Volume 23
Number 1
2002

[Table of Contents]

Estimating the population at risk for Ontario Workplace Safety and Insurance Board-covered injuries or diseases

Abstract

Difficulty in quantifying the population at risk for a work-related injury or disease limits the usefulness of workers’ compensation data for surveillance. This article presents a method of obtaining estimates of the Ontario Workplace Safety and Insurance Board (OWSIB)-covered workforce using the Canadian Labour Force Survey (LFS). The method involves extracting that class of worker most likely to be insured by the OWSIB and using actual hours worked to estimate full-time equivalents at risk. Compared to population at risk estimates readily available from published tables, the refined crude estimate was 26% lower and ranged from 15 to 79% lower depending on the age group. The percentage decrease from published estimates was generally greater for women compared to men, particularly in the 25 to 39 year age categories. Consequently, the method of deriving population at risk estimates should be considered when comparing rates across sexes, ages, industries or occupations.

Key words: denominators, Labour Force Survey, Ontario Workplace Safety and Insurance Board coverage, population at risk, work-related injury or disease rates

Introduction

Statement of Problem

A major limitation to using workers’ compensation data for the surveillance of work-related injuries or diseases is the difficulty in generating denominators for the calculation of accurate rates.^1–4 In Ontario, the workers’ compensation system is funded by premiums paid by employers. The premium is dependent on the nature of the business, the employer’s health and safety record, and the size of the payroll, not the number of full-time equivalent workers to be insured.⁵ For these reasons, an estimate of the insured population in Ontario is not readily available. In an attempt to produce an estimate, some researchers have relied on government census data, which overestimates the number of full-time equivalent workers at risk because the defined employed population includes full-time, part-time and temporary workers and workers not covered by workers’ compensation.³

In Canada, it is estimated that 20 to 30% of the workforce is not covered by workers’ compensation.⁶ Furthermore, since women are more likely than men to hold part-time or temporary positions, a greater overestimation of the at-risk population is likely for women relative to men.³ Other investigators^1,7,8 have used Statistics Canada’s Labour Force Survey (LFS) data. For example, Ashbury¹ used LFS published estimates of the employed population of Ontario, which are overestimates because they include full- and part-time workers, unpaid family workers and workers not covered by the Ontario Workplace Safety and Insurance Board (OWSIB).

Brooker et al.⁷ and Rael et al.⁸ improved upon Ashbury’s¹ method by using the LFS to obtain estimates for employed, paid workers. These estimates would exclude unincorporated business owners and unpaid family help, the former not likely to be insured and the latter definitely not insured by the OWSIB. Although the authors did not detail the mechanics of extracting this class of worker, it is clear that the actual hours worked by employed, paid workers was not utilized in quantifying the at-risk population. Consequently, an employed, part-time, paid worker would contribute the equivalent of an employed, full-time, paid worker to the population at risk estimate, producing an overestimate.

Rael⁹ found injury rates calculated using employed, paid workers were equivalent to those calculated using employed, paid hours to derive the employed, paid workers. However, Rael’s⁹ population at risk estimates were for males aged 15 to 64 in the Ontario construction industry, a group not likely to contain many part-time, paid workers. Thus, this equivalency may not be consistent across sex, industry or occupation and suggests that a method that adjusts for actual hours worked by employed, paid workers would be more appropriate for most applications.

Relevant Background Information

Canadian Labour Force Survey

The Canadian LFS is a monthly household survey that utilizes a multilevel sampling strategy to collect labour market activity information on those 15 years of age or older. Since July 1995, the national sample size has been 52,350 households, of which 15,858 are in Ontario. Data is collected every month in the week following the reference week, which is defined as the entire calendar week, Sunday to Saturday, usually containing the 15^th day of the month. Specifically excluded from the survey’s coverage are residents of the Yukon and Northwest Territories, Aboriginal people living on reserves, full-time members of the Canadian Armed Forces and inmates of penal institutions. These groups together represent an exclusion of approximately 2% of the Canadian population aged 15 or over. ^10,11 The LFS is a large, regular survey which, via its multistage sampling strategy and extensive quality control measures, is the most readily available valid and reliable source of information on the working age population of Canada.

Ontario Workplace Safety and Insurance Board Coverage

Although most businesses in Ontario that employ workers, including family members and sub-contractors, must register with the OWSIB within 10 days of hiring their first full- or part-time worker, registration is voluntary for banks, trust companies, insurance companies and other financial institutions; computer programmers; private healthcare practices, such as those of doctors and chiropractors; veterinary work; dentistry; law offices; trade unions; private daycare establishments; travel agencies; recreational and social clubs, such as golf or health clubs; educational and recreational camps; churches; theatres with live performances; broadcasting stations; motion picture productions; photographers; barbers, hair salons and shoeshine stands; taxidermy; and funeral directing and embalming. An employer not mandatorily covered is almost always permitted to apply for coverage, but the OWSIB may apply conditions. Sole proprietors, independent operators, partners and executive officers are not required to have personal coverage.⁵ Consequently, the employed labour force is not an accurate estimate of the population at risk of an OWSIB-covered injury or disease.

