Does emergency department use and post-visit physician care cluster geographically and temporally for adolescents who self-harm? A population-based 9-year retrospective cohort study from Alberta, Canada

In Canada, several recently publicized clusters of deaths by suicide among young people [1, 2] have focused attention on ‘point clusters’—unusually high numbers of suicides occurring in a close geographic location and brief time period. Such clusters have also been documented globally for individuals of all ages, [3‐7] although clustering has been found to be up to four times more common among adolescents and young adults than among other age groups [3]. The study of self-harm clusters—clusters of non-fatal self-poisoning or self-inflicted harm irrespective of suicidal intent—is uncommon even though the average lifetime prevalence of self-harming behavior is much higher among adolescents (ranging from 17–39 % in adolescence) than an outcome of death by suicide (9.0 per 100,000 for adolescents aged 15 to 19 years) [8‐10]. While deliberate self-harm often occurs in the absence of suicidal intent, it is considered a clear sign of emotional distress that may result in accidental death or serious injury. Compared to other age groups, young people are more likely to report suicidal thoughts and self-harm, elevating their risk for a suicide attempt or death by suicide at a later date [11]. Non-suicidal self-harm has also been shown to be predictive of future suicide attempts among adolescents with treatment-resistant depression [12]. Further, in a recent study, Swanson and Colman found that exposure to suicide is associated with increased suicidal ideation and attempt, [13] results that highlighted self-harm risks are spatially and temporally bound.

The study of self-inflicted harm clusters in the context of health care seeking has yet to be conducted, but presents an opportunity to identify important spatio-temporal trends in care, particularly among young people who may be at greater risk for negative health and psychological outcomes. We focused this study on spatial-temporal trends in emergency department (ED) visits made by adolescents who self-harm and post-ED follow-up visits made to physicians. Among young people who self-harm, approximately one in eight will present to an ED for related care [14‐18]. These individuals are considered to be at higher risk for subsequent mortality compared to those who do not present to the ED for care [19] and have a higher prevalence of mental disorders among them [14]. Thus, for these young people, the period immediately following an ED visit is an important time for risk reduction and psychiatric stabilization [20‐22].

Recent large-scale studies indicate that 44–75 % of ED visits made by young people for self-harm result in discharge home [17‐19, 23‐25]. This disposition is considered appropriate for those who are not actively suicidal, do not have access to lethal means, and have a responsible adult to ensure their safety [26]. For these young people, referral to urgent outpatient mental health care may be recommended as follow-up to the ED visit [20]. This recommendation is based on known vulnerabilities of this population, [14, 16] and clinical acumen that follow-up mental health care can promote and sustain the child’s safety, address psychosocial support needed by the child and family in the post-crisis period, and further explore mental health and coping needs of the child and family. Several studies have demonstrated, however, that receipt of follow-up services does not occur for the majority of young people, [25, 27] and that ED visit rates for self-harm and post-ED follow-up visit rates after such visits vary geographically. In accordance, in this study, we extracted population-based data to examine emergency mental health care and follow-up care for adolescents (age 15–17 years) in Alberta, Canada. We describe the adolescents who (1) presented to the ED for self-harm, and (2) presented to the ED for self-harm but did not have a mental health-related physician follow-up visit within 14 days after an ED visit. Using a statistical surveillance technique, we identified geographic areas with higher numbers than expected of adolescents defined by (1) and (2).

Recent Canadian and United States investigations of time trends in ED visits for intentional self-harm among adolescents have documented changes. In Ontario during 2002 to 2011, incident rates decreased between 2002/03 to 2006/07, but not thereafter, for adolescents aged 12 to 17 years [38]. In the United States, rates per 1,000 increased from 2.57 to 4.53 during 1993 to 2008 for adolescents aged 15–19 [39] while for those aged 10 to 18 years a National Trauma Data Bank study showed ED visits for self-harm increased from 2009 to 2012 [40]. Our nine-year Canadian population-based study showed the overall rate of ED visits by sex declined during the study period. The observed sex by time interaction for ED visits for self-harm may reflect the gender paradox noted by others in the literature; although not fully understood, it may be that girls seek more opportunities for intervention over time [41, 42]. The observed sex differences in rates of those adolescents without a 14-day physician follow-up (rates 2011, girls versus boys: 225.3 versus 80.2 per 100,000 adolescents) may also reflect the gender paradox. As this study was not designed to investigate reasons for the sex by time interaction or sex differences in health care utilization, further study is warranted and may contribute to understanding the gender paradox observed among young people who deliberately self-harm.

A novel contribution of our study to the literature is our use of a spatial scan to identify geographical-temporal clusters that had higher numbers of adolescents who presented to the ED for self-harm than expected by chance. These clusters are different than the ‘point clusters’ of suicide in that the lack of disaggregated data does not permit specific locations to be used and the focus was on self-harm rather than suicide. Nonetheless, this identification of clusters does provide insights in to the time frame and geographic locations where statistically higher numbers of adolescents have presented to EDs for self-harm. Follow-up to this body of literature is now necessary to determine if rates reflect clinically important changes to health care utilization (e.g., changing access to other services; diversion of adolescents who self-harm to other types of settings such as community supports, employee and family assistance programs, self-help supports), changes to health insurance status (e.g., Affordable Health Care for America Act), and/or whether the type or severity of self-harm is changing and reflected though ED visits.

