Objective To investigate differences and similarities between three sources of work-related injury information: workers compensation claims, emergency department (ED) presentation data and hospital admissions data.
Methods This population-based, retrospective descriptive analysis of non-fatal, work-related injuries of workforce participants in Victoria, Australia, has compared data from workers compensation claims and ED presentation and hospital admission data sets for the period 2004–2011. Work-related injury case frequency and rate were compared across study years according to gender, age, geographical location and injury type. Injury rates were expressed as cases per million hours worked.
Results Rates of hospital admissions for treatment of work-related injury increased over the study period, compared with decreasing rates of injury in compensation claims and ED data. The highest rate of injuries to younger workers was captured in ED data. There was greater capture of musculoskeletal injuries by workers’ compensation data, and of open wound and burn injury by the ED data. Broad similarities were noted for temporal trends according to gender, for the distribution of cases across older age groups and for rates of fracture injuries.
Conclusions These study findings inform use of workers’ compensation, ED presentation and hospital admission data sets as sources of information for surveillance of work-related injuries in countries where these types of data are routinely collected. Choice of data source for investigation of work-related injury should take into consideration the population and injury types of interest.
Statistics from Altmetric.com
What this paper adds
Multiple sources of information are available for the purpose of work-related injury surveillance, but few studies have compared the relative capture of work-related injury burden by different datasets in relation to injury or population level characteristics.
This study found the temporal trend in injury risk differed across three sources of work-related injury data: workers’ compensation and emergency department data showed declines in injury risk, while hospital admissions data showed an increased risk of injury.
Emergency department data captured the highest rate of injuries to younger workers and the greatest proportion of open wound and burn injuries; the highest proportion of musculoskeletal injuries was captured in workers' compensation data, and rates of fracture injury were similar across all three datasets.
Use of different data sources for work-related injury surveillance may lead to different conclusions regarding trends in injury rate and burden across labour force subgroups.
Despite recognition by the International Labour Organisation that a safe and healthy working environment is a fundamental human right, the global burden of disease, injury and death due to work-related activities and exposures remains considerable.1 It is estimated that world-wide, nearly one million work-place accidents occur each day, and work-related accidents and diseases lead to more than 2.3 million fatalities each year.1 ,2 Continuing improvement in the capacity to meet current and emerging challenges to the safety of workers is therefore a priority. The development and evaluation of strategies to prevent work-related injuries requires the ongoing systematic collection of information about the number and type of injuries, the characteristics of injured workers, and the circumstances in which injuries occur. Ideally, such surveillance data will be generalisable to the population of interest, reliable and continuous over time, allowing trends in injury burden to be monitored and high-risk workforce subgroups to be identified.3
In many countries, including Australia, information about work-related injuries is routinely collected in the form of workers’ compensation claims, and in administrative records of hospital emergency department (ED) presentations, and hospital admissions. However, each of these information sources is likely to represent the burden of work-related injury differently. For example, workers’ compensation claims will be for work-related injuries that require hospital treatment, as well as those that do not. In addition, some claims will be for work-related injuries treated in community-based settings, while others will be for injuries for which healthcare has not been sought. ED visits and hospital admissions records will capture all work-related injuries that require emergency or hospital treatment, including those that do not result in the submission or acceptance of a workers’ compensation claim. Finally, there will be an unknown proportion of work-related injuries treated by community-based providers or not at all, for which workers’ compensation is not claimed, and which will therefore not be represented in any of these data sets. While the proportion of all work-related injuries captured by each of these data sets is likely to differ, comparison of trends in injury frequency and rate between data sets will provide insight into consistency of underlying factors influencing this capture. This information will in turn inform the interpretation of surveillance information derived from the different data sources. For instance a comparison of two work-related injury data sources in Canada has found relative consistency between rates of injury over time, suggesting each are a valid source of surveillance data.4
While the temporal trends in all injuries might be similar across data sources, the types of injuries represented in workers’ compensation and hospital-based data sets would be expected to differ. For instance, ED data is likely to capture a larger proportion of superficial injuries, including open wounds and burn injuries than workers’ compensation data, as these injuries may not all result in the minimum number of days absence from work to be eligible for workers’ compensation, but often require immediate medical care. Conversely, workers’ compensation data may capture a higher proportion of musculoskeletal injuries than emergency or hospital admission data, as these injuries often do not require emergency or hospital services. The rate of fractures should be similar for workers’ compensation and hospital data sets as these injuries often require hospital-based treatment and can lead to extended periods of work absence.
