Correlates and predictors of suicidal ideation and substance use among adults seeking substance use treatment with varying levels of suicidality

1. Introduction

Substance use and suicidality are significant public health concerns with high comorbidity and overlapping risk factors. Substance use disorder rates are high in the U.S., with past year and lifetime prevalence rates of 13.9% and 29.1% for alcohol use disorder, respectively, and 3.9% and 9.9% for drug use disorders, respectively (Grant et al., 2015, 2016). Research has conceptualized suicidality as a continuum progressing from thoughts of death, active suicidal ideation (SI; e.g., thoughts of killing oneself), to intent and/or planning, attempts, and death (Silverman et al., 2007). Suicide attempts include any self-directed injury with any intent to die, regardless of the lethality of the injury (Nock et al., 2008). National data indicate that 10.7 million adults (4.3%) experienced serious thoughts of suicide and 1.4 million (0.6%) made a suicide attempt in the past year (Substance Abuse and Mental Health Services Administration [SAMHSA], 2019a). These rates are even higher among those with substance use disorders (Poorolajal et al., 2016). There is also a clear gap in integrated substance use and mental health treatment, as only 68% of all substance use treatment facilities offer any type of mental health services (SAMHSA, 2019b). In 2018, 15.3% of adults aged 18–25 and 7.0% of adults aged 26 and older demonstrated a need for substance use treatment, yet only 6.3% of adults 18–25 and 12.9% of adults 26 or older who needed treatment received any specialty substance use treatment. Only 7% of all adults with both a mental health disorder and substance use disorder received both mental health and substance use treatment.

Rates of SI and suicide attempts are substantially higher among adults seeking substance use treatment compared to the general population. Roughly a third (32.7%) of adults who use illicit drugs and were recruited from drug treatment and other healthcare clinics reported SI (Arribas-Ibar et al., 2017). Among adults who misused stimulants, opiates, or both at least weekly in the prior year, 32% (n = 129) reported a history of suicide attempt and 18% (n = 72) had multiple past attempts (Darke et al., 2010). In contrast, general population rates of lifetime history of SI and suicide attempt are 13.5% and 4.6%, respectively (Kessler et al., 1999). A meta-analytic review found that individuals with a history of suicide attempt reported greater severity of substance use problems and mental distress than those with no attempt history, particularly if they also reported an unintentional overdose (Bohnert et al., 2010).

A limitation in prior research examining suicidality and substance use problems has been the confounding of SI and suicide attempts. The majority of individuals with SI do not go on to engage in suicide attempts (Nock et al., 2010), yet SI independent of suicide attempts is associated with significant distress and lower mental health–related quality of life (Fairweather-Schmidt et al., 2016). The ideation-to-action framework suggests the factors and processes that contribute to SI and to the progression from SI to attempts and death are substantially different (Klonsky & May, 2015). Research in the past decade supports the idea that risk factors for SI differ among those with and without a history of suicide attempt (Nock et al., 2012, 2016). Substance use disorders, along with anxiety disorders, post-traumatic stress disorder, and sexual trauma history are all more common in individuals with attempts and differentiate those with SI and attempts, with moderate-sized effects (May & Klonsky, 2016).

The field has a poor understanding of and ability to predict suicidality. One contributing factor to this lack of knowledge is that base rates of suicide attempts are relatively low; thus, large samples are required to improve understanding of suicidality among adults in substance use treatment. Elucidating key differences between individuals without suicidality, with SI, and with suicide attempts entering substance use disorder treatment has the potential to improve clinical decision-making in that setting. In addition, despite research indicating that SI and suicide attempts are conceptually distinct phenomena with different correlates and risk factors, studies examining the link between suicidality and substance use problems have generally not distinguished between SI and attempts (e.g., Arribas-Ibar et al., 2017; Darke et al., 2010; Darke et al., 2015). Although suicidality and substance use problems likely share common risk factors (e.g., impulsivity; Anestis et al., 2014; Wit, 2009) and underlying features (e.g., affect dysregulation; Khantzian, 1997; Law et al., 2015), important differences may distinguish individuals who misuse substances and experience SI from those who misuse substances and attempt suicide.

Cross-sectional research on associations between suicidality and substance use problems offers insight into the clinical presentation of individuals with suicidality entering substance use disorder treatment. Existing literature highlights increased substance use frequency and trauma among those with SI. Correlates of SI among men and women include current opioid misuse (Ashrafioun et al., 2017). Studies specifically of women have also identified history of childhood physical and sexual abuse, rape after age 15, post-traumatic stress disorder, depressive disorder symptoms (Vaszari et al., 2011), alcohol use disorder, age of onset of trauma symptoms, and use of psychiatric medication (Resko et al., 2018) as correlates of SI. Among Spanish adults in treatment for substance use disorders, recent aggression perpetration, receiving psychological treatment, and earning income illegally (e.g., selling illicit substances) were associated with SI and/or suicide plans (Arribas-Ibar et al., 2017). In a separate sample of Spanish adults treated in substance use treatment clinics, individuals with either SI or attempts also reported more severe substance use problems and psychological symptoms and were more likely to receive inpatient (compared to outpatient) treatment, relative to those without suicidality (López-Goñi et al., 2018). Other cross-sectional research has exclusively focused on suicide attempts, and not SI, in individuals with heroin and stimulant misuse (Darke et al., 2010) and intravenous drug use (Havens et al., 2004).

Although this body of research suggests that individuals with suicidality present to treatment with greater psychological and substance use symptoms, we are unaware of any research that evaluates the impact of suicidality on response to substance use treatment. Individuals with comorbid depression and alcohol use disorders for example do not respond well to substance use treatment (Riper et al., 2014), perhaps because psychiatric comorbidities are not directly treated and gains from substance use treatment do not necessarily generalize to psychiatric symptoms. In contrast, SI does not worsen treatment outcomes for cognitive behavioral therapy for depression (von Brachel et al., 2019), perhaps because the cognitive skills and behavioral activation of cognitive behavioral therapy address both SI and depressed mood. It is likely then that individuals with SI and suicide attempts do not progress as well in substance use treatment as their counterparts without suicidality.

Longitudinal studies in substance use treatment samples have primarily focused on prospective relationships between substance use problems and suicide attempts, with mixed results (e.g., Bakken & Vaglum, 2007; Britton & Connor, 2010; Darke et al., 2007; Darke et al., 2015). Bakken and Vaglum (2007) found that substance use frequency at baseline and follow-up did not predict presence of a suicide attempt during the subsequent 6 years; however, Britton and Connor (2010) found that substance use frequency, specifically daily or more frequent cocaine use, predicted suicide attempt in the year following substance use treatment. Because SI and attempts have different risk factors (e.g., Nock et al., 2016), these longitudinal findings about attempts may not generalize to SI. Past SI severity tends to be the strongest predictor of future SI (e.g., Ribeiro et al., 2016), and past suicide attempts are the strongest predictor of suicide death (Bostwick et al., 2016). The disconnect between predictors of SI and predictors of attempts is particularly important for clinicians treating substance use disorders to intervene before individuals transition from SI to suicidal behavior. To our knowledge, no research has investigated predictors of SI over time in individuals receiving substance use treatment, which limits our ability to identify and intervene with individuals who may be at increased risk for recurrent SI during and after substance use treatment.

