A well-established close relationship exists between self-harm and suicide in older adults, yet the detailed longitudinal outcome data needed to guide pathways for identification of at-risk individuals and intervention are lacking. The big data study of Cheung and colleagues extends our knowledge of self-harm in older adults by moving beyond simply identifying risk factors for self-harm cross-sectionally, to ranking their relative importance, and longitudinally examining the incidence of self-harm in a large community-dwelling cohort (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). This brings us a step closer to understanding factors contributing to self-harm in older people. However, we already know much about these population-level risk factors (Troya et al., Reference Troya2019). Research in this field needs to delve deeper, and explore the clinical trajectories and broader outcomes of older adults who self-harm in combination with qualitative data derived from the individuals themselves.
In older adults, there are common risk factors for self-harm and suicide, and self-harm is more likely to be lethal with greater reported intent to die (Fassberg et al., Reference Fassberg2019; Schmutte et al., Reference Schmutte, Olfson, Xie and Marcus2019). However, not all people who self-harm do so with suicidal intent, and the two groups may be distinct (DeJong et al., Reference DeJong, Overholser and Stockmeier2010), with the outcome of death following self-harm sometimes related to external factors rather than intent or another meaning of the behavior (Linden and Barnow, Reference Linden and Barnow1997). Nonetheless, knowledge of the risk factors for self-harm provides opportunities to understand a proportion of suicides and effectively intervene.
The epidemiological risk factors for self-harm in older adults have been well described (Troya et al., Reference Troya2019), and include physical illness, functional disability, mental illness (especially depression; alcohol misuse), and social issues (such as loneliness, social isolation, financial difficulties, and interpersonal problems). However, these potential risk factors are common in older adults in general, most of whom do not self-harm. Further, most of these data have been derived from quantitative research and hospital-based cohorts, rather than from community-dwellers and the older adult themselves (Troya et al., Reference Troya2019; Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b). Thus it is difficult for clinicians, especially in primary care, where the majority of older adults receive their healthcare (both physical and mental health), to know who and when to prioritize for intervention (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018a).
The New Zealand-based retrospective cohort study of Cheung and colleagues utilized data from over 93,000 community-dwelling older adults (aged 60+) who had a comprehensive geriatric assessment using interRAI-Home Care (HC) over a 5-year period (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The interRAI-HC assessment is required for all older adults who want to access public home-based care services or placement in long-term residential aged care facilities. Thus their study population was a group of older adults needing support services and likely to have physical illness and disability; known risk factors for self-harm and suicide (Fassberg et al., Reference Fassberg2016) and therefore most relevant for selective prevention. Importantly, all self-harm was captured, regardless of intent, which they rightly highlight may be difficult to determine (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The 24 people who died by suicide were excluded from the study, including two individuals who had self-harmed in the study period. This may be considered a limitation of the study if our fundamental interest in self-harm is about intervening to prevent suicide. Over the 5-year period, 251 people (0.27%) had self-harmed at least once (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The incidence of non-fatal self-harm was 160.39/100,000 person years, highest in the first year of follow-up and the “young” old (age 60–79); with a standardized incidence ratio of 5.12.