Given the different coverage and likely differences in types of injuries represented in each of these data sets, there is a need to better understand potential differences and similarities between compensation claim and hospital-based data sets for the purpose of work-related injury surveillance. Accordingly, the objective of this study, set in the Australian state of Victoria, was to investigate the differences and similarities between three different sources of work-related injury information (workers’ compensation claims data, administrative records of hospital ED presentations and hospital admissions) by comparing the frequency, rate and type of work-related injuries represented in each of these data sets for the period 2004–2011.
An observational, descriptive study of work-related injuries of Victorian workforce participants was conducted using injury data from three independent sources: workers’ compensation claims data held in the Compensation Research Database (CRD); ED presentations data in the Victorian Emergency Minimum Dataset (VEMD); and hospital admissions data held in the Victorian Admitted Episodes Dataset (VAED), for the period 1 January 2004 to 31 December 2011. This time-frame was chosen because a maximum and consistent number of hospitals have reported data to the VEMD since the beginning of this period. In the state of Victoria the number of workers in the labour market is currently in excess of 2.8 million,5 and coverage of workers’ compensation in this jurisdiction is approximately 85% of all labour market participants.6 These factors allow for one of the first examinations of population level surveillance of work-injuries using three different data sources, in a jurisdiction with a high level of workers’ compensation coverage and a universal healthcare system. Approval to conduct this study was obtained from the Monash University Human Research Ethics Committee.
Compensation Research Database
The CRD holds data derived from workers’ compensation claims submitted to WorkSafe Victoria, a statutory authority responsible for managing the Victorian workers’ compensation scheme.7 As mentioned above, approximately 85% of all Victorian workers are insured with the WorkSafe Insurance scheme, with the remainder predominantly consisting of self-employed workers, workers employed by self-insured employers and employees of the Federal Government.6 This database includes de-identified data about injured workers, the nature and circumstances of their injuries and details of their occupation and the industry within which they work. In Victoria, work injuries are compensable if they result in 10 or more days absence from work, or if healthcare costs are above a certain threshold value.8 Claim level data was extracted from the CRD according to variables that described characteristics of the injured person (age group, gender, local government area of residence), date of injury affliction, claim type and injury characteristics (nature of the injury and part of body injured).
Victorian Emergency Minimum Dataset & Victorian Admitted Episodes Dataset
The VEMD and VAED hold data regarding Victorian hospital ED presentations and hospital admissions, respectively. These data sets contain deidentified demographic, administrative and clinical data reported by hospitals at the individual level.9 ,10 The VAED is composed of all records of admissions to all Victorian public and private hospitals, while the VEMD consists of records of ED presentations to all Victorian public hospitals that have a 24-hour ED. The VAED is coded to the International Statistical Classification of Diseases (ICD) and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM).11 All Australian citizens and permanent residents are eligible for free treatment in public hospitals through Australia's publically funded universal health insurance scheme.12
Estimates of work-force participation
We used two sources of denominator data to calculate work-related injury rates because of the different scope of coverage for workers’ compensation claims and hospital visits (ie, all Victorian workers can attend a hospital for treatment, but only 85% can access workers’ compensation through WorkSafe Victoria). Denominator data for the ED and hospital admissions data were estimated from the Australian Bureau of Statistic's Victorian Labour Force Survey over the relevant time period.5 For the workers compensation data, a custom tabulation from SafeWork Australia was used that included a restriction to exclude self-employed workers and federal government employees from the labour market counts, as these groups are not eligible to report injuries to WorkSafe Victoria. It should be noted that a small number of companies and organisations are not required to report injuries to WorkSafe Victoria. These companies are either approved self-insurers or are covered under federal legislation. Given these companies are spread across industries it is not possible to exclude them from the WorkSafe denominator estimates. Therefore, our estimates of rates using workers’ compensation data will be slightly underestimated.