1.1. Current study

The purpose of these secondary analyses is to fill the gaps in knowledge in: 1) baseline differences in the full spectrum of suicidality (no SI, SI but no attempts, and suicide attempts) in a substance use disorder treatment sample; 2) predictors of SI (not only suicide attempts) over time; and 3) the impact of suicidality on substance use treatment outcomes and durability of improvement over time. The study has three aims. Aim 1 is to investigate demographic, clinical, and substance use differences among adults with varying levels of lifetime suicidality initiating substance use treatment: individuals with no lifetime suicidality, recent (i.e., past month) SI, past (i.e., prior to the past month) SI, recent suicide attempt, and past suicide attempt. Based on prior research, we hypothesize that female (SAMHSA, 2019a) and unemployed (Blakely et al., 2003) adults will be overrepresented in the recent SI group and both attempt groups. We expect that recent SI and recent attempt groups will be younger (SAMHSA, 2019a) and report greater severity of mental health, physical health, and substance use problems at baseline than those with past and no suicidality. Aim 2 is to identify predictors of SI immediately post-treatment and at 3 and 6 months post-treatment. We expect that individuals with any history of suicidality and those with the highest levels of baseline mental health and substance use severity will experience elevated risk of SI prospectively. Aim 3 is to understand the longitudinal relationship between baseline suicidality and substance use outcomes post-treatment and at 3 and 6 months post-treatment. We expect that those with recent SI and both attempt groups will show significantly shorter duration of abstinence during treatment and higher severity of substance use at follow-ups compared to the other groups.

2. Method

2.1. Sample

The sample included individuals from five randomized controlled trials comparing the efficacy of standard intensive outpatient treatment alone to standard intensive outpatient treatment plus contingency management (CM). All five trials included English-speaking adults aged 18 or older seeking treatment for substance use disorders at community-based intensive outpatient clinics (N = 1,147) recruited from two states in New England. Research staff collected data across the five trials between September 1998 and February 2009. In four trials, participants met criteria for a DSM-IV cocaine or opioid abuse or dependence diagnosis (American Psychiatric Association, 1994), while one trial (Petry, Weinstock, & Alessi, 2011) required an alcohol, cocaine, or opioid diagnosis. The majority of the sample met criteria for a DSM-IV diagnosis of cocaine abuse or dependence (88.5%, n = 1,015), half met criteria for DSM-IV alcohol abuse or dependence (53.7%, n = 616), and about a third met criteria for DSM-IV opiate abuse or dependence (31.4%, n = 360). Individuals with recent (i.e., past month) SI or suicide attempt were eligible if they were receiving intensive care for suicidality, while researchers excluded those with untreated and severe active SI and referred them to a higher level of care. We excluded a total of 16 individuals across studies for untreated and severe active SI; all were from a single trial (Petry, Weinstock, & Alessi, 2011). The five trials had similar baseline and follow-up assessment procedures and inclusion/exclusion criteria, justifying the combination of studies for optimal power to detect differences between suicidality groups. Further descriptions of inclusion and exclusion criteria are in reports of primary treatment outcomes (Petry et al., 2004; Petry et al., 2005; Petry et al., 2006; Petry, Weinstock, & Alessi, 2011; Petry et al., 2012).

2.2. Procedures

All participants provided written informed consent prior to initiating study procedures. Baseline assessments included self-report measures, a semi-structured clinical interview, and toxicology screens (i.e., urine drug screen, breathalyzer). Follow-up assessments were similar to those at baseline. We randomly assigned participants to usual care at the intensive outpatient program or usual care plus contingency management (CM). CM is a psychosocial treatment based on operant conditioning, in which rewards are provided contingent upon a specific, confirmed behavior (e.g., negative urine drug screens, treatment attendance). Objective verification (e.g., urine drug screen, breathalyzer) is a requirement for reinforcement of abstinence. Petry et al. (2004) compared the efficacy of two CM reinforcement magnitude options ($80 vs. $240 possible vouchers) versus usual care only. Petry et al. (2005) compared CM with a prize-based system, CM with a voucher system, and usual care only. Petry et al. (2006) compared three conditions: CM reinforcing abstinence, CM reinforcing completion of activities related to treatment goals, and usual care only. Petry, Weinstock, and Alessi (2011) compared CM for attending clinic group therapy sessions versus usual care only. Petry et al. (2012) tested two CM for abstinence conditions with different reinforcement magnitudes ($250 or $560 in possible prizes), CM for attendance with $250 in prizes possible, and usual care only, depending on baseline urine toxicology results for cocaine. The university’s Institutional Review Board approved all study procedures.

2.3. Measures

2.3.1. Suicidal ideation.

To capture a broad range of SI and account for individual differences in response to suicide assessments, we used items from two measures to assess SI. The Addiction Severity Index (ASI; McLellan et al., 1980) is a structured clinical interview with strong psychometric properties and support for use with patients in substance use disorder treatment (Leonhard et al., 2000; Rosen et al., 2000). At baseline, we measured recent (past month) and past (prior to the past month) SI with the ASI question “Have you had a significant period of time (that was not a direct result of drug/alcohol use) [when you] experienced serious thoughts of suicide?” and specified if SI occurred in the past month or prior to the past month. We also measured recent SI was via the Brief Symptom Inventory (BSI; Derogatis & Spencer, 1982) item, “In the past week, how much were you distressed by thoughts of ending your life?” rated on a scale of 1 “not at all” to 5 “extremely”. The ASI item may have captured more severe SI and planning, while the BSI item could include those who were distressed by SI “a little”; SI endorsement therefore included a range of SI severity. We coded baseline SI as past (lifetime, but not past month on the ASI or past week on the BSI) or recent (in the past month on the ASI or past week on the BSI). At follow-ups, we coded SI as present if participants reported SI in the past month on the ASI or past week on the BSI.

2.3.2. Suicide attempt.

An item from the ASI measured lifetime suicide attempt (i.e., prior to the past month) and recent suicide attempt (i.e., in the past month) with the question, “Have you had a significant period of time (that was not a direct result of drug/alcohol use) in which you have attempted suicide?”