A unique and important contribution from our study was the identification of geographic and temporal variations in the number of adolescents who were without 14-day physician follow-up after an ED visit for self-harm. Although follow-up mental health care following the ED visit would be recommended and an expected positive outcome designed to promote further risk reduction and psychiatric stabilization, [26] of the nearly 4,000 adolescents with ED visits for self-harm in our study, about 55 % did not have a physician follow-up visit for mental health care after the ED visit. This result is important to note because there are benchmarks for post-ED follow-up care, including increasing the rates of follow-up and coordination between health care service points, as part of the national strategy for suicide prevention in Canada [43]. We identified two potential geographical-temporal clusters where follow-up was absent and, as well, some potential geographic clusters during the 2009/2010 and 2010/2011 fiscal years. The potential clusters were the same for the analyses of adolescents with ED visits for self-harm and those with ED visits but without physician follow-up mental health care. These identified clusters were mainly in the north and central areas of Alberta and may be real or spurious. The sRHAs represented by the clusters included both urban and rural populations and also contain the majority of the First Nations Communities within Alberta. These findings could indicate that youth in the Northern or Central parts of the province (that are more rural) may be more vulnerable to mental health problems leading to self-harm (e.g., anxiety, mood disorders) than youth in other parts of the province. Other information would be required to verify if this is the case and to determine potential factors for the higher ED visit rates in these clusters (e.g., time of presentation – which would impact access to services; social and economic disparities such as unemployment, low income, housing difficulties; access and transportation issues; [44] number of prior self-harm events; First Nation status). Based on these results, prospective research is needed to determine outcomes associated with adolescents who receive physician follow-up following ED-based care for self-harm compared to those who do not. Investigation of the role and impact of follow-up with non-physician community supports should also be considered for future studies. This aspect is important because availability of community mental health non-physician supports is often far greater than physician services.

Although not central to our study, similar to other studies of emergency health care use for self-harm among adolescents, more ED visits were made by females for self-harming behavior [24, 38]. These consistent findings suggest that school-based mental health prevention efforts should address self-harming behavior among young females by discussing when and how to seek help and who they could talk to about their behavior.

We acknowledge that our study has several limitations. First, like all cluster detection methods, the spatial scan cannot determine if an identified cluster is clinically important or not. For example, a clinically important cluster could be due to higher severity of self-harm in adolescents requiring emergency care or an area with less availability of other health services. In contrast a spurious cluster can result from, for example, variation in coding practices. This limitation is true for all cluster detection analyses and further targeted epidemiological research is needed to determine if a detected cluster is clinically relevant. Second, although the KN spatial scan works well for identifying circular and primary clusters that are close in proximity, it can miss secondary or irregularly shaped clusters. Third, our case definitions of an adolescent with at least one ED visit for self-harm during the study period or an adolescent without follow-up do not include all adolescents who self-harm. For example, the definitions exclude any adolescents who self-harm that seek health services outside the ED. Further, it is not possible to distinguish between suicidal and non-suicidal acts when self-harming behaviors are identified by ICD diagnostic codes. Fourth, our study includes several other potential limitations: (1) we have assumed that the sRHA of residence has not changed over time, (2) the time of the ED presentation could impact which services were available in the ED to assist with discharge planning, and (3) only data for physician follow-ups were available. The latter is important to note, especially in large geographic rural areas of Alberta (such as Northern Alberta), because there is a general shortage of psychiatrists in these areas. A large proportion of the psychiatrists serving these areas is based in urban centres and provides travelling clinics to rural sites. Accordingly, the majority of mental health follow-up services for residents in those areas may be with general physicians (e.g., a family physician) or with non-physicians (e.g., private counselling, community mental health clinics, employee and family assistance programs, other community services, self-help groups or online services). Unfortunately, data for non-physician mental health professional services are not captured in the databases and such data are not available for analysis. Finally, we have restricted our study to 15 to 17 year olds and our conclusions would be limited to this age group. Notwithstanding these limitations, our study is based on a long study period with large, population-based databases that are comprehensive and complete with respect to physician follow-up data.

Using large, population-based databases over nine fiscal years, we identified geographic and temporal variations in the numbers of adolescents aged 15–17 presenting to EDs for self-harm and the number of adolescents presenting to EDs without 14-day physician follow-up for mental health care. The potential clusters identified may represent geographic areas with higher harm severity or a lower availability of non-ED health services. The clusters are not all likely to have occurred by chance. Further investigation and intervention is warranted to document and address the unmet needs of adolescents presenting to the ED with self-harm.

ED, emergency department; ACCS, ambulatory care classification system; RHA, sub-regional health authority; CIs, confidence intervals