Psychiatric Comorbidities and Risks of Death, Nonfatal Overdose, and Self-Harm in Opioid Agonist Therapy: A Retrospective Cohort Study

Abstract

Introduction

Psychiatric disorders are common among patients with opioid use disorder (OUD). Opioid agonist therapy (OAT) is an effective treatment for OUD, but outcomes for patients with psychiatric comorbidities remain understudied. This study aimed to examine the connections between psychiatric comorbidities and outcomes in individuals undergoing OAT.

Methods

We retrospectively collected data from the Wellbeing Services County of Southwest Finland clinical registry. The cohort included 1,235 patients receiving OAT from January 1, 2004, to April 15, 2021. Patients attending OAT, psychiatric diagnoses, overdoses, and self-harm were identified from the registry data using ICD-10 codes. In cases of death, the register also included the date of death. We analyzed associations between psychiatric comorbidities and outcomes using Cox regression for death and logistic regression for nonfatal overdose and self-harm.

Results

We identified the prevalence of the following psychiatric comorbidities: non-substance use disorder (non-SUD) psychiatric disorders (39.0%), psychosis (17.0%), mood disorders (29.1%), anxiety disorders (26.2%), personality disorders (20.7%), and other SUDs not related to opioids (other SUDs) (61.4%). After adjusting for all covariates, we found an association between alcohol use disorder (AUD) (hazard ratio = 1.54, 95% confidence interval = 1.06–2.22) and a higher risk of death during OAT. Additionally, non-SUD psychiatric disorders (overdose: odds ratio [OR] = 1.64, 1.14–2.36; self-harm: OR = 1.50, 1.03–2.18), psychosis (overdose: OR = 1.89, 1.29–2.79; self-harm: OR = 1.87, 1.26–2.78), and other SUDs (overdose: OR = 2.61, 1.81–3.83; self-harm: OR = 2.56, 1.74–3.82) were associated with higher odds of both overdose and self-harm.

Conclusion

The study emphasizes the need for screening and secondary preventive measures for psychosis and other SUDs (particularly AUD) in OAT settings.

Introduction

Opioid use disorder (OUD) is one of the most prevalent drug use disorders worldwide and contributes to a substantial global burden of disease, according to a study using the Global Burden of Disease Study 2023 data [1]. A national cohort study conducted in Sweden has identified it as the strongest risk factor among substance use disorders (SUDs) for death by suicide during a follow-up period from 2003 to 2016 [2].

Previous systematic reviews have reported psychiatric disorders such as psychotic, mood, anxiety, and personality disorders (PDs), as well as non-opioid SUDs, to be many-fold more prevalent in persons with OUD than in the general population [3, 4]. Multiple retrospective studies have shown these comorbidities to be associated with adverse outcomes in OUD, such as death [5], drug overdose [6], and self-harm [7].

Opioid agonist therapy (OAT) is an established and effective treatment for OUD, offering the benefit of reducing the risk of death [8], overdose [9], and self-harm [10]. Nevertheless, the prognosis for patients in OAT with psychiatric comorbidities remains uncertain due to the inherent complexity of this population [11]. A previous systematic review has reported that patients with comorbid psychotic disorders had poorer treatment retention during OAT [12]. To our knowledge, there is a paucity of research examining the effect of psychiatric comorbidities including non-opioid SUDs on treatment outcomes, including self-harm [13–15], overdose, and death [16–20] during OAT. Among these studies, only a few were longitudinal, with follow-up periods ranging from short term (1 year) to long term (15 years), and used administrative or clinical samples of individuals receiving OAT or methadone maintenance treatment (MMT) [16–20]. Most of them have exclusively focused on methadone therapy [14, 16, 17, 19, 20], and specific psychiatric subtypes, particularly psychosis, are understudied. The chronic relapsing nature of both OUD and psychiatric disorders [21, 22], and the widespread use of buprenorphine treatment, especially in Finland [23, 24], calls for longitudinal studies with sufficient durations of follow-up to further knowledge on the impact of specific psychiatric comorbidities, including psychosis, on outcomes during OAT.

In this study, we aimed to examine the trajectories of patients in OAT presenting with different psychiatric and substance use comorbidities. We used clinical register data on individuals undergoing OAT in Southwest Finland, with a median follow-up of 9 years. Several comorbidities, such as non-SUD psychiatric disorders, psychosis, mood disorders, anxiety disorders, PDs, and other SUDs not related to opioids, were examined. The outcomes included death, self-harm, and nonfatal overdose.

