Opioid Use as a Proximal Risk Factor for Suicidal Behavior in Young Adults

Abstract

Introduction:

There is a concomitant rise in suicide rates with the prevalence of opioids involved in overdose deaths, especially among adolescents and young adults. However, there are limited studies on whether opioid use prospectively predict suicidal behavior in youth.

Methods:

Our sample included 183 psychiatric patients (18–30 years) admitted for a suicide attempt (SA), have current suicidal ideation (SI), and psychiatric controls without ideation or attempt (PC). Suicidal behavior was assessed using the Columbia-Suicide Severity Rating Scale. We also recruited a healthy control group (HC; n=40). Patients and controls were followed over a year. ANOVA, regression, and cox regression were used.

Results:

SA [β=0.87, CI (0.1–1.6), p=0.02] and SI [β=0.75, CI (0.03–1.5), p=0.04] were significantly more likely than HCs to have used opioids in the past year at baseline. Opioid use was associated with increased anxiety symptoms [β=0.75, CI (0.001–1.5), p=0.05], PTSD symptoms [β=3.90, CI (1.1–6.7), p=0.01], and aggression [β=0.02, CI (0.01–0.04), p=0.02]. Opioid use in the month prior to hospitalization predicted SA at 6 months [OR=1.87, CI (1.06–3.31), p=0.032].

Conclusions:

Opioid use is a proximal predictor for SA. These findings may help clinicians better identify patients at risk for suicidal behavior, allowing for more personalized treatment approaches.

INTRODUCTION

Substance use disorders (SUDs) and opioid use disorder in particular have become a growing public health problem in the United States. While the use of opioids in the treatment of chronic non-cancerous pain is debated, opioids remain among the top 10 most prescribed medications in the US due to their unmatched ability to alleviate acute pain (Fuentes et al. 2018). However, they carry the potential for addiction. Currently, opioids are the drugs most often involved in overdose deaths with greater than 120% increase in overdose deaths involving opioids between 2010 and 2018, reaching approximately 46,802 (National Institute on Drug Abuse (NIDA) 2020). The death rate from opioids, both prescription and illicit opioids, in adolescents and young adults aged 15–24 years increased by 15.4% yearly between 2013 and 2015 (Ali et al. 2019). Suicide rates in general have also risen, by 39.3% from 2000 to 2018 placing it among the top 10 leading causes of death in the US (National Center for Injury Prevention and Control, CDC 2020b). In 2018, suicide was the second leading cause of death among adolescents and young adults (National Center for Injury Prevention and Control, CDC 2020a). Adolescence and young adulthood is a high-risk period for the onset of suicidal ideation and attempt (Kessler et al., 1999).

Prior history of suicidal behavior is the most important risk factor for death by suicide (Bostwick et al., 2016). Additional risk factors for both death by suicide and suicidal behavior have been identified in adolescents and young adults and these include sex, sexual orientation, race, low socioeconomic status, negative life events in childhood and adversity, as well as psychiatric risk factors such as mental illness, depression, impulsivity, hopelessness, and substance use (Hawton et al., 2012; Cha et al., 2018; Melhem et al., 2019). However, meta-analysis of longitudinal studies from the past 50 years of research found that these risk factors did not improve the prediction of suicidal behavior as most studies did not examine the combined effect of multiple risk factors (Franklin et al., 2017). Another meta-analysis showed that suicide risk assessment based on the presence of several risk factors did not actually improve the prediction of death by suicide (Large et al., 2016). In our recent 15-year longitudinal study, we found improved prediction of suicide attempt among high-risk youth because they were the offspring of parents with mood disorders and found age younger than 30 years to be one of the predictors of suicide attempt in these offspring (Melhem et al., 2019). Most of our predictors were identified within ~8 months of suicide attempt highlighting the importance of the temporal characterization of risk factor assessment for suicidal behavior.

While there is a frequently reported relationship between opioid use and suicidal behavior (Ashrafioun et al. 2017; Darke and Ross 2002; Ilgen et al. 2016; Wilcox et al. 2004; Roy 2010; Kazour et al. 2016; Degenhardt et al. 2011) with 14 times increased risk of death by suicide among opioid users (Darke and Ross 2002; Degenhardt et al. 2011). There is limited data examining drug poisoning deaths from opioid use in adolescents and young adults. However, the death rate from opioids increased yearly by 15.4% between 2013 and 2015 in this population (Ali et al. 2019). Another study reported an increase in child and adolescent mortality rate from opioids of 268.2% between 1999 and 2016 (Gaither et al. 2018). However, it is not clear whether opioid use specifically or substance use disorders in general increases risk for suicidal behavior. Studies reported alcohol use disorder (AUD) and acute alcohol use to be associated with increased suicidal ideation (SI) and behavior in adolescents, young adult, and adult samples (Boenisch et al. 2010; Darvishi et al. 2015; Flensborg-Madsen et al. 2009; Sher et al. 2009; Wilcox et al. 2004). Higher rates of suicidal behavior and death by suicide were also associated with tobacco in adolescents, young adult, and adult samples (Berlin et al. 2015; Boden et al. 2008; Bohnert et al. 2014; Evins et al. 2017; Miller et al. 2000; Poorolajal and Darvishi 2016), cocaine (Artenie et al. 2015), and amphetamines in adults (Artenie et al. 2015; Youssef et al. 2016) and adolescent samples as early as 14 years of age and older (Marshall et al. 2011). The literature on whether marijuana is associated with increased risk for suicidal behavior in youth is mixed (Borowsky et al. 2001; Kung et al. 2003; Pedersen 2008; Rylander et al. 2014), although a recent large comprehensive review reported a moderate association between cannabis use, SI, suicide attempt (SA), and death by suicide (National Academies of Sciences, Engineering, and Medicine and Committee on the Health Effects of Marijuana: An Evidence Review and Research Agenda 2017). The use of marijuana has been also found to be associated with increased health risk behaviors including injury and suicidal ideation and attempt in adolescents (Schauer et al. 2020). Finally, SUDs were reported to be important risk factors for SA in the months and days preceding SA in adult patients (Walsh et al. 2017). Most of the aforementioned studies are cross-sectional in nature and do not examine opioid and substance use as predictors for future suicidal behavior in high-risk youth psychiatric patients.