2.3.3. Substance use, abstinence, and substance use severity.

We examined demographic characteristics (gender, age, race, employment, income, and marital status); and clinical characteristics: years of substance use, frequency of past month substance use, lifetime incarceration (no/yes), length of most recent voluntary abstinence from primary substance of difficulty and length of longest abstinence, and substance use treatment history. A checklist assessed alcohol, cocaine, and heroin abuse and dependence, based on items from the Structured Clinical Interview for DSM-IV (First et al., 1996). Urine toxicology and breathalyzer screens occurred at baseline and twice weekly during treatment (for 12 weeks postrandomization). Urine toxicology screens tested for cocaine, alcohol, and opioids in Petry (2004, 2005, 2006, 2012) and those substances plus methamphetamine (only positive in < 1% of samples) in Petry, Weinstock, & Alessi (2011). We calculated longest duration of abstinence as the number of consecutive weeks without positive or unexcused missed urine toxicology and alcohol breathalyzer screens during treatment (range 0–12 weeks). Samples that were positive for one or more of the targeted substances interrupted the string of abstinence. ASI composite scores for alcohol and drug use severity include items assessing substance use frequency, withdrawal symptoms, and need for treatment, and we assessed for them at baseline and follow-ups.

2.3.4. Mental health.

The BSI (Derogatis & Spencer, 1982) is a 53-item self-report measure of psychiatric symptoms answered on a 5-point Likert scale from “never” to “extremely”. The Positive Symptom Distress Index (PSDI) is the sum of BSI items rated with non-zero responses divided by the sum of the distress reported on those items. This index provides the average level of distress that a person experiences, regardless of how many symptoms the participant endorses, meaning that endorsement of items related to suicidality would not necessarily raise scores. Global indices of the BSI have demonstrated sound psychometrics in samples with substance use (Benishek et al., 1998). ASI items on mental health included the number of lifetime psychiatric outpatient treatments, number of lifetime psychiatric and medical inpatient hospitalizations, and history of sexual, verbal/emotional, and physical abuse (yes/no). We captured psychosocial functioning at baseline via ASI composite scores for medical, employment, legal, family/social, and gambling problems. The Quality of Life Inventory (Frisch et al., 1992) measures satisfaction with (rated −3 to 3) and importance of (rated 0 to 2) 17 domains (e.g., health, work, family); each domain score is the product of its satisfaction rating and importance rating, and the total score is the average weighted (by importance) satisfaction of domains rated as important (not zero). Adequate psychometric properties have been established in community (Frisch et al., 1992) and clinical samples (McAlinden & Oei, 2006).

2.4. Data analytic plan

We created five mutually exclusive and hierarchical groups based on endorsement of relevant ASI and BSI items: no suicidality (negative on all suicidality items), past SI, recent SI, past attempt, and recent attempt. The past SI group included those who reported lifetime, but not past month SI, and no attempts. The recent SI group included all participants who reported past month SI and had no attempts. The past suicide attempt group had a lifetime history of suicide attempts, but no attempts in the past month. The recent suicide attempt group included individuals with attempts in the past month. Those with past suicide attempt tended to also report past SI, and those with recent suicide attempt tended to also report recent SI, but our classifications prioritized the most severe indicator (recent over past; attempts over SI).

ANOVAs for continuous variables and chi square tests for nominal or dichotomous variables examined suicidality group differences on demographic characteristics and baseline mental health and substance use indicators. We compared all five suicidality groups to each other. ANOVAs also tested suicidality group differences in substance use disorder treatment outcome (i.e., longest duration of abstinence). For ANOVAs, Bonferroni-corrected post-hoc tests determined differences between groups with significant main effects. In chi square tests, z-tests compared column proportions with Bonferroni-adjusted p-values. Corrected t and F values are presented when tests of homogeneity were significant. We calculated effect sizes for all analyses (partial eta squared [η_p²] for ANOVAs, Phi [ϕ] Cramer’s ν for chi-square tests) and interpreted them based on recommendations from Cohen (1992).

For longitudinal analyses, individuals without suicidality served as the reference group. Two logistic regressions tested models predicting later SI. First, we tested baseline suicidality group membership as a predictor of SI during any of the follow-ups (post-treatment, 3 months post-treatment, 6 months post-treatment). Then, a second logistic regression tested whether suicidality group membership predicted SI during any of the follow-ups, when controlling for variables that significantly differentiated suicidality groups at baseline. For substance use outcomes, a set of repeated measure ANOVAs assessed alcohol and drug use severity (ASI scores) over time, from baseline, end of treatment, 3 months post-treatment, and 6 months post-treatment. According to Little’s chi-square statistic (Little, 1988), continuous data were not missing completely at random, χ² (37)= 191.99, p < .001. To manage the likelihood that data were missing systematically (e.g., individuals who returned to using substances during the follow-up period may have been more likely to miss follow-up assessments), we used maximum estimation procedures. With missing data accounted for in the model, effects were all similar in size; however, two effects were statistically significant with missing data accounted for, but not in models with available data only. Therefore, longitudinal analyses are conservatively presented with all available data only.

3. Results

Over half of the sample reported no history of suicidality (58.1%, n = 666). Of the total sample, 9.9% (n = 113) reported recent (i.e., past month) SI and 5.7% (n = 65) reported past SI without a history of attempts. In total, 2.3% of the sample (n = 26) reported a recent suicide attempt and 24.1% (n = 277) reported a history of suicide attempt. Most of the sample with lifetime suicide attempts also reported lifetime SI (100% of those with recent attempt; 96% of those with past attempt). To our knowledge, one participant died by suicide during the five clinical trials, but after the 9-month timeline included in the current study. We had assigned this participant to CM; this participant did not report any lifetime suicidality at baseline and completed all four data points included in this study.

3.1. Baseline differences in suicidality groups

3.1.1. Demographic differences between suicidality groups.

Suicidality groups differed on gender, race, employment status, and income with effect sizes ranging from small to medium (see Table 1). Females were significantly more likely to report a past suicide attempt than males and were over-represented in the no suicidality group. Black participants were significantly less likely to report a past suicide attempt and were over-represented in the no suicidality group compared to white participants and those who identified their race as “Other”. Individuals who identified their race as “Other” were significantly less likely to report no suicidality compared to white or Black participants. Those with past suicide attempt were significantly more likely to be unemployed compared to those with no suicidality. Participants with past SI and past suicide attempt reported about half of the income of those with no suicidality and recent SI. Suicidality groups did not significantly differ in age, marital status, or education.

Table 1.