Methods

Data Sources

The data analyzed in this study were collected retrospectively from the clinical registry of the Wellbeing Services County of Southwest Finland (previously the Hospital District of Southwest Finland) via the Auria Clinical Informatics data pool. The register contains health-related data on all patients treated in specialized healthcare services in Southwest Finland, provided by Turku University Hospital (TYKS), including the secondary care central hospitals in Salo, Loimaa, and Uusikaupunki. We extracted the data in April 2021, covering the period from January 1, 2004, to April 15, 2021. We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline [25] (online suppl. Table S1; for all online suppl. material, see https://doi.org/10.1159/ear/acrag001).

Study Sample

The sample used in this study included patients who received OAT in Southwest Finland during the reference period from 2004 to 2021. We identified patients using the International Statistical Classification of Diseases, 10th edition (ICD-10) diagnosis code F11.22 for opioid dependence. This code applies to patients with opioid dependence enrolled in OAT. The sample included all patients who had received this diagnosis during the reference period.

The target population of this study is people who have received maintenance treatment for OUD in Finland. During the reference period, the Addiction Psychiatry Unit at TYKS was the main provider of clinical assessment for OAT in the Southwest region of Finland. Hence, the sample is a comprehensive account of all patients in the region who underwent assessment to receive OAT during that period.

Psychiatric Diagnoses

Psychiatric diagnoses, as recorded in the register data according to ICD-10 criteria, were collected for each participant. These contain psychiatric diagnoses set at TYKS between January 1, 2004, and April 15, 2021. Psychiatric diagnoses set at the same date as the first OAT diagnosis or after were included in the data. The analysis excluded diagnoses set before the first OAT diagnosis. We mainly explored the associations between the following psychiatric comorbidities and outcomes: (1) non-SUD psychiatric disorders; (2) psychosis; (3) mood disorders; (4) anxiety disorders; (5) PDs; and (6) other SUDs.

Comprehensive diagnostic assessment utilizing the Structured Clinical Interview for DSM-IV (SCID) or Mini-International Neuropsychiatric Interview (MINI) has been an established practice in the Addiction Psychiatric Unit. These clinical interviews have been conducted face-to-face by trained clinicians, and the register data contain the outputs of these interviews. Online supplementary Table S2 shows the detailed psychiatric diagnostic codes.

Outcomes: Death, Self-Harm, and Nonfatal Overdose

Outcome data were available until April 15, 2021. Death encompassed all mortal outcomes attributable to any cause within the designated reference period. Self-harm includes the commonly used term “suicide attempt” [26]. Regarding ICD-10 codes, this included the following diagnostic codes: X60–X84, Z91.5, Y87.0, or Z72.8. Nonfatal overdose generally includes both intentional and unintentional cases, and it is referred to as overdose in the following text. We classified overdose as ICD-10 codes: T36–T50, X40–X49, Y19, X60–X69.

Covariates

We gathered the following demographic and clinical information based on previous research and data availability: sex (male, female); age at first OAT enrollment (<30 years, ≥30 years); year of first OAT enrollment (2004–2006, 2007–2010, 2011–2016, or 2017–2021), which can be used to adjust for the cohort entry year; place of residence (Turku, other than Turku), which may serve as a proxy for the OAT treatment setting, as residents of Turku might be more likely to receive OAT outside the university hospital, potentially influencing patient characteristics and treatment outcomes [27, 28]; OAT medication records in the medicine registers: (1) a record for methadone as OAT medication only, (2) buprenorphine as OAT medication only, (3) a record that both methadone and buprenorphine had been used at some point during OAT; Charlson Comorbidity Index (CCI; 0, 1, ≥2), and history of self-harm or overdose (yes, no). We established categorizations for age at first enrollment in OAT and the year of first OAT based on their distributions. Additionally, the categorization for the CCI variable was derived from prior research [29]. The three category variables were intended solely for descriptive statistical purposes. In subsequent analyses, we used age at first OAT enrollment, the year of first OAT enrollment, and CCI as continuous variables.

Statistical Analysis

We summarized and contrasted clinical characteristics across groups formed by dichotomizing the six psychiatric conditions or outcomes in OAT. The Wilcoxon rank-sum test was used for continuous variables, while Pearson’s chi-squared test or Fisher’s exact test was applied for categorical variables. Factors associated with at least one exposure and at least one outcome (p < 0.1), based on single-predictor analyses, were included in multi-predictor regression analyses. Means and standard deviations characterized normally distributed continuous variables, whereas medians and interquartile ranges (IQRs) were used for non-normally distributed continuous variables. Categorical variables are shown with their respective frequencies and percentages.

