Objectives This study aims to investigate mortality and cancer incidence of Australian male volunteer firefighters and of subgroups of firefighters by duration of service, era of first service and the number and type of incidents attended.
Methods Participating fire agencies supplied records of individual volunteer firefighters, including incidents attended. The cohort was linked to the Australian National Death Index and Australian Cancer Database. standardised mortality ratios (SMRs) and standardised incidence ratios (SIRs) for cancer were calculated. Firefighters were grouped into tertiles by duration of service and by number of incidents attended and relative mortality ratios and relative incidence ratios calculated.
Results Compared with the general population, there were significant decreases in overall cancer incidence and in most major cancer categories. Prostate cancer incidence was increased compared with the general population, but this was not related to the number of incidents attended. Kidney cancer was associated with increased attendance at fires, particularly structural fires.
The overall risk of mortality was significantly decreased, and all major causes of death were significantly reduced for volunteer firefighters. There was evidence of an increased mortality from ischaemic heart disease, with increased attendance at fires.
Conclusion Volunteer firefighters have a reduced risk of mortality and cancer incidence compared with the general population, which is likely to be a result of a ‘healthy-volunteer’ effect and, perhaps, lower smoking rates.
- Volunteer firefighter
- cancer incidence
Statistics from Altmetric.com
What this paper adds
Studies of the health of volunteer firefighters have not previously been published.
Exposure metrics were developed using the number and type of incidents attended for the majority of firefighters in the study.
Australian male volunteer firefighters have a reduced risk of mortality and cancer incidence compared with the general population.
Death from ischaemic heart disease was associated with increased attendance at fires in internal analyses.
Prostate and kidney cancer incidences were raised, and the latter was associated with attendance at incidents.
There are a large number of volunteer firefighters in Australia, and this study examines the mortality and cancer incidence of male volunteer firefighters. The findings for female volunteer firefighters are presented elsewhere.
In Australia, there are currently an estimated 180 000 volunteer firefighters providing service to rural and some periurban areas. The major urban areas are covered by a much smaller number of full-time and part-time paid firefighters employed by different fire agencies for each state and territory. In most states, the volunteer service is organised by the fire agency, but in two of the most populous states, volunteers have separate agencies. These two separate volunteer agencies and three agencies that have paid and volunteer branches contributed volunteer records to the study. Two smaller states and a territory did not contribute volunteer firefighter records.
During firefighting, firefighters are exposed to a number of hazardous materials. Pyrolysis products contain a mixture of particulates, toxic gases and fumes including many carcinogens.1–5 Previous firefighter studies have been criticised for their lack of exposure assessment,6 the duration of employment has been the most usual measure of extent of exposure.7 8 Little has been published relating risk to attendance at fires or other incidents.
There is little or no existing published evidence about the mortality or cancer risk for volunteer firefighters. The scientific literature indicates that the overall mortality rate in employed firefighters is usually lower than that of the general population, which is probably explained by the healthy-worker effect.6 9 Studies conducted in the USA have shown that there is an increase in cardiac mortality associated with career firefighting, particularly close to an alarm or turn-out.10–12
There is evidence that paid firefighters in Australia have an increased risk of some cancers,13 14 in particular prostate and melanoma. Higher than expected cancer rates have been found in meta-analyses of international studies.6 15 Several studies have shown an increased risk of the following cancers in career firefighters: testicular cancer,7 prostate cancer, non-Hodgkin lymphoma and multiple myeloma. Other studies have reported that career firefighting may be associated with the following cancers: digestive system,10 16 17 testicular cancer,7 18 prostate cancer,17–20 mesothelioma,16 20 malignant melanoma,17 20 non-Hodgkin lymphoma6 and multiple myeloma.10 17 20
There are likely to be several key differences between the mortality and cancer incidence of paid and volunteer firefighters. Volunteers are likely to attend fewer fires per month than paid firefighters. The pattern of exposure will be different for volunteer and paid firefighters. The fires attended by volunteer firefighters in Australia are more likely to be bushfires, which can last for several days and occur mainly in the summer months. This means that exposure may be much more seasonal for volunteers than for paid firefighters, and the exposures experienced are different from structural fires. Before joining an agency, volunteer firefighters complete a self-declaration/disclosure of requested medical conditions, but there is no formal fitness assessment. They are thus much less highly selected than paid firefighters and will also have a different age distribution to paid firefighters, having a proportion over-retirement age. In addition, in the past, volunteer firefighters were less likely than paid firefighters to be equipped with individually issued respiratory protective equipment.
