Introduction
Age-specific suicide rates in Australia are highest among individuals over the age of 45 years. In 2022, the highest increase in suicide rates was reported in males aged 55–64 years, in 2021, the highest rate was for males aged 85 years and over (Australian Institute of Health and Welfare, 2022). Females aged 45–49 years represent a second peak in lifetime rates of sex- and age-specific hospitalizations for intentional self-harm (ISH), after their first peak in young adulthood (Australian Institute of Health and Welfare, 2020). Nonfatal ISH is relatively steady for all ages in males (Australian Institute of Health and Welfare, 2020). Higher rates of fatal intent in ISH have been reported in those aged over 60 years compared with younger persons, and a greater proportion of non-fatal ISH outcome has been described for females, compared with males (Hawton and Harriss, Reference Hawton and Harriss2008).
Among people with psychiatric disorders, mortality has been shown to be up to 10-fold higher from physical disorders as it has from suicide (Kisely et al., Reference Kisely, Xiao, Lawrence and Jian2014). Australian research findings identified mandatory community-based mental health treatments among an inception cohort of people with psychiatric disorder to significantly reduce all-cause mortality up to 3 years following index presentation (Steve et al., Reference Steve, Neil, Jianguo, David, Sandra and Elizabeth2013).
The role of community mental health (CMH) care in reducing rates of suicidal behaviors has been previously explored (Boulet et al., Reference Boulet, Gagnon, Nadeau, Mowbray and Mercier2022, Lapierre et al., Reference Lapierre2011, Okolie et al., Reference Okolie, Dennis, Simon Thomas and John2017, Rihmer et al., Reference Rihmer, Rutz and Pihlgren1995, Vasiliadis et al., Reference Vasiliadis, Lesage, Latimer and Seguin2015); however, access to CMH care is not consistent. The quality and consistency of healthcare are known to contribute to poor health and excess mortality among people with mental disorders (Druss et al., Reference Druss, Von Esenwein, Compton, Rask, Zhao and Parker2010). When able to be provided in a timely manner following ISH, there is evidence that it is a crucial point for early intervention for suicidal ideation, particularly within male populations and older ages where suicidality may not be so explicit (Almeida et al., Reference Almeida2012, De Leo et al., Reference De Leo, Draper, Snowdon and Kõlves2013, Deisenhammer et al., Reference Deisenhammer, Huber, Kemmler, Weiss and Hinterhuber2007, Schou Pedersen et al., Reference Schou Pedersen, Fenger-Grøn, Bech, Erlangsen and Vestergaard2019).
Emergency departments are often not ideal places to provide mental health care, and their primary focus often being the physical treatment for the consequences of the self-harm. The first priority is to stabilize the patient’s physiology, which may require hospital admission. Treating clinician referrals to CMH care from the Emergency Department or hospital inpatient ward prior to discharge are vital, with care coordination and follow-up more likely to ensure engagement. Older adults attending ED’s more frequently report social disconnection than their younger counterparts (Kandasamy et al., Reference Kandasamy, Platts-Mills, Shah, Van Orden and Betz2018); in rural areas both social isolation and poorer access to high-quality CMH care can further impair outcomes (Lecamwasam et al., Reference Lecamwasam, Gupta and Battersby2022).
The primary aim of the study was to examine the impact of CMH care following index hospital-treated ISH on all-cause mortality and measure the impact of age by comparing the middle-aged (45–64), young-old (65–85) and “old” old (85+). A secondary aim was to describe patterns of CMH care surrounding index hospital-treated ISH.