We calculated the crude mortality rate per 1000 person-years (PYs) and its 95% confidence interval (CI). We measured follow-up duration, expressed in PYs, from the date of initial OAT entry to either the date of death or, for censored subjects, to April 15, 2021. We computed the standardized mortality ratio as the ratio of observed to expected deaths. We derived expected deaths from PYs in our cohort and from Finland mortality rates, stratified by sex, age, and calendar year, obtained from the Human Mortality Database. We calculated the years of potential life lost (YPLL) for each fatality by subtracting the age at the time of death from 75 years for individuals younger than 75 years. This criterion is widely employed in previous studies [30].

This study examined the association between six psychiatric comorbidities and risk of death in OAT through separate Cox regression analyses. We estimated hazard ratios (HRs) with 95% CIs to present the results. We employed Schoenfeld’s scaled residuals to verify the proportional hazards assumption; if the models failed to meet it, we used Prentice weights from the coxphw package in R to estimate average HRs [31]. Cox regression analyses on self-harm and overdose were not applicable because the diagnosis data were not obtained from nationwide registers, and the censoring point remains indeterminate. We conducted separate logistic regression analyses to assess the association between psychiatric comorbidities and subsequent self-harm and overdose in OAT, yielding odds ratios (ORs) with 95% CIs.

Each regression analysis employed a hierarchical modeling approach: the crude model used no covariates; model 1 adjusted for demographic variables (sex, age at first enrollment in OAT, year of first OAT, and place of residence); model 2 additionally included clinical variables (OAT medication, CCI, history of self-harm or overdose); and model 3 was further mutually adjusted for all six psychiatric comorbidities. We used the variance inflation factor to evaluate multicollinearity, with values below 5 indicating no multicollinearity [32].

We conducted post hoc analyses of significant findings. First, we examined the influence of the subtype of other SUDs, specifically alcohol use disorder (AUD) and sedative use disorder, on the risk of death among individuals undergoing OAT. This investigation was prompted by the observation that alcohol and benzodiazepines are the most prevalent substances involved in opioid-related fatalities [33, 34]. Second, we analyzed associations between combinations of psychosis and other SUDs (both psychosis and other SUDs, other SUDs only, psychosis only, and neither condition) and outcomes (self-harm and overdose) using Fisher’s exact tests and logistic regression.

Sensitivity analyses were performed by restricting the sample to individuals with no prior recorded diagnoses of self-harm for the self-harm outcome and no previous recorded overdose diagnoses for the overdose outcome. This approach reduces confounding bias related to the substantial predictive value of a history of overdose or self-harm for future events [35, 36], while also acknowledging that such diagnoses may recur and do not necessarily represent new incident cases. Additionally, supplementary sensitivity analyses for self-harm and overdose outcomes were performed by restricting the sample to individuals with explicitly documented OAT medication data. Furthermore, the robustness of the findings regarding the death outcome was assessed by replacing the Cox regression model with a logistic regression model and by conducting the analysis among individuals with at least 1 year of follow-up. Moreover, for all three outcomes, the stability of the results was further verified in sensitivity analyses using covariates chosen via a stepwise procedure based on the Akaike information criterion (AIC). Exploratory analyses on the death outcome were conducted concerning OAT medication and follow-up duration, using simplified models due to the limited number of events. Specifically, these analyses focused on individuals with available OAT medication data in administrative records to assess missing data bias, and additional analyses were performed with follow-up censored at 5 years among those with at least 1 year of follow-up to further examine potential biases related to the year of cohort entry. Data analyses were performed using R version 4.4.1. For all hypothesis tests, a two-tailed p value below 0.05 was deemed indicative of statistical significance.

Results

Demographic and Clinical Characteristics of the Study Population

The patient cohort comprised all individuals (n = 1,235) who received OAT at TYKS from January 1, 2004, to April 15, 2021. The median follow-up duration was 9.36 years (IQR: 3.87–13.45), corresponding to 10,906.03 PYs, as shown in online supplementary Figure S1. Of these patients, 171 (13.8%) died, with a median age at death of 37.78 years (IQR: 31.83–43.50). The median time from the initial OAT to death was 5.24 years (IQR: 2.12–8.37). The overall crude mortality rate was 15.68 per 1,000 PYs (95% CI: 13.42–18.21). The standardized mortality ratio was 8.49 (95% CI: 7.28–9.82), indicating that the observed number of deaths was approximately 8 times the expected number in the general Finnish population, matched by age, gender, and calendar year. The total YPLL before age 75 was 5,991.79. Three deaths were recorded among individuals over age 75. The mean YPLL per death was 35.67 years. Additionally, 300 patients (24.3%) received care for an overdose, and 279 patients (22.6%) received care for self-harm during OAT.

