Article Text
PDF
Abstract
Objectives Given the tendency for young people to show elevated work injury rates, this study sought to examine trends in young worker and adult compensation claim rates in a Canadian province.
Methods Workers' compensation records and labour force survey data from 1991 to 2007 were used to compute claim rates by age group. To examine the contribution of work-related factors to claim rates by age group, multivariate analyses included industry and job tenure.
Results Descriptive analysis showed that age groups had different rates of declines over the time period. Multivariate analyses showed that claim rate declines for older adults were greatest prior to 1999. Young workers showed the largest declines after 1999. There was no indication that changes in industry or job tenure accounted for the trends in claim rates among older or younger workers.
Conclusions This study is one of the first to show a convergence in youth and adult workers' compensation claims in a North American jurisdiction. Ruling out work-related factors increases the possibility that systemic interventions may have contributed to the convergence. This provides policy makers in occupational health and safety with empirical data to guide targeting of resources.
- Occupational health
- adolescent
- age differences
- time trends
- workers' compensation
- occupational health practice
- injury
Statistics from Altmetric.com
- Occupational health
- adolescent
- age differences
- time trends
- workers' compensation
- occupational health practice
- injury
What this paper adds
The inverse relationship between age and non-fatal work injuries has been one of the most consistent findings in the occupational health and safety literature in the last two decades.
The present findings document a recent and steeper decline in work injury rates among adolescent and young adult workers' compensation claims requiring wage replacement.
As work-related factors have been ruled out, systemic interventions may have contributed to the delayed decline in work injury rates among young workers.
These data demonstrate that the elevated work injury risk among young workers is not static and may vary due to other labour market factors not assessed and/or broad-based prevention initiatives.
Introduction
Paid employment is a common part of the lives of Canadian adolescents (ie, 15–19 years old) and young adults (ie, 20–24 years old). Employment figures show that 44.7% of adolescents and 70.9% of young adults hold jobs at some stage each year.1 Many young workers have part-time, temporary jobs, particularly during the school year when they are juggling both school and work.2 For example, 71.9% of Canadian employed adolescents and 27% of young adults worked less than 30 h per week.3 Canadian adolescents are most likely to hold sales and service jobs (75.5%), with some representation in manual and goods-producing jobs (eg, construction, agriculture; 15.5%).3 By young adulthood, Canadians are less concentrated in sales and service jobs (50.4%) and more likely to hold manual and goods-producing jobs (22.4%).
As regards age restrictions for hazardous jobs, in Ontario, where the present study took place, youth aged 16 years and older can work in the logging, construction and manufacturing industries but are prohibited from working in underground mines or as window cleaners.4 While there are no specific hazardous orders for minors as in the USA, general clauses in the Ontario Health and Safety Act require due diligence of the employer in terms of training and providing a safe workplace.
The inverse relationship between age and non-fatal work injuries has been one of the most consistent findings in the occupational health and safety (OHS) literature over the last two decades.5 The two most recent reviews focused on age-related differences in work injuries6 7 described 14 population-based, cross-sectional studies (eg, studies reporting on workers' compensation claims across multiple industries) that provided unadjusted work injury rates for at least one age group for under 25-year-old workers. In 10 of these studies, adolescent and young adult workers had higher work injury rates than their adult counterparts. This elevated risk was more marked and more consistent for young males than for young women (for citations of the 14 studies, see the introduction of Breslin et al8). The four studies not showing elevated youth injury rates did not adjust for the increased rates of part-time work among youth (ie, lower duration of exposure to hazards of younger workers relative to older workers).9 10 A general limitation of these 14 descriptive studies was that very few study designs took into account differences in occupational mix across age groups, with young workers generally being employed in lower skilled jobs and more physically demanding occupations.8
To date, only four population-based, cross-sectional studies of age-related differences in work injury have examined in a multivariate fashion the degree to which work factors such as industrial sector account for the elevated risk of adolescent and young adult workers.8 11–13 Adjusting for work variables such as job title can be seen as an indirect measure of different hazard exposures between jobs.14 These multivariate analyses provide useful information on the extent to which youth, being concentrated in more physically demanding jobs, can account for the elevated risk of young workers. Despite many methodological differences among these cross-sectional studies and some variability in findings, most indicate that a substantial part of the elevated injury risk experienced by youth appears to be due to differences in the types of jobs young people and adults hold (and the associated hazard exposures).
