Further test of a census approach to compiling data on fatal work injuries.

"There is no single agreed upon estimate of the number of occupational fatalities in the United States." This startling conclusion from the National Research Council's Panel on Occupational Safety and Health Statistics underscores a gap in safety statistics.(1) Annual estimates of work-related deaths in the United States vary widely, ranging from about 3,500 to 10,000.(2)

In 1988, the Bureau of Labor Statistics and the Texas Department of Health compiled data on fatal occupational injuries occurring in 1986.(3) Four sources were used: death certificates, workers' compensation reports, Occupational Safety and Health Administration fatality reports, and survey forms submitted in the BLS Annual Survey of Occupational Injuries and Illnesses. The study concluded that collecting occupational injury fatality data from multiple sources is feasible on a retrospective basis. The results also emphasized the need to substantiate the work relationship for fatalities in which only one source identifies the death as work related. The results of the study prompted the question: "Is a multiple source system of compiling fatal injury data feasible on a more timely basis?" Early identification of work hazards could lead to more timely prevention efforts. Also, timeliness is important in maximizing respondent recall to verify work relationship and in reducing the number of nonrespondents resulting from relocation.

In 1990, the Bureau contracted with the Texas Workers' Compensation Commission and the Colorado Department of Health to determine whether a more timely comprehensive count of fatal occupational injuries is feasible. Data on fatalities occurring in 1990 were collected over the April 1990-May 1991 period. This report summarizes the results of that test study determining the number of workplace fatalities in Texas and Colorado and the circumstances surrounding the incidents.

Test methods

Data collection began by first identifying administrative agencies within the States of Colorado and Texas responsible for investigating or maintaining records on work-related fatalities. Sources included death certificates, State workers' compensation claims, motor vehicle traffic accident fatality reports, and coroner or medical examiner reports. In addition, information on work-related fatalities was provided by Federal agencies, such as the Department of Labor's Occupational Safety and Health Administration, Mine Safety and Health Administration and Employment Standards Administration. (The Employment Standards Administration provided information on fatalities to civilian Federal employees and to longshore and harbor workers.)

Source documents were matched using the decedent's name and other identification to avoid duplication. Cases were reviewed for work relationship according to the Bureau of Labor Statistics Recordkeeping Guidelines for Occupational Injuries and Illnesses. For a fatality to be considered work related, the decedent must have been employed that is, working for pay or other compensation) at the time of the event. The criteria presume work relationship if the injury occurred on the employer's premises. Injuries occurring in the company parking lot or in company-owned recreational facilities are considered work related only if the worker was engaged in a work-related activity. Injuries occurring off the employer's premises are considered work related if the employee was engaged in a work-related activity, was present at the location as a condition of employment, or was in travel status and engaged in work or travel functions. Injuries occurring while traveling to and from work are not considered work related.

To further ensure an accurate count, each fatality was substantiated by either two or more independent source documents or a source document and a followup questionnaire.(4) (See table 1.)

A followup questionnaire was mailed to the employer or other contact (excluding next-of-kin) who may have knowledge about the circumstances of the incident. The questionnaire also was used to collect necessary information missing from the source document. Nonresponse to the questionnaire and/or inconsistent data entries resulted in further followup by tele one.

For each fatality case, approximately 50 data elements were coded describing the demographic characteristics of the decedent and the circumstances surrounding the event. Demographic characteristics include the decedent's age, sex, race or Hispanic origin, and occupation. Data on circumstances surrounding the event include industry and size of establishment, date of injury, type of incident, location of the incident, and objects or machinery involved. The data record for each case also contained various control information, such as source documents received for a case, the date source documents were received, and if a followup was conducted.

