The calculation of rates of occupational injury claims is essential to identify groups at high risk, yet limitations of denominator data have often restricted our capacity to do this. Haggar-Guénette's method of using Statistics Canada's data on paid workers from the Labour Force Survey as denominators has been expanded by incorporating information from the Census. The method is illustrated by calculating denominators for male construction industry workers within the province of Ontario. Information for paid workers employed in construction occupations was derived from the Labour Force Survey to produce denominators for those aged 15-64, overall and by 10-year age groups. Census data on the distribution of construction occupational roles were applied to produce denominators both overall, then simultaneously by age and by occupational role. Advantages and disadvantages, including the limitations or biases due to the differing sources for denominators and numerators are identified.

Key words: Canada; Census data; denominator calculation; Labour Force Survey; methods; occupational injury rates; Ontario

Introduction

It is not surprising that studies of occupational injury statistics often do not include any information on the population at risk of injury-denominator options can be complex and confusing. There are many sources of denominator data, and various classification systems are used to define the categories of workers.

The most common denominator data comes from routine monitoring of individuals or employers, often based on payroll or personnel records. Government departments require submission of this information from companies to produce statistics related to their responsibilities. For example, Labour Canada requests that firms with 20 or more employees provide the number of person-hours worked and/or the number of full-time employee equivalents, as well as injury information, to produce annual reports of injury statistics.¹ Intermittent national surveys such as the quinquennial Census and the monthly Labour Force Survey capture changing trends in population and labour force characteristics, respectively. Each of these sources of denominators may use different classification systems to define the categories, although Statistics Canada strives for consistency across surveys.

Classification systems used to define denominators may reflect the characteristics of workers and the basis of "exposure" to occupational risk: these include the industry, the occupation and the Workers' Compensation Board "rate group." The rate group reflects the assessed risk of the firm, referring to the proportion of payroll that employers pay for their compensation insurance premiums.

When denominators are developed within a company, the basis of exposure may be the total number of people working, the total number of hours worked or the number of full-time equivalent workers-all generally derived from payroll cost estimates. Other possible denominators, available through Statistics Canada, include the entire labour force (i.e. all those available to work), the employed population or "paid workers." A paid worker is someone who is on a company payroll, and thus it excludes unincorporated business owners and unpaid family help.

Although there are diverse denominators available, they are not always known or available to those interested in developing measures of occupational risk. Studies that do include denominators often are restricted in their usefulness. Small firm-based studies, crucial for a company to understand the distribution and determinants of injuries to their own personnel, may not be generalizable to the larger population. On the other hand, provincial or national injury rates may be based on denominators for broad industrial or occupational groupings.

Methods

Haggar-Guénette's Method

A national study of Canadian occupational injury rates ² derived denominators for Canadian injury rates from Statistics Canada's Labour Force Survey and used numerators from the National Work Injuries Statistics Program with data provided by the provincial Workers' Compensation Boards. Using these, national injury rates were developed by sex and either major industry group, occupation or age.

The study ² showed that injury rates were highest for occupations in forestry and logging, machining and materials handling and for the industries of forestry, construction and manufacturing. Rates for men were higher than for women and decreased with increasing age beyond 20 years, whereas rates for women increased slightly with increasing age. Haggar-Guénette speculated that the differences in injury rates by age and sex may have been due to the differential distribution of workers in various occupations and industries, but her analysis was restricted to the broad categories described.

An Enhancement of the Method

To address this limitation, our paper briefly describes the use of paid workers from the Labour Force Survey data as denominators,² extending this approach to a provincial context and incorporating the distribution of occupational roles from the Census.³ In this way, more refined denominators can be developed, including both age and occupational role information.

These denominators can be used with numerators from Workers' Compensation Board data to develop injury rates and to identify the relative contributions that age and occupation make to the risk of occupational injury.

The steps in denominator preparation are presented here, with an example of Ontario construction workmen, using the two Statistics Canada sources of the data.

Results

Denominator Data from the Labour Force Survey

Statistics Canada's Labour Force Survey gives national estimates of the number of paid workers, which can be used to develop denominators. Each month, the survey is done in about 48,000 households across Canada, and each household remains in the survey for six months.⁴ The sampling strategy is based on a complex set of factors, reflected in a single final "weight" calculated for each respondent. This weight is critical in the development of denominators because it denotes the number of individuals represented by each respondent: it is stored with the raw electronic data and provided to any user of the data.

Information is collected on household characteristics and on labour force activity of those over age 18. This includes, among other things, whether each person is working, how many hours a week, what type of work and the type of company.