Methods
We included presentations for ISH by people aged 45 years or older to any NSW ED or hospital inpatient department (“hospital treated”) with an arrival date between January 1, 2014, and December 31, 2019. A longitudinal whole of population-level dataset was analyzed to examine patterns of CMH care prior to and following a hospital-treated, nonfatal index episode of ISH among people aged 45 years or older in NSW and all-cause mortality during the study period following index ISH. As the data were longitudinal and contained any presentation to hospital for any reason during the time period, as long as they had at least one episode of ISH, where there was more than one episode for an individual, the index was taken as the first occurring in chronological order. Day only admissions for psychiatric care were not included in the count of psychiatric hospital admissions following ISH. Using a unique person identifier allocated by the NSW Centre for Health Record Linkage, we sequentially linked records at the individual level across the Admitted Patient Data Collection (APDC), Emergency Department Data Collection (EDDC), Cause of Death Unit Record File (CODURF), Registry of Births, Deaths, and Marriages (RBDM), and CMH datasets.
Data collection and cleaning
Data were cleaned using SAS Enterprise Guide (v7.1). See Supplementary 1 for a full description of data cleaning. The APDC and EDDC datasets were used to identify all individuals aged 45 years or older with an index event of hospital-treated, nonfatal ISH in NSW, occurring during the study period (Supplementary 2). Hospital-treated ISH incidents were identified using two methods, given the fact that ED diagnosis codes alone have low sensitivity for detecting suicidal thoughts or behaviors. First, standard self-harm codes (X60-X84, Y87.0) were extracted using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD10-AM) – an international system used to classify diseases, injuries, and related health problems (Australian Consortium for Classification Development, 2017). Suicidal ideation was identified by the ICD-10-AM code R45.81. Second, NSW Health adopts an enhanced system for routine reporting of self-harm or suicidal ideation presentations (Sara et al., Reference Sara2023). Systematized Nomenclature of Medicine Clinical Terms (SNOMED) terms were first transformed to ICD10-AM codes using mapping tables supplied by the Australian Digital Health Agency, and further codes denoting self-harm were added by NSW Health. These additional diagnostic codes covered specific substances often used in deliberate self-harm (paracetamol, sedatives, and antidepressants) as an indication of self-poisoning. Additionally, this method used regular expressions such as “self-harm,” “suicidal,” or “overdose” across the presenting problem text. Self-harm and suicidal ideation presentations identified in these ways were considered as a single category, as they both have significant association with suicide and also require CMH assessment and intervention (Large et al., Reference Large, Corderoy and McHugh2021).
Given ED patient records do not record the external cause codes from the ICD10-AM dictionary, whereas the admitted patient records do, the principal diagnosis given within the ED dataset was used to classify the main reason for admission for ED only presentations, and the APDC dataset used to classify the same for admitted persons.
Measurements
The index data were evaluated to describe the cohort, using demographic characteristics isolated at the index series including age category at index (45–64, 65–84, and 85+), sex (male or female), Indigenous status which included Aboriginal and/or Torres Strait Islander (yes or no), marital status (never married, married/defacto, divorced, widowed, and unknown), and principal diagnosis at index ISH presentation, derived using ICD10-AM and SNOMED codes (poisoning, mental disorder/suicidal ideation, injury, substance use, chronic condition, and other/missing). Chronic conditions included cancer, neurodegenerative disorders such as Parkinson’s disease, gastrointestinal disorders such as alcoholic liver disease, and cardiac conditions such as cardiomyopathy. ICD10-AM codes used to identify chronic conditions are included in Supplementary 3.
Key outcome measures were all-cause death following (but not including) index presentation, identified by linking the CODURF and RBDM data (Supplementary 4), CMH presentation within 14 days following index and hospital (ED or inpatient), readmission within 14 days following index. Survival was determined when persons had no RBDM/CODURF record in the study period, and all persons remaining alive on December 31, 2019, were right censored.
The underlying cause of death was extracted from ICD-10 codes listed as cause of death in CODURF datasets; for this study, we considered mortality from all causes, not just suicide mortality.