Demographic differences in total sample and by suicide group (N = 1,147)

	Total Sample	No Ideation or Attempts	Past Suicidal Ideation	Recent Suicidal Ideation	Past Suicide Attempt	Recent Suicide Attempt
	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)
Total	100 (1147)	58.1 (666)	9.9 (113)	5.7 (65)	24.1 (277)	2.3 (26)	F/X2	Effect Size	Interpretation
Gender							42.40***	.19	Medium
Male	52.6 (603)	58.3 (388)a	56.6 (64)a	61.5 (40)a	36.8 (102)a	34.6 (9)a
Female	47.4 (544)	41.7 (278)b	43.4 (49)a	38.5 (25)a	63.2 (175)b	65.4 (17)a
Race							34.67***	.17	Medium
White	47.4 (544)	45.8 (305)a	46.0 (52)a	50.8 (33)a	51.6 (143)a	42.3 (11)a
Black	40.1 (460)	45.5 (303)b	36.3 (41)a	36.9 (24)a	30.0 (83)b	34.6 (9)a
Other	12.5 (143)	8.7 (58)c	17.7 (20)a	12.3 (8)a	18.4 (51)a	23.1 (6)a
Age	36.52 (9.16)	36.66 (9.30)	36.55 (8.96)	38.17 (9.20)	36.08 (8.90)	33.54 (8.41)	1.42	---	---
Employment Status							31.17***	.17	Medium
Employed/Student	65.3 (749)	70.1 (467)a	60.2 (68)a	73.8 (48)a	52.7 (146)a	76.9 (20)a
Not Employed	34.7 (398)	29.9 (199)b	39.8 (45)a	26.2 (17)a	47.3 (131)b	23.1 (6)a
Income1	$9,403 (15,358)	$11,127 (17,367)a	$6,444 (10,695)b	$11,767 (15,878)a,c	$5,922 (10,458)b,c	$9,306 (13,548)a,b,c	7.23***	.02	Small
Relationship Status							10.79	.07	Small
Single	56.6 (649)	59.5 (396)a	54.0 (61)a	50.8 (33)a	52.3 (145)a	53.8 (14)a
Widow/divorce/separated	26.9 (309)	23.4 (156)a	29.2 (33)a,b	32.3 (21)a,b	32.9 (91)b	30.8 (8)a,b
Married/cohabitating	16.5 (189)	17.1 (114)a	16.8 (19)a	16.9 (11)a	14.8 (41)a	15.4 (4)a
Education (years)1	11.90 (1.95)	11.96 (1.84)	11.81 (2.19)	12.39 (2.08)	11.71 (2.03)	11.81 (2.21)	1.87	---	---

Note.

^*p < .05

^**p < .01

^***p < .001

¹F test reported with Welch’s correction due to significant homogeneity test.

^a-cSuicidality groups with different letters have percentages or means that are significantly different at p<.05.

Superscripts indicating post-doc group differences are only presented for analyses with significant main effects. Employment includes lull and part time jobs, students, and active military; unemployment includes disability and restricted enviromnents. Effect sizes are partial eta squared for ANOVAs and Phi/Cramer’s ν for chi square tests. For dichotomous variables, percentages and sample sizes are presented; means and standard deviations are presented for continuous variables.

3.1.2. Physical and mental health differences in suicidality groups.

Overall, individuals with recent SI, past suicide attempt, and recent suicide attempt reported worse mental and physical health and more service utilization than those with no suicidality and past SI (see Table 2). Groups differed in prior psychiatric inpatient hospitalizations; quality of life; psychiatric distress; emotional, physical and sexual abuse; and psychosocial problems; effect sizes ranged from negligible to large. Those with past and recent suicide attempts reported significantly more psychiatric inpatient hospitalizations than those with past SI, recent SI, and no SI. Individuals with no SI reported the highest quality of life; individuals with no and past SI were not significantly different. Participants with past suicide attempt reported significantly lower quality of life compared to those with no SI. Those with recent SI or attempts reported similar quality of life, and both groups were significantly worse than the no SI group. Group differences in lifetime history of experiencing emotional/verbal, physical, and sexual abuse were as follows: individuals with past SI and past suicide attempts both reported higher likelihood of emotional/verbal abuse and physical abuse than those with no suicidality, recent SI, and recent attempts. Individuals with past SI, past suicide attempt, and recent suicide attempt reported similarly higher likelihood of sexual abuse compared to those with no suicidality and those with recent SI. On psychiatric symptom distress, adults with recent SI, past suicide attempts, and recent suicide attempts reported similarly high levels of distress from psychiatric symptoms compared to those with past SI and no suicidality, who reported similarly lower levels of psychiatric distress. On family/social problems, individuals with recent SI reported the highest levels of family/social problems, and those with past and recent suicide attempt reported higher levels of family/social problems than those with no suicidality. Group differences on medical and employment problems were significant but negligible in size. Remaining effects by group were not significant.

Table 2.

Physical and mental health for total sample and by suicide group at baseline (N = 1,147)

	Total Sample	No Ideation or Attempts	Past Suicidal Ideation	Recent Suicidal Ideation	Past Suicide Attempt	Recent Suicide Attempt
	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)
	100 (1147)	58.1 (666)	9.9 (113)	5.7 (65)	24.1 (277)	2.3 (26)	F/χ2	Effect Size	Interpretation
Health Treatments
Psychiatric inpatient hospitalizations1	1.11 (3.18)	.33 (1.30)a	.94 (1.75)a	.72 (1.32)a	2.87 (4.57)b	4.12 (10.45)b	24.15**	.08	Small
Psychiatric outpatient treatments1	1.51 (3.84)	.86 (1.92)a	1.84 (3.76 )a	1.09 (1.67)a	3.03 (6.49)b	1.62 (2.10)a,b	9.46***	.03	Small
Medical inpatient hospitalizations	3.08 (7.97)	2.69 (5.76)	2.65 (3.21)	2.65 (3.21)	4.24 (13.04)	3.62 (7.27)	2.02	.01	Negligible
Quality of life	1.27 (2.06)	1.54 (1.99)a	1.19 (2.18)a,b	.21 (1.92)c	1.00 (2.07)b	.31 (2.56)b,c	10.13***	.03	Small
Psychiatric distress1	1.82 (.71)	1.66 (.65)a	1.79 (.69)a	2.23 (.62)b	2.08 (.72)b	2.29 (.83 )b	27.88***	.09	Small-Med
Emotional abuse	52.4 (600)	41.8 (278)a	63.7 (72)a	55.4 (36)a	72.2 (200)a	53.8 (14)a	79.53***	.26	Large
No emotional abuse	47.6 (546)	58.2 (387)b	36.3 (41)b	44.6 (29)a	27.8 (77)b	46.2 (12)a
Physical abuse	36.0 (412)	25.8 (171)a	51.3 (58)a	27.7 (18)a	55.2 (153)a	46.2 (12)a	89.39***	.28	Large
No Physical abuse	64.0 (733)	74.2 (493)b	48.7 (55)b	72.3 (47)a	44.8 (124)b	53.8 (14)a
Sexual abuse	21.8 (250)	11.6 (77)a	33.6 (38)a	18.5 (12)a	40.8 (113)a	38.5 (10)a	113.25***	.31	Large
No sexual abuse	78.2 (896)	88.4 (588)b	66.4 (75)b	81.5 (53)a	59.2 (164)b	61.5 (16)b
ASI Problems
Medical1	.25 (.35)	.22 (.33)a	.25 (.36)a,b	.26 (.34)a,b	.33 (.39)b	.29 (.40)a,b	4.32**	.01	Negligible
Employment1	.74 (.28)	.74 (.27)a,b	.76 (.27)a	.64 (.29)b	.76 (.27)a	.66 (.37)a,b	2.73*	.01	Negligible
Legal	.14 (.22)	.14 (.21)	.15 (.22)	.12 (.22)	.14 (.22)	.23 (.26)	1.36	---	---
Family/social1	.18 (.22)	.14 (.20)a	.17 (.22)a,b	.28 (.24)c,d	.23 (.23)b,c	.25 (.23)a,b,c,d	10.98***	.04	Small
Gambling1	.02 (.06)	.02 (.04)	.02 (.04)	.03 (.07)	.02 (.07)	.02 (.05)	.76	---	---
Incarceration	55.6 (613)	56.9 (365)	57.1 (64)	38.7 (24)	55.9 (147)	54.2 (13)	7.70	.08	Small-Med

Note.