In addition to having cross-sectional information on age differences in work injury rates, time trends are also useful in establishing the temporal patterning between predictors and outcomes. It is also helpful to have a temporal understanding of how the age-related relative risks occurred. That is, over time youth may come to have a higher relative risk, for example, because their work injury rate is stable while older adults' rates are declining. A different aetiological process may underlie a pattern where the elevated risk due to young workers' work injury rate is increasing in the context of a stable older worker rate.
Time trends for work injury rates in Ontario are of particular interest because descriptive data have recently shown sharper declines in the number of lost-time claims for young workers 15–24 years of age in Ontario than for older workers.15 This differential reduction in the number of claims may be taken as evidence of effective interventions targeting young worker safety. For example, in Ontario, several young worker health and safety initiatives began in 1999/2000, including OHS learning expectations being added to the curriculum in high school and a social marketing campaign that included TV and radio ads to increase young people's awareness of work safety.16 In addition, there has been an enforcement initiative (not specific to youth) where labour inspectors target the firms with the poorest work injury statistics, which may preferentially employ younger workers.17
However, counts of work injuries may simply reflect a reduction in the number of young people who are at risk of a work injury, because either fewer are employed or they are working fewer hours during the week. Other labour market changes that might be related to a reduction in claim rates include differential shifts over time in the types of industries young workers are employed in, or changes in the numbers of young workers who are within the first month of a new job, with both these factors being previously associated with increased risk of workplace injuries.13 18 19
Accordingly, the objectives of the present study were to examine trends in the rates of lost-time claims in Ontario by age groups over a 15-year period (1991–2007). Specifically, we sought to describe the rates in lost-time claims over the time period by age groups. We also evaluated whether the age/gender trends observed differed significantly from one another. Finally, we examined whether temporal shifts in the industries workers were employed in or the proportion of new workers in each age/gender group could explain these differences.
Methods
This study utilised lost-time claims reported to the Ontario Workers Safety and Insurance Board (WSIB) with an accident date between January 1991 and December 2007. The WSIB is the principal provider of workers' compensation in Ontario and covers approximately 65% of the labour force. Those excluded (or not specifically included) from coverage with the WSIB include the self-employed, domestic workers and the majority of the finance industry, as well as other minor industry groups (eg, shoe shining, barber shops).20 Information provided with each claim reported to the WSIB includes the date of the accident, the gender and age of the claimant, the industry they were employed in (using the Canadian Standard Industry Classification 1980 (SIC80)) and the employment start date of the worker, from which each claimant's job tenure when injured can be calculated.
Claim rates across the time period of interest were calculated using labour force estimates generated from the Canadian Labour Force Survey (LFS). The LFS follows a complex, rotating panel sample design to efficiently estimate monthly changes in the Canadian labour force.21 For the purpose of our analyses, we restricted the LFS estimates to only Ontario respondents, and those working in industrial groups with complete coverage by the WSIB (for further information on this method, see Smith et al22). To account for the different employment hours commonly worked by men and women and across age groups, all denominator counts were presented as full-time equivalents (FTEs), using the usual number of hours worked as reported by each respondent in the week of the LFS. One FTE is equal to 2000 h of work per year. FTEs and lost-time claim counts were estimated across each age by gender by industry by job tenure group.
The number of lost-time claims reported to the WSIB between 1991 and 2006 totalled 1 783 722. Of these claims, 3.7% were missing information on industry and could not be used. To ensure congruence between our numerators and denominators, we removed claims from industry groups with voluntary or mixed coverage, or those covered under an alternate schedule (Schedule 2) with the WSIB (N=530 649 (29.7%) lost-time claims), leaving 1 186 503 lost-time claims. Of this sample, 4103 lost-time claims (0.3%) were missing information on age or gender. These claims were more likely to be from years that were earlier in the time series and from claimants in the agriculture, forestry and fishing, information, arts, entertainment, recreation and accommodation/food industries (compared to retail trade), and least likely to be from the mining or health care industries. This left a final sample of 1 182 400 lost-time claims.