For the purpose of this study, a case was defined as a compensation claim, or an ED presentation or a hospital admission for an episode of non-fatal, intentional or unintentional acute work-related injury of a Victorian resident aged 15 years or over that occurred during the period 1 January 2004 to 31 December 2011.
For the purpose of case selection from each of the data sets, injury was defined as physical damage to the body resulting from acute exposure to intolerable levels of energy, or lack of a vital element such as oxygen, with this damage being quickly apparent.13
A total of 427 995 workers’ compensation claims for which the date of injury affliction occurred during the study period were initially extracted from the CRD. Claims that did not satisfy our injury definition (diseases or disorders resulting from repeated or long-term occupational exposures) were then excluded (n=186 638). The claims most frequently excluded using these criteria were: non-traumatic back pain, strain and lumbago (n=67 900) and non-traumatic muscle strain (n=23 772). Also excluded were minor claims (where less than 10 days of wage replacement or healthcare below the pecuniary threshold is expected, as the reporting obligations for these injuries differs from standard claims; n=113 866), claims for fatal injuries (n=332), and claims submitted by non-Victorian residents (n=2182), leaving a final number of 124 977 eligible claims.
A total of 231 964 VEMD cases were identified that had an ICD-10-AM primary diagnosis in the range S00–T75, T79 (community injury) and either the activity when injured variable was coded to ‘working for income’ or the compensable status variable was coded to ‘Work Safe’. Repeat visits for treatment of an existing injury, deaths in hospital and non-Victorian residents were then excluded (n=8975), giving a final number of 222 989 ED cases.
A total of 56 715 VAED cases were identified that had an ICD-10-AM principal diagnosis in the range S00–T75, T79 (community injury) and either an activity code of ‘While working for income’ (U73.0) or the compensable status variable was coded to ‘Victorian WorkCover Authority—VWA’. Transfers within or between hospitals, readmissions to the same hospital within 30 days, deaths in hospital and non-Victorian residents were then excluded (n=6797), giving a final number of 49 918 admissions.
We identified cases from the VEMD and VAED based on both compensable status and activity code, as relying on activity code alone may lead to underestimates of work-related presentations in hospital-based data sets.14 The proportions of cases identified using these different methods are shown in online supplementary appendix 1. Classification codes selected for comparison of injury characteristics across the three databases are supplied in online supplementary appendix 2.
Initial analyses compared the number of work-related injury cases identified in the CRD, the VEMD and the VAED across study year, gender, age and geographical location groups. Age-specific and gender-specific injury rates were then estimated within each data source. Given differences in labour market participation (ie, to account for part-time work) rates of injury cases were presented as injuries per million hours worked within each age and gender category. A final series of analyses compared the rates of injuries across type of injury groups. As the age distribution of WorkSafe insured workers may differ slightly from the whole population of work-force participants (given that older workers are more likely to work for the government or be self-employed15), age adjusted rates were calculated by direct standardisation, using the ABS denominator data as the standard age distribution.16 CIs for injury rates and rate ratios were calculated using the formula derived by Breslin et al.17 Data analysis was conducted using Stata V.12.1 software.18
Table 1 presents the distribution of work-related injury cases from the three data sources across years, gender, age group and geographical location. The largest number of cases was observed in the ED visits (VEMD), followed by workers’ compensation claims (CRD), and hospital admissions (VAED) data. While the number of claims in the CRD declined between 2004 and 2011 (by 6.5%) the number of injuries presenting to EDs (VEMD) and admitted to hospital (VAED) increased over time, with increases in hospital admissions being larger than for ED presentations. This led to increases in the ratio of VEMD and VAED cases relative to CRD claims over the time period. Males constituted the majority of injury events in all databases, although the proportion of males was higher in the VEMD and the VAED than in the CRD. The VEMD contains a larger proportion of injuries from younger workers when compared with the other data sets: 74% of cases in the VEMD data were aged less than 45 years, compared with 64% in the VAED and 58% in the CRD. The CRD contained the highest proportion of injuries from residents of metropolitan Melbourne, with the VEMD containing the lowest proportion of injuries from residents of metropolitan Melbourne.