More details about the study methodology are provided in the technical report at the Monash University website.21
The aims of this study were to examine mortality and cancer incidence among male volunteer firefighters and in different subgroups based on the duration of firefighting service, era of first service and the number and type of incidents attended.
The cohort comprised all male volunteers from five fire agencies enrolled on or after the date on which the agency’s roll was complete and who had ever held an active firefighting role.
Volunteers from two volunteer agencies and from fire agencies that also included paid firefighters (one territory and two states) were included in this study. These agencies supplied records of all individual volunteers, including name, date of birth and service history. The personnel records provided varied by agency, started at various dates ranging from 1998 to 2000. A date was chosen for each agency when the roll was complete for all volunteers at that agency, and follow-up of all volunteers started at that date and continued until 30 November 2011 for cause of death and to 31 December 2010 for cancers. The period of service for volunteers included all periods holding active firefighting roles, that is, those before and after follow-up started. If service was complete before follow-up started, the firefighter was not included in the cohort.
The records were loaded into a secure database and checked for quality and completeness. There were 22 records with missing name details, 32 124 with missing birth dates and 1723 duplicate records. The service histories were checked with the relevant agency, and only the periods for which a firefighting role was identified were included. Each agency was sent a list of unique role titles extracted from their human resources (HR) data and asked to identify the volunteer designations that involved turning out to fight fires. This was to ensure that volunteers who had only ever held administrative or technical roles but never held an active firefighting role were not included in the cohort. All individuals who had ever held a firefighting role, however, were included. Periods of service as a junior (under 18 years of age) were also excluded.
Firefighters who had both paid employment and volunteer service were excluded by cross-checking the names and dates of birth against records in the parallel study of paid firefighters. This resulted in a cohort of 163 094 male volunteer firefighters.
Records of incidents attended by each volunteer firefighter were also supplied by the agencies. Incident data collection is carried out by all agencies, and deidentified data are contributed to a national database held by the peak body, the Australasian Fire Authorities Council. Data are collected in accordance with a national Australian Incident Reporting System (AIRS) Standard Manual, which sets out a standard coding and recording protocol. Fire agencies started incident data collection for volunteers at different times between 1998 and 2011. Incident data are collected contemporaneously about all the events to which each individual volunteer firefighter are dispatched, including fires, rescues and non-emergency events such as false alarms and are coded into an incident type. The following categories were drawn from specific AIRS codes for analysis: all incidents, all fires, and then structural fire, landscape fires and vehicle fires.
Follow-up for the incident analyses commenced from the date when each agency’s incident data collection commenced. The cumulative number of incidents was calculated for each volunteer firefighter for each person-year of follow-up. Some firefighters started volunteering before incident data were collected, however. For these male volunteers (n=84 812), the mean number of incidents per year for the years where incident data were available was allocated to the years for which data were unavailable back to the year of the start of service.
Male volunteer firefighters with no recorded incidents and who had left before the commencement of incident data collection at that agency were excluded from the incident analysis (n=1224). Firefighters from an agency who only commenced recording incidents in 2011 were excluded from the incident analysis, as it was likely to have a substantial underrecording of attendance (n=17 358). Thus, there were 144 512 male firefighters in the internal analyses by incident.
To obtain mortality and cancer outcomes, the cohort was linked to the National Death Index (NDI) and the Australian Cancer Database (ACD), both held by the Australian Institute of Health and Welfare (AIHW). At the time of linkage, the NDI was nationally complete to 30 November 2011 for cause of death coding. The ACD was nationally complete to 31 December 2010, except for one state, which was complete only until the end of 2009.
The AIHW used a probabilistic program to identify possible matches with records based on personal identifiers such as surname, first name and birth date. The underlying cause of death was coded to ICD-9 (InternationalClassification of Diseases, Ninth Revision)22 or ICD-10,23 and cancer incidence records coded to ICD-10. Each potential match was scored as to the probability of it being a true match. The possible death matches were each reviewed by two members of the study team, and disagreements were examined by a third member before a final decision on which records were to be accepted as true matches was made. The AIHW provided the researchers with a deidentified list of cancer cases that were scored as highly certain matches.