To use the Labour Force Survey information, the 12 data tapes for 1989 were procured from Statistics Canada. A subset of the data was extracted for the Ontario construction industry, for men aged 15-64 who reported themselves as paid workers and employed in construction occupations. As Haggar-Guénette did at the national level, we produced monthly point estimates of the number of paid workers within 10-year age groups, then averaged the 12 estimates for the annual number of employed paid workers (see Table 1). These simple steps represent an expansion of Haggar-Guénette's method, producing a provincial-level denominator for construction occupations within the construction industry. The next section describes further enhancements to the method, which permit the development of denominators by more refined categories related to various roles within construction occupations.

Census

The occupational role distribution was derived from Canada's national 1986 Census. It surveyed every household in Canada, and one in five households received a more extensive set of questions.⁵ This 20 percent sample provides detailed information about occupation, and the data on the entire labour force are published for each province, sex and age group, by detailed occupation.⁶ Complete (4-digit) occupational codes for Ontario males aged 15-64 who worked in construction occupations were transcribed onto a spreadsheet.

Occupational Role Groupings

Occupational role groupings are defined by the question of interest and by size considerations. In our case, we were interested in how the risk of occupational injury might be affected by the category of workers. Therefore we used the Standard Occupational Classification codes to assign construction workers to one of the three occupational roles of foremen and inspectors, tradesmen and their apprentices, or labourers.

The decision about grouping for adequate size in each cell was done on an iterative basis, recognizing that Statistics Canada requires a minimum cell size of 4 000. The distribution of workers within these occupational roles was reviewed to ensure that cells had adequate numbers. When rates are to be calculated using these denominators, a review of the distribution of injuries (for the numerator) must also be done to ensure that enough events occur in each cell to preclude unstable estimates. Limitations imposed by these criteria will be presented in the section on disadvantages.

Spreadsheet formulae were used to sum the labour force participants and calculate their proportions. Table 2 summarizes the Census information, showing the proportions of workers within each occupational role overall and by 10-year age groups.

Combining the Data

Finally, the Census proportions (Table 2) were applied to the average monthly numbers of paid workers from the labour force (Table 1), to infer denominators that did not previously exist; these are shown in Table 3.

The right-hand column of Table 3 is the number of Ontario construction workmen estimated for the three selected occupational roles, and this information permits the calculation of rates by occupational role. The bottom row provides the estimates of workmen in each age group, and the internal cells provide estimates for age and occupational role simultaneously. By applying these potential denominators to comparably stratified data on occupational injury, it is possible to develop occupational injury rates that pin-point whether the real risk of occupational injury lies with particular age groups or occupational roles. These two steps, of applying the method to provincial data and then stratifying the data synchronously on age and occupational role, represent enhancements of Haggar-Guénette's method of calculating denominators.

Discussion

Assumptions

This method of applying Census role information to Labour Force Survey data on paid workers requires three assumptions to be made. The occupational role distribution is assumed to be consistent:

• across all industries in which those in construction occupations work;
• over time, that is, 1986 through 1989; and
• by employment status, that is, unemployment does not differ by occupational role.

Publicly available forms of both the Census and the Labour Force Survey (which collect such information from a sample of the population) do not provide information at the level of disaggregation that is required. Census data on occupational roles and age distribution are published only at the level of the whole labour force, within a given province. On the public-use data tapes, Statistics Canada suppresses detail below the level of the major group (i.e. it does not report beyond the first two digits) of the Standard Occupational Classification, rounds all data to the nearest 1 000 and suppresses any estimates below 4 000.

Thus, it is not possible to procure an age breakdown by occupational roles for male paid workers in construction occupations, employed within the construction industry. It is worth noting that if one had such a data set, one would not need to use the method detailed in this paper, because one would, ipso facto, have the exact denominators for identifying the population at risk of occupational injury.

Advantages

This method has a number of advantages for the researcher. The data are relatively inexpensive, and they are available and accessible. Statistics Canada provides technical documentation and telephone enquiry support services. A further benefit is that data quality assurance is a priority of Statistics Canada, and their two surveys are designed to be compatible.

Disadvantages

There are some disadvantages to this method of combining Labour Force Survey and Census data to develop finer detail in denominators for occupation. The method can only be applied to relatively large occupational groups: Statistics Canada requires suppression for cells with less than 4 000 and rounding to the nearest 1000 for all cells. Another disadvantage is that Census occupational role tabulation, although not difficult, is tedious and time-consuming. Electronic sources of Census data may improve this-however, 1991 Census publications do not provide data in a format equivalent to the 1986 information used here.

