Caution: Some people may find parts of this content confronting or distressing.
Please carefully consider your needs when reading the following information about suicide and self-harm. If this material raises concerns for you contact Lifeline on 13 11 14, or see other ways you can seek help.
The information included here places an emphasis on data, and as such, can appear to depersonalise the pain and loss behind the statistics. The AIHW acknowledges the individuals, families and communities affected by suicide each year in Australia.
Aboriginal and Torres Strait Islander readers are advised that information relating to Indigenous suicide and self-harm is included.
The AIHW supports the use of the Mindframe guidelines on responsible, accurate and safe suicide and self-harm reporting. Please consider these guidelines when reporting on statistics on the monitoring of suicide and self-harm.
The AIHW National Mortality Database (NMD) contains records for deaths in Australia from 1964 to 2020. The database comprises information about causes of death and other characteristics of the person, such as sex, age at death, area of usual residence and Indigenous status.
The Cause of Death Unit Record Files are provided to the AIHW by the Registries of Births, Deaths and Marriages in each state and territory and the National Coronial Information System (managed by the Victorian Department of Justice). The cause of death data are compiled and coded by the Australian Bureau of Statistics (ABS) to the International Statistical Classification of Diseases and Related Health Problems (ICD) and maintained at the AIHW in the NMD. Registration of deaths is the responsibility of the Registry of Births, Deaths and Marriages in each state and territory.
To improve the quality of data, the ABS annually revises the causes of death for coroner-referred deaths to reflect the latest available information. This process applies to deaths registered after 1 January 2006. Deaths registered in 2017 and earlier are based on the final version of cause of death data; deaths registered in 2018 are based on the revised version; and deaths registered in 2019 and 2020 are based on the preliminary version. Revised and preliminary versions are subject to further revision by the ABS. For a more detailed description of the coverage and processing of deaths data, including deaths certified by the coroner, refer to the Explanatory Notes in ABS Causes of death, Australia (ABS Catalogue No. 3303.0), which is available from the ABS website
In the NMD, both the year in which the death occurred and the year in which it was registered are provided. Year of registration has been used for the purposes of monitoring deaths by suicide. Deaths based on the year the death occurred have also been presented; however, as some deaths at the end of each calendar year may not be registered until the following year, year of death information for the latest available year (2020) is generally an underestimate of the actual number of deaths that occurred in that year.
The data quality statements underpinning the AIHW NMD can be found on the following ABS internet pages:
For more information on the AIHW NMD see National Mortality Database and About National Mortality Database.
The Aboriginal and Torres Strait Islander status of a deceased person is captured through the death registration process; however, it is recognised that not all such deaths are captured through these processes, leading to under-identification. Also, data on deaths by suicide in Indigenous people have been compiled by jurisdiction of usual residence for New South Wales, Queensland, Western Australia, South Australia and the Northern Territory only. Data for Victoria, Tasmania and the Australian Capital Territory have been excluded in line with national reporting guidelines.
Data for patients who were hospitalised with intentional self-harm injuries are sourced from the AIHW’s National Hospital Morbidity Database (NHMD). Most of the data used for the monitoring of hospitalisations for intentional self-harm are from 2008–09 to 2020–21. For each reference year, the NHMD includes all hospitalisations for patients who were discharged between 1 July and 30 June.
The NHMD is a compilation of episode-level records from admitted patient morbidity data collection systems in Australian hospitals. It is a comprehensive data set that has records for all episodes of admitted patient care from essentially all public and private hospitals in Australia.
The data supplied are based on the National Minimum Data Set (NMDS) for Admitted Patient Care and include administrative, demographic, clinical and length of stay data, as well as data on the diagnoses of the patients, the procedures they underwent in hospital and external causes of injury and poisoning.