The hierarchy of risk factors for non-fatal self-harm identified through multivariate analysis were depression (hazard ratio [HR] = 3.02), at-risk alcohol use (HR = 2.38), and bipolar disorder (HR = 2.18) (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Significant effects were also reported for demographic factors such as younger age and being separated/divorced, and clinical features such as fair/poor self-rated health, daily tobacco smoking, and moderate to severe depressive symptoms. The top three risk factors of depression, at-risk alcohol use and bipolar disorder are unsurprising, but importantly confirm similarities between community-based populations and the existing research which is largely derived from hospital-based cohorts. While depression is an accepted risk factor for self-harm in older adults (Troya et al., Reference Troya2019), its impact may have been overestimated in this study as this factor included people both receiving treatment and monitored but untreated. On the other hand, also moderate depressive symptoms were associated with an increased risk for subsequent self-harm. This is important to note as subclinical depressive symptoms are very common in older people. By contrast, the importance of alcohol misuse as a risk factor is likely underestimated in Cheung’s study, as the interRAI-HC only considers five or more drinks in a single sitting in a fortnight as “at risk” (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Thus alcohol dependency (and daily drinking up to 4 standard drinks/day) may not be captured as “at risk”. With ageing-related changes in pharmacokinetics and pharmacodynamics, greater propensity for drug and disease interactions, and higher rates of cognitive impairment in older adults, it is imperative that we move away from “ageless” safe drinking guidelines (National Health and Medical Research Council, Reference Health2009) to a more nuanced approach (Health Promotion Agency, 2018). Although bipolar disorder is associated with high rates of attempted and completed suicide in younger adults, it is less commonly mentioned as a risk factor in older adults and findings are mixed (Schaffer et al., Reference Schaffer2015). However, the strength of the association between this risk factor and self-harm identified by Cheung and colleagues highlights the need to improve the detection and management of bipolar disorder in older adults; an important opportunity to prevent self-harm (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Cancer – usually a risk factor for self-harm and suicide (Fassberg et al., Reference Fassberg2016; Henson et al., Reference Henson, Brock, Charnock, Wickramasinghe, Will and Pitman2019) – was surprisingly identified as a protective factor in the New Zealand study. This finding should be interpreted with caution and is difficult to explain without unpacking the data at an individual level. As the authors point out, perhaps the older adults with cancer felt their needs were being met after the assessment for homecare support or placement (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). This concords with our own qualitative work demonstrating the importance of unmet needs to self-harm in older adults (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b), but we agree that it requires further exploration. The observed protective effect of severe functional dependence against self-harm (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) may reflect a physical inability to self-harm, or cognitive deficits accompanying functional impairment (such as occurring in dementia) which limit ability to develop intent, plan, and enact self-harm (Draper, Reference Draper2015). However, the severity of cognitive impairment was not protective in the multivariate model (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The explanation for this association remains unclear, but may be better understood using qualitative methodology.
Almost in parallel, a community-based study in Australia longitudinally explored factors associated with self-harm and suicide in older adults (Erlangsen et al., Reference Erlangsen2020). This postal survey-based general population study – therefore with implications for primary/universal prevention – followed-up almost 103,000 people aged 65+ for a mean of 8.5 years. In contrast to the New Zealand study (Cheung et al., Reference Cheung, Chai, Troya and Luo2020), there was a focus on suicide as well as self-harm. The rate of self-harm was 21.9/100,000 person years, much lower than in Cheung et al.’s study (Cheung et al., Reference Cheung, Chai, Troya and Luo2020); perhaps unsurprising as the community-based cohort of the latter study was at higher risk given recruitment on the basis of need for support services. Other key study findings included 109 deaths by suicide and a rate of suicide of 12.5/100,000 person years (Erlangsen et al., Reference Erlangsen2020). Rates of self-harm were higher in people who rated their health as poor/fair (also found by Cheung and colleagues), or reported higher levels of physical limitations, poor vision, or male erectile problems (Erlangsen et al., Reference Erlangsen2020). Interestingly, alcohol misuse was not identified as a risk factor for self-harm or suicide in the Australian cohort, perhaps because the study measure relied upon self-report and was designed to assess whether the participant followed safe-drinking levels; already identified as problematic above. Similar to Cheung and colleagues’ study, psychological factors were also associated with self-harm, in particular high distress on the Kessler-10, and past diagnosis or currently receiving treatment for depression or anxiety (Erlangsen et al., Reference Erlangsen2020). Having a more active social network was associated with less risk of self-harm. The authors rightly highlight this is a complex factor to untangle as there are bidirectional relationships between core risk factors for self-harm of poor physical and mental health and social connectedness and loneliness.