Table 2 presents the distribution of number and rate per million hours worked of work-related injury cases for each study year, by gender and data source. In addition, the percentage changes in frequency and rate of injury from 2004 to 2011 are shown. Across all study years and among both males and females, the highest injury rates were observed for ED presentations (VEMD), followed by workers’ compensation claims (CRD) and then hospital admissions (VAED). For each data source, rates of injury were higher for males than females for each study year. Rates of claims in the CRD, and rates of ED presentations in the VEMD both decreased between 2004 and 2011, with this decrease being more pronounced for the CRD (a 22.3% decrease) than the VEMD (a 4.7% decrease). In contrast, the rate of hospital admissions (VAED) increased over the study period by 12.5%. For both the CRD and VEMD, the decrease in rate of injury cases was greater for males than for females. For the VAED, the rate increase was greater for females than for males. Additional regression analyses comparing the change in risk of injury across year of injury confirmed these findings, with a stronger negative trend in injury risk observed in the CRD, a weaker negative estimate observed in the VEMD and a weak positive estimate observed in the VAED (results not shown, but available from the authors on request).
Figure 1 presents a comparison of rates per million hours worked of work-related injury cases for each of the three data sources, by 5-year age group. For workers younger than 45 years, the highest rates of work-related injury were observed for ED presentations (VEMD), followed by workers compensation claims (CRD) and then hospital admissions (VAED). The difference between data sources was most pronounced for the 15–19 years age group, with the rate of ED presentations (VEMD) being three and a half times greater than the rate of compensation claims (CRD). Age-specific rates of ED presentations (VEMD) declined with increasing worker age, ranging from a high of 16.9 cases per million hours worked for the 15–19 years age group to 3.9 cases per million hours worked for the 55–59 years age group. In contrast, after a decrease in rate from 4.8 to 3.0 cases per million hours worked for the age groups 15–19 and 25–29 years respectively, rates of compensation claims (CRD) increased with increasing worker age, reaching a peak of 5.3 claims per million hours worked for workers aged 60 years or over. The rates of compensation claims and ED presentations were broadly similar for workers aged 45 years and over.
Table 3 presents the distribution of the number and age-adjusted rate per million hours worked of work-related injury cases by injury type, for each of the three data sources. In addition, the ratios of VEMD and VAED rates relative to CRD age-adjusted rates are shown. The rate of musculoskeletal injury cases was highest in the compensation claims data (CRD), and least in hospital admissions (VAED) data. The rate of ED presentations (VEMD) for open wound and burn injuries was almost three times greater than the rate of compensation claims (CRD) and five times greater than hospital admissions (VAED) for this injury category. Of all injury types, fractures was the category for which rates were most similar between the CRD, VEMD and VAED data sets.
This study has compared work-related injury information held in workers compensation, ED presentation and hospital admission data sets through estimation of trends in injury frequency and rate according to age group, gender and type of injury, for workers in Victoria, Australia. We observed more work-related injury ED presentations than compensation claims over the study period and hospital admissions accounted for the least number of cases. This frequency profile was expected as all workers are eligible to seek healthcare in a public hospital ED under a universal health insurance scheme, regardless of injury severity. Eligibility for workers’ compensation however requires satisfaction of eligibility criteria regarding days of work lost and total medical costs. Hospital admissions are likely to represent the most serious injuries which will be a smaller subgroup of all work-related injuries.