Individuals were followed up from the date of first enrolment as a volunteer or the date from which the HR data for the relevant agency was considered complete, whichever was the later. The date from which the agency data were considered complete was based on information from the relevant agency and examination of the yearly death rates by agency.
Any cancers and/or deaths that occurred after each individual’s cohort start date were included in the analyses. Follow-up continued until the date of death or the end of follow-up based on when the national cancer and death data were complete. Only the site of a primary cancer was included in the analysis. However, if a person was diagnosed with more than one primary cancer, then all of these cancers were included in the analyses, as the person can stay at risk of another cancer after the first cancer was diagnosed.
Australian population data were used to calculate the expected numbers of deaths and cancers for each firefighter group based on sex and calendar year and 5-year age groups.24
Standardised mortality ratios (SMRs) and standardised incidence ratios (SIRs) for cancer were calculated. SMRs and SIRs were also calculated for volunteer firefighters grouped by duration of service (>3 months to <10 years, 10–20 years and 20+ years) and by era of when they were first enrolled (pre-1970, 1970–1994, post-1995). Firefighters who had been enrolled for less than 3 months were excluded from duration analyses, because their firefighting exposure would have been minimal.
Analyses involving internal comparisons were carried out using Poisson regression modeling, with an offset term set to the number of person-years. Relative risks (relative mortality ratio (RMR) and relative cancer incidence ratio (RIR)) were calculated, comparing the risks for longer service duration with that of the shortest duration (>3 months to <10 years). Incident analyses commenced follow-up from the year that incident data collection commenced or the date of the first enrolment as a volunteer, whichever was the later. There were a substantial proportion of firefighters who did not have any incidents recorded (but were volunteers over the period that the incident data were being collected), so the SMRs, SIRs, RMRs and RIRs were recalculated after exclusion of these individuals (n=61 021).
All volunteer firefighters who were born in or before 1925 (n=1250) were removed, and the SMRs were recalculated as a sensitivity analysis. The overall SMR and overall cancer SIR for volunteers who started after their agency’s commencement date (incident volunteers) were compared with the SMR and SIR for those who had volunteered prior to observation (prevalent volunteers).
Incident groups were formed for each of the five incident categories based on the cumulative incidents total for each person-year for each firefighter. The log-normal distribution of incidents attended suggested that tertiles might result in dilution of any effect for those who had attended a large number of incidents. The cut points chosen were those defined by the tertiles from a parallel study of paid full-time firefighters (see online supplementary tables S1 and S2).13 The baseline group of 99 276 volunteer firefighters each attended ≤383 recorded incidents (≤220 fires), while group 2 included 5178 volunteers who each attended between 383 and 1053 incidents (>220 to 576 fires). Group 3 included 1795 volunteers who had attended more than 1053 incidents (>576 fires). The baseline group, for example, with total incidents of ≤383 attendances, was then used as the comparison group in internal analyses. Trend tests were conducted for the internal analyses involving duration of employment by assigning each duration of service category a numeric value (0, 1 or 2) and using this in a Poisson regression model as a continuous variable and reporting its p value.
All analyses were adjusted for age and calendar year and were undertaken using Stata software.25 The Wald-based method was used to derive 95% CIs.
Ethics approval for the study was granted by the Human Research Ethics Committees of Monash University, the State and Territory Cancer Registries, AIHW and the National Coronial Information Service.
The male cohort description is shown in table 1. The average age at the end of follow-up was 48.7 years and 63% first volunteered after 1995.
There were 4647 deaths matched on the NDI and 7057 cancers matched on the ACD among the 163 094 male volunteer firefighters. There were 177 agency-notified deaths not found on the NDI among volunteer firefighters. A further 35 volunteer firefighters were matched to the NDI before the relevant start date or before they were supposed to have commenced, indicating errors in the HR records, and were not included in the cohort analyses.
The overall risk of mortality and that of almost all major causes of death were significantly reduced for volunteer firefighters (table 2), whether those who attended incidents were included or not. The only cause of death that was significantly increased was death from fire.
There was little difference in the SMRs with the era of first enrolment (online supplementary table S3) or with increasing service duration in external (online supplementary table S4) or internal analyses (online supplementary table S5). Where trends were identified among volunteers, they appeared to be related to reducing risk with increasing service duration.