Table 1 and online supplementary Table S3 show the characteristics of patients in OAT. The cohort consisted of 820 males (66.4%) and 415 females (33.6%), with half residing in Turku. The median age at first OAT diagnosis was 29.20 years (IQR: 24.68–35.71). Most patients had a record of buprenorphine medication (93.1%), 129 patients (16.4%) had a record of methadone medication, and 450 patients (36.4%) had an unknown OAT medication type. All patients had at least one psychiatric contact, and the majority (97.4%) had addiction psychiatry contact at any time during OAT. Approximately 59.1% of the patients were documented as having a medical comorbidity, and 24.3% of them were documented as having a history of self-harm or drug overdose prior to their initial OAT.

Table 1.

Characteristics of the study population by outcomes in OAT

Characteristic	Overall	Self-harm			Drug overdose			Death
	N = 1,235	no	yes	p value1	no	yes	p value1	no	yes	p value2
		N = 956	N = 279		N = 935	N = 300		N = 1,064	N = 171
	N (%)	N (%)	N (%)		N (%)	N (%)		N (%)	N (%)
Sex				0.121			0.018			0.007
Female	415 (33.6)	332 (34.7)	83 (29.7)		331 (35.4)	84 (28.0)		373 (35.1)	42 (24.6)
Male	820 (66.4)	624 (65.3)	196 (70.3)		604 (64.6)	216 (72.0)		691 (64.9)	129 (75.4)
Age at first OAT diagnosis				<0.001			<0.001			0.014
<30 years	664 (53.8)	485 (50.7)	179 (64.2)		470 (50.3)	194 (64.7)		587 (55.2)	77 (45.0)
≥30 years	571 (46.2)	471 (49.3)	100 (35.8)		465 (49.7)	106 (35.3)		477 (44.8)	94 (55.0)
Year of first OAT				<0.001			<0.001			<0.001
2004–2006	309 (25.0)	218 (22.8)	91 (32.6)		215 (23.0)	94 (31.3)		255 (24.0)	54 (31.6)
2007–2010	314 (25.4)	217 (22.7)	97 (34.8)		216 (23.1)	98 (32.7)		252 (23.7)	62 (36.3)
2011–2016	321 (26.0)	252 (26.4)	69 (24.7)		249 (26.6)	72 (24.0)		275 (25.8)	46 (26.9)
2017–2021	291 (23.6)	269 (28.1)	22 (7.9)		255 (27.3)	36 (12.0)		282 (26.5)	9 (5.3)
Place of residence				0.002			0.002			0.188
Turku	614 (49.7)	453 (47.4)	161 (57.7)		442 (47.3)	172 (57.3)		521 (49.0)	93 (54.4)
Other than Turku	621 (50.3)	503 (52.6)	118 (42.3)		493 (52.7)	128 (42.7)		543 (51.0)	78 (45.6)
Methadone	129 (16.4)	74 (12.8)	55 (26.8)	<0.001	69 (12.3)	60 (26.9)	<0.001	113 (15.6)	16 (25.8)	0.038
Buprenorphine	731 (93.1)	545 (94.0)	186 (90.7)	0.116	530 (94.3)	201 (90.1)	0.037	677 (93.6)	54 (87.1)	0.064
Addiction psychiatry contact	1,203 (97.4)	928 (97.1)	275 (98.6)	0.167	908 (97.1)	295 (98.3)	0.247	1,039 (97.7)	164 (95.9)	0.192
CCI				<0.001			<0.001			<0.001
0	505 (40.9)	428 (44.8)	77 (27.6)		417 (44.6)	88 (29.3)		445 (41.8)	60 (35.1)
1	526 (42.6)	395 (41.3)	131 (47.0)		388 (41.5)	138 (46.0)		472 (44.4)	54 (31.6)
2 or more	204 (16.5)	133 (13.9)	71 (25.4)		130 (13.9)	74 (24.7)		147 (13.8)	57 (33.3)
History of self-harm or overdose	300 (24.3)	211 (22.1)	89 (31.9)	<0.001	200 (21.4)	100 (33.3)	<0.001	268 (25.2)	32 (18.7)	0.067

OAT, opioid agonist therapy; CCI, Charlson Comorbidity Index.

¹ Pearson’s chi-squared test.

² Pearson’s chi-squared test; Fisher’s exact test.

Table 2 and online supplementary Table S4 present the prevalence of recorded psychiatric disorders among OAT patients. Overall, 899 (72.8%) had at least one recorded diagnosis of any of the following six psychiatric comorbidities: non-SUD psychiatric disorders, psychosis, mood disorders, anxiety disorders, PDs, or other SUDs. Other SUDs constituted the most prevalent recorded psychiatric comorbidity subgroup, accounting for 61.4%. Within this subgroup, administrative records identified sedative use disorder in 36.8% of cases, stimulant use disorder in 27.0%, AUD in 22.8%, cannabis use disorder in 13.7%, and polysubstance use disorder in 21.1%. Additionally, we observed the following subgroups in the records: non-SUD psychiatric disorders (39.0%), psychosis (17.0%), mood disorders (29.1%), anxiety disorders (26.2%), and PDs (20.7%).