The purpose of the NMDS for Admitted Patient Care is to collect information about care provided to admitted patients in Australian hospitals. The scope of the NMDS includes episodes of care for admitted patients in all public and private acute and psychiatric hospitals, free standing day hospital facilities, and alcohol and drug treatment centres in Australia. Hospitals operated by the Australian Defence Force, corrections authorities and in Australia’s off-shore territories are not in scope but may be included. Hospitals specialising in dental, ophthalmic aids and other specialised acute medical or surgical care are included.
episode of care: The period of admitted patient care between a formal or statistical admission and a formal or statistical separation, characterised by only one care type (see care type and separation). METeOR identifier: 268956.
separation: The process by which an episode of care for an admitted patient ceases. A separation may be formal or statistical. METeOR identifier: 327268.
formal separation: The administrative process by which a hospital records the cessation of treatment and/or care and/or accommodation of a patient.
statistical separation: The administrative process by which a hospital records the cessation of an episode of care for a patient within the one hospital stay.
The criteria used to describe intentional self-harm hospitalisations reported in Suicide & self-harm monitoring is described in the Codes and classifications section.
States and territories are primarily responsible for the quality of the data they provide. However, the AIHW undertakes extensive validations on receipt of data, checking for valid values, logical consistency and historical consistency. Where possible, data in individual data sets are checked with data from other data sets. Potential errors are queried with jurisdictions, and corrections and resubmissions may be made in response to these queries. Except as noted, the AIHW does not adjust data to account for possible data errors or missing or incorrect values.
The most recent Data quality statement for Admitted Patient Care is available in METeOR. The Data Quality Statement contains information on other changes that may affect interpretation of the data for the relevant year.
While the Indigenous status data in the Admitted Patient Care NMDS for all states and territories are considered of sufficient quality for statistical reporting, separations for Aboriginal and Torres Strait Islander people are generally under-enumerated. In 2011–12, about 88% of Indigenous Australians were identified correctly in hospital admissions data, and the ‘true’ number of separations for Indigenous Australians was about 9% higher than reported. Caution should be used in the interpretation of Indigenous status data because of the under-enumeration overall and differences in under-enumeration among the states and territories. The quality of the data for private hospitals is not known, but likely to be poor.
The National Ambulance Surveillance System (NASS) is a new public health monitoring system providing timely and comprehensive data on intentional self-harm (including suicidal behaviours with self-injurious intent), mental health, and alcohol and drug harms in the community. Data for the National Ambulance Surveillance System (NASS) are compiled by Turning Point in partnership with Monash University and are sourced from paramedic electronic patient care records provided by Australian state and territory-based ambulance services. As part of the National Suicide and Self-harm Monitoring Project, the AIHW has contracted Turning Point through Monash University to develop the National Ambulance Surveillance System (NASS) for self-harm related attendances. Self-harm (suicidal ideation, suicide attempt, self-injury) related modules from the NASS are reported here.
The ambulance attendance data includes 1-month per quarter snapshots from New South Wales (NSW), Victoria (Vic), Tasmania (Tas) and the Australian Capital Territory (ACT) from 2018 to 2021, and Queensland (Qld) from 2020 to 2021. AIHW began receiving monthly snapshot data for NSW, Vic, Qld, Tas and the ACT from January 2021.
Information is obtained and coded through manual scrutiny of de-identified electronic patient care records (ePCRs), including paramedic clinical assessment, patient self-report, information from third parties and other evidence at the scene, such as written statements of intent (including social media, text messages and written notes), as recorded by paramedics. Intent of self-harm behaviours derived from the ePCR may be from either stated or physical evidence, or where there is evidence but the patient may have denied the behavioural intent (Lubman et al. 2020).
Self-harm related ambulance attendances are included if self-harm occurred in the preceding (past 24 hours) or during the ambulance attendance, with 4 categories of self-harm related ambulance attendances defined and coded as:
Suicide, suicide attempt and suicidal ideation are considered mutually exclusive; however, self-injury could be simultaneously coded with any other self-harm case category.
The number of attendances related to suicide is under-represented as ambulances do not attend all deaths. Furthermore, when they do attend there may be insufficient information to determine suicidal intent at the scene.
Methods of suicide, suicide attempt or suicidal ideation are coded, as are methods of self-injury and categories of suicidal ideation preparation (planned, unplanned and unknown if planned) using a modified ICD-10 coding framework.