Online supplementary figure 1 shows the distribution of the number of incidents attended by individual male volunteer firefighters in the incident analyses. The mean cumulative value for recorded incidents at the end of follow-up was 71.15 (median 5.4), but there were a small number of firefighters who have attended many more incidents, and the maximum cumulative values were over 5000 (n=68). About 29% of male volunteer firefighters did not have any recorded incidents. About half (53%) of all incidents that volunteers attended were fires, 13% of all incidents were structural fires, 27% were landscape fires and 6% were vehicle fires.
After excluding those who had no recorded incidents, internal analyses using cut-offs based on paid firefighters and compared with the baseline group showed no consistent pattern of increased risk of death with increasing attendance for overall mortality or for most major causes of deaths (malignancies, nervous system, respiratory, digestive and injury) (table 3). There were, however, significantly increased RMRs for circulatory disease, more specifically for ischaemic heart disease (IHD), for attendance at incidents overall and for all types of fires. Some other RMRs were significantly increased but did not show a monotonic rise across the groups. Attendance at vehicle fires was associated with significantly increased circulatory relative mortality and, possibly, mortality from all malignancies, although the increase was not monotonic.
The volunteer records contained some very elderly volunteer firefighters, who may have died before death registries commenced in 1980 but remained on the agency records, and if included in the cohort would result in an overestimation of the expected number of deaths. Removing volunteer firefighters who were born in or before 1925 resulted in 590 volunteer firefighter deaths being removed, and the recalculated SMR was 0.54 (95%CI 0.52 to 0.55), which made a negligible difference to the overall SMRs.
There were 177 agency-notified deaths not found on the NDI among volunteer firefighters. If these individuals had been matched, the overall SMR for volunteer firefighters would have been 0.56 (95%CI 0.54 to 0.57), which means that the exclusion of these deaths made a negligible difference.
Compared with the Australian population, the overall incidence of cancer and the incidence of most major cancer categories were significantly reduced (table 4), and the risk estimates did not change when those who had not attended incidents were excluded. Prostate cancer was in excess, however, whether or not those who had attended incidents were included (table 4), and the risk increased with increasing service duration (table 5).
The prevalent volunteers’ SMR (0.53, 95% CI 0.52 to 0.55) was very similar to that for the incident volunteers’ SMR (0.55, 95% CI 0.51 to 0.58). The overall cancer SIR for prevalent volunteers’ SIR (0.85, 95% CI 0.82 to 0.87) was also similar compared with that for incident volunteers’ SIR (0.91, 95% CI 0.86 to 0.95).
Internal analyses showed no significant trend with duration for overall cancer incidence for volunteer firefighters, and the risk estimates did not change when those who had not attended incidents were excluded (table 5). There was a trend of increasing testicular cancer with increasing service among all volunteers, but when those who had not attended incidents were removed, the risk decreased. The overall cancer SIR was not increased with the era of first service; before 1970, the SIR was 0.84 (95% CI 0.81 to 0.88), and after 1995, it was 0.88 (95% CI 0.85 to 0.92) (online supplementary table S6).
There was no clear increase in risk of overall cancer incidence with increasing incident attendance group for volunteer firefighters (table 6). Firefighters showed an increased risk of kidney cancer with increasing attendance at structural and landscape fires. Incident attendance was not associated with increases in other cancer types. Lung cancer was significantly reduced compared with the general population, and longer service was associated with a trend in the reduction of lung cancer.
This is the first study to examine the cancer and mortality risk of volunteer firefighters. Volunteer firefighters have reduced mortality and cancer incidence compared with the general population. This study was sufficiently powered, allowing analyses for specific cancers by types of fire attended. The only SMR that was significantly increased was death from fire. These deaths occurred during two major bushfires and may not have been service related. Volunteers are more likely than the general population to reside in rural areas and therefore be at higher risk of bushfires when off duty.