Purpose of Present Research

The purpose of this research is twofold. First, to detail how the LFS can be used to improve the accuracy of the population at risk estimates needed to generate crude and specific rates of OWSIB-covered injuries or diseases. Second, to demonstrate the degree to which population at risk estimates vary depending on the method of derivation from the LFS.

Methods

Statistics Canada provides a public use microdata LFS file for those wishing to undertake their own analyses. The files for the 12 months in 1997 were obtained and data for the province of Ontario was extracted. First, the employed labour force was calculated. This is an estimate that has been used previously¹ due to its availability from regularly published Statistics Canada tables. It will be used as a baseline for comparison with more refined methods.

To estimate the employed labour force in Ontario for 1997, the frequency of “labour force status” values equal to “employed, at work” or “employed, absent from work” were calculated for each of the 12 months using the final weights provided. These 12 monthly estimates were then averaged to produce an annual estimate for the employed labour force. This employed labour force estimate includes the self-employed as well as employees; full-time as well as part-time employed; unpaid family workers; and the employed who were not at work during the reference week due to factors such as illness or disability, personal or family responsibilities, vacation, or labour dispute.¹¹ To produce a measure that would reflect the actual hours worked, an annual estimate of the employed, full-time equivalents was calculated. The actual hours worked per week at all jobs by the employed labour force was calculated for each of the 12 months using the “actual hours per week at all jobs” variable and the final weights provided. These 12 estimates were then averaged to produce an annual estimate of the actual hours worked per week at all jobs by the employed labour force. This annual estimate for the employed labour force was multiplied by 52 weeks and divided by 2,000 hours (assuming a 40 hour work week for 50 weeks out of the year) to produce an annual estimate of the employed, full-time equivalents for Ontario during 1997. This estimate includes the self-employed, who are not automatically covered by the OWSIB; unpaid family workers who are not covered by the OWSIB; and employees, that is, individuals who collect wages or salary and are usually covered by the OWSIB.

To remove the self-employed and unpaid family workers from the annual estimates of the employed labour force and employed, full-time equivalents, the above procedures were repeated after using the “class of worker, main job” variable to extract public and private employees from the employed. To examine the effect of sex and age on the variability of the estimates, sex and age-specific annual estimates of the employed labour force and employed, full-time equivalents were calculated using the “sex” and “age group” variables to extract the appropriate data.

Results

Table 1 presents age-specific employed and employed, full-time equivalent annual estimates for the labour force as a whole and the employee subgroup. After extracting employees from the employed labour force and using the “actual hours per week at all jobs” to calculate full-time equivalents, the crude annual estimate of the at-risk population insured by the OWSIB, 4,014,181 employee full-time equivalents, was 26% lower than the employed labour force value of 5,412,868 employed persons (Table 1) readily available from published tables. This difference ranged from a low of 15% in the 25 to 29 age group to a high of 79% in the 70 plus age group. The sex-specific data (Tables 2 and 3) demonstrated similar trends but the percentage difference between the employed labour force and employee, full-time equivalent estimate was always greater for women except in two age categories: 60 to 64 and 70 plus years, where the sex differences in the percentage change were minimal. The female to male percentage change ratio was greatest in the 25 to 39 year age categories, ranging from 1.93 to 2.44.

Discussion

This research supports concerns about the overestimation of the full-time equivalent workforce at risk for an OWSIB-covered injury or disease and the differential overestimation in women relative to men, which can occur with the use of data readily available in published tables.^1,3

Limitations in Method of Estimation

Although the method presented above attempts to make estimates more accurate, these refined estimates have limitations. First, although most businesses in Ontario that employ workers must register within 10 days of hiring their first full- or part-time worker, registration is voluntary for some.⁵ The extent to which these businesses voluntarily choose to insure themselves is not known. If the tendency is low, even the refined population at risk estimates will be excessive, particularly in certain industry or occupational groups. It was not possible to produce reasonable lower limits on the population at risk estimates specifically examined (Tables 1–3) by removing those businesses in Ontario for which registration is voluntary due to the crude method that the LFS uses to code industries and occupations. For example, veterinarians are classified in the LFS industry code “agriculture” which includes all agricultural and related services such as livestock farms; other animal specialty farms; field crops; horticultural specialities; combination farms; and services incidental to agriculture, where veterinary services would be classified.¹² Hence, attempting to remove veterinarians would remove many others not exempt from coverage.

Second, the variable “actual hours per week at all jobs” has a 99-hour limit. Consequently, those working greater than 99 hours per week would not have their additional hours included in the full-time equivalents at-risk estimate. Since the percentage of employees with “actual hours per week at all jobs” greater than or equal to 99 hours was 0.09 for 1997, this limit will be negligible. The final limitation is the error introduced by multiple job holders. Since the “class of worker” is based on the “main job” those employees who are self-employed outside of their main job would inappropriately have these additionally worked hours added to their employee hours worked. Conversely, those who have main jobs classified as self-employed or unpaid family work, but have secondary jobs as public or private employees, would not have their employee hours worked included in the employee, full-time equivalents at risk estimate. Since only 4.9% of the employed labour force in Ontario during 1997 were multiple job holders and the main job accounted for 98.3% of the actual hours worked, it is likely that this error will have a negligible effect on subsequent calculated rates.