Analysis
Lost-time claims and FTEs were estimated across 10-year age groups, with the exception of respondents aged 15–24 years, who were divided into those 15–19 years of age and those 20–24 years of age. This is because the labour market attachment and range of occupations worked among 15–24 year olds is more varied than among older 10-year age categories. Given that the lost-time claim rates were broken down by age, gender, industry (four groups: primary industry and construction; manufacturing, wholesale trade and transport; retail trade; and other service industries) and job tenure categories (1 month, 2–4 months, 5–12 months, and >12 months), there were 102 observations for each of the 16 years.
An original descriptive analysis examined the monthly lost-time claim rates for each age by gender group between 1991 and 2007 using box plots developed by Tukey.23 This exploratory analysis technique is useful for comparing key components between population groups without making any assumptions of the underlying statistical distribution. Within each box plot, the bottom and top of each box represents the 25th and 75th percentile, the notches represent the median, and the end protrusions represent the outliers. The spacing between these various indicators within each box helps one to compare various levels of dispersion between and within age groups.
Poisson regression models then examined the time trends between 1991 and 2007 for each age and gender group using PROC GENMOD in SAS. In these models, the number of claims was the dependent variable and the log of the associated FTE count was the offset. An initial set of age- and gender-stratified models compared trends in claim rates between 1991 and 1998 and between 1999 and 2007. Our reason for selecting this cut-point was because of the introduction of legislation that led to increased provincial OHS activity. We expected that the impact of this activity should have become apparent from 1999 onwards. To test this, separate regressions for each age/gender group had year entered as a continuous variable, time block (1991–1998 vs 1999–2007) as a dichotomous variable and the interaction between year and time block. A second series of analysis then examined if adjusting for differences in job tenure or industry mix within age group across the time period could account for any differences observed in the trend in claim rates across age groups. All analyses were conducted in SAS.24
Results
Figure 1 shows the lost-time claim rate for males by age group. The Tukey box plots represent the distribution of monthly lost-time claim rates averaged for each year. Between 1991 and 1995, the injury rates for 15–19 year olds rose and the injury rates observed for 20–24 year olds and 25–34 year olds declined, so that these groups convergence and stay stable for the 15–24 year olds until about 2000. Another clear observation among the older age groups (ie, 35–44, 45–54 and 55+) is the consistent decline across the time period. In 2001, there is a drop in rates among all three age groups under 35, with somewhat steeper declines among the younger age groups. This leads to another convergence, so that in about 2006 all age groups have about the same lost-time claim rates.
Tukey box plots for male claim rates of monthly observations grouped by age, annually from 1991 to 2007. (A): 15–19, 20–24 and the 25–34-year-old male age categories; (B): 35–44, 45–54 and 55+ year-old male age categories. Box plots are useful for displaying the differences between populations without making any assumptions of the underlying statistical distribution. Note the clear distinction between the box plots for the 15–19-year-old age category and the box plots for the other age categories between the years 1991 and 1997. FTE, full-time equivalent.
To determine whether the claim rate slopes were different before and after 1999, we conducted Poisson regressions on the lost-time claim counts (with the relevant subpopulation of workers included as an offset) stratified by age and gender to produce average year-to-year changes in injury rates. The initial predictors in this model were year as an interval and time period (ie, 1991–1998 vs 1999–2007). We describe in table 1 whether the interaction term was significant for each age/gender regression, and also show the calculated β estimates before and after 2000, so that the nature of the interaction can be seen more clearly. In tables 1 and 2, the β estimates represent the year-to-year change in injury risk within the time period. Exponentiation of the β estimate provides the relative risk from year to year.
Unadjusted β estimates from separate regression stratified by gender and age group of rates of lost-time claims submitted to the Workers' Compensation Board of Ontario from 1991 to 2007
Adjusted* β estimates from separate regression stratified by gender and age group of rates of lost-time claims submitted to the Workers' Compensation Board of Ontario from 1991 to 2007
As shown in table 1, the year by time period interactions for all age groups were significant. Follow-up regressions stratified by each gender and age group show that the nature of the time trend before and after 1999 differed by age group. Specifically, 15–19-year-old and 20–24-year-old males showed minimal declines prior to 1999, and steeper declines in the 1999–2007 period. The reverse was true for the older males, particularly those over 35 years of age. A similar but less marked pattern was seen among younger and older women over the same time periods.