Statistical analysis
All statistical analyses were conducted using STATAv17 (STATACorp, 2017). We calculated descriptive statistics for all measures and compared crude differences between the groups of interest, namely those who received CMH care within 14 days of index ISH presentation with those who did not. Pearson’s chi-squared test of independence was used to determine p-value statistics for binary or categorical variables, concordant with appropriate data sizes. The rank-sum (Mann–Whitney U) test was used to compare two independent groups of numerical data with non-normal distributions. The statistical significance level for all tests of association was set at p = 0.05. Longitudinal data was used to derive history of CMH care prior to index ISH hospitalization, as well as derive the grouping variable for the survival analysis: CMH within 14 days from index or not (1, 0). The data were set up for survival analysis utilizing the stset function to specify the time variable and censoring information in a survival analysis.
Cox proportional hazards (PH) regression analysis was used in survival analysis to investigate the relationship between the predictor variables and the time until death from any cause following the index ISH hospital attendance. We used the Breslow method to handle tied event times in this model. The Breslow method assigns a weight to each individual at the time of an event, based on the number of individuals at risk just before the time, then using these weights to calculate the partial likelihood, which is used in the Cox PH regression model. As a regression assumption this was evaluated using the Schoenfeld goodness-of-fit test to examine whether the residuals of the Cox PH model were independent of time. Variables entered into the first iteration of the Cox PH regression model were attendance at CMH services within 2 weeks of index discharge, sex, age category in years at index, marital status at index, principal diagnosis at index, readmission within 14 days from index, and previously known to CMH care. The Bayesian Information Criterion (Dziak et al., Reference Dziak, Coffman, Lanza, Li and Jermiin2020) (BIC) was used to compare the different Cox PH models in terms of their goodness-of-fit and complexity. The BIC is based on the maximization of the likelihood function and considers the number of parameters in the model and the sample size. The variable removed in the most parsimonious model according to the BIC was “previously known to CMH care.”
Results
Index ISH events
There were 24,544 persons aged 45 years and over identified with an index event of ISH (not ending in death), occurring between 2014 and 2019 inclusive. There were slightly more males than females, and the age category 45–64 years represented the majority of cases; and 5.1% (n = 1,255) of persons under study were aged 85 years and older. Under half required hospital admission at index (n = 10,404, 42.4%), and of these, 3248 (31.2%) within a psychiatric unit. Of those requiring admission at index, the largest age group was the 45–64 years (n = 7843, 75.4%), with around 20% of the 65–84 years age group receiving admission (n = 2103, 20.2%), and less than 5% of the 85+ years age group (n = 458, 4.4%). Admission to a psychiatric unit was also experienced much less in the older age groups (45–64 years (n = 1794, 80.3%), 65–84 years (n = 400, 17.9%), and 85+ years (n = 39, 1.7%). The administrative data does not specify further the type of ward a patient has received treatment in. Table 1 describes the characteristics of individuals at their index presentation with ISH.
Table 1. Characteristics of individuals at their index presentation for intentional self-harm (n = 24,544)
*Includes physical injuries, burns, and asphyxia.
a If admitted to hospital, not day only attendance.
b Includes neurological conditions (e.g. Parkinson’s disease), cancer, gastrointestinal disorders (e.g. alcoholic liver disease).
Community mental health care
CMH attendance prior to the index presentation is reported in Table 1. Older individuals were less likely to have had any engagement with CMH care prior to their index presentation for ISH (OR 0.96, p < 0.001). Just over half of the study index population received CMH care within 14 days following index hospital presentation (n = 13752, 56.0%). Individuals who had been previously known to CMH care, were 8.2 times more likely to attend CMH post-discharge within 2 weeks of their index hospital presentation (p < 0.001, 95% CI 7.76, 8.71). Table 2 compares characteristics of individuals who were treated at CMH services within 2 weeks from index hospital with those who were not. Post index follow-up care at CMH services was more likely seen among younger, female persons (n = 7543, 54.8%, p < 0.001), those with mental disorders or suicidal ideation at index (n = 1505, 10.9%, p < 0.001), and those who received admission to hospital at their index presentation (n = 7877, 57.2%, p < 0.001).