In this longitudinal study, we examined the relationship between opioid use and suicidal behavior in high-risk young adult psychiatric patients; the relationship of opioid use with clinical correlates of suicidal behavior (e.g., depression, hopelessness, impulsivity, aggression); and whether opioid use predicts suicidal behavior prospectively in this population. While opioid use was our primary focus, we also assessed use of other substances to determine whether opioid use was unique in terms of its cross-sectional and prospective relationships to suicidal behavior in young adults.

MATERIALS & METHODS

Sample:

The sample was derived from a longitudinal study examining biological markers in the hypothalamic-pituitary-adrenal (HPA) axis and inflammatory pathways for suicidal behavior, “Promising Biomarkers for Suicidal Thoughts and Behavior in Youth, the PROmiSe Study”. Psychiatric patients, aged 18–30 years, in inpatient or outpatient treatment for psychiatric disorders were included. Inpatients were recruited from Western Psychiatric Hospital at University of Pittsburgh Medical Center (UPMC) and outpatients were recruited through the Clinical Translational Science Institute (CTSI) Participant Research Registry, which includes ~230,000 participants recruited at points of routine clinical care at UPMC, from outreach events, and MyChart, a patient health portal app. The sample consisted of inpatients admitted for suicide attempt (SA, n= 67), those with suicidal ideation with no prior history of SA (SI, inpatients n=72; outpatients n=2) and psychiatric controls without current suicidal ideation or SA (psychiatric controls, inpatients n=34, outpatients n=8). For inpatients, a member of their treatment team approached them about the study and those who were interested and able to provide informed consent were recruited to the study. We also recruited healthy controls (HCs, n=40) with no personal history of psychiatric disorders and suicidal behavior, or family history of severe psychiatric disorders (e.g., bipolar disorder, schizophrenia) from the CTSI Research Participant registry. While healthy controls did not meet criteria for substance use disorders by study design, they could still use opioid and other substances and hence wanted to compare the extent to which youth psychiatric patients were different from a healthy control group on opioid and other substance use. Because the aims of the study from which these data were generated were to examine biological markers in the HPA axis and inflammatory pathways for suicidal behavior, our exclusion criteria for all groups included chronic inflammatory conditions and the use of medications that may affect inflammatory measures (e.g., NSAIDs, steroids, immunosuppressants and chemotherapy). HCs were also excluded if they had a prior history of childhood trauma, which affects these biological pathways. All participants were recruited in accordance with the University of Pittsburgh Institutional Review Board (IRB).

Assessments:

Lifetime history of suicidal ideation and attempt were assessed using the Columbia Suicide Severity Rating Scale or C-SSRS (Posner et al. 2011). The Structured Clinical Interview for DSM-5 disorders (SCID-5) was administered to assess lifetime history and current psychiatric disorders. We also obtained consent to access medical records. Psychiatric diagnoses used in our analyses were based on information from the SCID-5 and medical records. To assess the severity of psychiatric symptoms, the Patient Health Questionnaire (PHQ9) (Kroenke et al. 2001), Generalized Anxiety Disorder scale (GAD7) (Spitzer et al. 2006), PTSD Checklist for DSM-5 (PCL-5) (Blevins et al. 2015), Beck Hopelessness Scale (BHS) (Beck et al. 1974), Suicidal Ideation Questionnaire (SIQ) (Reynolds 1991), Barratt Impulsiveness Scale (BIS) (Barratt 1965; Stanford et al. 2009) and the Aggression Questionnaire (Buss and Perry 1992) were used to measure the severity of self-reported depression, anxiety, and PTSD symptoms, hopelessness, suicidal ideation, impulsivity, and aggression respectively. Substance use were assessed using the Drug Use Screening Inventory-Revised (DUSI-R), which has been shown to have high reliability and validity (Tarter and Kirisci 1997; Kirisci et al. 1995; Siewert et al. 2004). The DUSI-R quantifies substance use on a scale of 1 to 5 with 1=0 times, 2=1–2 times, 3=3–9 times, 4=10–20 times, and 5= more than 20 times. The DUSI-R assesses the frequency of use for each substance over the past month and past year. Participants were followed at 3, 6, and 12 months after baseline. We used the Longitudinal Interval Follow-Up Evaluation (LIFE) to record symptom variations over weekly intervals since the last assessment to determine remission and onset of new disorders and suicidal ideation and behavior. In addition to the LIFE to track suicidal behavior prospectively, we also included information collected from medical records. SA at follow-up was defined using a narrow definition of actual SA, defined as an attempt with a clear intent to die; and a broader definition including actual SA and suicide related behaviors of interrupted, aborted, and ambiguous attempts as defined on the C-SSRS. Non-suicidal self-injurious behavior was not included in our broad definition of attempt. We also assessed socioeconomic status (SES) using the MacArthur Subjective Social Status scale (Adler et al., 2000).

