Summary of findings
Using multiple Danish national registers, we examined whether excess mortality due to suicide exists among people treated for DUD, and identified risk factors associated with completed suicide between 2000 and 2010. We also compared the prevalence of suicide between individuals treated for DUD and the general population.
Patients who had been in treatment for DUD, but who did not seek psychiatric care in the past year had more than a sevenfold increase in risk of suicide after admission to treatment, in comparison to age- and gender matched individuals from the general population without a history of recent psychiatric care. This is a substantial increase in risk, which is in line with findings from psychological autopsies [
5], and the 95% confidence interval between five and eight clearly indicates a statistically significant difference.
Patients who had been in treatment for a DUD with past psychiatric care had a more than 13 times higher risk of committing suicide compared with gender or age-matched individuals with no history of treatment for substance use disorders and psychiatric care in the general population. However, the high relative elevation must be considered in light of the low base-rate of completed suicide, and it must be remembered that among the individuals that we tracked for up to ten years, less than 1% took their own lives (compare also [
32]).
In addition, it should be noted that people from the general population with a recent history of psychiatric care had practically the same elevation in risk as patients with both a history of treatment for DUD and recent psychiatric care (SMR = 13.6).
Our findings highlight some key risk factors for suicide among people seeking treatment for DUD. We found that younger age, past year history of psychiatric care, use of opioids, and use of alcohol were all associated with increased risk of suicide. Use of cannabis was associated with a lower risk of suicide. Past psychiatric care was associated with a higher risk of suicide in our cohort. As would be expected, mental health problems were associated with an elevated suicide risk in both the DUD cohort and among the controls, a finding also reported by Cavanagh et al. [
33].
Our finding that opioid use was a strong predictor of completed suicide is consistent with other studies [
7,
34,
35], even if those other studies have assessed suicidal ideation and attempts, rather than completed suicide. In our context, opioid use was part of a drug problem that had led to treatment, meaning that our findings may not be relevant for opioid use among pain patients (compare [
36]).
Our finding that alcohol use was a strong predictor of completed suicide is also consistent with other studies [
37,
38]. For instance, one in five people who committed suicide in an Australian psychological autopsy study were found to have an alcohol use disorder [
39]. In addition, alcohol intoxication is associated with methods of increased lethality when attempting suicide, i.e. methods that have a higher risk of a fatal outcome [
40]. However, this study adds to the literature by showing that even among people with other DUD, alcohol is an independent contributor to suicide risk.
Our finding that cannabis was associated with lower risk of completed suicide was unexpected [
41,
42]. It is possible that third variable confounding underlie these negative correlations. However, some research suggests at least one active component in cannabis, namely cannabidiol [
43], can have beneficial effects on substance use disorders, by reducing drug seeking behaviour and symptoms of anxiety [
43]. This may in turn reduce the risk of completed suicide in the context of multiple types of DUD. It is also possible that the general loss of initiative associated with cannabis use may indirectly influence suicidal behaviour as well [
44].
Our findings of no association between self-harm and completed suicide is contrary to other studies (see 12). It is possible that we did not capture self-harm with sufficient precision, or that our patient group did not present with self-harm at hospital-based clinics, including emergency departments, but rather with symptoms of intoxication or withdrawal.
Implications for practice
There is growing evidence that patients with DUD who experience mental health problems may be helped by interventions that are integrated with substance abuse treatment. At least one meta-analysis [
45], as well as more recent clinical trials [
46‐
48], indicate that both mental health problems and substance use disorders are receptive to psychotherapy as a mode of treatment. Evidence from large-scale trials indicate that integrated services for people with co-morbid substance use disorders and mental health problems can be implemented in routine clinical settings [
49].
Furthermore, there is evidence that antidepressants can be helpful for people with co-morbid depression and substance use disorders, even if the effects are larger when patients are abstinent before being treated [
50,
51] and the quality of the evidence is mixed [
52]. In this context, benefits and risks should be carefully weighted: There is a risk that antidepressants may become part of a lethal cocktail of substances leading to a fatal overdose [
53]. However, this relationship is complicated, as depression and anxiety disorders may themselves be associated with overdose risk, and this risk may be attenuated by treatment with antidepressants that is continuous over a longer period of time [
54].
Finally, other studies show that mental health problems can be validly assessed among people undergoing treatment for DUD with use of self-report instruments [
55‐
57]. As such, identification of co-existing psychopathology should be highlighted even more as a potential first step towards suicide prevention.
Strengths and limitations
Some limitations must be noted for this study. First, as with any register-based study, we were not able to provide direct quality control over the process of data collection. Secondly, and perhaps more importantly, suicides by poisonings may be difficult to discriminate from overdoses [
58]. This could especially lead to an under-estimation of the association between opioid use and suicide, as opioids are the drugs that are primarily involved in accidental poisonings [
58].
In the present study, the case definition of suicide only included ICD-10 causes of death codes for intentional self-inflicted poisoning or injury (X60-X84), and sequelae of self-harm (Y87.0). A systematic review from 2012 concluded that suicide deaths are generally under-reported [
59]. In this review, the level of under-reporting varied between different primary studies, but high quality studies tended to report less under-reporting than studies of poorer quality. As such, it is likely that our case definition leads to lower bound estimates of suicide deaths and may dilute estimated associations. A potential solution to under-reporting is redistribution of ICD-10 death codes which may contain suicide deaths (such as undetermined intent injury codes (Y10-Y34), and exposure to undetermined factors (X59) [
2]). However, a recent Norwegian national registry study covering death certificates for the years 2005–2014 reported that redistribution of X59 codes which constituted 26% of all injury deaths, only changed the suicide estimates by 2 percentage points (60). In the same study, 12% of all injury deaths were assigned undetermined intent injury codes (Y10-Y34). The Norwegian and Danish registries of Causes of Death share many similarities, use the same coding system (ICD-10), and are both of high quality. It is therefore unlikely that redistribution of death codes would change our overall findings to a great extent, perhaps with exception of actual intentional over-doses being assigned as undetermined intent (Y10-Y15). Finally, while it was a strength to include a matched control group in this study, because it allowed us to compare the prevalence of suicide between individuals who attended DUD treatment and individuals who did not, we cannot exclude the possibility of selection bias due to the chosen matching strategy.