Table 2.

Psychiatric comorbidities of the study population by outcomes in OAT

Psychiatric condition	Overall	Self-harm			Drug overdose			Death
	N = 1,235	no	yes	p value1	no	yes	p value1	no	yes	p value2
		N = 956	N = 279		N = 935	N = 300		N = 1,064	N = 171
	N (%)	N (%)	N (%)		N (%)	N (%)		N (%)	N (%)
Non-SUD psychiatric disorders	482 (39.0)	311 (32.5)	171 (61.3)	<0.001	303 (32.4)	179 (59.7)	<0.001	416 (39.1)	66 (38.6)	0.901
Any psychosis	210 (17.0)	116 (12.1)	94 (33.7)	<0.001	110 (11.8)	100 (33.3)	<0.001	189 (17.8)	21 (12.3)	0.077
Psychotic disorders	172 (13.9)	101 (10.6)	71 (25.4)	<0.001	96 (10.3)	76 (25.3)	<0.001	157 (14.8)	15 (8.8)	0.036
Any SIP	96 (7.8)	43 (4.5)	53 (19.0)	<0.001	35 (3.7)	61 (20.3)	<0.001	83 (7.8)	13 (7.6)	0.928
Any mood disorder	360 (29.1)	244 (25.5)	116 (41.6)	<0.001	246 (26.3)	114 (38.0)	<0.001	310 (29.1)	50 (29.2)	0.978
Any depression	299 (24.2)	206 (21.5)	93 (33.3)	<0.001	206 (22.0)	93 (31.0)	0.002	256 (24.1)	43 (25.1)	0.758
Any bipolar	68 (5.5)	45 (4.7)	23 (8.2)	0.023	45 (4.8)	23 (7.7)	0.059	56 (5.3)	12 (7.0)	0.351
Any anxiety disorder	323 (26.2)	223 (23.3)	100 (35.8)	<0.001	222 (23.7)	101 (33.7)	<0.001	290 (27.3)	33 (19.3)	0.028
Any PD	256 (20.7)	152 (15.9)	104 (37.3)	<0.001	155 (16.6)	101 (33.7)	<0.001	226 (21.2)	30 (17.5)	0.268
Antisocial PD	127 (10.3)	67 (7.0)	60 (21.5)	<0.001	68 (7.3)	59 (19.7)	<0.001	113 (10.6)	14 (8.2)	0.331
Borderline PD	131 (10.6)	78 (8.2)	53 (19.0)	<0.001	79 (8.4)	52 (17.3)	<0.001	120 (11.3)	11 (6.4)	0.056
Other SUDs	758 (61.4)	525 (54.9)	233 (83.5)	<0.001	510 (54.5)	248 (82.7)	<0.001	652 (61.3)	106 (62.0)	0.859
Alcohol	281 (22.8)	172 (18.0)	109 (39.1)	<0.001	171 (18.3)	110 (36.7)	<0.001	220 (20.7)	61 (35.7)	<0.001
Cannabis	169 (13.7)	111 (11.6)	58 (20.8)	<0.001	108 (11.6)	61 (20.3)	<0.001	156 (14.7)	13 (7.6)	0.013
Sedative	454 (36.8)	277 (29.0)	177 (63.4)	<0.001	270 (28.9)	184 (61.3)	<0.001	406 (38.2)	48 (28.1)	0.011
Stimulant	333 (27.0)	192 (20.1)	141 (50.5)	<0.001	184 (19.7)	149 (49.7)	<0.001	299 (28.1)	34 (19.9)	0.025
Multidrug	260 (21.1)	132 (13.8)	128 (45.9)	<0.001	128 (13.7)	132 (44.0)	<0.001	222 (20.9)	38 (22.2)	0.686

PD, personality disorder; SUD, substance use disorder; SIP, substance-induced psychosis; other SUDs, other substance use disorders not related to opioids.

¹ Pearson’s chi-squared test.

² Pearson’s chi-squared test; Fisher’s exact test.

Among patients with at least one documented psychiatric comorbidity, 72.7% had two or more, while only 28 patients (2.3%) in the total sample had records of all six comorbidities. Psychosis frequently coexisted with other psychiatric conditions in the records; only three patients (1.4%) with psychosis presented with isolated psychosis, whereas 205 patients (27.0%) diagnosed with other SUDs had no records of the additional five psychiatric conditions. More details are provided in online supplementary Figure S2.