Statistical analyses:

We compared SA, SI, psychiatric controls and HCs on demographics, clinical characteristics and opioid and substance use at baseline using chi-square tests and analysis of variance (ANOVA). We only included substances endorsed by at least 10% of the sample in our analyses. We conducted post-hoc pairwise comparisons and applied a Bonferroni correction to correct for multiple comparisons. Our alpha level was 0.05 and post-hoc comparisons were corrected at α=0.05/6=0.008. Linear regression analyses were conducted to compare groups at baseline (SA, SI, psychiatric controls, and HCs) with use of opioid and each of the other substances in the past year and past month, one model at a time, while controlling for sex, age, race, prior history of SA and other diagnoses variables that were significantly different between groups. Opioid use and each substance in the past year and past month were the dependent variables in these models, one at a time, and group (SA, SI, psychiatric controls, and HCs) was the independent variable. Group was used as a dummy coded variable with HCs as the reference category. To examine the relationships of opioid use with clinical correlates of suicidal behavior at baseline, we used Pearson’s correlation and linear regression with opioid and substance use in the past year and past month at baseline as the dependent variable and each clinical correlate at baseline as the independent variable controlling for group, age, sex and race. We did not control for SES due to missing data in 15% (34/223) of the sample; however, we conducted sensitivity analyses on the subset without missingness controlling for SES and similar results were obtained. For all of the regression models, analyses were conducted with and without HCs since we selected HCs to have no prior history of psychiatric diagnoses and suicidal behavior and expected them to have lower levels of psychiatric symptoms, by study design, compared to psychiatric patients. When excluding HCs, psychiatric controls were used as the reference category.

To examine whether opioid and substance use at baseline predict suicide attempt prospectively, we first conducted univariate analyses comparing those with and without suicide attempt at 6 and 12 months on demographics, clinical characteristics, and opioid and substance use at baseline using chi-square and t-tests. These analyses were restricted to psychiatric patients because we did not have any suicidal events among HCs. We then conducted logistic regression to examine whether opioid use in the past year and past month at baseline predicted SA at 6 and 12 months follow-ups, which were the dependent variables in these models. We controlled for age, race, sex, and history of prior SA in these models. In addition, we controlled for other clinical covariates that were statistically significantly associated with SA. We selected final models to be the most parsimonious models excluding other covariates that were not significant. We used the narrow and broad definitions of SA for these analyses as two separate outcomes and only participants who have been in the study for at least 6 and 12 months were included, respectively. We did not have enough events for the narrow (n=1) and broad (n=6) definitions of SA at 3 months and hence we did not conduct separate analyses for the 3-months SA outcome. Similar analyses were conducted to examine the relationships of opioid and substance use in the past month and past year at baseline with time to onset of SA using cox regression analyses. Time to onset of SA for the year following baseline was our outcome, which constitutes the year post hospital discharge for most participants, a high-risk period for suicidal behavior. These analyses included all participants and accounted for censored observations. All statistical analyses were done using StataSE version 15.1.

RESULTS

Characteristics of the sample:

Our sample had a mean age of 23.3 ± 3.7 years, was equally distributed by sex (50.9%, n=114, males), and 74.8% (n=166) identified as White. SA, SI, psychiatric controls, and HCs were similar with respect to sex and age (Table 1). HCs showed significantly higher SES than other groups. While there was an overall significant difference between groups on race and impulsivity, none of the post-hoc comparisons were significant after correcting for multiple comparisons. SA, SI, and psychiatric controls were significantly different at baseline from HC with higher scores on self-reported depression, anxiety, hopelessness, PTSD, and aggression. SA and SI were similar and significantly different from psychiatric controls with higher scores on depression and anxiety symptoms and on suicidal ideation. SA were significantly different from psychiatric controls and HCs with higher scores on PTSD symptoms. SA, SI and psychiatric controls were similar on all primary and comorbid psychiatric diagnoses with the exception of mood disorder. SA and SI were similar on rates of mood disorders, (70.2% and 81.7%, respectively) but only SI was significantly different from psychiatric controls (54.8%) after correcting for multiple comparisons. SA were more likely to have a lifetime history of SA (64.1%) compared to SI (45.7%), who in turn were more likely to have a history of SA than psychiatric controls (19.5%) (Table 1).

Table 1.

Demographics, clinical characteristics and frequency of opioid and substance use in those who attempted suicide, subjects with suicidal ideation, psychiatric patients without SI or SA, and healthy controls