There was evidence of an association between attendance at fires and an increased risk of death from circulatory disease for volunteers who had attended incidents. The risk of IHD was significantly increased with increased attendance at fires. Exposure to outdoor pollution, such as small particles, has been shown to be a trigger for cardiac arrests.26 There is evidence from Australia that exposure to bushfire smoke is associated with increased risk of out-of-hospital cardiac arrests.27
There was also an increased risk of developing kidney cancer with increasing exposure to fires, particularly structural fires. Kidney cancer has been found in excess in other studies of paid firefighters.6 16 28 29
Volunteer firefighters were diagnosed with prostate cancer more commonly than the Australian population, and there was an increase with years of service, which showed a significant trend and suggests that the risk may be service related, but the risk did not increase with increasing incidents attended. Elevated rates of prostate cancer have been observed in several other studies of firefighters, including the Australian study of paid firefighters,13 and a meta-analysis found a summary risk estimate of 1.28 (95% CI 1.15 to 1.43).6 Recent US studies reported an SIR for prostate cancer of 1.03 (95% CI 0.98 to 1.09)16 and an OR of 1.5 (95% CI 1.3 to 1.7)17 and an increased risk for the age category of 30–49 years (SIR 2.59, 95% CI 1.34 to 4.52) for paid firefighters.20
In recent years, there has been a large increase in prostate cancer diagnosis because of the availability of screening tests. However, to explain the increased rates of prostate cancer in volunteer firefighters resulting from a diagnostic bias, firefighters would have to have been more likely than the general population to be screened. The fire agencies involved in this study do not offer prostate cancer screening to firefighters, but they may carry out other health promotion activities. This may make firefighters seek health screening more readily than the general population.
Lung cancer was reduced in this study, and risk decreased with increasing years of service, showing a statistically significant trend. It was also not increased in most previous studies.15 30 However, a significant excess in lung cancers (SIR 1.12, 95% CI 1.04 to 1.21) was found in a recently reported cohort of US paid firefighters.16 A follow-up study of career firefighters in Nordic countries also observed an excess of lung cancer, mainly lung adenocarcinoma among older firefighters (SIR 1.90, 95% CI 1.34 to 2.62).20
Those who volunteer to become firefighters in Australia appear to have a reduced mortality and cancer incidence compared with the general population. This suggests that a ‘healthy-volunteer’ effect is operating in the same way as the healthy-worker effect was shown in the Australian paid firefighters study.13 The mortality of occupational cohorts is typically lower than that of the general population. This ‘healthy-worker’ effect9 31 has been shown in other firefighter cohorts.32–36 The reduction in observed cancers may also be a result of an equivalent ‘healthy-volunteer’ effect. Volunteer firefighters do not have to pass the strict entry requirements for fitness applied to paid firefighters, and they are only required to disclose certain pre-existing medical conditions; however, members of the community who are ill are probably less likely to volunteer than those who are well. Further, individuals may become volunteers after normal retirement age, cancer rates increase with age and it is probable that those with cancer (or other illnesses) would be less likely to volunteer creating a ‘healthy-volunteer’ selection effect.
The study design reduced the impact of the healthy-worker survivor effect by identifying a follow-up commencement date at which there was a complete roll of volunteer firefighters. This means that firefighters who may have subsequently left because of ill health were included in the cohort. The average age of the cohort at the start of follow-up was 38, and the average age at the start of volunteering was 31, which suggests that few volunteers have been lost.
There was a concern that volunteers who were not recorded as having turned out to an incident but were still registered as volunteers were inactive because of ill health. Volunteers may have remained a member of the organisation, for example, taking part in social and fundraising activities but no longer be fit enough to turn-out. When those individuals with no recorded incidents were removed from the analyses, the risk estimates did not change.
In an occupational cohort, the HR records can be expected to be accurate for most individuals in terms of the start and end dates of employment. The records of volunteer agencies, in the absence of the need for salary payment, may not be as accurate. There may be no formal process for resignation from the service, resulting in apparently immortal volunteers. The presence of a number of very elderly individuals in the records did not appear to have greatly affected the SMR, and the removal of these born before 1925 had little effect. All volunteers are followed up until death, and all deaths are identified through the national registration scheme. The effect of a lack of resignation would be to extend the years of service, but analysis by incidents attended is not affected. Further, the overall SMR and cancer SIR for incident volunteers did not differ from those of prevent volunteers. This suggests that any selection bias is small. A small proportion of records were missing names, and about 10% of the volunteer records that were supplied were missing birth dates, which suggest that the records were reasonably complete.
The contemporarily collected and individually attributable incident data are a major strength of this study and a noted limitation of previous studies of firefighters. Most previous studies have used duration of employment as a surrogate for exposure. Duration of employment has been shown to be inaccurate as an estimate of exposure.37 The incident data in this study identified the type and numbers of incidents including fires attended for each individual. It was not possible to do analyses by attendance duration for individual firefighters. It is likely that landscape fires may go on for days or even weeks, which is rare for structural or vehicle fires. A simple count of the incidents does not consider length of exposure. The removal of volunteers who had not attended incidents had little effect on the cancer incidence or mortality risk estimates compared with the general population.
