Role of migrant factors in work-related fatalities in Australia.

The role of migrant factors in work-related fatal ities in Australia. Scand J Work Environ Health 1994;20:364-70. OBJECTIVES - The importance of workers' language and migration characteristics to safety in the work environment has been debated but remains unclear. This study examined the role of these fac tors in the occurrence of work-related fatalities in Australia. METHODS - The study was based on an investigation of all work-related fatalities occurring in Aus tralia during 1982-1984. Denominators for each year were obtained according to gender and coun try-of-birth census data from the 1981 and 1986 national censuses, interpolated and adjusted accord ing to annual labor force survey estimates for the period 1981 to 1986 to indicate the true movement of the employed civilian labor force over the study period. RESULTS - Of 1211 decedents identified with known country of birth, 333 were born outside of Aus tralia. The overall fatality incidence per 100000 person-years in the employed civilian labor force was 7.12 [95% confidence interval (95% CI) 6.36--7.88], which is similar to that of Australian-born persons, 6.56 (95% CI 6.12-6.99). However, fatality incidences in rural and mining occupations were significantly increased among overseas-born persons when they were compared with Australian-born persons. Mortality ratios standardized separately for occupation and age showed significantly elevat ed mortality for duration of residence of less than five years, particularly for persons of non-English speaking background. These values converged to the Australian rate with increasing duration of resi dence.CONCLUSION - This study suggests that factors related to country of birth (eg, language) and dura tion of residence of overseas-born workers are important determinants of safety at work in Australia.

that it has a relatively new and large migrant population, Au stralia provides an ideal setting for the study of the effect of migrant status and language in general. Unfortunately, information on actual ethnic origin is often not available in either numerator or denominator data, but country-of-birth information is accepted as a more objective and reliable measure and can be used as a proxy (12).
Migrant status alone is not a factor which should cause differential occupational health and safety outcomes. Adverse outcomes are more likely to be related to variables associated with migrant status, such as lack of English language proficiency, or occupation, and they are more likely to be modified by other factors, such as duration of residence. Treating Australi an-born persons as one group makes it pos sible to study factors specific to migration, and these factor s can then be compared with the Australian-born group being used as reference. This paper describes work-related fatalities in light of these factors.

Subjects and methods
Language and migrant factors were examined as part of a larger study of work-related fatalities in Australia, where deaths attributable to work-related injury during 1982-1984 were studied. The study method has been described in detail elsewhere (13) and is briefly outlined in this report.
A list was obtained of all deaths which occurred in the study period and which had been assigned codes of the International Classification of Diseases, revision 9 (lCD 9), corresponding to external causes of death , but excluding suicide and medical misadventure. The list contained 16 246 case s, and coroners ' files were found for IS 462 (95.2 %) of these cases. Research officers examined the files and assessed which of them met the study definitions of work-related fatalities. Data were extracted from the coroners' files for these cases, recorded on data forms , and subsequently entered into computer files for analysis. The original study classed 1738 cases as work-related, of which 1278 were defined as coming from the employed civilian labor force, after the exclusion of 266 fatalities that occurred on the road during journeys to and from work.
There were 65 (5.1%) cases for which country of birth was not stated and two cases later assessed as not meeting study definitions of work-relatedness. This study is therefore based on 1211 fatalities that occurred among persons of 51 different nationalities . For most of these nationalities, however, the number of fatalities was too small for meaningful rates to be derived. The number of fatalities among the women was small: 34 among Australian -born persons and 13 among overseas-born persons. Analyses excluding women did not change significantly the comparisons of the variables of interest between different migrant groups. Therefore all of the reported rates include both genders. Data on industry was missing for 133 (11.0 %) persons. In addition, denominator data was not available for industry. Information on duration of residence was missing for 28 (18.5%) persons born in the main English-speaking countries and for 23 (12 .6%) person s born in non-English-speaking countries.
The employed civilian labor force consisted of persons who worked for pay, profit, or payment-inkind in a job or busines s or on a farm (including employees, employers and self-employed persons) and persons who worked without pay for IS h or more per week in a famil y business or on a farm. Persons 15 years of age or younger and members of the permanent armed forces were excluded.