While the purpose of the study was to compile a complete count of workplace injury fatalities in Colorado and Texas, these two States also collected information on fatal occupational illnesses that were recognized as work related or that occurred at work. (There is not necessarily a causal relationship implied for deaths resulting from illnesses occurring at work.) A comprehensive count of fatalities resulting from occupational illnesses cannot be produced from current data largely because of the latency period associated with many occupational illnesses, as well as the difficulty in associating illnesses with exposures in the workplace. However, the illness data provide valuable insights on prolonged exposures in the workplace that may lead to premature death.(5)

Proflie of fatalities

Motor vehicle accidents and homicides were leading causes of work injury deaths in both Colorado and Texas during 1990. (See table 2.) Trucking accounted for a large proportion of the motor vehicle accidents. Homicides occurred most often to executives and other managers and to sales and service workers. Food stores and restaurants were the establishments primarily affected.

Overall, about a third of the total injury fatalities occurred in the construction and transportation industries. (See table 3.) Agriculture, forestry, and fishing and services also accounted for a large proportion of the fatalities in Colorado. Similarly, clustering is evident when one looks at major occupational groups of those who were fatally injured at work. Precision production and craft workers (which include construction trades workers); transportation and material moving workers including truck drivers); and handlers, helpers, and laborers (including those in construction) accounted for a majority of the fatalities. In addition, Colorado had a large percentage of fatalities in the farming, forestry, and fishing occupation.

In Texas, a relatively large proportion of the occupational injury fatalities occurred among Hispanics, who account for a large proportion of the workers in that State.(6)

While focusing on fatal injuries, the Bureau also attempted to count the number of heart attacks occurring at work that did not result from a traumatic incident (such as a fall or motor vehicle accident). Records showed such a roster to be less than complete because administrative records under-count heart attacks and other illnesses that are difficult to relate to the job. An industry distribution of heart attacks which occurred in 1990 at the workplace was reported. (See table 6.) Although not based on a census, such fatal illness data can be useful in studying the full spectrum of workers who die on the job.

Program expansion

The pilot study demonstrates that it is feasible to collect current occupational injury fatality data using multiple source documents. In fact, multiple data sources are essential in compiling a complete count of occupational injury fatalities. The total occupational injury fatality counts of 74 in Colorado and 508 in Texas differ from counts compiled by a system which uses a single source system.(7) In Colorado, the State's vital statistics system is both timely and comprehensive, while in Texas, the State workers' compensation system is timely as an initiating source, but is not as comprehensive as the State's vital records system.(8)

Texas made extensive use of the followup questionnaire to solicit missing information about the fatality and to verify that the fatality was work related. Both States used newspaper articles, which often provide demographic information about the decedent as well as circumstances about the fatal events, to further substantiate that a fatality was work related.

While data were being collected for the pilot study, BLS and the two States met to discuss various issues. Consensus was that the methods used in the pilot study were effective in collecting timely occupational fatality information. It was recommended that the program be implemented nationwide, that the parties pursue additional sources of occupational fatality information, and that additional data elements useful to accident prevention be collected.

In February 1991, the Census of Fatal Occupational Injuries was established as an ongoing data collection program. Its purpose is to develop a framework for improving the validity, usefulness, and efficiency of compiling a database of work-related injury fatalities.

Currently, the Bureau has made cooperative agreements with 33 States to conduct data collection in calendar year 1991. Data from these participating States should be available by mid- 1992. It is anticipated that by 1992, the remaining States will join in compiling a data base to establish a nationwide Census of Fatal Occupational Injuries program.

The following information concerning injury fatalities was added to the 1991 program: worker activity at the time of the fatal event; decedent's tenure in the occupation; decedent's usual lifetime industry and occupation; results of any toxicology tests that were performed on the decedent; and worker impairments such as language barriers or hearing impairments.

Information on workplace fatalities was requested from additional Federal agencies, for example, the Department of Transportation and the National Transportation Safety Board.

In January 1991, participating States were trained and provided with manuals to standardize the coding of information on work-related fatalities, which will facilitate the analysis of data among States. A quality assurance program, including a procedure for the review of data coding, was implemented to ensure consistency among States. These quality control measures, combined with the participation of the remaining States in 1992, will enable the Census of Fatal Occupational Injuries program to provide the detailed and accurate data needed to develop strategies for preventing fatal workplace injuries.