Table 2. Comparison of index presentation characteristics of individuals who were treated at Community Mental Health (CMH) Services within 14 days versus not
*Includes physical injuries, burns, and asphyxia.
a Aboriginal or Torres Strait Islander.
b Includes neurological conditions (e.g. Parkinson’s disease), cancer, gastrointestinal disorders (e.g. alcoholic liver disease).
We also evaluated outcomes from index presentation for non-fatal ISH and found hospital re-presentation also more common among individuals who received CMH care within 14 days from index (21.7%, p < 0.001) compared with those who did not receive CMH care within 14 days (15.4%). Death following index, for the 1317 persons who died during the follow-up study period, was more frequent among those who did not receive CMH care within 14 days (n = 784, 7.3%) compared with those who did receive CMH care within 14 days from index (n = 533, 3.9%). Breakdown by age group shows this was more frequent among older ages than younger. Table 3 describes these outcomes following index presentation, compared between CMH care within 14 days or not.
Table 3. Comparisons of outcomes of individuals who were treated at Community Mental Health (CMH) Services within 14 days versus not
Our Cox proportionate hazards survival analysis evaluated 24,544 subjects over 72,116.8 person years of time at risk. There were 1,317 deaths yielding a crude mortality rate of 18.25 deaths per 1000 person years (95% CI 17.00–19.50). The adjusted Cox PH survival regression analysis found that females had a lower risk of death (HR 0.56, 95% CI 0.51–0.63) than males, those aged 85 years and over had 6.17 times the risk of death compared to those aged 45–64 years (HR 6.17, 95% CI 5.26–7.23) and persons aged 65-84 years had almost 3 times the risk of death compared with persons aged 45–64 years (HR 2.96, 95% CI 2.62, 3.35). Individuals with a chronic condition as the principal diagnosis at index ISH had 3.31 times the risk of death, compared to individuals who had a principal diagnosis of self-poisoning at index ISH (95% CI 2.41, 4.54).
The particular focus of this study was the receipt of CMH care following index presentation for ISH. The adjusted hazard ratio of death of 0.66 (95% CI 0.58, 0.74) demonstrated that, holding all other variables constant, CMH care within 2 weeks following index offered 34% lower risk of death than those who did not receive CMH care within this time period.
Using the Schoenfeld goodness-of-fit test to assess the proportional hazards assumption that the hazard ratio should remain constant over time, gave a p-value of 0.079, indicating that the Cox PH model was appropriate for the data. Table 4 displays the univariable and multivariable hazard ratios from survival regression analysis; this being the most parsimonious model with the lowest BIC.
Table 4. Univariable and multivariable hazard ratios from survival regression analysis
*Includes physical injuries, burns, and asphyxia.
a Community Mental Health treatment within 2 weeks from index.
b Includes neurological conditions (e.g. Parkinson’s disease), cancer, gastrointestinal disorders (e.g. alcoholic liver disease).
Discussion
We evaluated all-cause mortality following index ISH presentation among 24,544 individuals aged 45 years and over in the Australian state of NSW, between 2014 and 2019, and particularly focused on their engagement with CMH care services both prior to and within the first 14 days following this index presentation. Our study population had a crude mortality rate of 18.25 deaths per 1000 person years (95% CI 17.00–19.50). The oldest age group (85 years and over), males, and those with chronic conditions demonstrated higher risk of death following index ISH. Chronic conditions such as neurological disorders and cancer cause significant suffering and debility, not only increasing risk of suicidal behaviors (Fässberg et al., Reference Fässberg2016) but also leading to death as an outcome of the condition.
Readmission and mortality outcomes differed according to CMH attendance. Those who attended CMH within 14 days were more likely to be readmitted to ED and/or the hospital, presumably due to detection of clinical deterioration by the CMH clinician. They also had lower all-cause mortality outcomes than those who did not attend. This might in part relate to the higher rates of 14-day readmissions that this group experienced, readmissions that could well have addressed issues that might otherwise have been lethal. Yet, overall, those that did require readmission within 14 days were at increased risk of dying.