Limitations of the Canadian Labour Force Survey

The LFS assumption that the reference week is representative of the whole month has been challenged.¹³ Webber¹³ was concerned that annual estimates of worked hours may be biased if labour disputes or public holidays occurred disproportionately in the 12 reference weeks relative to the remaining 40 weeks in a year. For example, certain important public holidays, such as Thanksgiving, Good Friday, Easter Monday and Remembrance Day, can fall in the LFS reference week. Conversely, the mid-month location of the reference week precludes the remaining important public holidays from ever falling in that time period. For this reason, the LFS estimate will sometimes exceed the true weekly average of worked hours for the month, and sometimes will be less. To examine the combined effect of labour disputes and public holidays, Webber¹³ used statistics published by Labour Canada on person-days lost through work stoppages by month, data from the Pay Research Bureau, and data from the LFS on the magnitude of hours lost due to holiday. The annual actual hours worked estimates, adjusted for labour dispute and holiday effects, ranged from 1.5% below to 1.5% above the unadjusted estimates. Webber¹³ concluded that the unadjusted survey results could be preferred on the basis of ease of calculation and the small effect of adjustment relative to the errors inherent in the original data, but for data users interested in measures of year-to-year changes in aggregate annual actual hours worked, the adjustments have a substantial impact.

Summary and Conclusions

This research provides evidence that regularly published labour force survey estimates overestimate the population at risk for an OWSIB-covered injury or disease. The degree of overestimation was demonstrated to vary with sex and age and certainly varies across occupations and industries. A method was presented to obtain more accurate estimates of the at-risk population. This method extracts those employed individuals who are most likely to be insured by the OWSIB and uses the actual hours worked to estimate full-time equivalents at risk. Although there is no gold standard to establish the veracity of the derived estimates, certainly rate comparisons across age groups and sexes within industry and occupational groups would be more valid if discrepancies in the actual hours worked between the sexes and age categories were acknowledged in the population at risk estimates. In conclusion, when utilizing the LFS to derive estimates of the population at risk over time, any changes in the methods employed by the LFS, the OWSIB policy on mandatory coverage, or the tendency towards voluntary registration should be acknowledged.

References

1. Ashbury F. Occupational repetitive strain injuries and gender in Ontario. JOEM 1995; 37[4]:479–485.

2. Franklin G, Haug J, Heyer N, Checkoway H, Peck N. Occupational carpal tunnel syndrome in Washington state, 1984–1988. Am J Public Health 1991; 81:741–746.

3. Sprout J. The gender differences in upper-extremity occupational repetitive strain injuries in Manitoba. Winnipeg: University of Manitoba; 1997.

4. Yassi A, Sprout J, Tate R. Upper limb repetitive strain injuries in Manitoba. Am J Ind Med 1996; 30:461–472.

5. Garth D. Workers’ Compensation in Ontario Handbook. Toronto: Butterworths; 1999.

6. Manga P, Broyles R, Reschenthaler G. Occupational health and safety issues and alternatives. Ottawa: Economic Council of Canada; 1981.

7. Brooker A, Frank J, Tarasuk V. Back Pain Claim Rates and the Business Cycle. Soc Sci Med 1997; 45(3):429–439.

8. Rael E, Badley E, Frank J, Shannon H. Using labour force survey and census data to generate denominators for occupational injury rates: an application and expansion of Haggar-Guenette’s method. Chronic Dis Can 1996; 17(3/4):87–91.

9. Rael E. An epidemiological study of the incidence and duration of compensated lost time occupational injury for construction workmen, Ontario, 1989: An assessment and application of Workers’ Compensation Board and Labour Force data. Toronto: University of Toronto; 1992.

10. Statistics Canada. Methodology of the Canadian Labour Force Survey. [71-526-XPB]. Ottawa: Minister of Industry; 1998.

11. Statistics Canada. Guide to the Labour Force Survey: January 1997. Ottawa: Statistics Canada; 1997.

12. Statistics Canada Standards Division. Standard Industrial Classification 1980.Ottawa: Statistics Canada, Standards Division; 1980.

13. Webber M. Estimating total annual hours worked from the Canadian Labour Force Survey. Ottawa: Statistics Canada; 1983.

Author References

Dianne Zakaria, James Robertson and John Koval Department of Epidemiology and Biostatistics, University of Western Ontario

Joy C MacDermid, Hand and Upper Limb Center, St. Joseph’s Health Center, London, Ontario

Kathleen Hartford, Lawson Health Research Institute and Department of Epidemiology and Biostatistics, University of Western Ontario

Correspondence: Dianne Zakaria, Department of Epidemiology and Biostatistics, University of Western Ontario, Kresge Building, London, Ontario, N6A 5C1; Fax: (519) 661-3766; E-mail: diannez@biostats.uwo.ca

[Previous] [Table of Contents] [Next]