For more information see Lubman et al.2020.
Data are collected for operational rather than monitoring or research purposes with paramedics only recording information that they either observe or is provided to them by the patient or bystanders, and which they deem clinically relevant to patient care. It is possible that relevant information with respect to self-harm or mental health variables is not recorded, or similar events may not be recorded consistently by different paramedics over time.
The Multi-Agency Data Integration Project (MADIP) is a partnership among Australian Government agencies to develop a secure and enduring approach for combining information on healthcare, education, government payments, personal income tax, and demographics (including the Census) to create a comprehensive picture of Australia over time (ABS 2018). The key MADIP datasets used in this analysis were:
In order to identify socioeconomic factors associated with deaths by suicide in Australia, 2011 Census and 2011 to 2017 Causes of Death data were linked to the ABS Person Linkage Spine (Spine). The Spine is comprised of all persons in the Medicare Enrolments Database, Personal Income Tax or Social Security and Related Information data sets at any point between 2006 and 2016 (ABS 2019). As the baseline population, 2011 Census was considered a closed population and several assumptions were made about this population. These include:
Table 1 shows the linkage coverage of Census 2011 and deaths by suicide from the ABS Causes of Death. The Estimated Residential Population of Australia at 30 September 2011 was 22.43 million people (ABS 2021). Of these, 20,739,159 were accounted for in the Census 2011, noting that the Census 2011 started in August 2011. In total, the linked Census 2011 population was 16,700,062 (74.4% of the total Australian population of September 2011). According to the National Deaths Index, there are 17,306 deaths by suicide from September 2011 to December 2017, of which 11,580 (67%) deaths by suicide were linked to the linkable Census 2011 data. Suicide was defined by ICD-10 external cause codes X60–X84 and Y87.0
Total (n)
Linked (n)
Linked (%)
ERP(a) at Sept 2011
22,432,771
16,700,062
74
Deaths by suicide(b)
17,306
11,580
67
To address the issue of unlinked deaths by suicide and 2011 Census records, an imputation weighting technique was used. This section describes the method used to develop these weights, which involved a three-staged approach.
First stage: imputing weights to scale up the Census population. The ABS historical ERP for 31 December 2011 by states, sex and 5-year age groups were used to derive weights by these demographic characteristics, based on the assumption that there were no significant differences in the age distribution of the population. The derived weight was applied at the person level for each record of Census that has ABS Person Linkage Spine (Spine) information to enable analysts to weight the analyses to the 31 December 2011 total ERP.
Unlike the original ABS research paper (ABS 2016) describing the creation of a linked data set between 2011 Census and deaths registered in the following 13 months, the imputation method did not calculate weights by Indigenous and non-Indigenous populations. Also, note that Diplomatic personnel resident in Australia have not been excluded from total ERP.
Second stage: suicide weights were calculated by using all deaths by suicide from 2011 to 2017 by states and territories, sex and 5-year age groups. Suicide weights were then applied at person level to only those linked Census records with suicide information. This made it possible to weight the analyses to all deaths by suicide (18,848) from 2011–2017.
An issue with applying suicide weights is that suicide weights are slightly higher when compared with population weights applied in the first stage. As such, the combined weights of the linked records with both 2011 Census and suicide information when aggregated, the weighted ERP will be slightly higher than that of 31 December 2011. Hence the need for a scale down adjustment factor.
Third stage: Finally, a scale down adjustment factor, derived based on total ERP, linked deaths by suicide and all deaths by suicide, was applied at the person level to only Census records without linked death by suicide information. Hence the weights of the Census population with or without linked death by suicide information, aggregated to the 31 December 2011 ERP (22,340,025).
Australian residents in the 2011 Census, weighted to 31 December 2011 estimated resident population (ERP) and linked to ABS Causes of Death data from 2011 to 2017 created a binary outcome of either died by suicide (ICD 10 external cause codes X60–X84, Y87.0) or not. Note that deaths by suicide used in this analysis are based on year of occurrence. These may differ from deaths by suicide data used in other AIHW publications which are based on year of registration. In addition to the closed population assumptions noted above, due to data quality issues the age in this analysis is at the time of the 2011 Census except for those who have died by suicide.
