planing mills/millwork. Among females, industries with a prevalence ratio of >2.0 were nursing and personal care facilities, beauty shops, and motor vehicle equipment manufacturing.

The estimated overall prevalence of cases of work-related back pain with at least one lost workday was 4.6 percent of the working population. Furthermore, the proportion of back pain cases reporting lost workdays was 50 percent higher among those who attributed their pain to work than among those who did not. The investigators grouped industries according to the most workdays lost (overall average of 9.2 days for work-related cases) and according to the highest number of workdays lost per lost-workday case. Nationally, back pain of any type was estimated to account for 149 million lost workdays, the majority of which (68 percent) were associated with work-related back pain. The industry with the most lost workdays was construction for men (1.76 million days) and elementary and secondary schools for women (760,000 days). Examining severity of the back pain cases among those with work-related pain, the industry with the highest average lost workdays was the electronic computing equipment industry (29 days/case) for men and the U.S. Postal Service (61 days/case) for women. Both sexes were ranked high for grocery stores, hospitals, banking, and eating and drinking establishments. However, there is substantial variation by gender. For example, in the U.S. Postal Service, men reported average lost workdays of only 1.9 days per case, one-thirtieth of that found among women.

Survey Results for Hand Discomfort

The survey questions about hand discomfort elicited information about carpal tunnel syndrome, tendinitis and related syndromes, and arthritis. Analyses of these self-reports indicate that annually among working adults:

1.

1.87 million report having carpal tunnel syndrome, of which over one-third report that their health care provider diagnosed their condition as carpal tunnel syndrome and half of these were believed to be work-related (Tanaka et al., 1995, 1997).

2.

588,000 report having tendinitis or related syndromes, of which 28 percent were labeled as work-related by a health care provider (Tanaka, Petersen, and Cameron, in press; Tanaka et al., 1997).

3.

Almost 2 million active or recent workers were estimated to have hand-wrist arthritis that caused a major change in work activities, jobs, or missed workdays among almost 20 percent (Dillon, Petersen, and Tanaka, in press, Tanaka et al., 1997).

The cases in which one or more of these three disorders confirmed by a health care practitioner were evaluated for associations with a number of risk factors, including age, race, gender, and work requiring repetitive bending and twisting. All three conditions were significantly more common among those whose work required bending and twisting of the hands and wrists (Dillon, Petersen, and Tanaka, in press; Tanaka, Petersen, and Cameron, in press; Tanaka et al., 1995, 1997).

Annual Survey of Occupational Injuries and Illnesses

The Annual Survey of Occupational Injuries and Illnesses survey is the only resource that currently provides ongoing, consistent reporting of injuries and illnesses that permits examination of trends over time. Therefore, it is the only national resource that can be used for the surveillance goal of tracking trends in injury and illness rates. Trend analysis is useful both to identify new risks and to evaluate efforts to reduce known risks. The survey has limitations that are well recognized.

Limitations of Part I

1.

The survey represents only private industry, excluding approximately 25 percent of the workforce, primarily government workers, those who are self-employed, those who work on small farms, and those employed in places with fewer than 10 individuals.

2.

In order to gain cooperation from the reporting establishments, BLS protects the confidentiality of results by not making them publicly available if (a) the estimates for the industry are based on too few reporting units, (b) average employment in the industry is too small (generally < 10,000), (c) statistical estimates do not meet minimum reliability criteria, or (d) there is any other way that publication might disclose confidential information.

3.

The sample size of 165,000 is too small to permit estimates at any level below the national level except for large aggregated industry groupings. This is not the case for the subcategory of mining, since the Mine Safety and Health Administration requires reporting of all injuries and illnesses, not just a sample of these events.

4.

The illness coding system is not ICD-based. Rather, it utilizes very crude collapsing of unlike conditions and is determined by a person with no specific training for the task.

5.

The demographic characteristics of individuals are not reported, and only industry, not occupation or task, is collected, limiting the description of the injuries and illnesses with adequate specificity.

Limitations of Part II

In addition to the first three limitations listed for Part I:

1.