Previous studies have reported increased risk of all-cause mortality in individuals who self-harm particularly in the first year (Kuo et al., Reference Kuo, Gunnell, Chen, Yip and Chen2012, Ostamo and Lönnqvist, Reference Ostamo and Lönnqvist2001, Steeg et al., Reference Steeg2018). An Australian study reported that mortality risk in the 12 months after self-harm hospitalization was nearly ten times higher than in a matched non-injury comparison cohort (Mitchell and Cameron, Reference Mitchell and Cameron2018). Our findings that mortality risk was increased in males, older people, and in the presence of chronic conditions is consistent with previous studies (Kuo et al., Reference Kuo, Gunnell, Chen, Yip and Chen2012, Mitchell and Cameron, Reference Mitchell and Cameron2018, Mitchell et al., Reference Mitchell, Draper, Harvey, Brodaty and Close2017b, Ostamo and Lönnqvist, Reference Ostamo and Lönnqvist2001).
There are a number of possible explanations for the association between reduced all-cause mortality and CMH attendance within 14 days. The process of establishing a connection between the person who self-harms and CMH might in itself encourage further connections with primary and secondary health care and thus enable treatment of physical and mental disorders. Persons aged 50 years and over who are hospitalized for self-harm were found to have higher rates of chronic physical disorders such as neurological disorders, malignancies, diabetes, respiratory disorders, liver disease, and pain as well as mental disorders than persons hospitalized for other injuries (Mitchell et al., Reference Mitchell, Draper, Harvey, Brodaty and Close2017a). Therefore, this is a population requiring regular medical attention for optimal management.
It is also possible that those that attended CMH were a self-selected group in better overall health and/or with better adherence to medical treatments for their various comorbidities. They were also younger and more likely to be female. Thus, the lower mortality outcomes might largely relate to these factors rather than the intervention.
Older people in this study were less likely to have had contact with CMH prior to the index presentation than those under the age of 65, with the proportion known to CMH (36.3%) being similar to that reported in a systematic review of self-harm in older adults (41.3%) (Troya et al., Reference Troya2019). This may be of importance, as we found that the oldest group in this study (85+ years) had more than 6 times greater risk of death than the 45–64 years age group (HR 6.17 (5.26, 7.23), adjusted). At the index presentation, psychiatric hospitalization decreased by age with less than 20% of persons aged 65–84 years receiving psychiatric admission, and only 1.7% of persons 85 years and over receiving psychiatric admission, a finding that contrasts to the experience in the UK where those aged 65 years and over had the highest rates of psychiatric hospitalization (Kapur et al., Reference Kapur2015), and much lower than in a national study in the US where 77% of persons aged 65 years and over who presented to an emergency department with deliberate self-harm in 2015 were hospitalized (Schmutte et al., Reference Schmutte, Olfson, Xie and Marcus2019).
Among the total population, 56% had contact with CMH care in the 14 days following the index presentation. Attendance at CMH within 14 days of discharge was more likely to occur in females, persons under the age of 65, those who were divorced/separated, persons with pre-index CMH attendance, and in those who were hospitalized. Older people that had attended CMH before the index presentation had similar rates of post-discharge CMH attendance as younger people indicating the importance of having an established relationship with the CMH service and replicating findings in the US (Schmutte et al., Reference Schmutte, Olfson, Xie and Marcus2019). An earlier Australian study of all-age adults admitted to hospital for self-harm found lower rates of CMH care follow-up in the 30 days following the index presentation (41%) than the current study, but also found that follow-up rates were lower for older adults (Spittal et al., Reference Spittal, Shand, Christensen, Brophy and Pirkis2017).This difference in follow-up rates is not necessarily explained by the different sampling frame (admitted persons vs emergency department presentations and admitted persons), as in the current study, being admitted to hospital increased the likelihood of CMH care follow-up. One possible explanation is that there has been an overall increase in CMH care follow-up after hospital-treated self-harm. Those who self-harmed had a 21 percent higher health service use in the 12 months pre and post the self-harm admission (Mitchell and Cameron, Reference Mitchell and Cameron2018).