^*p<.05

^**p<.01

^***p<.001.

¹Welch’s correction used with significant homogeneity test.

^a-cGroups with different letters (e.g., a vs b) are significantly different at p<.05. Letters differentiating groups are only presented for significant main effects.

ASI = Addiction Severity Index. Med = Medium. Effect sizes were partial eta squared for ANOVAs and Phi/Cramer’s V for Chi square tests. Psychiatric distress = number of positive psychiatric symptoms. For dichotomous variables, percentages and sample sizes are presented; means and standard deviations are presented for continuous variables.

3.1.3. Substance use differences in suicidality groups.

Suicidality groups did not differ on years of use of any substance (i.e., alcohol to intoxication, cocaine, marijuana, heroin, or polysubstance use). However, groups differed in baseline past month days of use of alcohol to intoxication, cocaine, marijuana, and multiple substances, with null to small effect sizes. Individuals with recent SI reported the most days of alcohol use to feel an effect compared to all other groups. Those with a recent suicide attempt reported more days of marijuana use in the past month compared to those with no suicidality and past SI. Individuals with recent SI or attempt reported the most days of cocaine and polysubstance use compared to all other groups and higher than those with no suicidality and past SI. Individuals with recent SI reported using alcohol to feel an effect, cocaine, and polysubstance use 2.5 to 4.4 more days than other groups. Between-group differences in severity of alcohol and drug use were significant and small in size. Individuals with recent SI reported significantly worse severity of alcohol use compared to those with no suicidality, past SI, and past suicide attempts. Those with recent suicide attempts and recent SI reported similarly higher levels of drug use severity and both reported significantly more severe drug use than those with no suicidality and past SI (see Table 3).

Table 3.

Substance use for total sample and by suicide group at baseline (N = 1,147)

	Total Sample	No Ideation or Attempts	Past Suicidal Ideation	Recent Suicidal Ideation	Past Suicide Attempt	Recent Suicide Attempt
	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)	% (n)/M(SD)
	100 (1147)	58.1 (666)	9.9 (113)	5.7 (65)	24.1 (277)	2.2 (26)	F/χ2	Effect Size	Interpretation
Years of Use
Alcohol intoxication	10.18 (10.27)	10.07 (10.30)	11.67 (11.13)	11.02 (10.27)	9.92 (9.95)	6.96 (8.60)	1.23	---	---
Cocaine	9.73 (7.87)	9.58 (7.85)	10.43 (8.07)	9.30 (8.22)	9.85 (7.73)	10.38 (8.24)	.98	---	---
Cannabis1	8.52 (8.78)	8.10 (8.53)	8.92 (8.82)	8.67 (8.63)	9.24 (9.51)	9.35 (7.09)	.95	---	---
Heroin1	2.62 (5.98)	2.48 (5.69)	3.54 (6.94)	1.91 (5.94)	2.84 (6.35)	1.73 (4.44)	1.25	---	---
Poly-substance use	10.61 (8.69)	10.16 (8.43)	12.07 (10.12)	10.05 (9.35)	11.21 (8.49)	10.63 (8.28)	1.68	.01	Negligible
Past Month Days of Use
Alcohol intoxication1	2.82 (6.04)	2.98 (6.36)a	2.08 (5.23)a	5.63 (7.88)b	2.03 (4.78)a	3.62 (5.77)b	4.38*	.02	Small
Cocaine1	3.88 (6.70)	3.92 (6.62)ab	2.27 (5.64)a	6.65 (8.10)c	3.55 (6.59 )a,b	6.58 (7.81)b,c	5.28***	.02	Small
Cannabis1	1.86 (5.64)	1.61 (5.20)a	.78 (3.35)a	2.89 (6.78)a,b	2.37 (6.59)a,b	5.04 (8.35)b,c	4.32**	.01	Small
Heroin	.91 (3.95)	.91 (3.89)	.59 (3.51)	1.09 (3.56)	1.03 (4.38)	.81 (3.53)	.28	---	---
Poly-substance use1	3.21 (6.15)	3.29 (6.16)a,b	1.70 (4.92)a	5.78 (7.55)c	3.10 (6.20)a,b	5.77 (7.17)b,c	5.31***	.02	Small
Alcohol Use Severity1	.21 (.24)	.20 (.24)a	.21 (.22)a,b	.32 (.29)c	.21 (.24)a,b	.27 (.26)a,b,c	2.88*	.01	Small
Drug Use Severity	.16 (.10)	.15 (.10)a	.15 (.09)a	.20 (.10)b	.17 (.10)a,b	.21 (.09)b	6.65***	.02	Small
Substance Use Treatments
Alcohol detox	1.89 (9.13)	2.19 (11.37)	2.05 (5.02)	.49 (1.75)	1.61 (4.72)	.31 (.74)	.83	---	---
Drug detox1	2.44 (6.05)	2.16 (5.80)	3.34 (6.35)	1.95 (6.30)	2.94 (6.52)	1.46 (4.94)	1.68	.01	Small
Alcohol treatments	3.93 (11.04)	4.06 (13.12)	3.71 (6.53)	3.05 (5.34)	4.13 (8.08)	1.35 (2.43)	.52	---	---
Drug treatments	5.99 (9.02)	5.58 (9.26)	6.79 (8.04)	5.46 (8.13)	6.88 (8.87)	5.04 (10.24)	1.37	---	---
DSM-IV Substance Diagnosis
Alcohol Dependence	53.7 (616)	51.4 (342)	61.9 (70)	58.5 (38)	54.9 (152)	53.8 (14)	5.32	.07	Small
Cocaine Dependence2	85.8 (984)	83.6 (557)	90.3 (102)	83.1 (54)	88.4 (245)	100 (26)	10.70*	.10	Small - Med
Opiate Dependence	33.6 (336)	32.8 (189)	36.3 (37)	25.5 (14)	37.1 (91)	25.0 (5)	4.19	.07	Small
Prior abstinence (length in months)
Last abstinence	15.36 (27.87)	14.9 (25.80)	18.21 (34.36)	18.21 (34.37)	15.71 (25.74)	17.05 (41.73)	.16	---	---
Longest abstinence1	23.42 (32.66)	21.96 (31.44)	23.87 (30.98)	30.17 (43.67)	24.34 (31.76)	31.63 (46.30)	.87	---	---
Ever abstinent	84.7 (972)	83.2 (554)	88.5 (100)	81.5 (53)	88.1 (244)	80.8 (21)	5.71	.07	Small
Overdose on drugs	21.4 (257)	15.2 (101)a	23.0 (26)b	6.2 (4)a	37.5 (104)b	42.3 (11)b	74.11***	.25	Large
Positive Toxicology	20.9 (239)	21.7 (144)a	9.7 (11)b	24.6 (16)b	22.7 (63)a	19.2 (5)a	9.92*	.09	Small

Notes:

^*p<.05

^**p<.01

^***p<.001.