Over the period 2011 to 2017, Australia recorded more than 18,800 deaths by suicide of people who were in the 2011 Census. This resulted in a cumulative incidence of about 84 per 100,000 people during the 7-year period. The cumulative number and incidence of deaths by suicide that occurred over the 7 years varies considerably by sex, educational attainment and labour force status.
All data are subject to some level of uncertainty. For the data presented in this analysis the sources of uncertainty include:
Linkage error: Uncertainty is introduced when there is error in linking data sets. The data used in this report carries some risk of linkage error. An attempt has been made to reduce this error through imputation weighting process but some uncertainty remains.
Timeliness of data: Some of the data used in this analysis is Census data collected in August 2011. A person’s education status and employment status can change over time, particularly for certain population groups. The use of out-of-date information introduces a source of error to the analysis.
Randomness in the number of deaths by suicide that occur in a given time period, 2011–2017: The number of deaths by suicide that occur in a given time period fluctuate, even if the underlying population risk remains the same. The exact distribution of the counts is unknown. With deaths by suicide being a rare event it is often assumed that the counts follow a Poisson distribution. If this is the case then the relative level of uncertainty due to randomness decreases as the number of deaths by suicide increase.
The MADIP datasets used in this modelling are outlined in the Data section of these Technical notes. In this analysis, only people aged 25 to 64 years in the linked 2011 Census have been included, representing, over 9 million people in the 2011 Census and 7,000 deaths by suicide from 2011 to 2017. This age group was chosen because most deaths by suicide occur between these ages and because of the relative stability of socioeconomic factors over time (such as level of education) among this age group. While suicide is the leading cause of death among people aged 15 to 24 years, people in this age group were excluded from the modelling because of their lack of socioeconomic stability.
Missing values have been excluded from this analysis. Educational attainment has the highest proportion of missing values (5.5%). Unlike with the cumulative suicide risk estimations, the data used in the regression modelling has not been weighted.
To identify modelling predictors and explore their association with suicide deaths, an extensive literature review of social factors was carried out. This included earlier analyses published by AIHW, which showed deaths by suicide varied by factors such as employment and educational attainment.
Socioeconomic factors identified from the 2011 Census were used as predictors and deaths by suicide as the outcome variable. A total of 10 factors were included:
Two modelling approaches were tested: Poisson regression and competing-risks regression (as described by Fine & Gray 1999). For Poisson regression, counts of the outcome variable with the value 1 for deaths by suicide and 0 for those who did not die by suicide were created and data aggregated by socioeconomic factor.
For the competing-risks regression, the influence of other causes of death is considered. This is because people who died from any other causes (such as cancer and coronary heart disease) are no longer at risk of dying by suicide.
Sex-stratified and Indigenous-stratified multivariate models were also fitted to investigate the associations within males and females, and within Indigenous and non-Indigenous people. Due to data quality issues including small sample sizes, Indigenous-stratified models have not been published. Univariate and multivariate models (including quasi-Poisson to deal with slight overdispersion) were also refitted. The coefficients obtained were back transformed so they could be interpreted as rate ratios (for Poisson models) and subhazard ratios (for competing-risks models). Analysis was conducted using R (glm package) and Stata (version 16) software.
Of the models tested, competing-risks regression, a method that accounts for people being censored from the risk set because of a competing cause, was used to estimate the risk of death by suicide and the selected socioeconomic factors. Univariate, multivariate and sex-stratified competing-risks models were developed. Generally, competing-risks regression models can be regarded as an extension of the Cox proportional hazards model, where subjects who experience competing events (deaths from other causes) are adequately counted as not having any chance of dying by suicide.