Death

In Cox regression analyses, both the unadjusted and adjusted models demonstrated that there was no definitive evidence supporting an association between the six psychiatric comorbidity subgroups and the risk of death in OAT. Figure 1 shows the results of the full model.

Fig. 1.

Results of multi-predictor regression analyses on the associations between psychiatric comorbidities and outcomes, including death, nonfatal overdose, and self-harm. ^aFor weighted Cox regression models, estimates = average HRs; for logistic regression models, estimates = ORs. ^bWeighted Cox regression models were used for analyzing the death outcome. ^{c, d}Logistic regression models were employed to analyze nonfatal overdose and self-harm outcomes. ^{b, c,d}Mutually adjusted for all six psychiatric comorbidity subgroups – namely, non-SUD psychiatric disorders, psychosis, mood disorders, anxiety disorders, PDs, and other SUDs – based on adjustments for demographic variables (sex, age at first OAT, year of first OAT, and place of residence) as well as clinical variables (OAT medication, CCI, and history of self-harm or overdose). HRs, hazard ratios; ORs, odds ratios; PDs, personality disorders; OAT, opioid agonist therapy; CI, confidence interval; SUD, substance use disorder; other SUDs, other substance use disorders not related to opioids; CCI, Charlson Comorbidity Index.

The results of the multi-predictor Cox regression analysis remained consistent when the model was transitioned to a logistic regression framework, restricted to participants with follow-up periods exceeding one year, or employing covariates chosen through a stepwise procedure based on the AIC. Nevertheless, simplified models indicated that comorbid anxiety disorders were associated with a reduced risk of death in exploratory analyses that reanalyzed the data with follow-up censored at five years among individuals with at least one year of follow-up (adjusted HR = 0.28, 95% CI = 0.10–0.75) and restricted the analysis to individuals with explicitly documented OAT medication data (adjusted HR = 0.46, 95% CI = 0.24–0.91).

Given the potential association between alcohol and sedatives and opioid-related death, we conducted post hoc analyses. These analyses revealed that comorbid AUD was linked to a 1.54-fold increased hazard of death among patients undergoing OAT (adjusted HR = 1.54, 95% CI = 1.06–2.22), whereas there was no clear evidence of an association between sedative use disorder and the risk of death (adjusted HR = 0.94, 95% CI = 0.63–1.40). Results of detailed analyses regarding the risk of death in OAT are presented in online supplementary Tables S5 and S8.

Drug Overdose

In logistic regression analyses of overdose outcomes, all six psychiatric comorbidity subgroups showed a higher likelihood of drug overdose during OAT in both the unadjusted models and those adjusting for demographic and clinical variables. As observed in Figure 1, when additionally considered all comorbid psychiatric disorders collectively, the subsequent psychiatric comorbidity subgroups were associated with an elevated risk of overdose in OAT, with ORs ranging from 1.64 to 2.61 times: non-SUD psychiatric disorders (adjusted OR = 1.64, 95% CI = 1.14–2.36), psychosis (adjusted OR = 1.89, 95% CI = 1.29–2.79), and other SUDs (adjusted OR = 2.61, 95% CI = 1.81–3.83). However, the previously observed positive associations of comorbid mood disorders (adjusted OR = 0.92, 95% CI = 0.64–1.30) and anxiety disorders (adjusted OR = 0.89, 95% CI = 0.62–1.27) with overdose were no longer evident, and the association between PDs and overdose was attenuated (adjusted OR = 1.23, 95% CI = 0.86–1.77).

Sensitivity analyses were conducted by excluding patients with a recorded history of overdose and by using covariates selected via a stepwise procedure based on the AIC. These analyses yielded comparable results, supporting the robustness of the findings. However, an additional sensitivity analysis excluding individuals with unknown OAT medication types in the administrative records weakened the association between comorbid non-SUD psychiatric disorders and overdose (adjusted OR = 1.49, 95% CI = 0.96–2.31). Results of detailed analyses on overdose in OAT are shown in online supplementary Tables S6 and S8.

Self-Harm

Similar to overdose, in logistic regression analyses of self-harm outcomes, all six psychiatric comorbidity subgroups demonstrated an increased risk of self-harm in patients undergoing OAT, observed in both unadjusted and adjusted models that account for demographic and clinical factors. As shown in Figure 1, when all psychiatric comorbidities were further included in the same model, the associations of self-harm with the following subgroups in OAT persisted: non-SUD psychiatric disorders (adjusted OR = 1.50, 95% CI = 1.03–2.18), psychosis (adjusted OR = 1.87, 95% CI = 1.26–2.78), PDs (adjusted OR = 1.52, 95% CI = 1.06–2.18), and other SUDs (adjusted OR = 2.56, 95% CI = 1.74–3.82). Nevertheless, the association with mood disorders weakened (adjusted OR = 1.13, 95% CI = 0.79–1.61), and the original positive association with anxiety disorders was no longer observed (adjusted OR = 0.95, 95% CI = 0.66–1.38).