Demographics	Subjects who attempt suicide (SA) n=67	Subjects with suicidal ideation (SI) n=74	Psychiatric patients without SA or SI n=42	Healthy Controls n=40	Test	df	p
Sex, % females (n)	56.7% (38)	46.0% (34)	36.6% (15)	55.0% (22)	x2 = 5.0	3	0.17
Race, % White (n)	64.1% (n=41)	80.3% (57)	69.1% (29)	87.5% (35)	x2 = 9.2	3	0.03
Age, mean ± SD	23.3 ± 3.7	24.3 ± 4.1	23.9 ± 3.4	24.9 ± 3.6	F = 1.9	3, 219	0.14
SES**	−0.3 ± 0.7 a	−0.2 ± 0.8 a	−0.1 ± 0.7 a	0.5 ± 0.7 b	F = 8.9	3,185	<0.001
Clinical correlates of suicidal behavior
Depression symptoms, mean ± SD	21.1 ± 8.6a	22.5 ± 8.6a	12.3 ± 8.5b	0.9 ± 1.6c	F = 82.7	3, 210	< 0.001
Anxiety symptoms, mean ± SD	12.4 ± 5.2a	13.1 ± 5.2a	8.6 ± 5.2b	0.7 ± 1.3c	F = 67.7	3, 210	< 0.001
Hopelessness, mean ± SD	10.6 ± 2.5a	10.3 ± 2.1a	10.1 ± 2.6a	8.4 ± 1.9b	F = 8.62	3, 210	< 0.001
Suicidal ideation, mean ± SD	83.6 ± 38.2a	88.4 ± 33.3a	20.6 ± 23.0b	1.2 ± 3.1 b	F = 106.7	3, 210	< 0.001
PTSD symptoms, mean ± SD	45.9 ± 17.9a	39.7 ± 19.1a, b	29.1 ± 19.8b	1.4 ± 1.9c	F = 57.9	3, 210	< 0.001
Impulsivity, mean ± SD	69.0 ± 8.0	68.1 ± 7.9	65.5 ± 8.3	64.2 ± 4.8	F = 3.9	3, 174	0.01
Aggression, mean ± SD	0.56 ± 0.1a, b	0.57 ± 0.1 a	0.47 ± 0.1 b	0.33 ± 0.1 c	F = 34.1	3, 204	< 0.001
Psychiatric Diagnoses
Anxiety disorders, % (n)	40.4% (23)	33.3% (20)	22.6% (7)	-	x2 = 2.8	2	0.24
Mood disorders, % (n)	70.2% (40) a, b	81.7% (49) a	54.8% (17) b	-	x2 = 7.3	2	0.03
Psychotic disorders, % (n)	7.0% (4)	6.7% (4)	12.9% (4)	-	FET*	2	0.55
PTSD, % (n)	24.6% (14)	20.0% (12)	22.6% (7)	-	x2 = 0.4	2	0.84
Substance use disorder, % (n)	35.1% (20)	41.7% (25)	51.6% (16)	-	x2 = 2.3	2	0.32
Lifetime history of suicide attempt	64.1% (41)a	45.7% (32)b	19.5% (8)c	-	x2 = 20.0	2	< 0.001
Substance use in past year
Alcohol, mean ± SD	3.1 ± 1.3	2.9 ± 1.4	3.2 ± 1.3	2.6 ± 0.9	F=1.6	3, 202	0.20
Stimulants, mean ± SD	1.4 ± 0.9	1.4 ± 1.0	1.5 ± 1.1	1.1 ± 0.3	F=2.0	3, 202	0.12
Cocaine, mean ± SD	1.6 ± 1.1 a	1.5 ± 1.0 a	1.4 ± 0.8 a, b	1.0 ± 0.2 b	F=4.1	3, 202	0.01
OTC medications, mean ± SD	2.4 ± 1.5 a	2.3 ± 1.3 a, b	2.3 ± 1.5 a, b	1.7 ± 0.9 b	F = 2.7	3, 202	0.05
Opioids, mean ± SD	1.6 ± 1.3	1.5 ± 1.3	1.4 ± 1.2	1.0 ± 0.0	F=2.5	3, 202	0.06
Prescription pain killers, mean ± SD	1.5 ± 1.0	1.5 ± 1.0	1.7 ± 1.4	1.1 ± 0.5	F = 2.3	3, 202	0.08
Marijuana, mean ± SD	3.4 ± 1.7 a	3.0 ± 1.8 a	3.0 ± 1.7 a	1.3 ± 0.8 b	F=15.3	3, 202	< 0.001
Cigarettes, mean ± SD	3.0 ± 1.9 a	3.1 ± 1.9 a	2.8 ± 1.8 a	1.3 ± 0.7 b	F=11.3	3, 202	< 0.001
Substance use in past month
Alcohol, mean ± SD	2.4 ± 0.9	2.4 ± 1.3	2.6 ± 1.2	2.4 ± 0.9	F=0.8	3, 202	0.48
Cocaine, mean ± SD	1.4 ± 1.0	1.4 ± 1.0	1.3 ± 0.7	1.0 ± 0.0	F = 2.3	3, 202	0.08
OTC medications, mean ± SD	1.9 ± 1.3	1.7 ± 1.1	1.7 ± 1.3	1.3 ± 0.6	F=2.4	3, 202	0.07
Opioids, mean ± SD	1.4 ± 1.1	1.4 ± 1.1	1.3 ± 1.0	1.0 ± 0.0	F = 1.9	3, 202	0.12
Prescription pain killers, mean ± SD	1.3 ± 0.7	1.2 ± 0.7	1.4 ± 1.1	1.0 ± 0.0	F=2.6	3, 202	0.05
Marijuana, mean ± SD	2.8 ± 1.7 a	2.6 ± 1.8 a	2.9 ± 2.8 a	1.2 ± 0.5 b	F=10.4	3, 202	< 0.001
Cigarettes, mean ± SD	2.6 ± 1.8 a	2.8 ± 1.9 a	2.7 ± 1.8 a	1.2 ± 0.8 b	F=8.3	3, 202	< 0.001

^*Fisher’s Exact Test. Letter superscripts reflect post-hoc comparisons;

^**Z-scores

Relationship between opioid use, other substance use, and suicidal behavior at baseline:

The frequency of past year and past month opioid use was similar between all groups (Table 1). All groups were also similar on past year use of alcohol, amphetamines/stimulants and prescription pain killers. SA, SI, and psychiatric controls were similar on past-year cocaine use but only SA and SI were significantly different from HCs [1.6±1.1 vs. 1.5±1.0 vs. 1.4±0.8 vs. 1.0±0.2, F=4.1, df (3,202), p=0.01]. SA were significantly different from HCs on past year use of over-the-counter medications [2.4±1.5 vs. 2.3±1.3 vs. 2.3±1.5 vs. 1.7±0.9, F=2.68, df (3,202), p=0.048]. SA, SI and psychiatric controls were similar but significantly different than HCs on the frequency of past-year marijuana and cigarettes.