Because the Labour Force Survey data tapes are so large, extraction of data from the large data sets must be done on a mainframe computer, either by the researcher or by Statistics Canada. This difficulty may be circumvented by requesting either a more restricted extraction of observations or summary tabulations of data. However, such requests for restriction of data require careful planning, and they are not necessarily cheaper than ordering the entire data tape.

Calculation of Rates

For epidemiologists, the ultimate objective of developing better denominators is to produce better rates for events of interest. Ideally, it would be important to test this method by comparing any rates devised against, say, national counts of full-time equivalents of employed persons. In the absence of an established surveillance system, however, we need to do two things. The results of any new rates should be compared to other available studies, even if they are only small-scale. Our other obligation, when combining denominators with numerators from other sources to produce rates, is to verify whether the two are compatible.

As an example of what sorts of differences might be found in such a comparison, consider the case of a study (not shown here) where we combined the denominators developed above from the (national) Statistics Canada sources of the Labour Force Survey and the Census, with numerator data on occupational injuries (in this case, provincial data on Ontario construction workmen employed in construction industries).

The occupational classification system used by the provincial and national agencies differed: Statistics Canada used the Standard Occupational Classification 1980,⁷ and the Ontario Workers' Compensation Board used the Canadian Classification and Dictionary of Occupations.⁸ Manual comparison of the coding categories revealed discrepancies in the coding for some labourers. A decision was made to exclude some labourers' injuries from the rate calculations, meaning that their injury rates were underestimated. This example demonstrates how important it is to recognize potential bias arising from resolution of any differences between the classification systems.

Applying this Method to Other Sources of Data

This method, or adaptations of it, could be applied to other provincial or national sources of data where information about labour market activity is included. Such sources could be secondary data sources of morbidity, mortality or health care utilization, or it could be surveys of attitudes or behaviours. Personal identifying information is not required, as this is not a formal data linkage. Numerator information may be in the form of individual observations (with weighting information if the respondents were not randomly sampled), or it may be stratified for meaningful groups (e.g. by age, sex and occupational group).

Conclusion

This method combined data from two separate sources to produce occupational groupings within an industry, both overall and by 10-year age groups. The potential denominators were based on paid worker information from the Labour Force Survey, supplemented by occupational role distribution information from the Census. The method demonstrated here could be applied to numerators from Workers' Compensation Board claim data or from any other provincial or national measures of occupational morbidity or mortality.

The use of two sources of denominator data enhances our capacity to understand better the distribution of risk by occupation. Given the ever-increasing constraints on resources, this method may help others to make the best use of available data.

Acknowledgement

This work was supported by the Institute for Work and Health. We thank Cynthia Haggar-Guénette for clarifying and confirming the details of her method.

References

1. Labour Canada. Occupational injuries and their cost in Canada. Ottawa: Supply and Services Canada, 1991; Cat L151-2238/91B.
2. Haggar-Guénette C. Work injuries in Canada, 1982 to 1986. In: Statistics Canada. Employment, earnings and hours: January 1988. Ottawa: Supply and Services Canada, 1988; Cat 72-002.
3. Rael EGS. 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 [dissertation]. Toronto (Ont): University of Toronto, 1992.
4. Singh MP, Drew JD, Gambino JG, Mayda F. Methodology of the Canadian Labour Force Survey 1984-1990. Ottawa: Supply and Services Canada, 1990; Statistics Canada Cat 71-526.
5. Statistics Canada. User's guide to 1986 Census data on labour force activity. Ottawa: Supply and Services Canada, 1990; Cat 99-111E.
6. Statistics Canada. 1986 Census. The nation series. Population and dwelling characteristics: occupation. Ottawa: Supply and Services Canada, 1989; Cat 93-112: Table 2.
7. Statistics Canada (Standards Division). Standard occupational classification 1980. Ottawa: Supply and Services Canada, 1981; Cat 93-565E.
8. Employment and Immigration Canada. Canadian classification and dictionary of occupations. Ottawa, 1971.

Elizabeth GS Rael, Institute for Work and Health, 250 Bloor Street East, Suite 702, Toronto, Ontario M4W 1E6; and Department of Preventive Medicine and Biostatistics, University of Toronto, Toronto, Ontario
Elizabeth M Badley, Arthritis Community Research and Evaluation Unit, Wellesley Hospital, Toronto, Ontario; and Wellesley Research Institute, Clinical Epidemiology Division, Wellesley Hospital
John W Frank, Institute for Work and Health, Toronto; and Department of Preventive Medicine and Biostatistics, University of Toronto, Toronto, Ontario
Harry S Shannon, Institute for Work and Health, Toronto; and Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario

[Table of Contents] [Next]