The results remained unchanged in the sensitivity analysis, which employed covariates selected via a stepwise procedure based on the AIC. The sensitivity analysis, which excluded patients with a documented prior history of self-harm, demonstrated that comorbid PD was no longer significantly associated with self-harm (adjusted OR = 1.52, 95% CI = 0.98–2.34) within the context of OAT. Furthermore, an additional sensitivity analysis, limited to individuals with known information regarding OAT medication type, revealed that the associations of non-SUD psychiatric disorders (adjusted OR = 1.41, 95% CI = 0.90–2.20) and PDs (adjusted OR = 1.39, 95% CI = 0.91–2.11) with self-harm were reduced. The results of detailed analyses concerning self-harm in OAT are provided in online supplementary Tables S7 and S8.

Associations of Psychosis and Other SUD Combinations with Outcomes

Fisher’s exact tests demonstrated that the proportions of overdose and self-harm varied significantly across the four combination groups of psychosis and other SUD comorbidities (both other SUDs and psychosis, other SUDs only, psychosis only, and neither condition) (overdose: p < 0.001; self-harm: p < 0.001). Patients recorded with both other SUDs and psychosis exhibited the highest proportions of overdose (52.8%) and self-harm (50%) in OAT. Conversely, patients without these two comorbidities in the administrative records showed the lowest proportions of overdose (10.5%) and self-harm (9.4%) (see Fig. 2). We used logistic regression to further assess associations of psychosis and other SUD combinations with outcomes. As shown in the regression results summarized in Table 3, individuals documented with both other SUDs and psychosis had a higher risk of self-harm and overdose than the other three groups in both unadjusted and adjusted models. In the unadjusted models, they had approximately 3.03 to 10 times greater odds of self-harm and 3.13 to 9.09 times greater odds of overdose compared to individuals in the other three groups. In the adjusted models, the odds were approximately 2 to 4.76 times higher than those of the other groups for both self-harm and overdose.

Fig. 2.

Grouped bar charts displaying the percentages of patients who received care for overdose and self-harm across four mutually exclusive groups in OAT: neither other SUDs nor psychosis, psychosis only, other SUDs only, and both other SUDs and psychosis. OAT, opioid agonist therapy; SUD, substance use disorder; other SUDs, other substance use disorders not related to opioids.

Table 3.

Associations of combinations of psychosis and other SUDs with overdose and self-harm in OAT

	Self-harm1		Overdose1
	unadjusted OR (95% CI)	adjusted OR (95% CI)2	unadjusted OR (95% CI)	adjusted OR (95% CI)2
Combinations of psychosis and other SUDs
Both	1 (ref.)	1 (ref.)	1 (ref.)	1 (ref.)
Psychosis only	0.15 (0.04–0.41)	0.23 (0.06–0.67)	0.18 (0.06–0.45)	0.26 (0.08–0.72)
Other SUDs only	0.33 (0.23–0.47)	0.50 (0.33–0.76)	0.32 (0.23–0.45)	0.50 (0.34–0.76)
Neither	0.10 (0.07–0.16)	0.21 (0.12–0.36)	0.11 (0.07–0.16)	0.21 (0.12–0.36)

SUD, substance use disorder; other SUDs, other substance use disorders not related to opioids; CI, confidence interval; OR, odds ratio; OAT, opioid agonist therapy; PDs, personality disorders; CCI, Charlson Comorbidity Index.

¹ Logistic regression model, reporting OR for self-harm and overdose.

²Adjusted models included combinations of psychosis and other SUDs, sex, age at first OAT, year of first OAT, place of residence, OAT medication, CCI, history of self-harm or overdose, non-SUD psychiatric disorders, mood disorders, anxiety disorders, and PDs.

OAT Medications and Outcomes

Individuals who have had both methadone and buprenorphine medications at some point during their OAT demonstrated approximately twice the risk of overdose and self-harm compared to use of buprenorphine medication alone. Additionally, use of methadone only was also associated with a roughly twofold increased risk of overdose compared with use of buprenorphine only. We provided additional details in online supplementary Table S8.

Discussion

In our study of patients attending OAT in Southwest Finland, comorbid non-SUD psychiatric disorders, psychosis, and other SUDs showed an association with an increased risk of overdose and self-harm. Patients recorded with both psychosis and other SUD diagnoses constituted the highest risk group for overdose and self-harm compared to other combinations of psychosis and other SUDs. Furthermore, patients with comorbid AUD in the records had a higher likelihood of death.