When controlling for covariates (age, race, sex, lifetime history of suicide attempt, mood disorders), SA [β=0.87, 95% CI (0.1–1.6), p=0.02] and SI [β=0.75, 95% CI (0.03–1.5), p=0.04] were significantly more likely than all other groups to have used opioids in the past year. SA [β=2.07, 95% CI (1.2–3.0), p<0.001], SI [β=1.52, 95% CI (0.6–2.4), p=0.001], and psychiatric controls [β=1.42, 95% CI (0.6–2.3), p=0.002] were more likely to have used marijuana in the past year compared to HCs. Similarly, SA [β=1.87, 95% CI (0.9–2.9), p<0.001], SI [β=2.08, 95% CI (1.1–3.1), p<0.001], and psychiatric controls [β=1.89, 95% CI (0.9–2.8), p<0.001] were more likely to have used cigarettes in the past year compared to HCs (Table 2). When excluding HCs, SA, SI, and psychiatric controls were similar on the use of opioids and other substances in the past year (Table S1).

Table 2.

Regression analyses examining the relationship between past-year and past month opioid and substance use with group controlling for covariates

	Alcohol		Stimulants		Cocaine		OTC Medications		Opioids		Prescription Pain Killers		Marijuana		Cigarettes
Past Year	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p	Coeff 95% CI	p
Age	0.05 0.001–0.1	0.05	0.02 −0.03–0.1	0.48	0.03 −0.01–0.1	0.10	−0.03 −0.1–0.03	0.32	0.04 −0.01–0.1	0.14	−0.01 −0.1–0.04	0.77	0.01 −0.1–0.1	0.67	0.08 0.01–0.2	0.02
Race*, White	0.37 −0.1–0.8	0.12	0.13 −0.2–0.5	0.47	0.05 −0.3–0.4	0.76	0.34 −0.2–0.9	0.20	0.31 −0.1–0.8	0.18	0.11 −0.3–0.5	0.58	−0.49 −1.1–0.1	0.09	0.13 −0.5–0.8	0.68
Sex**, Female	−0.11 −0.5–0.3	0.60	−0.06 −0.4–0.3	0.73	0.01 −0.3–0.3	0.96	−0.02 −0.5–0.4	0.93	−0.25 −0.7–0.1	0.21	−0.04 −0.4–0.3	0.84	−0.75 −1.2–−0.3	0.003	−0.58 −1.1–−0.04	0.04
Lifetime history of attempt	0.26 −0.2–0.8	0.29	−0.03 −0.4–0.4	0.86	0.32 −0.1–0.7	0.09	0.29 −0.3–0.8	0.29	0.16 −0.3–0.6	0.50	0.30 −0.1–0.7	0.15	−0.11 −0.7–0.5	0.72	0.40 −0.3–1.1	0.22
Mood Disorder	0.24 −0.3–0.8	0.36	0.03 −0.4–0.4	0.87	−0.05 −0.4–0.3	0.78	0.17 −0.4–0.7	0.55	−0.18 −0.7–0.3	0.47	−0.15 −0.6–0.3	0.51	0.04 −0.6–0.7	0.91	−0.43 −1.1–0.3	0.22
Group***														<		<
SA	0.45 −0.3–1.2	0.24	0.50 −0.1–1.1	0.10	0.52 −0.1–1.1	0.07	0.51 −0.3–1.4	0.23	0.87 0.1–1.6	0.02	0.32 −0.3–1.0	0.33	2.07 1.2–3.0	0.001	1.87 0.9–2.9	0.001
SI	0.21 −0.5–1.0	0.58	0.46 −0.1–1.1	0.12	0.56 0.003–1.1	0.05	0.37 −0.5–1.2	0.38	0.75 0.03–1.5	0.04	0.38 −0.3–1.0	0.24	1.52 0.6–2.4	0.001	2.08 1.1–3.1	<0.001
Psych	0.66 −0.1–1.4	0.07	0.40 −0.2–1.0	0.16	0.28 −0.3–0.8	0.31	0.58 −0.2–1.4	0.15	0.57 −0.1–1.3	0.11	0.59 −0.03–1.2	0.06	1.42 0.6–2.3	0.002	1.89 0.9–2.8	<0.001
Past month
Age	0.06 −0.004–0.1	0.04	- -	-	0.03 −0.003–0.1	0.08	0.04 −0.01–0.1	0.16	0.06 0.01–0.1	0.01	0.01 −0.02–0.04	0.56	0.02 −0.04–0.1	0.51	0.06 −0.01–0.1	0.11
Race* White	0.17 −0.3–0.6	0.47	- -	-	−0.13 −0.2–0.4	0.39	0.05 −0.4–0.5	0.83	0.15 −0.2–0.5	0.43	−0.18 −0.5–0.1	0.18	−0.52 −1.1–0.1	0.08	0.27 −0.3–0.9	0.40
Sex** Female	−0.04 −0.4–0.4	0.84	- -	-	0.10 −0.2–0.4	0.45	0.20 −0.2–0.6	0.28	−0.10 −0.4–0.2	0.54	−0.03 −0.3–0.2	0.79	−0.70 −1.2–−0.2	0.01	−0.74 −1.3–−0.2	0.01
Lifetime history of attempt	0.29 −0.2–0.8	0.23	- -	-	0.27 −0.1–0.6	0.10	0.22 −0.2–0.7	0.32	0.26 −0.1–0.6	0.19	0.23 −0.1–0.5	0.10	0.15 −0.5–0.8	0.62	0.62 −0.02–1.3	0.06
Mood Disorder	−0.05 −0.6–0.5	0.84	- -	-	−0.01 −0.3–0.8	0.98	−0.01 −0.5–0.5	0.96	0.15 −0.3–0.6	0.46	0.19 −0.1–0.5	0.21	0.20 −0.4–0.8	0.52	−0.37 −1.0–0.3	0.41
Group***			-	-
SA	0.40 −0.4–1.1	0.30			0.26 −0.2–0.8	0.31	0.50 −0.2–1.2	0.16	0.37 −0.2–1.0	0.22	−0.004 −0.4–0.4	0.99	1.31 0.4–2.3	0.01	1.33 0.3–2.3	0.01
SI	0.08 −0.7–0.8	0.82			0.40 −0.1–0.9	0.11	0.26 −0.4–0.9	0.44	0.27 −0.3–0.9	0.37	−0.03 −0.5–0.4	0.88	0.84 −0.1–1.8	0.07	1.68 0.7–2.7	0.001
Psych	0.23 −0.5–0.9	0.53			0.21 −0.3–0.7	0.37	0.46 −0.2–1.1	0.17	0.25 −0.3–0.8	0.39	0.23 −0.2–0.6	0.28	1.38 0.5–2.3	0.002	1.61 0.7–2.6	0.001