Strengths and limitations
Our study had several strengths. The age distinctions provided in this study provide further insight into the variability in the population by age structure. While some of the findings presented here are consistent with previous studies in Australian cohorts (Clapperton et al., Reference Clapperton, Dwyer, Millar, Tolhurst and Berecki-Gisolf2021, Pillans et al., Reference Pillans, Page, Ilango, Kashchuk and Isbister2017), the distinction between persons aged 65–84 and 85+ years is unique to this study.
Furthermore, by using longitudinal data linkage, we were able to describe a large population of anonymized trajectories of care by persons diagnosed with ISH both before and after their index ISH hospitalization, where they had received CMH care. When combined with smaller-size qualitative studies into a mixed-method approach, data linkage can provide the grounds for high-quality research. Future research should aim to incorporate qualitative literature in order to determine the nuance of some of the findings presented here.
Our study also had some limitations. Principally, this study did not have access to primary healthcare consultation data, obtainable from the Medicare Benefits Schedule. It is possible that some individuals had an existing relationship with their General Practitioner, psychologist, or specialist psychiatric consultant and attended these services instead of CMH care.
The administrative datasets used in this study do not identify the suicidal or non-suicidal intent of the self-harm incident, due to the limitations of the ICD-10-AM in accurately identifying this distinction. Although non-suicidal self-injury is less prevalent in older than younger adults, it may represent approximately one-quarter of self-harm presentations in older adults (Schmutte et al., Reference Schmutte, Olfson, Xie and Marcus2020). Nevertheless, non-suicidal self-harm is a risk factor for suicide (Grant et al., Reference Grant2019). A further limitation of the data is the likelihood that some episodes of self-harm presenting to the ED will have been captured in the data: the reason for ED care is recorded as a single final diagnosis by ED clinicians, along with a free-text field identifying the presenting problem. This limitation has been somewhat mitigated in the ED dataset used for this study by annotating cases using the text fields recorded at triage to improve diagnostic sensitivity (Grant et al., Reference Grant2019).
Finally, it should be noted that these data are accepted as correlational, and despite the methodological rigor of these analyses, a causal relationship cannot be assigned. There are complex variable interrelationships, as well as other factors not able to be identified within administrative data that could produce both post-incident CMH involvement and mortality.
Implications for future research
This is one of the largest studies to analyze associations between CMH care following non-fatal ISH; and the largest in Australia.
Ongoing research should aim to further isolate confounders to CMH care, socioeconomic variables, as well as addressing the limitations outlined in this paper, such as examining the impact of care received following hospital-treated self-harm from other services, e.g. general practitioners, psychologists, psychiatrists, and dedicated suicide aftercare services.
Conclusion
CMH care following an episode of hospital-treated self-harm in those aged 45 years and over may be protective against all-cause mortality. This finding may be a result of the protective effects of CMH care, and/or the possibility that those who engaged with CMH care had better overall health and were more likely to engage in other types of healthcare. Younger people and females were more likely to engage with CMH care, indicating that we are missing the opportunity to engage older males, one of the highest risk groups in Australia, in community care following self-harm.
Source of funding
This work was supported by funding from the Paul Ramsay Foundation.
Description of authors’ roles
FS (among others) obtained the research funding. BD conceived of the study; MW undertook the initial literature review. LNS cleaned the datasets and conducted the analysis with oversight from BD and FS. LNS and MW drafted the original manuscript. All authors contributed to its review, editing, and interpretation of findings. MW undertook Honors research (University of New South Wales Faculty of Medicine and Health) within this project.
Acknowledgments
We gratefully acknowledge the peer review ahead of submission by Dr Grant Sara, NSW Health.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1041610223000959.