¹Welch’s correction used with significant homogeneity tests.

^a-cGroups with different letters (e.g., a vs b) have percentages or means that are significantly different at p<.05. Letters differentiating groups are only presented for significant main effects.

Med = Medium. There were no significant differences in proportions of SI groups who met criteria for cocaine dependence. Effect sizes were η_p² for ANOVAs and Φ/ ν for Chi square tests. For dichotomous variables, percentages and sample sizes are presented; means and standard deviations are presented for continuous variables.

The main effect of suicidality group on cocaine dependence diagnosis was small to medium in size, but pairwise group differences were not significant. Large differences in unintentional overdose showed that those with past SI, past suicide attempt, and recent suicide attempt had greater likelihood of overdose history than those with recent SI and no suicidality. In addition, those with past SI reported a lower likelihood of testing positive for any substance at baseline and those with recent SI displayed a higher likelihood of testing positive compared to those with no suicidality, past suicide attempt, and recent suicide attempt. Remaining effects by group were not significant.

3.2. Prediction of SI over time

Table 4 presents the results of a logistic regression predicting SI at any of the three follow-up assessments (i.e., post-treatment, 3 months post-treatment, and/or 6 months post-treatment). Almost 1 in 5, or 18.4% (n = 195) reported SI at any follow-up, specifically 7.8% (n = 89) at post-treatment, 7.5%, (n = 86) at 3 months post-treatment, and 6.5% (n = 75) at 6 months post-treatment. Individuals with past SI, recent SI, past attempt, and recent attempt were more likely to report SI at any follow-up compared to those who did not report suicidality. As expected, those with recent suicide attempts displayed the largest effect size for later SI, such that the odds of an individual with a recent suicide attempt reporting later SI was 8.0 times higher than the odds for participants with no suicidality.

Table 4.

Logistic regression predicting suicidal ideation over time (n = 1,061)

	OR	OR 95% CI		SE	p
Model 1 – Suicidality group predicting suicidal ideation during the follow-up
Suicidality (No suicidality = reference)
Past Suicidal Ideation	2.45	1.39	4.30	.29	.002
Recent Suicidal Ideation	5.49	2.94	10.25	.32	< .001
Past Suicide Attempt	5.61	3.80	8.23	.20	<.001
Recent Suicide Attempt	8.00	3.35	19.14	2.08	<.001
Model 2 – Suicidality group and baseline characteristics predicting suicidal ideation during follow-up
Suicidality (No suicidality = reference)
Past Suicidal Ideation	2.31	1.29	4.14	.30	.005
Recent Suicidal Ideation	3.63	1.88	7.03	.34	<.001
Past Suicide Attempt	4.63	3.02	7.10	.22	<.001
Recent Suicide Attempt	5.85	2.32	14.76	.47	<.001
Female Gender	1.06	.72	1.56	.20	.78
Sexual Abuse	1.12	.73	1.72	.22	.60
Incarceration	.89	.62	1.28	.19	.89
Past month drinking to intoxication frequency	1.03	1.0	1.06	.02	.08
Past month poly-substance use frequency	1.0	.96	1.03	.02	.76
Medical Problems	1.48	.93	2.37	.24	.10
Family/social Problems	1.46	.66	3.24	.41	.35
Psychiatric symptom distress	1.74	1.34	2.25	.13	<.001
Contingency Management Treatment Group	.70	.49	1.0	.18	.048

Note. Results presented with all available data (n = 989 at post-treatment, n = 927 at 3 month, n = 921 at 6 month-post treatment; total N = 1,061 since 170 (14.8%) of participants missed only one assessment point).

Next, a model tested if baseline suicidality group, demographic, health, substance use indicators, and treatment group predicted future SI. After controlling for demographic, health, and substance use indicators, individuals with past SI, recent SI, past suicide attempt, and recent suicide attempt were significantly more likely to report SI in the future compared to those without suicidality. In the context of the significant effects of suicidality group, baseline psychiatric symptom distress significantly increased the likelihood of future SI and being assigned to contingency management (compared to usual care alone) significantly decreased the likelihood of future SI.

3.3. Suicidality group differences in substance use outcomes

3.3.1. Longest duration of abstinence.

Suicidality groups did not significantly differ in longest duration abstinence during treatment, F(4, 1142) = .92, p = ns, η_p² < .01.

3.3.2. Alcohol use severity.

Repeated measures ANOVA showed a significant medium-sized main effect of time (F = 48.23, p < .001, η_p² = .06) and small suicidality group interaction (F = 2.06, p < .05, η_p² = .01) on severity of alcohol use overtime (see Figure 1); the main effect of suicidality group was not significant (F = 1.65, p = ns, η_p² = .01) with all available data, but was significant yet still small in size in a parallel model that estimated missing data. Overall, alcohol use severity decreased over time. Those with past suicide attempts reported higher alcohol use severity than those without suicidality at post-treatment (B = .03, p = .05), 3 months post-treatment (B = .04, p = .01), and 6 months post-treatment (B = 06, p < .05). The time by suicidality group interaction indicates that those with past suicide attempts displayed less improvement in alcohol use severity over time compared to other groups. At 6 months post-treatment, those who reported recent SI at baseline reported higher alcohol use severity compared to those without suicidality (B = .03, p < .05), indicating that those with recent SI were less likely to maintain improvement in alcohol use at 6 months post-treatment compared to those with no suicidality.

Figure 1.

Repeated measures ANOVA examining differences in suicidality groups’ alcohol use severity over time

Note. Analyses with missing data estimated with maximum estimation procedures showed similar effect sizes for time, suicidality group, and their interaction. Tx = treatment.

3.3.3. Drug use severity.

Repeated measures ANOVA showed a significant medium-sized main effect of time (F = 89.84, p < .001, η_p² = .11) and small main effect of suicidality group (F = 5.96, p < .01, η_p² = .03) on severity of drug use over time; the time by suicidality group interaction was not significant (F = 1.89, p = .ns, η_p² = .01; see Figure 2) with all available data, but was significant and similar in size with missing data modeled. Those with a past suicide attempt did not display as large of an improvement in drug use severity over time and remained significantly higher than those with no suicidality at post-treatment (B = .03, p = .001), 3 months post-treatment (B = .03, p = .001), and 6 months post-treatment (B = .03, p < .001). Although those with recent SI and recent attempts reported higher baseline drug use severity compared to those with no suicidality and past SI, adults with recent SI and recent attempts both demonstrated similarly low severity of drug use by post-treatment and throughout follow-ups.