^*Compared to Non-White;

^**Compared to males;

^***Compared to healthy controls

None of the groups were significantly different on opioid use in the past month compared to HCs. However, SA [β=1.31, 95% CI (0.4–2.3), p=0.01] and psychiatric controls [β=1.38, 95% CI (0.5–2.3), p=0.002] were more likely than all other groups to have used marijuana in the past month. SA [β=1.33, 95% CI (0.3–2.3), p=0.01], SI [β=1.68, 95% CI (0.7–2.7), p=0.001], and psychiatric controls [β=1.61, 95% CI (0.7–2.6), p=0.001] were also significantly more likely to smoke cigarettes in the past month compared to HCs. When excluding HCs, SA, SI, and psychiatric controls were similar on past month use of opioid and other substances after controlling for covariates (Table S1).

Relationship between opioid use and clinical correlates of suicidal behavior:

Table S2 and S3 presents the correlations of past-year and past-month opioid use with clinical correlates of suicidal behavior, with and without HCs. When controlling for covariates, anxiety symptoms [β=0.75, 95% CI (0.01–1.5), p=0.05], PTSD symptoms [β=3.91, 95% CI (1.1–6.7), p=0.01]; and aggression [β=0.02, 95% CI (0.01–0.04), p=0.02] were significantly associated with opioid use in the past year (Table 3); however, opioid use in the past month was not significantly associated with any of the clinical correlates (Table 3). Tables S4 and S5 present the correlations between opioid and substance use in the past year and past month, with and without HCs.

Table 3.

Regression analyses examining the relationship of opioid use in the past year and past month with clinical correlates of suicidal behavior among psychiatric patients

	Coefficient	95% CI	P
Opioid use in the past year
Depression Symptoms	0.80	−0.4–2.0	0.19
Anxiety Symptoms	0.75	0.01–1.5	0.05
Suicidal Ideation	−0.58	−5.5–4.3	0.82
PTSD Symptoms	3.91	1.1–6.7	0.01
Hopelessness	0.08	−0.3–0.4	0.63
Impulsivity	−0.07	−1.2–1.1	0.91
Aggression	0.02	0.01–0.04	0.02
Opioid use in the past month
Depression Symptoms	0.53	−1.0–2.1	0.50
Anxiety Symptoms	0.56	−0.4–1.5	0.24
Suicidal Ideation	−0.74	−6.9–5.4	0.81
PTSD Symptoms	3.29	−0.2–6.8	0.07
Hopelessness	0.13	−0.3–0.6	0.54
Impulsivity	−0.40	−1.8–1.0	0.56
Aggression	0.02	−0.002–0.1	0.08

Opioid and other substance use as predictors of suicide attempt prospectively.

Tables S6 and S7 present the comparisons between those with and without suicide attempt at 6 and 12 months, respectively, using the narrow and broad definitions. At the univariate level, there were no significant differences between those with and those without a suicide attempt at 6 and 12 months on opioid and substance use in the past year and past month at baseline. When controlling for sex, age, race, lifetime history of SA, up to and including the baseline attempt for the SA group, and clinical characteristics that were significantly associated with SA at follow-up (Tables S6–S7), opioid use in the past month at baseline significantly predicted actual SA at 6 months [OR=1.87, 95% CI (1.1–3.3), p=0.03] (Table 4). Cox regression analyses were conducted to examine the univariate relationships between covariates and time to onset of attempt (Table S8). Opioid use in the past year and past month at baseline did not predict time to onset of attempt using the narrow and broad definitions at the univariate and multivariate levels although there was a HR of 1.4 (95% CI 0.88–2.23) for past month opioid use at baseline predicting time to onset of actual attempt that did not reach statistical significance (p=0.154). The severity of self-reported depression symptoms, anxiety disorders, and prior history of suicide attempt were important predictors of time to onset of attempt (Table 4).

Table 4.