Unlike a similar study in Canada, where healthcare and workers’ compensation arrangements are similar to those in Australia, rates of injury in each of our databases were not relatively consistent over time.4 We observed a decline in the rate of compensation claims, a shallower decline in the rate of ED presentations and an increase in the rate of work-related hospital admissions over the observation period. While the sharper decline in workers’ compensation claims relative to ED data could suggest successful primary and secondary prevention strategies in Victoria which have resulted in either less severe injuries occurring, or workers being able to return to work before 10 days of absence have accrued, this does not explain the increase in hospital admissions for work-related injuries over this time period. As such, it would appear that there has been an increase in the proportion of work injuries that are less severe (needing emergency healthcare, but not 10 days of absence from work) and injuries that are more severe (requiring hospital admission) from 2004 to 2011 in Victoria.
Data derived from the Australian Bureau of Statistics Multipurpose Household Survey suggests that the self-reported work-related injury rate in Australia declined from 64 injuries per 1000 employed persons in 2005–2006 to 53 injuries per 1000 employed persons (a decline of 17%) in 2009–2010.19 In these studies a work-related injury included injuries, illnesses and disease, including those that do and do not require time off work. However, the proportion of injured employees who applied for workers’ compensation increased from 37% to 43% over this time period, suggesting that decreased claiming behaviour is not likely to be a major factor underlying the observed decrease in claim frequency over the study period in that study.19
The observed increase in number of work-related injury ED presentations over the study period may reflect general increases in hospital service utilisation over this time period. A study investigating the utilisation of public hospital ED healthcare in Melbourne, Australia, has reported a persistent rise in ED demand by all age groups during the decade to 2008–2009, over and above that explained by population growth alone.20 It found that almost half of all patients presenting to Melbourne EDs during 2008–2009 were of working age, and that a majority of presentations were for semiurgent or non-urgent conditions that could be managed in a community-based healthcare setting. Reduced accessibility of community-based care and increased cost to the patient have been suggested as factors contributing to this increased ED presentation trend.21
Potential explanations for the observed increase in rate of work-related hospital admissions over the study period include improvements in documentation of work-related injury admissions in the VAED data set, changes in hospital admissions policy (eg, the proportion of less severe injuries that stay longer than 4 h, defined as admissions in Victoria,22 may have changed over time), or a reduced impact of occupational health and safety strategies on the incidence of the most serious work-related injuries. Future examinations of injury rates should attempt to tease out trends in injuries of different severity, to investigate whether the difference in the temporal trends of hospital admissions versus workers’ compensation reflects true differences in the rates of injury across injury severity groups. This information could then be used for the development of more targeted prevention strategies for these most severe injuries.
A disparity was observed between compensation claims and ED presentations for the younger age groups. The rate of ED presentations was 3.5 times greater than the rate of compensation claims for the 15–19 years age group. Other studies have also reported a higher incidence of ED visits of younger workers than older workers for treatment of work-related injuries.4 ,23 ,24 One investigation of merged ED and compensation data sets has found younger workers were relatively more likely to have injuries treated at an ED that were not also recorded in the compensation data set,23 and another comparing ED and compensation data sets independently reported a higher ratio of ED visit incidence to compensation claim incidence for younger workers.4 The higher prevalence of younger workers in ED data observed in this study may be due to younger workers sustaining a greater proportion of less severe injuries not eligible for compensation than older workers. Data from Australia and elsewhere have shown that young workers are more likely to suffer from wounds and burns, and that these injuries—unless severe—would generally not reach the eligibility thresholds for compensation.17 ,25 Younger workers may also recover from the same injury more quickly than older workers, leading to less time off work and fewer claims for compensation, as they have not reached the minimum eligibility criteria.26 ,27 Finally, it is possible that a larger proportion of younger workers do not lodge compensation claims, even if they are eligible. It has been reported that in Australia during 2009–2010, 63% of injured workers aged less than 25 years did not apply for workers’ compensation, compared with only 58% of older workers, with 10% of young workers not being aware of workers’ compensation or not knowing they were insured.19
We observed differences in the proportions of injury types captured by the three data sources. The highest rate of open wound and burn injuries was documented in emergency presentation data, the highest rate of musculoskeletal injuries was documented in the compensation claims data and rate of fracture injuries was similar across all three data sets. While these observations support the findings of other comparisons between workers’ compensation and ED data,4 ,23 and between workers’ compensation and hospital admission data,28 this study is the first to make these observations across three data sources for labour market participants across an entire jurisdiction with both universal health insurance and a relatively high percentage of the labour market covered by a single workers’ compensation system.