Although incident reporting is compulsory, it is unclear whether all incidents and all volunteer attendees at incidents are recorded. After examination of the records and discussion with one fire agency, their volunteer firefighters were excluded from the incident analyses, because there appeared to have been under-reporting of attendance. It has been suggested that volunteers may join when single, become less active if they have small children and then become more active again when they retire from paid employment. The use of incident data will therefore be a more accurate portrayal of attendance than years as a volunteer.
This records-based study could not assess potential confounders for some outcomes such as individual lifestyle factors, including ethnicity, smoking, alcohol consumption, diet or non-firefighting job exposures experienced, for example, in previous and current jobs. Neither race nor ethnic origin was identified in agency records, nor are they captured in the AIHW cancer or mortality data and, therefore, could not be investigated. Aboriginal and Torres Strait Islanders make up approximately 2% of the Australian population. Genetic factors play a part in the risk of some diseases; for example, those with paler skin38 are more susceptible to melanoma. For melanoma, sun exposure is an important risk factor, but no information was available on this for the cohort members on an individual basis. Many of the volunteers live in rural areas, so they are likely to include a higher proportion of men with outdoor occupations, such as farming or forestry, than in the general population. They would therefore be likely to have more sun exposure than the general population, but no excess melanoma was identified, although this has been seen among paid firefighters.17
Tobacco smoking is known to be a significant cause of IHD, and data from one volunteer agency suggest that the male volunteers are less likely to smoke than the general population. Mortality from IHD, chronic obstructive pulmonary disease and lung cancer incidence are all reduced compared with the general population, which may be a result of this lower smoking rate. However, it was not possible to get smoking data on the cohort on an individual basis, which is a limitation of the study.
The power of this study to determine risk associated with cancer or specific causes of death was limited by the relatively young age of the cohort: The mean age at the end of follow-up was under 50 years, and the comparatively short period of follow-up is less than 10 years in most cases.
This is a cancer incidence study that provides a more complete picture of cancer diagnosis than cancer mortality studies. As with earlier diagnoses and better cancer treatments, cancer is not necessarily fatal. Cancer diagnosis registration is mandatory in all Australian states and territories, and registration is virtually complete.39 The exception is for non-melanotic skin cancers, which are not collected. All primary cancers are collected by the ACD, and about 8% are cancers that arose after the primary cancer. It is possible that some of these may be treatment related, but these cannot be identified in the national reference data (nor, therefore, in the cohort); therefore, it was not possible to carry out a sensitivity analysis to assess their impact.
Male volunteer firefighters have significantly decreased mortality and overall cancer incidence compared with the general Australian population. This is likely to be a result of a healthy-volunteer effect and, possibly, lower smoking rates. The pattern of mortality findings are similar to that of paid firefighters, but their SIRs are lower. All cancers, prostate cancer and melanoma were found to be elevated in paid firefighters, prostate cancer was the only cancer found in excess for volunteers.
There was an increase in IHD, which was related to attendance at incidents. Kidney cancer may also be incident related, and prostate cancer was elevated, but this was not related to incidents attended.
While this study has some strengths, including the contemporary incident data, the very large cohort size and the ability to access nationally complete death and incident cancer databases, there are some limitations, such as the relatively young age of the cohort and no information being available about individual lifestyle factors such as jobs held and individual smoking status.
Other studies of volunteer firefighters should ensure they correctly identify and include only firefighters who turn out to incidents. Our data show that not all volunteers will attend incidents, and almost 8% of those who had turned out had attended only one incident.
Continued follow-up of the cohort is recommended, because the cohort is relatively young, with small numbers of deaths and cancers to date, and further follow-up time will increase statistical power, especially for the type of incident analyses.
The study was funded by the fire agencies through the national council and industry peak body, the Australasian Fire and Emergency Service Authorities Council.
Monash University acknowledge the assistance of the participating agencies in compiling the data for the cohort. We thank the members of the advisory committee and the technical reference group for their assistance. We also acknowledge the assistance provided by the Australian Institute of Health and Welfare for the timely linkage of the cohort and provision of reference data.
Contributors All authors contributed to the design and execution of the study and to the paper.
Competing interests None declared.
Ethics approval Monash University Human Research Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
Collaborators Australian Institute of Health and Welfare.
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.