Country of birth and language group (as persons born in English-or in non-English-speaking countries) were coded for each decedent to be consistent with denominator data supplied by the Australian Bureau of Statistics. Using the Bureau's classification, we defined persons from English-speaking countries as those born in the United Kingdom and Ireland, the United States, Canada, New Zealand, and South Africa (12). Occupations were coded according to the Classification and Classified List of Occupations (CCLO) (14) Special care was given to the construction of denominator data . The Australian Labor Force Survey provides information on country of birth, and information on duration of residence can also be obtained on request from the Bureau of Statistics. The Labor Force Survey is based on a multi-stage area sample of private dwellings covering about two-thirds of one percent of the population. Given these size limitations, sampling error is an important issue. Data from the Labor Force Survey, when stratified by a few variables, provide very smalI numbers, which are subject to significant standard errors (4). The attraction of the Labor Force Survey is that data are available on a monthly basis, with language and migrant status data published yearly.
Census data appeared as the most attractive denominator when only the employed labor force was selected and the armed forces were excluded, because theoretically there should be no sampling errors . The main disadvantage of using census data was that only data for 1981 and 1986 were available. Preliminary studies on data from the Labor Force Survey and census showed an increase in persons in the employed civilian labor force between 1981 and 1986, but the Labor Force Survey showed a decline below the linear interpolation of the census data for the period 1983-1984, which corresponded to a time of lower economic activity in Australia. It was therefore necessary to adjust the linear interpolation of the census data according to the Labor Force Survey data. The issue was further complicated by the observation that not all ethnic groups behaved similarly to the overalI population. Some ethnic groups increased in number, while others decreased, in the labor force during this period. Important gender differences were also observed. Therefore, specific denominator s were obtained for each of the gender and country-of-birth combinations available from the census data and then adjusted according to the gender and country-of-birth yearly movements indicated by the Labor Force Survey.
Data were analyzed with the SAS (statistical analysis system) package (15). Confidence intervals were calculated for incidence estimates with the use of the Poisson distribution (or its normal approximation for numbers greater than 100) ( 16). Confidence intervals for indirect standardization were obtained as described by Rothman (17).

Results
During the years 1982 to 1984, of 1211 work-related fatalities identified among persons with known country of birth, 878 (72.5%) occurred among Australianborn worker s, 151 (12 .5%) among workers born in the main Engli sh-speaking countries, and 182 (15.0%) among workers born in non-English-speaking countries. Persons born overseas had slightly higher fatality rates than Australian-born persons [rate ratio 1.09, 95% confidence interval (95% CI) 0.96-1.23]. In particular, elevated rates were ob- The mean age at death for the Australian-born group was 38.2 (SD 14.7) years. Persons born in the main English-speaking countrie s had a mean age of 39.3 (SD 13.3) years at death , while for migrant s born in non-English-speaking countries the mean age was 42.0 (SD 11.7) years at death. There was an increasing trend of fatality with age which was more marked for the Australian -born persons (trend X.2= 46.1, P<O.OOO I). The trend was significant for the non-Engli sh-speaking group (trend X , 2 = 8.9, P =0.003), but marginal for the Engli sh-speaking group (trend X/ = 3.1, P =0.08).
The number of death s and the fatality rate decreased for all persons combined for the years 1982 to 1984. The overall incidence per 100 000 personyears was 7.35 for 1982, 6.60 for 1983, and 6.16 for 1984. This decreasing trend was heavily influenced by the Australian-born group, which had a statistically significant decreasing trend (trend X,2= 12.8, P =0.0003). For migrants born in English-speakin g countries, and more particul arly for migrants born in non-English-speaking countries, there was a nonsignificant increase in incidence in the three years under study.