Estimated coefficients of competing-risks models can be interpreted in a similar way as coefficients estimated from a Cox model, except that they estimate the effect of certain covariates in the presence of competing events. Note that the transformed coefficients are known as subhazard ratios, similar to hazard ratios estimated in Cox regression. The subhazard ratio can be interpreted as a rate ratio (Henan 2010), but here we are considering the relative change in rates of the event in those subjects who are either currently event-free or who have previously experienced a competing event (Austin & Fine 2017). For simplicity and ease of understanding, coefficients in this report are referred to as hazard ratios.
Researchers from the Australian National University’s Centre for Social Research and Methods (CSRM), in close collaboration with the AIHW, have extended the analysis Regression risk models for selected census variables. An extract from the Multiagency Data Integration Project (MADIP) was used. For this analysis, the following MADIP datasets were utilized. 2011 Census information, 2011-2016 Personal income tax (PIT) records, 2011-2016 Social security and related information (SSRI) and 2011-2016 cause of death data. The data linkage process was identical to the linkage process described in the Data section of these Technical notes.
The following set of explanatory variables were used to capture the social and economic factors. More specifically, the CSRM used the time-invariant 2011 Census data to capture social factors and time-variant PIT and SSRI to capture economic factors of suicide.
Variable
Source
Type
1. Highest level of education
2011 Census
Social
2. Age
3. Sex
4. Indigenous status
5. Need for assistance with core activities
6. Household composition
7. Total income
2011 – 2016 PIT
Economic
8. SSRI
2011 – 2016 SSRI
In addition to the variables presented in Table 1, the following variables were derived using Census, PIT and SSRI data.
The MADIP dataset is a longitudinal dataset therefore, utilizing longitudinal regression methods allows us to control for unobserved individual heterogeneity across the time period. For example, longitudinal regression methods would allow us to control for unobserved measures such as behavioural differences and cultural differences across individuals. Given that suicide is a complex individual decision and may not be entirely explained by observed variables, controlling for these unobserved heterogeneity would allow us to obtain unbiased estimates.
In this analysis, two competing longitudinal regression methods were tested – a random effects logistic model and a population-averaged logistic model. Both models were estimated with robust standard errors to account for heteroskedasticity.
Given that the random effects logistic model is dependent on the strong assumption that the underlaying variation have no serial correlation, the population-averaged logistic model was selected as the primary regression method due to its robustness to serial correlation (Hill et al., 2010). Furthermore, a likelihood-ratio test was also undertaken to decide between the two models. The result suggested that the population-averaged model was more appropriate than the Random effects model. Equation (1) and (2) outline the population-averaged model.
Setting up the dataset as a panel allowed us to account for individual heterogeneity. However, given that suicide is a rare event, explanatory variables with large number of categories (for example – occupation) were excluded from the longitudinal analysis. This was primarily due the low number of suicides each year, which in turn made the models more sensitive to variables with a large number of categories. Given this drawback of panel data, a cross-sectional analysis was also conducted as a part of the sensitivity analysis.
1 Year 2011 was not included in the analysis.
Hill, R.C., Griffiths, W.E. and Lim, G.C., 2010. Principles of econometrics. pp 537-560. John Wiley & Sons.
In addition to the NMD, the Australian Defence Force (ADF) suicide monitoring analysis used the following data sources:
The NDI is managed by the AIHW and contains person-level records of all deaths in Australia since 1980 obtained from the Registrars of Births, Deaths and Marriage in each state and territory. Its use is confined to data linkage studies approved by the AIHW Ethics Committee for health and medical research. NDI records are supplemented with cause of death information from the NMD (AIHW 2018). The NDI is used in linkage with the Personnel Management Key Solution (PMKeyS) and Defence Suicide Database (DSD) to create the linked PMKeyS–NDI data set used in analysis of deaths by suicide in the ADF population.
PMKeyS is a Defence staff and payroll management system that contains information on all people with ADF service on or after 1 January 2001 (when the system was introduced). This database contains demographic and service information at a given point in time and is linked to the NDI to identify deaths, including deaths by suicide, in the 3 ADF service status groups.