Figure 2.

Repeated measures ANOVA examining changes in suicidality groups’ drug use severity over time

Note. Analyses with missing data estimated with maximum estimation procedures showed similar effect sizes for time, suicidality group, and their interaction. Tx = treatment.

4. Discussion

This study adds to the literature that documents high rates of SI and attempts among substance use treatment patients. Despite excluding participants who were at immediate suicide risk and recruiting in outpatient (rather than inpatient) clinics, a quarter of the sample reported a lifetime history of suicide attempt and roughly 1 in 10 reported past month SI. Rates of suicidality in this sample were higher than the general population, particularly in comparison to past year SI rates of 4.3% (SAMHSA, 2019a) and similar to research with adults reporting weekly substance use (Darke et al., 2010). High rates of suicidality in substance use treatment samples indicate a need for further research to investigate temporal relationships among SI, suicide attempts, and substance use disorders to understand if suicidality interventions could reduce the likelihood of developing substance use disorders or vice versa.

Results of this study improve understanding of the correlates of varying levels of suicidality among adults in outpatient substance use disorder treatment. Consistent with prior literature documenting higher rates of suicide attempts in females compared to males (Crosby et al., 2011; Nock et al., 2008), this study found that women were more likely to report a past suicide attempt than men. However, in contrast to higher rates of SI in women in the general population (Nock et al., 2008) but in line with substance use treatment samples from Spain (López-Goñi et al., 2018), this study found that men and women receiving outpatient substance use disorder treatment did not significantly differ in regard to past or recent SI at intake. The discrepancy in gender differences may be because this study examined mutually exclusive suicidality groups, whereas prior research conflated SI and suicide attempts. In terms of racial differences, Black/African American adults were less likely to report suicide attempts than white adults, consistent with national prevalence rates of suicide attempt by race (National Center for Injury and Prevention, 2015). In line with our hypotheses, individuals with past suicide attempts were more likely to be unemployed, relative to those with no suicidality. Unemployment shows particularly strong relationships with suicide death (Breuer, 2015), particularly for adults aged 55–64 (Lin & Chen, 2018), potentially due to decreased connection to important people, values, and life purpose (see Klonsky & May, 2015).

We found key differences between suicidality groups at baseline; in general, individuals with recent SI and any suicide attempt history reported more severe mental health, physical health, and substance use problems. Individuals with recent SI reported the highest levels of past month alcohol, cocaine, and polysubstance use, and adults with recent SI and recent attempts reported more days of marijuana use in the past month than those with no suicidality and past SI. Given the particularly strong associations between suicidality (Caetano et al., 2013; Norström & Rossow, 2016) and chronic marijuana use (Borges et al., 2016), more use of these substances among those with recent suicidality is unsurprising. Suicidality group differences in DSM-IV cocaine dependence diagnosis were also significant, though pairwise differences were not. This overall main effect may be driven by adults with recent suicide attempt, 100% of whom met DSM-IV criteria for cocaine dependence. This finding is also in line with research documenting relationships between daily or more frequent use of cocaine use and suicide attempt in the following year (Britton & Connor, 2010) but highlights the overlap in cocaine use and recent SI. Overall, patterns in substance use at intake indicate that adults in outpatient substance use treatment with past month SI and lifetime history of suicide attempt may engage in more substance use than patients with no suicidality and SI prior to the past month.

No substance use variables differentiated individuals with past and recent SI from those with past and recent suicide attempts. Individuals with past SI tended to be more similar to those with no suicidality in terms of substance use, past month use, and substance use treatments. Those with past SI were also significantly less likely than all other groups to have a positive toxicology screen for cocaine, alcohol, or opioids at baseline. Those with past SI reported lower severity of family problems, employment problems, and psychiatric distress than those with recent SI and both attempt groups; this is in line with prior research that showed those who recover from SI tend to have lower severity of depressive symptoms, anxiety, and hopelessness (Zhang et al., 2015) and higher social support (Teismann et al., 2016) than those with continued SI. The only variable that differentiated individuals with any SI from those with any suicide attempts was number of prior inpatient psychiatric hospitalizations, which were higher in both attempt groups than those with no suicidality and past SI. Overall, those with recent SI tended to present more similarly to those with suicide attempt histories than to those with past SI, possibly because those with more persistent and recent SI have higher psychopathology and substance use, and according to the ideation-to-action framework (Klonsky & May, 2015), are more likely to transition to suicide attempts.

Past month substance use frequency was higher at baseline for those with recent SI and lifetime suicide attempts. Similar to previous work examining the impact of psychiatric symptom severity on response to usual care and CM (Weinstock et al., 2007), suicidality group membership did not impact longest period of abstinence during treatment based on urine toxicology results, which is a clinically relevant outcome because of its strong association with long-term substance use outcomes (e.g., Petry, Ford, & Barry, 2011). Adults with and without suicidality achieved similar abstinence during treatment, despite higher rates of positive urine toxicology and more severe self-reported substance use frequency and severity at baseline. Substance use treatment, particularly CM, may bolster treatment outcomes to the point that these baseline severity indicators do not significantly impact patients’ ability to achieve and maintain abstinence during treatment, but gains may not be maintained post-treatment. Although the overall sample showed significant decreases in alcohol and drug use severity from baseline to 6 months post-treatment, adults with past suicide attempts continued to report higher severity difficulties with alcohol use problems and drug use problems post-treatment than all other groups. Other studies have also found that individuals with a history of suicide attempt have difficulty maintaining treatment gains. For example, individuals in substance use treatment with a history of suicide attempts (Wojnar et al., 2006), particularly impulsive suicide attempts (Wojnar et al., 2008), had worse outcomes with respect to alcohol relapse and time to relapse.

Inconsistencies between urine/breath toxicology–based results and remaining outcomes have several possible explanations. One explanation is differences in measurement; results of urine/breath toxicology tests are objective but measured only the presence or absence of substance use, while self-reported substance use severity on the ASI reflects the perception and impact of substance use including frequency, withdrawal symptoms, and need for treatment. Time frames also differed. We calculated longest consecutive duration of abstinence with twice weekly urine drug tests and alcohol breathalyzers during 12 weeks of treatment that assesses substance use in approximately the past 48–72 hours, while we captured self-reported measures of past month substance use severity at end of treatment, 3 months, and 6 months post-treatment. Since severity also includes money spent on alcohol and drugs, and individuals with past attempts were more likely to be unemployed than those with no suicidality, it is possible that some effects were driven by those with past suicide attempts having difficulty securing stable employment or other financial problems. As disruptions in psychosocial functioning are robust risk factors for suicide (Batty et al., 2018), individuals with recent SI and past attempts may also be more susceptible to negative consequences of employment, social, financial, or physical problems related to alcohol and drug use.