The findings of this study should be interpreted in light of a number of limitations. None of the data sets included in the present analysis were originally intended for the purpose of research. For example, not everyone with a work-related injury who is eligible for compensation will submit a claim. Similarly, information about cases of work-related injury resulting in ED presentation or hospital admission is limited by the accuracy and completeness of recorded data; for instance, ED data recording may take place in hurried, stressful circumstances.29 The dates used to group claims by year differ between workers’ compensation data sets and hospital bases data sets. In the workers’ compensation data we used the date of the injury, while we used the date of presentation or admission in the hospital-based data. While these dates will be similar for severe injuries, the timing between the two dates may differ for less severe injuries (eg, musculoskeletal injuries that eventually require surgery). This comparison of data sets has been limited to the parameters of age, gender and type of industry. While it would be of interest to also compare other criteria such as mechanism of injury, occupation and industry, there is insufficient information about these in the hospital-based data for meaningful cross-matching with the compensation data.
Despite these limitations, this study has a number of strengths. Cases of work-related injury have been extracted from data sources that have a comprehensive capture of work-related injury for the whole population of Victoria. In addition, selection criteria were used to avoid double counting of cases associated with multiple ED presentations or with multiple hospital admissions for the same injury, thereby allowing a comparison of number and rates of injury cases from three independent data sources. Extensive matching of injury codes and data fields was also conducted, allowing comparison of a harmonised set of injury types across the three data sources. The selection of cases from the VAED has included the additional step of selection according to compensable status as the activity code ‘while working for income’ has been reported to be poorly coded in hospital-treated data sets.14 The three data sets explored in this study are not mutually exclusive, that is, the same case of injury could be represented in two or all three data sets. There is currently no unique identifier for all three data sets in Victoria to allow investigation of this overlap through merging of data sets. The value of this study is that by comparing data sets with one another, the potential advantages and weakness of each as a source of information about work-related injury can be investigated. Future studies should investigate the feasibility of data linkage between the CRD, the VEMD and the VAED.
This study informs future use of workers’ compensation, ED presentations and hospital admissions data sets as sources of information for surveillance of work-related injuries in jurisdictions where these three types of data are routinely collected. Use of the different data sources for work-related injury surveillance may lead to different conclusions regarding trends in injury burden and rate across gender, age group and injury type subgroups. Furthermore, choice of data source for work-related injury surveillance should take into account the population and injury types of interest. Where capture of all injuries is a research priority then ED data may offer advantages over compensation data, as many injuries that do not exceed the compensation eligibility thresholds will require treatment in an ED setting. Alternatively, if a focus is on musculoskeletal conditions, then workers’ compensation may offer advantages over hospital-based data. Indeed, comparison of data from multiple sources, while understanding the limitations of each, is likely to provide the most complete picture for work-related injury surveillance purposes.
Data was provided by the Institute for Safety, Compensation and Recovery Research (ISCRR), and by the Victorian Injury Surveillance Unit based within the Monash Injury Research Unit of Monash University.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors PS and JAM conceived the study idea. PS, JAM, AJC and LMD contributed to the study method and analysis design. AJC extracted injury data for analysis. JAM conducted data analyses and prepared the initial and final manuscripts. AJC and PS contributed to manuscript preparation and review. LMD, EMM and MRS critically commented on the manuscript.
Funding This research was partially funded through a Discovery Early Career Researcher Award to PS from the Australian Research Council. JAM was partially funded through a Larkins Fellowship from Monash University to PS. AJC was supported by a Victorian Injury Surveillance Unit core grant from the Victorian Department of Health.
Competing interests None.
Ethics approval Monash University Human Research Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.