Footnotes

ACKNOWLEDGMENT: The authors thank the staffs of the Texas Workers' Compensation Commission and the Colorado Department of Health for compiling the 1990 data and for their invaluable assistance in refining data collection procedures.

1 See the National Research Council's Counting Injuries and Illnesses in the Workplace: Proposals for a Better System (Washington, National Academy Press, 1987).

2 Fatality figures from three agencies are compared: the Bureau of Labor Statistics (BLS), the National Institute for Occupational Safety and Health (NIOSH), and the National Safety Council (NSC). The BLS Annual Survey of Occupational Injuries and Illnesses estimated that there were 3,600 work-related fatalities during 1989. The National Safety Council estimated 10,100 work-related fatalities for that year, while NIOSH's National Traumatic Occupational Fatality estimate was 6,400 for 1985, the latest year for which those data are available.

Besides estimates for occupational fatalities, the BLS annual survey produces national estimates of nonfatal occupational injuries and illnesses from a sample of about 280,000 private sector establishments. Estimates of occupational fatalities derived from the annual survey data are lower than other estimates because of various exclusions from survey coverage: the self-employed, public sector employees, and employees of private households. Annual survey fatality estimates also exclude fatalities for workers in establishments with fewer than 11 employees. Deriving figures for occupational fatalities--a relatively rare event--from a sample survey may also play a part in the low figures. See Occupational Injuries and Illnesses in the United States by Industry, 1989, Bulletin 2379 (Bureau of Labor Statistics, 1991 .

National Safety Council estimates cover unintentional (homicides and suicides are excluded) injury deaths of persons in the civilian work force, 14 years and older, with the exception of private household workers. See National Safety Council, Accident Facts: 1989 Edition.

The NIOSH estimates cover traumatic injuries (intentional and unintentional) of persons 16 years and older identified on the death certificate as occurring "at work." See National institute for Occupational Safety and Health, National Traumatic Occupational Fatalities: 1980-1985, March 1989.

3 See Janice Windau and Donna Goodrich, "Testing a census approach to compiling data on fatal work injuries," Monthly, Labor Review, December 1990, pp. 47-49. Previous studies have shown that some traumatic fatalities identified by a single source document are nonwork related.

4 See Windau and Goodrich, "Testing a census approach."

5 In this study, an occupational injury is defined as "any injury such as a cut, fracture, sprain, amputation, etc., which results from a work accident or from a single instantaneous exposure in the work environment." An occupational illness is "any abnormal condition or disorder, other than one resulting from an occupational injury (that is, resulting from an instantaneous event), caused by exposure to environmental factors associated with employment." With the exception of table 6, tables exclude occupational fatalities due to illness. Data for these fatalities are incomplete due to the difficulty of determining their association with workplace exposures. See Harvey J. Hilaski, "Understanding statistics on occupational illnesses," Monthly Labor Review, March 1981,pp.25-29.

6 Both the Colorado and Texas death certificate forms include information on whether the decedent was of Hispanic origin.

7 Data are based on the State where the incident occurred. Eight fatalities were excluded from the tables because the incident resulting in death occurred outside the State. Also, there may be some work-related fatalities omitted because the death occurred in another State. In a national program, this omission would not occur.

8 State workers' compensation reports fail to capture a census of fatal occupational injuries occurring in the State because of incomplete coverage of workers. State workers' compensation programs generally exclude injuries and illnesses to the self-employed, employees of small farms, Federal Government employees, private household and seasonal employees, longshore and harbor workers, and railroad employees.

While death certificates cover all deaths occurring in the State, only those death certificates identifying a fatal injury as work related (that is, with the "injury at work" box marked "yes") were used as an initiating source in the study. Because there is not a standard definition of work relationship in the vital statistics program, some fatalities, particularly those resulting from motor vehicle accidents, may not be considered work related by persons completing the death certificate.

See Nancy Stout and Catherine Bell, "Effectiveness of Source Documents for Identifying Fatal Occupational injuries: A Synthesis of Studies," American Journal of Public Health, June 1991, pp. 725-28.