These two longitudinal studies use different approaches (interRAI-HC versus postal survey) linked to hospital and mortality data to study different community-based populations of older adults, highlighting risk factors for self-harm relevant to tertiary/selective (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) and primary (Erlangsen et al., Reference Erlangsen2020) prevention. Linked health data can be a powerful method to evaluate the effectiveness and quality of health services, the uptake of treatments and the impact of screening and prevention programs. However, common limitations include missing data (e.g. incomplete recording or lack of interaction with health services), inaccurate data (e.g. when the hospital record of an individual could not be accurately linked due to insufficient identifying information), and low data quality (Harron et al., Reference Harron2017). While a big data approach offers the opportunity to track patients over time and across services (including pre-, during, and post-treatment), it cannot tell us the reasons behind discovered associations as the data is limited by a lack of depth. In contrast, mixed-methods and inclusive research offer the opportunity to deliver much needed answers to how and why risk factors for self-harm develop in older people. We agree with the recent call for greater presence of qualitative research in psychiatry by Galasinski (Reference Galasinski2020). While quantitative research methods predominate in mental health research, it is qualitative research that offers insight into the complexity of clinical associations and meanings of quantitative findings. Qualitative data can identify important aspects of the environmental, social, and organizational context that cannot be detected in fixed data sources. Currently we have an incomplete understanding of the associated clinical and social factors and trajectories of older adults who self-harm in terms of service responses and pathways of care, or outcomes such as cognition, function, place of residence, and mortality. Given how frequently identified risk factors for self-harm occur in older adults in general, without these data we cannot say what works and for whom in relation to prevention of self-harm in late life.
Conclusions
Recent longitudinal community-based studies of older adults who self-harm such as those led by Cheung (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) and Erlangsen (Erlangsen et al., Reference Erlangsen2020) have helped identify and rank risk factors for self-harm. However, in order to move from identification of risk factors to action, there is a need for large-scale epidemiological data which describes the clinical care pathways, trajectories, and outcomes of older adults who self-harm. This must be combined with more in-depth understanding of the underlying reasons for self-harm from the unique perspective of the older adult and their carers, informed by qualitative methods (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b). A simultaneous focus on the big and the small may meaningfully guide much needed intervention in late life self-harm.
You have
Access
A well-established close relationship exists between self-harm and suicide in older adults, yet the detailed longitudinal outcome data needed to guide pathways for identification of at-risk individuals and intervention are lacking. The big data study of Cheung and colleagues extends our knowledge of self-harm in older adults by moving beyond simply identifying risk factors for self-harm cross-sectionally, to ranking their relative importance, and longitudinally examining the incidence of self-harm in a large community-dwelling cohort (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). This brings us a step closer to understanding factors contributing to self-harm in older people. However, we already know much about these population-level risk factors (Troya et al., Reference Troya2019). Research in this field needs to delve deeper, and explore the clinical trajectories and broader outcomes of older adults who self-harm in combination with qualitative data derived from the individuals themselves.
In older adults, there are common risk factors for self-harm and suicide, and self-harm is more likely to be lethal with greater reported intent to die (Fassberg et al., Reference Fassberg2019; Schmutte et al., Reference Schmutte, Olfson, Xie and Marcus2019). However, not all people who self-harm do so with suicidal intent, and the two groups may be distinct (DeJong et al., Reference DeJong, Overholser and Stockmeier2010), with the outcome of death following self-harm sometimes related to external factors rather than intent or another meaning of the behavior (Linden and Barnow, Reference Linden and Barnow1997). Nonetheless, knowledge of the risk factors for self-harm provides opportunities to understand a proportion of suicides and effectively intervene.
The epidemiological risk factors for self-harm in older adults have been well described (Troya et al., Reference Troya2019), and include physical illness, functional disability, mental illness (especially depression; alcohol misuse), and social issues (such as loneliness, social isolation, financial difficulties, and interpersonal problems). However, these potential risk factors are common in older adults in general, most of whom do not self-harm. Further, most of these data have been derived from quantitative research and hospital-based cohorts, rather than from community-dwellers and the older adult themselves (Troya et al., Reference Troya2019; Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b). Thus it is difficult for clinicians, especially in primary care, where the majority of older adults receive their healthcare (both physical and mental health), to know who and when to prioritize for intervention (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018a).