To examine the extent that work-related factors might account for the different age group trends, two additional covariates were included in the model: industry and job tenure. Table 2 presents the χ2 values for the age by time period interactions, adjusted for changes in industrial mix and the percentage of short-tenure employment across age and gender groups. Age-related differences in claim trends described above did not appreciably change with further adjustment for industry or job tenure.
Discussion
Many previous cross-sectional studies have shown elevated risks of work injuries among young workers aged 15–24 years.8 25 26 The present study makes several important contributions to understanding young workers and occupational injury. First, these trend data show that the work injury rates of teenagers and young adults is relatively variable over time. For example, in the early 1990s, 15–19-year-old males in Ontario showed lower rates compared to older adults. In the last year of our time trend, 2007, we observed a marked reduction in the elevated injury risk of 15–24 year olds. There were several years in the late 1990s and early 2000s where the young worker rate was static while older workers' rates continued to decline. There is no obvious mechanism that would help explain this differential pattern of decline, but our analyses again suggest that it was not labour market factors like shifts in industries or in the proportion of new workers in each age group.
Second, this jurisdiction is unique in showing a reduction over time in work injury risk (as assessed by lost-time claim rates) for young workers compared to older workers. Recent reviews indicate that most jurisdictions continue to show elevated injury rates for young workers.7 This convergence pattern led us to examine whether there were age-related shifts in work factors that might explain the pattern. The claims data allowed us to examine type of industry and job tenure, two variables that are strong predictors of work injury.13 The multivariate analyses indicated that differential shifts in age groups with regard to industrial mix and proportion of new workers did not appear to explain this convergence of claim rates. This pattern of results does not provide direct evidence but leaves open the possibility that the set of youth-specific interventions begun in Ontario in 1999/2000 contributed to the steeper decline in rates among young workers. It should also be noted that starting in 2004–2008 there was a broader enforcement initiative by the Ontario Ministry of Labour.27 During the initiative, labour inspectors visited firms with unusually high numbers of compensation claims for their industry four times a year, focusing on workplace hazards to help firms reduce on-the-job injuries. Although these workplaces represented just 2% of all firms insured by the WSIB, they accounted for 10% of all lost-time injuries and 21% of injury costs in Ontario. To the extent that youth may be over-represented in firms with high lost-time claim rates, this general initiative may have differentially affected young workers. Unfortunately, to date no evaluation information has been provided for these initiatives. This underscores the need for more attention to be paid to examining the impact of these kinds of broad-based prevention initiatives. Short of a randomised trial, a more rigorous examination of the effect of these interventions would require more detailed information on the breadth and intensity of the interventions and a comparison jurisdiction where other work safety initiatives were employed.
It should be noted as well that there were no interventions targeting older workers at the start of the time period, yet their rates declined more rapidly than younger workers' rates in that time period, suggesting that non-intervention factors can lead to substantial reductions in work injuries as well. The decreasing rate of decline among older workers later in the time series could be the result of hazard reductions through regulation and technological changes becoming less frequent.28 Another possibility is that the recent focus on youth might carry with it an unintended consequence of targeting young workers in that older workers—and their employers—see themselves less as an at-risk population for work injuries.
These findings must be interpreted in light of the study limitations. This study uses aggregated claims data as the unit of observation, which makes it vulnerable to the ecological fallacy.29 Also, the use of temporal breakdowns of lost-time claims assumes no systematic differences in claim reporting practices over time. Also, lost-time claims are not a proxy for all work injuries in Ontario, independent of claim reporting practices. However, a previous analysis suggests that these trends are unlikely to be primarily due to changes in claim reporting practices.30 In addition, although we did adjust for industry and job tenure, to ensure stability in our denominator estimates these groups were quite broad, which may mask for detailed differences within these large labour market groups.
In sum, descriptive findings showed substantial differences in temporal trends in claim rates by age group. Specifically, young workers 15–24 years of age showed a much steeper decline in claim rates from 2000 to 2007 compared to older adults. These results highlight the need to further examine work injury rate trends and the factors influencing these trends. This type of information can provide policy makers in the occupational health and safety with empirical data to guide the targeting of resources.
References
Footnotes
Funding PS is supported by a New Investigator Award from the Canadian Institutes for Health Research.
Competing interests None.
Ethics approval This study was conducted with the approval of the University of Toronto, Health Sciences I ethics committee.
Provenance and peer review Not commissioned; externally peer reviewed.