The DSD is maintained by Defence and contains information on suspected and confirmed deaths due to suicide of personnel serving full time since 1 January 2000. Suspected and confirmed deaths by suicide are included in the database only on the advice of the ADF Investigative Service. Cases are confirmed by receipt of a coronial finding of death by suicide. This database is linked to the PMKeyS and NDI and records with a status of ‘confirmed’ are used to supplement cause of death information from the NDI for numbers of deaths by suicide only.
For further information see Technical notes of the National suicide monitoring of serving and ex-serving Australian Defence Force personnel: 2021 update.
Estimates of fatal (years of life lost, YLL) and non-fatal burden (years lived with disability, YLD) were sourced from the Australian Burden of Disease Study (ABDS) 2015. The ABDS 2015 used burden of disease analysis to measure the impact of 216 diseases and injuries on the health of the Australian population. The study provides a detailed picture of the burden of disease and injury in the Australian population in 2003, 2011 and 2015. It also includes estimates of the contribution made by selected risk factors on the disease and injury burden in Australia, and by socioeconomic areas for some risk factors.
The ABDS 2015 uses and adapts the methods of global studies to produce estimates that are more relevant to the Australian health policy context. The chosen reference period (2015) reflects the data availability from key data sources (such as the National Health Survey, deaths data, hospital admissions data and various disease registers) at the time of analysis.
Results from the study provide an important resource for health policy formulation, health service planning and population health monitoring. The results provide a foundation for further assessments.
Full details on the various methods, data sources and standard inputs used in the ABDS 2015 are available in Australian Burden of Disease Study 2015: methods and supplementary material.
The Victorian Suicide Register (VSR) contains data on all suicide deaths reported to the Coroners Court of Victoria (CCOV) since 2000 to the present. The Coroners Prevention Unit, a specialist investigative service for Victorian Coroners, manages the VSR since its implementation in 2012. The VSR operates with funding from the Victorian Department of Health.
VSR data is used to inform investigations into suspected suicides and support coronial recommendations to prevent similar deaths. VSR data are regularly shared with the Victorian Department of Health, AIHW, Victoria Police and other organisations involved in suicide prevention.
Data on the VSR is obtained and coded from materials gathered throughout the course of coronial investigation, including police notification of death, forensic reports (autopsy and toxicology), witness statements and medical records.
CCOV releases a range of suicide data reports from the VSR on their website, including information on suspected deaths by suicide on a monthly basis, and overviews of Aboriginal suicides in Victoria.
The Queensland Suicide Register (QSR) was established in 1990, containing records on all confirmed suicide deaths in Queensland from 1990 to 2017. The interim Queensland Suicide Register (iQSR) is a suicide surveillance system, providing real-time data on probable deaths by suicide from its development in 2011 to the present day. The Australian Institute for Suicide Research and Prevention (AISRAP) at Griffith University manages the databases, with funding from the Queensland Mental Health Commission.
The databases contain information from the Coroners Court of Queensland and Queensland Police Service, and checked against the National Coronial Information System. Data sources include post-mortem examination report (including toxicology), report of death from police, and psychological autopsy report.
AISRAP publishes an annual report on suicide in Queensland on their website, The Queensland Suicide Register (QSR).
The New South Wales (NSW) Suicide Monitoring and Data Management System contains data on all suspected and confirmed suicide deaths between 2019 to present. Established in October 2020, the system is a collaboration between the NSW Ministry of Health, Department of Communities and Justice, the State Coroner and NSW Police.
The suicide surveillance system contains initial police information of suspected suicide deaths from the JusticeLink information system, which is managed by NSW Department of Communities and Justice. The information is manually checked against other data sources, including coronial determination.
NSW Health publishes monthly reports on suspected deaths by suicide in New South Wales on their website, NSW Suicide Monitoring System - Towards Zero Suicides
This website needs JavaScript enabled in order to work correctly; currently it looks like it is disabled. Please enable JavaScript to use this website as intended.
We'd love to know any feedback that you have about the AIHW website, its contents or reports.
The browser you are using to browse this website is outdated and some features may not display properly or be accessible to you. Please use a more recent browser for the best user experience.