The New Zealand-based retrospective cohort study of Cheung and colleagues utilized data from over 93,000 community-dwelling older adults (aged 60+) who had a comprehensive geriatric assessment using interRAI-Home Care (HC) over a 5-year period (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The interRAI-HC assessment is required for all older adults who want to access public home-based care services or placement in long-term residential aged care facilities. Thus their study population was a group of older adults needing support services and likely to have physical illness and disability; known risk factors for self-harm and suicide (Fassberg et al., Reference Fassberg2016) and therefore most relevant for selective prevention. Importantly, all self-harm was captured, regardless of intent, which they rightly highlight may be difficult to determine (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The 24 people who died by suicide were excluded from the study, including two individuals who had self-harmed in the study period. This may be considered a limitation of the study if our fundamental interest in self-harm is about intervening to prevent suicide. Over the 5-year period, 251 people (0.27%) had self-harmed at least once (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The incidence of non-fatal self-harm was 160.39/100,000 person years, highest in the first year of follow-up and the “young” old (age 60–79); with a standardized incidence ratio of 5.12.
The hierarchy of risk factors for non-fatal self-harm identified through multivariate analysis were depression (hazard ratio [HR] = 3.02), at-risk alcohol use (HR = 2.38), and bipolar disorder (HR = 2.18) (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Significant effects were also reported for demographic factors such as younger age and being separated/divorced, and clinical features such as fair/poor self-rated health, daily tobacco smoking, and moderate to severe depressive symptoms. The top three risk factors of depression, at-risk alcohol use and bipolar disorder are unsurprising, but importantly confirm similarities between community-based populations and the existing research which is largely derived from hospital-based cohorts. While depression is an accepted risk factor for self-harm in older adults (Troya et al., Reference Troya2019), its impact may have been overestimated in this study as this factor included people both receiving treatment and monitored but untreated. On the other hand, also moderate depressive symptoms were associated with an increased risk for subsequent self-harm. This is important to note as subclinical depressive symptoms are very common in older people. By contrast, the importance of alcohol misuse as a risk factor is likely underestimated in Cheung’s study, as the interRAI-HC only considers five or more drinks in a single sitting in a fortnight as “at risk” (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Thus alcohol dependency (and daily drinking up to 4 standard drinks/day) may not be captured as “at risk”. With ageing-related changes in pharmacokinetics and pharmacodynamics, greater propensity for drug and disease interactions, and higher rates of cognitive impairment in older adults, it is imperative that we move away from “ageless” safe drinking guidelines (National Health and Medical Research Council, Reference Health2009) to a more nuanced approach (Health Promotion Agency, 2018). Although bipolar disorder is associated with high rates of attempted and completed suicide in younger adults, it is less commonly mentioned as a risk factor in older adults and findings are mixed (Schaffer et al., Reference Schaffer2015). However, the strength of the association between this risk factor and self-harm identified by Cheung and colleagues highlights the need to improve the detection and management of bipolar disorder in older adults; an important opportunity to prevent self-harm (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). Cancer – usually a risk factor for self-harm and suicide (Fassberg et al., Reference Fassberg2016; Henson et al., Reference Henson, Brock, Charnock, Wickramasinghe, Will and Pitman2019) – was surprisingly identified as a protective factor in the New Zealand study. This finding should be interpreted with caution and is difficult to explain without unpacking the data at an individual level. As the authors point out, perhaps the older adults with cancer felt their needs were being met after the assessment for homecare support or placement (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). This concords with our own qualitative work demonstrating the importance of unmet needs to self-harm in older adults (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b), but we agree that it requires further exploration. The observed protective effect of severe functional dependence against self-harm (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) may reflect a physical inability to self-harm, or cognitive deficits accompanying functional impairment (such as occurring in dementia) which limit ability to develop intent, plan, and enact self-harm (Draper, Reference Draper2015). However, the severity of cognitive impairment was not protective in the multivariate model (Cheung et al., Reference Cheung, Chai, Troya and Luo2020). The explanation for this association remains unclear, but may be better understood using qualitative methodology.