Individuals with past suicide attempt showed a decrease in drug use problem severity from baseline to post-treatment, but severity remained significantly higher compared to those with no suicidality immediately post-treatment and throughout the 6-month follow-up period. Psychiatric history might impede the rate of progress on drug/alcohol problem severity among adults with a history of suicide attempt. Individuals with suicide attempts tend to have other psychiatric comorbidities, including anxiety, traumatic stress, and conduct disorders (Nock et al., 2010). Those with suicide attempt history also reported more prior inpatient hospitalizations for psychiatric problems than those without suicidality. Overall, worse long-term outcomes among individuals with suicidality fits with prior literature that suggests that individuals with more severe psychiatric comorbidities do not respond well to substance use treatment (Riper et al., 2014), perhaps because these comorbidities are not directly treated.

Supporting prior research that shows that past suicidality is the most consistent and strongest predictor of later suicidality (Ribeiro et al., 2016), adults in this sample who reported any history of suicidality were more likely to report SI after substance use treatment ended compared to adults with no suicidality history. Those with recent attempts had the highest odds of subsequent SI and those with past SI had the lowest odds of subsequent SI compared to other groups who reported suicidality, supporting a continuum of suicidality from SI to attempts. When accounting for baseline characteristics that distinguished suicidality groups, psychiatric symptom distress was related to increased likelihood of SI, and random assignment to CM decreased likelihood of SI compared to usual care. The association between CM and decreased likelihood of SI in the 6 months post-treatment, even after controlling for baseline suicidality, is a novel finding, but it is supported by past research showing that CM decreases psychiatric symptom severity through reductions in substance use (Petry et al., 2013). Meta-analytic data show that psychotherapies that are not specifically focused on SI do not tend to reduce SI (Cuijpers et al., 2013; Tarrier et al., 2008). Thus, it is particularly compelling that CM, which does not focus on suicidality, reduced the likelihood of SI even when controlling for baseline suicidality and despite similar duration of abstinence during treatment among those with and without suicidality.

The high rates of suicidality and impact on substance use treatment response may be due to overlapping risk factors for substance use disorders and suicidality, such as trauma (Mandavia et al., 2016; Miller et al., 2013). Multiple theoretical models, including the tension reduction hypothesis (Greeley & Oei, 1999) and stress-response dampening hypothesis in relation to alcohol use disorders (Sher, 1987), and the self-medication hypothesis for cocaine and opiate use disorders (Khantzian, 1985; 1997) suggest a causal relationship between pre-existing psychiatric disorders and alcohol and other substance use disorders. Substance use disorders likely increase risk for suicide attempts by increasing the likelihood of predisposing (e.g., worsening social support; impulsivity; aggression) and precipitating factors (e.g., depressed mood; negative life events; employment difficulties; Lamis & Malone, 2012).

4.1. Limitations and future directions

Results should be interpreted in light of several limitations. First, this study considered SI and attempts as categorical grouping variables. Further research on severity of SI, including duration and frequency, and severity of intent and lethality of suicide attempts is warranted using comprehensive measures of suicidality, such as the Columbia Suicide Severity Rating Scale (Posner et al., 2011) and Beck Suicide Scale (Beck et al., 1979). More work using well-validated SI measures should be done to understand if standard substance use treatment and/or CM is associated with clinically meaningful reductions in SI severity. This sample comprised individuals in treatment for substance use disorder and may not generalize to the majority of individuals with substance use disorders, i.e., those who are not in treatment (SAHMSA, 2019). In addition, although we presented results here related to alcohol use, participants in the studies were not required to meet criteria for DSM-IV alcohol abuse or dependence to participate; thus, results may not generalize to samples of patients specifically seeking treatment for alcohol use disorders. However, data were from real-world patients treated at community-based substance use treatments centers and, thus, these patients are likely to be representative of the types of patients typically seen in such settings.

We collected data within clinical trials for substance use treatment specifically and so assessments focused on substance use outcomes; we measured certain behaviors relevant to suicidality research, including past overdose and experiences with trauma, with single items on the ASI. Although the wording of the item explicitly distinguished unintentional overdoses from suicide attempts, more comprehensive measures of overdose could improve understanding of unintentional compared to intentional overdose, given the strong relationships between these two phenomena (Bohnert et al., 2010). Researchers collected the data for the five trials between 10 and 20 years ago; given increases in suicidality particularly among young adults (Schulenberg et al., 2019), research should replicate these findings with more recent samples to understand possible cohort effects. Recent active and untreated SI was an exclusion criterion in all trials; however, we excluded only 16 adults for suicidality and they were all from only one study (Petry, Weinstock, & Alessi, 2011). As such, we may have less power to detect differences between the recent SI and other groups, and the recent SI group may reflect those who were willing or able to secure treatment for suicidality. Since adults with recent SI did not maintain improvements in alcohol use problems 6 months after treatment ended, more longitudinal work is needed to understand if adults with recent SI may return to baseline levels of alcohol use severity and how to address this possible lack of durability of alcohol use response (potentially with booster sessions).

4.2. Conclusion

This study further elucidates the differences between SI and attempts among adults seeking treatment for substance use problems. We cannot overstate the importance of suicide risk assessment and mental health services within substance use clinics. Findings from other studies suggest that evidence-based safety planning procedures (see Stanley et al., 2018; Stanley & Brown, 2012) may be indicated in substance use clinics as an adjunct to standard substance use treatment and for those with any suicidality, but especially for those with recent suicide attempts. More research should address poor long-term responses among individuals presenting to substance use treatment with suicidality. Additional interventions may be needed to reduce long-term severe substance use for adults with past suicide attempts and to maintain reductions in severity of alcohol use among adults with recent SI after initial active treatment ends. For example, offering dialectical behavior therapy (DBT) after initial response to substance use treatment to those with the greatest risk of continued SI—including on-call services in early stages of treatment, identification of reasons for living, and cultivating values and behaviors to foster fulfillment—may be helpful for treating adults with substance use disorders long-term. DBT has also shown small reductions in substance use among those with co-occurring borderline personality disorder and substance use disorders (Lee et al., 2015), yet research should explore if DBT is appropriate for general populations of individuals in substance use treatment. Treatment interventions that includes additional skills to address suicidality directly, are longer in duration, or include booster sessions after active treatment ends may be helpful for those with recent SI and lifetime suicide attempts.

Highlights.

• Suicidality rates are high among adults seeking substance use disorder treatment

• Suicidality groups differed in demographics, substance use, and health indicators

• Suicidality groups did not differ in drug/alcohol toxicology outcomes throughout treatment

• Adults with recent suicidal ideation did not maintain reductions in severity of alcohol problems at 6-months post-treatment

• Adults with past suicide attempts showed less improvement in drug use severity from post-treatment to 6 months post-treatment