Regression analyses for the relationship of opioid use in the past year and past month at baseline with suicide attempt at 6 and 12 months; and cox regression for time to onset of suicide attempt

	Opioid use in the past year						Opioid use in the past month
	Narrow SA			Broad SA			Narrow SA			Broad SA
Suicide attempt at 6 months	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Age	0.86	0.7–1.1	0.32	0.96	0.8–1.1	0.65	0.85	0.7–1.1	0.20	0.96	0.8–1.1	0.65
Race*, White	3.79	0.4–34.6	0.23	1.69	0.4–6.7	0.46	4.40	0.5–41.5	0.20	1.69	0.4–6.7	0.46
Sex**, Female	4.07	0.5–12.6	0.27	1.24	0.4–4.0	0.72	2.64	0.5–13.5	0.24	1.23	0.4–4.0	0.72
Lifetime history of attempt	2.50	0.5–13.4	0.29	2.21	0.6–7.6	0.21	2.33	0.4–12.8	0.33	2.21	0.6–7.6	0.21
Opioid use	1.47	0.9–2.4	0.12	1.30	0.9–1.9	0.17	1.87	1.1–3.3	0.03	1.30	0.9–1.9	0.17
Suicide attempt at 12 months
Age	1.03	0.8–1.3	0.79	1.06	0.9–1.3	0.54	1.00	0.8–1.2	0.96	1.05	0.9–1.3	0.56
Race*, White	4.24	0.5–39.5	0.20	0.38	0.1–1.8	0.22	4.65	0.5–45.6	0.19	0.39	0.1–1.8	0.22
Sex**, Female	5.19	0.9–30.7	0.07	1.14	0.3–4.9	0.86	5.32	0.9–32.2	0.07	1.13	0.3–4.8	0.87
Lifetime history of attempt	2.74	0.5–15.5	0.25	0.78	0.2–3.1	0.73	2.91	0.5–17.0	0.24	0.78	0.2–3.1	0.73
Opioid use	1.26	0.8–2.1	0.36	1.07	0.7–1.6	0.76	1.52	0.9–2.7	0.15	1.08	0.7–1.7	0.75
Anxiety disorders	-	-	-	5.65	1.3–24.9	0.02	-	-	-	5.57	1.3–24.8	0.02
PTSD diagnosis	-	-	-	4.60	1.1–19.5	0.04	-	-	-	4.59	1.1–19.5	0.04
	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p	HR	95% CI	p
Time to onset of suicide attempt
Age	0.98	(0.82, 1.16)	0.79	0.97	(0.86, 1.10)	0.66	0.96	(0.80, 1.15)	0.63	1.03	(0.90, 1.17)	0.70
Race*, White	5.15	(0.65, 41.0)	0.12	1.83	(0.60, 5.59)	0.29	5.30	(0.66, 42.3)	0.12	3.01	(0.67, 13.5)	0.15
Sex**, Female	2.00	(0.55, 7.31)	0.29	1.12	(0.46, 2.72)	0.80	2.26	(0.59, 8.60)	0.23	1.08	(0.41, 2.80)	0.88
Lifetime history of attempt	6.24	(1.30, 29.9)	0.02	2.65	(1.04, 6.77)	0.04	5.82	(1.20, 28.1)	0.03	1.75	(0.61, 5.00)	0.29
Opioid use	1.17	(0.77, 1.78)	0.45	1.08	(0.79, 1.50)	0.62	1.40	(0.88, 2.23)	0.15	0.66	(0.39, 1.12)	0.12
Anxiety disorders	-	-	-	-	-	-	-	-	-	3.24	(1.22, 8.64)	0.02
Depression symptoms	-	-	-	-	-	-	-	-	-	1.07	(1.01, 1.13)	0.02

^*Compared to non-White;

^**Compared to males

DISCUSSION

We found young adult patients admitted for suicide attempt to be more likely than those with suicidal ideation, psychiatric controls, and healthy controls to have used opioids in the year prior to hospitalization even after controlling for prior history of SA and other important covariates. Opioid use in the past year was associated with higher current anxiety and PTSD symptoms and aggression at baseline. In addition, opioid use in the past month at baseline predicted prospectively the narrow definition of actual SA at 6 but not 12 months. Thus, opioid use is a proximal predictor for SA. Anxiety and PTSD diagnoses predicted the broad definition of SA at 12 months; and the severity of self-reported depression symptoms, anxiety disorders, and prior history of suicide attempt were important predictors of time to onset of attempt. Finally, psychiatric patients were also more likely to use marijuana and smoke cigarettes in the past year and past month compared to healthy controls.

We discuss these findings in the context of the strengths and limitations of the study. This study included a large sample of young adult psychiatric patients across the spectrum of psychiatric diagnoses as well as suicidal ideation and behavior who were followed for a year (the year post-hospital discharge for most of the sample). The year post-hospital discharge is a high-risk period for suicidal behavior, with 67–83% of deaths by suicide occurring in the first 6 months following discharge (King et al. 2001; Yim et al. 2004). Our sample is comprised of young adults, 18–30 years of age, who have been previously shown to be at high-risk for suicidal behavior (Melhem et al., 2019; Piscopo et al., 2006). According to the National Survey on Drug Use and Health, the percent of US adult individuals with serious thoughts of suicide was found to be higher among those 18–25 years of age (8.3%) compared to other adult age groups, namely, 26–49 (4.1%) and 50 and older (2.6%), in 2015 (Piscopo et al. 2006). Adolescents and young adults are also at increased risk for death from opioid overdoses (Ali et al. 2019). Our longitudinal design is a strength allowing us to examine opioid use and other substances as predictors for suicidal behavior prospectively. Another strength of our study design is the focus on suicide attempt as an outcome with a detailed assessment of intent. A major limitation of studies examining the role of opioid use in risk for death by suicide is the difficulty identifying suicidal intent among those who died from opioid overdose and the potential misclassification of opioid overdoses as accidental deaths (Bohnert et al. 2010; Finkelstein et al. 2015). However, our results are limited by the reliance on self-reported measures of opioid use and other substance use and thus recall bias and underreporting of substance use could have affected our findings. We also did not collect data on the duration of opioid and substance use.