Large differences in incidence were observed for the major occupational classificat ions. These differences remained when occupation was stratified by country of birth (Australia, main English-speaking countries and non-Engli sh-speaking countri es), and, in some cases, stratification highlighted the occupational differences. Signifi cantly elevated rates were observed for the occupati onal categories "farmers,  Zealand. However, only those born in New Zealand and former Yugoslavia had rates statistically significantly highe r than those of Australi an-born persons and overseas-born persons as a group. Slightly higher fatality rates were observed for migrants born in non-English-speaking countries than for migrants born in English-speaking countries. However, the differences were not statisticall y significant (figure 1). fishermen, hunters, timber getters and related workers" (CCLO major group 4), and for "miners, quarrymen and related workers" (CCLO major group 5) for both non-Australian-born groups in comparison with the Australian-born group and, in particular, for migrants born in non-English-speaking countries. for group 4 and 2.37 (95% CI 1.22--4.62) for group 5 (tables I and 2). Mortality differ ences between industrial groups were highlighted when stratified by language group. For Australian-born persons, nearly 25% of all deaths occurred in the indu strial group "agriculture, forestry, fishing and huntin g." For other groups, particularly for migrants born in the main Engli sh-speaking countri es, the proportion of fatalit ies in this category was considerably lower. For migrants born in non-English-speaking countrie s a large proport ion of fatalities occurred in the industrial groups "manufacturing" and "construction." These two groups combined accounted for 40% of death s among migrants born in non-English-speak ing countri es. In contrast,   the same industrial groups were responsible for nearly 28% of the death s among migrants born in the main English-speaking countries, and 22% of the deaths among Australian-born persons (tab le 3 on page 367) .
The risk for migrants appeared to be modifie d by duration of residence in Australia, with significant excess risk in the first five years when compared with Australi an-born workers. This exce ss risk remained when differences in occupation and age were controlled with the use of standardized morta lity ratios (stand ard ization by either age or occupation indicating nearly identical pattern s). These differences were more marked among the migrants born in non-English-speaking countrie s (tab le 4).

Discussion
Langu age has been identified as one of the cruc ial factor s affecting safety at work . In the report of the Joint ILOIWHO Committee on Occup ational Health (18), the following two significant observations were made for language: (i ) migrants are adversely affected in the simplest aspects of everyday life and (ii) migrants are particularly affect ed at work, where the uni ntelligibility of verbal com mun ication is exacerbated by the work environment. In thi s case, work environment refers to physical aspects, such as background noise, and more importantly to personal aspects, such as background know ledge of work practice s and the work environment itself. In addition, preventive measures cannot alway s be communicated with ordin ary work language. Instead, they may require the communication of abstract knowledge demanding higher language proficiency (19).
In this study, given the basis of language class ification, the potential for misclassification was significant. Most persons born in the main English-speaking countries (and Australia) are likely to be proficient in English and are therefore probabl y classified correctly as Engli sh-speaking. However, it is likely that some persons born in non-Engli sh-speaking countries were also proficient in Eng lish and so were incorrectly classified as non-Engli sh-speaking. This miscla ssifi cation affect s only the measurement of 368 proficiency with English and occurs independently of the outcome and its measurement. If it is assumed that proficiency with Engli sh is a determinant of fatality risk , the inclusion of English speakers in the non-Eng lish-speaking group would have attenuated the estimate of the rate ratio of fatality (as compared with Australian-born). Therefore, at worst , the observed association may have been biased toward s the null, meaning that migrants born in non-Engli shspeaking countries who were truly non -English speakers may have had a higher risk than that actually observed in this study.
In their original paper on work-related fatalities in Australia, Harri son et al (13) reported a decreasing trend of fatalities ove r the three years 1982-1984. However, when stratified by country of birth, fatalities showed a significant decline among Australianborn persons, but no declining trend for persons of either English-speaking or non-English-speaking backgrounds. The impli cations of these findings are uncertain. Interpretation is difficu lt without rates calculated for a number of sequential years becau se of random fluctu ation , and unf ortunately data from other years are not currently available in Australia .
From demo graphic studies of migrants in Australia, it is well known that migrants differ greatly from Austra lian-born persons in their occupational distributions (5,6,20). If migrants are over-represented in dangerous occupations, a higher number of accidents and work-related fatalities would be expected. Stratified analysis by occupation is one way of removing the effect of confounding arising from differential occupati onal distributions of migrants and Australian-born persons . In the absence of major confounding , and presum ing that there was no relationship between migrant status and risk of fatality, similar fatality rate s would be expected for all language or migrant groups with in each occupational group . Howe ver, major occupational categ ories are too broad for this conclusion to be made unequivocally, as it is possibl e that occupational differences occur within major occupational categories. Such a detailed stratification was not possible in this study because the number of fatalities was not large enough . Therefore, if migrant s were concentrated in the most dangerous occupations within a major occupational group , higher fatality rates would be expected in these subgroups. However, these higher rates would not be observed when only the major occupational groups were considered. Given these limitations , the data showed important differenc es in some major occupati onal groups when Austr alian-born persons, migrants born in English-speaking countri es, and migrants born in non-English-speaking countries were compared.