Almost in parallel, a community-based study in Australia longitudinally explored factors associated with self-harm and suicide in older adults (Erlangsen et al., Reference Erlangsen2020). This postal survey-based general population study – therefore with implications for primary/universal prevention – followed-up almost 103,000 people aged 65+ for a mean of 8.5 years. In contrast to the New Zealand study (Cheung et al., Reference Cheung, Chai, Troya and Luo2020), there was a focus on suicide as well as self-harm. The rate of self-harm was 21.9/100,000 person years, much lower than in Cheung et al.’s study (Cheung et al., Reference Cheung, Chai, Troya and Luo2020); perhaps unsurprising as the community-based cohort of the latter study was at higher risk given recruitment on the basis of need for support services. Other key study findings included 109 deaths by suicide and a rate of suicide of 12.5/100,000 person years (Erlangsen et al., Reference Erlangsen2020). Rates of self-harm were higher in people who rated their health as poor/fair (also found by Cheung and colleagues), or reported higher levels of physical limitations, poor vision, or male erectile problems (Erlangsen et al., Reference Erlangsen2020). Interestingly, alcohol misuse was not identified as a risk factor for self-harm or suicide in the Australian cohort, perhaps because the study measure relied upon self-report and was designed to assess whether the participant followed safe-drinking levels; already identified as problematic above. Similar to Cheung and colleagues’ study, psychological factors were also associated with self-harm, in particular high distress on the Kessler-10, and past diagnosis or currently receiving treatment for depression or anxiety (Erlangsen et al., Reference Erlangsen2020). Having a more active social network was associated with less risk of self-harm. The authors rightly highlight this is a complex factor to untangle as there are bidirectional relationships between core risk factors for self-harm of poor physical and mental health and social connectedness and loneliness.
These two longitudinal studies use different approaches (interRAI-HC versus postal survey) linked to hospital and mortality data to study different community-based populations of older adults, highlighting risk factors for self-harm relevant to tertiary/selective (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) and primary (Erlangsen et al., Reference Erlangsen2020) prevention. Linked health data can be a powerful method to evaluate the effectiveness and quality of health services, the uptake of treatments and the impact of screening and prevention programs. However, common limitations include missing data (e.g. incomplete recording or lack of interaction with health services), inaccurate data (e.g. when the hospital record of an individual could not be accurately linked due to insufficient identifying information), and low data quality (Harron et al., Reference Harron2017). While a big data approach offers the opportunity to track patients over time and across services (including pre-, during, and post-treatment), it cannot tell us the reasons behind discovered associations as the data is limited by a lack of depth. In contrast, mixed-methods and inclusive research offer the opportunity to deliver much needed answers to how and why risk factors for self-harm develop in older people. We agree with the recent call for greater presence of qualitative research in psychiatry by Galasinski (Reference Galasinski2020). While quantitative research methods predominate in mental health research, it is qualitative research that offers insight into the complexity of clinical associations and meanings of quantitative findings. Qualitative data can identify important aspects of the environmental, social, and organizational context that cannot be detected in fixed data sources. Currently we have an incomplete understanding of the associated clinical and social factors and trajectories of older adults who self-harm in terms of service responses and pathways of care, or outcomes such as cognition, function, place of residence, and mortality. Given how frequently identified risk factors for self-harm occur in older adults in general, without these data we cannot say what works and for whom in relation to prevention of self-harm in late life.
Conclusions
Recent longitudinal community-based studies of older adults who self-harm such as those led by Cheung (Cheung et al., Reference Cheung, Chai, Troya and Luo2020) and Erlangsen (Erlangsen et al., Reference Erlangsen2020) have helped identify and rank risk factors for self-harm. However, in order to move from identification of risk factors to action, there is a need for large-scale epidemiological data which describes the clinical care pathways, trajectories, and outcomes of older adults who self-harm. This must be combined with more in-depth understanding of the underlying reasons for self-harm from the unique perspective of the older adult and their carers, informed by qualitative methods (Wand et al., Reference Wand, Peisah, Draper and Brodaty2018b). A simultaneous focus on the big and the small may meaningfully guide much needed intervention in late life self-harm.