We found SA to be more likely to have used opioid in the past year compared to all other groups cross-sectionally at baseline and opioid use in the past year was associated with higher current anxiety and PTSD symptoms and higher aggression at baseline. It is plausible that opioid use may be linked to risk of SA through increased anxiety and PTSD symptoms and aggression. Indeed, we have previously reported aggression to be a significant predictor of SA (Melhem et al. 2007) and greater aggression has been reported among opioid users with history of SA compared to users without SA (Trémeau et al. 2008). Opioid antagonists have been shown to reduce self-injurious behaviors (Sandman et al. 1990) as well as aggression in humans and mice (Puglisi-Allegra et al. 1982; Goodman and New 2000; Giacomuzzi et al. 2006). PTSD has previously been linked to opioid use and opioid use disorder in general (Mills et al. 2006; Meier et al. 2014; Hassan et al. 2017; Riblet et al. 2020) as well as in young adults (Mackesy-Amiti et al. 2015). There is also a well-documented association between PTSD and suicidal behavior (Krysinska and Lester 2010; Allan et al. 2019). It has been hypothesized that opioid use disorder may exacerbate suicidal behavior in those with PTSD, given the potentially lethal nature of opioids and increased impulsivity within the context of PTSD (Riblet et al. 2020). Previous studies also reported an association between anxiety and non-medical opioid use (Martins et al. 2012; Amari et al. 2011).

We also found that opioid use in the past month at baseline predicted actual SA at 6 months even after controlling for comorbid disorders and prior history of SA. Longitudinal studies characterizing the temporal relationship of opioid and substance use with suicide attempt in youth are scarce. Prior studies reported adolescent high school students who have misused prescription opioids in the past 12 months to be more likely to experience suicidal ideation, plan, and attempt during the past 12 months compared to those who have never misused opioids (Baiden et al. 2019). These results extend prior studies by highlighting the importance of opioid use as a proximal predictor for SA in youth in the 6 months post-discharge from psychiatric hospitalization. These findings did not extend to the broader definition of SA, which included interrupted, aborted, and ambiguous attempts. Opioid use may serve as a proxy for the severity of pain and it is possible that those with actual attempt comprise those with more severe suffering. This is indirectly supported by strong evidence of a relationship between chronic pain and suicide deaths (Ilgen et al. 2013). Adolescents and young adults with prior history of mental illness and a first episode of chronic pain are also 2.4 times more likely to be prescribed a chronic opioid (Richardson et al. 2012). Higher doses of opioids are usually prescribed due to inadequate pain control at lower doses, and higher doses of opioids are associated with increased risk for suicide (Ilgen et al. 2016). There is also a relationship between the dose of acetaminophen (another analgesic) and suicide, suggesting that the increased risk might be specific to the use of analgesics (Ilgen et al. 2016). However, we found no significant relationship between over-the-counter medications and suicide attempt in this study. Future studies are needed to investigate whether the nature of a person’s opioid use in terms of prescription vs. illicit drug use, dose, and duration moderate the prospective relationship of opioid use and SA in high-risk individuals.

We found anxiety and PTSD diagnoses to predict SA using the broad definition at 12 months only. These results are consistent with studies showing PTSD to independently predict re-admissions related to self-harm after discharge from psychiatric hospitalization (Mellesdal et al. 2015). Anxiety also has been shown to be an independent risk factor for suicidal behavior above and beyond other psychiatric comorbidities (Nock et al. 2010; Nock et al. 2009). We have also found the severity of self-reported depression symptoms, anxiety disorders, and prior history of suicide attempt to predict time to onset of attempt, consistent with our prior work showing the severity of self-reported depression symptoms to be an important predictor above and beyond diagnoses and other clinical characteristics (Melhem et al. 2007) Collectively, our results suggest an underlying shared psychiatric etiology with complex pathways for the relationship between opioid use and risk for suicidal behavior with anxiety, chronic pain, depression symptoms, and PTSD as some of these shared risk factors.

In addition to opioid use, marijuana and cigarette smoking in the past year and past month were also significantly increased in psychiatric patients compared to healthy controls. These results are consistent with studies reporting increased cigarette smoking in people with severe mental illness compared to the general population (Taylor et al. 2014). Studies have previously reported an association between smoking and increased risk for suicidal behavior and death by suicide (Poorolajal and Darvishi 2016). Marijuana use during adolescence has been linked to onset of mental illness and psychosis in specific (Bourque et al. 2018). A meta-analysis reported marijuana use in adolescence to be associated with increased risk for onset of depression (OR= 1.37 (95% CI, 1.16–1.62), anxiety [OR=1.18 (95% CI, 0.84–1.67)], suicidal ideation [OR=1.50, 95% CI (1.11–2.03)], and attempt [OR=3.46 (95% CI, 1.53–7.84)] in young adulthood with the highest odds being for risk of suicide attempt (Gobbi et al. 2019).

Our findings have important clinical implications. Clinicians need to monitor and treat opioid use in psychiatric patients at high-risk for suicidal behavior and to monitor and address suicidal ideation and behavior in opioid users. Clinicians also need to assess and treat comorbid psychiatric disorders and chronic pain that could increase risk for opioid use and suicidal behavior. Future longitudinal studies are also needed to examine the trajectories of opioid use prospectively and whether the risk for SA varies with these trajectories.