The largest concentration of workers was in the occupational group "tradesmen , production-process worker s and laborers." The den ominator data used in this study indicated that 25.3 % of the Australianborn persons in the employed civilian labor force were in this group, compared with 29.7% of migrants born in English-speak ing countries and 44.2% of migrant s born in non-English-speakin g countries. Similar associations between countr y of birth and proporti ons of worker s in this occupational group have been reported (2,5,6). However, the fatality inciden ces in this occupational group were not statistically significantly higher than the fatalit y inciden ce for all occupati ons in each of these three birthplace groups. In addition, these three birthpl ace groups were not statistically significantly different from each other.
This study showed a marked variation in fatalit y inciden ce with duration of resid ence for overseasborn per sons. In general terms, the incidence was highest for shorter durat ion of residen ce and then approached the rate for Australi an-born person s as duration of residence increased. This finding is consistent with those from other studies in which increasing duration of residence is associated with employment characteristic s that are increasingly similar to those of Australian-born person s (2, 3, 7). Standardization of the observed rates separately by age and occupation did not alter this pattern. Duration of residence is strongly associated with language proficiency, with better proficiency in English after a longer duration of residence. Therefore, if differences in fatalit y rates between Australian-born persons and non-English-speaking migrant s are due to language differences, then decre asing rates with increa sing duration of residence would be expected. Elevated rates among English-speaking migrants with a short duration of residen ce in Australia are presumably not due to language difficulties. However, they may have been related to different work practices or experience or to social stress such as isolation and insecurity (which applie s to non-En glishspeaking migrants as well). It is also possible that new arri vals, regardle ss of ethni c background , are placed initially in the most dangerous catego ries within each occupational group. Such features of the data would not be detected by this analysis, which was based on only a broad occupational grouping. This is one of the few published studies to use accurately determined, contemporaneous denominators to investigate the occupational experien ce of migrant Scand J Work En viron Health 1994, vol 20. no 5 workers. Its majo r strength was that it was population-based, with adequate information on the relevant ethnic variables. Adjustm ent of 1981 census data to allow for fluctuation s during the study period allowed an accurate calcul ation of rates. In fact, comparisons made using other (unadj usted) denominators showed differences important enough to alter the ranking of rates and to obscure short-term trends. The major limitations of the study were that the data were not assembled primarily for the study of language and migrant factors. In particul ar, accurate information on the language proficien cy of the decedent would have been useful.
In this study there was a significant limitati on in the interp retation of the result s that deri ved from wide confidence intervals (low power) resulting from small numerat ors. Althou gh the results showed consistently elevated rates for migrant s, in particular for migrant s born in non-English-speakin g countries, most of these results lacked statistical significance. Data from a greater number of years than was available for this study would be needed to resol ve this issue. Lack of stati stical significance, howe ver, should not be equated with equal rates for migrants and nonmigrants and should not stop specific attempts to improve the occupational health and safety of migrants. Thi s study has clearl y demon strated that fact ors related to language and to durati on of residence in Australia were strongly associated with a major health and safety outcome among migrants, namely, traumatic death as a result of work. Migrants born in non-Engli sh-speaking countries were particularly at risk in the first five years of residence in Australia, a finding suggesting that language is an important contributory factor to work-rel ated accidents. Migrants born in English-speaking countries were also at risk, a finding suggesting that other factors, such as work practices or experience, may also contribute to work-related accidents.
Target ing newly arrived migrants with specific information on the work practice s of the host country and instruction in occupational health and safety seems important in the light of these findings. In addition, for migrants whose native language is not that of the host country, relevant information in their own language, plus the teaching of the language of the host country in the workplace, may greatly improve their occupational safety.