The greater detail on each injury or illness that is provided in Part II is collected exclusively for lost-time injury or illness cases. OSHA, however, requires this level of detail to be recorded on all injuries and illnesses. BLS has neither the mandate nor the resources to collect and analyze the additional data, although such data would greatly enhance the usability of the database available for surveillance purposes.

2.

The more detailed nature and body part coding system is not ICD-based and the narrative is provided by an individual with no specific training for the task. Although the coding is done by trained BLS staff, the inputs are still not adequately standardized.

3.

While there are proper demographic characteristics associated with all reports as well as the occupation or task being performed, no comparable denominator data are collected. This prevents the calculation of rates for these better-specified conditions. A recent analysis of workplace fatality rates indicates that it is possible to address this problem (Ruser, 1998.)

4.

Data items are suppressed at a national level if the number of cases is fewer than five.

National Center for Health Statistics Surveys

These surveys provide a number of ways to characterize the health of the U.S. population. The different survey elements allow for the development of a comprehensive measurement of health as self-reported and as diagnosed by health care providers. When these reports are combined, the result is an excellent resource for use to describe the health of the workforce. There are some important limitations that could be corrected, however, as is seen in the NIOSH-funded 1988 supplement to the National Health Interview Survey:

Limitations

1.

For the interview surveys (National Health Interview Survey and the National Health and Nutrition Examination Survey), only the job title or industry is collected and then only for the preceding week, month, or 12 months. For the health care records (e.g., the National Ambulatory Medical Care Survey), no data are collected on job or industry.

2.

Despite the importance of musculoskeletal disorders in the population (both work-related and nonwork-related), the currently designed surveys collect little detail about these disorders. The added detail in the 1988 NHIS supplement presents a good example of how informative survey data can be when these disorders are better characterized.

Workers' Compensation Records and the Washington State Fund

The data available from the Washington State Fund are valuable because they have a well-defined population base to which they refer. However, while the Washington State Fund data are among the best, the occupation and industry distribution in one state does not provide an adequate national representation of either those occupations and industries and cannot provide information about occupations and industries not prevalent in Washington.

Data based on workers' compensation vary across states as each state applies differing definitions and criteria in determining the conditions that are compensable. The between-state differences are substantial, placing further limits on efforts to derive national estimates of burden from workers' compensation claims. There is evidence that such claims are an underestimate of work-related musculoskeletal disorders, suggesting some limitations of workers' compensation reporting for surveillance and for estimating the national burden of work-related musculoskeletal disorders (Biddle et al., 1998; Morse et al., 1998).

While some limitations of the system are administrative, other limitations should be noted. The decision to file a workers' compensation claim is determined by knowledge that the option exists (Sum, 1996). This varies by state and by the education or training of each worker and employer. A decision to file a claim also depends on how likely the claim is to be contested (Herbert, Janeway, and Schecter, 1999), workers' concerns about employer retribution if they file (Speiler, 1994; Pransky et al., 1999; Ochsner et al., 1998), and the alternatives available for payment of medical costs (medical insurance).

National Occupational Exposure Survey

The National Occupational Exposure Survey is an important national resource for hazard surveillance providing the data to describe the distribution of ergonomic risk factors by industry and occupation. The survey, however, was last carried out almost 20 years ago and is not likely to represent well the distribution of current exposures. It also does not provide exposure characterization of the risk factors present in emerging industrial sectors.

Medical Care Utilization Data

Data from medical care records could be particularly valuable in providing much greater detail about the nature, distribution, time course, and disability associated with musculoskeletal disorders. Such data, however, have important limitations:

1.

Medical confidentiality mechanisms need to be developed to allow use of these data for surveillance purposes and to protect the individual identity of those whose records are in the system.

2.

Although specific diagnostic information is recorded by ICD-9 coding, algorithms will need to be developed to ensure consistency in diagnostic practice across providers and determine reliability.

3.

Systems to match information on occupation or employer will need to be developed so that patterns of musculoskeletal disorders among different working groups can be examined.