Contrary to our expectations, there was no definitive evidence of an association between all six psychiatric comorbidity subgroups and risk of death. Similarly, our results generally corroborate a study conducted over a 15-year follow-up period at an MMT clinic in Israel, which indicated that survival was not markedly associated with the presence of any axis I and/or axis II psychiatric diagnoses [17]. This absence of significant association may be attributable to the anti-craving, anti-anxiety, antidepressant, and antipsychotic properties inherent to opioid agonists such as methadone and buprenorphine. These psychotropic effects potentially mitigate symptoms in patients suffering from comorbid psychiatric conditions [37, 38]. However, a prior cohort study in Canada demonstrated that non-SUD psychiatric comorbidity was associated with an increased risk of all-cause death among individuals undergoing OAT [18]. It is noteworthy that this study involved a substantial sample size of 55,924, which may have contributed to incidental significant associations. Another Canadian investigation further categorized psychiatric disorders, akin to our analysis, and our findings are partially consistent with this research, which observed no conclusive evidence of a significant association for schizophrenia and death, whereas mood disorders emerged as independent risk factors for death [16]. The discrepancy may be attributed to varying samples, outcomes, and types of OAT, as it focused on opioid-related mortality among patients receiving social assistance in MMT. Overall, more evidence is necessary to elucidate the risk of death among patients with psychiatric comorbidities undergoing OAT.

Our novel observation that the proportion of overdose or self-harm was over fifty percent among patients recorded as having both psychosis and other SUD comorbidities in OAT is noteworthy. This proportion exceeded that of other combinations of psychosis and other SUDs, and individuals in this group faced a heightened risk of overdose and self-harm relative to persons in the other three groups. This finding may be associated with a series of social determinants of health and clinical factors. First, individuals with both psychosis and other SUDs frequently encounter economic and social challenges, including violence, housing instability, homelessness, poverty, less education, and unemployment [39–41]. Previous studies have identified lower socioeconomic status as a risk factor for nonfatal overdose [42] and self-harm [14, 43]. Second, prior evidence indicated that patients with psychosis and SUDs exhibited more positive psychotic symptoms than those with psychosis alone [40, 44, 45]. Furthermore, positive psychotic symptoms, such as auditory hallucinations and persecutory ideas, have been reported to be associated with lifetime suicide attempts [46]. A prior study suggested that the co-occurrence of SUD with schizophrenia spectrum disorders was linked to about a fourfold increased risk of non-suicidal self-injury compared with schizophrenia spectrum disorders without comorbidity [47]. Third, the coexistence of psychosis and other SUDs is associated with a higher incidence of cognitive deficits [39, 40]. Patients with cognitive deficits may experience diminished judgment and decision-making abilities, which hinder their capacity to assess risks related to tolerance, dosing, and polydrug use, thereby potentially increasing the likelihood of overdose [48]. Fourth, patients with psychosis and other SUDs may experience heightened stigma due to the additive effects of stigmatization associated with both psychosis and other SUDs [49, 50]. Schizophrenia and AUD are generally perceived as unpredictable and dangerous and are particularly heavily stigmatized among mental health conditions [51, 52]. Sibley et al. [53] suggested that stigma may contribute to an increased risk of overdose through either a psychological distress pathway or a stigma avoidance pathway. Given the highly elevated risk of overdose and self-harm among patients documented as having both psychosis and other SUDs, it is imperative to identify this group utilizing various assessment tools, such as self-reports, collateral reports, physical examinations, and laboratory tests [40, 54]. Moreover, it is essential to provide supplementary care for these individuals within the OAT framework.

As anticipated, other SUDs were identified as the most common comorbidity within this cohort, constituting 61% according to administrative records. Similarly, previous studies have reported the prevalence of current SUD other than heroin as ranging from 59% to 69% among patients undergoing MMT [55, 56]. Prior research reported a significant relationship between other SUDs and impaired emotion regulation [57]; increased problem severity in employment, family/social, legal, and psychiatric domains [58]; decreased treatment retention [59, 60]; and reduced satisfaction and subsequent nonresponse to treatment [61–63] in OAT. Our results concerning the increased risk of self-harm and overdose among patients with other SUDs in OAT broaden these findings. Consequently, given the high prevalence of other SUDs and their associated consequences in patients undergoing OAT, the implementation of targeted pharmacological or psychosocial interventions for other SUDs during OAT is recommended.

We observed an increased risk of death among patients with recorded AUD undergoing OAT in this study. This finding is consistent with the results of a prior study that employed descriptive analyses, as well as another investigation concentrating on opioid-related death among patients receiving social assistance within the context of OAT [16, 64]. The association between AUD and death during OAT aligns with existing scientific understanding. Comorbid AUD may contribute to death either through predispositional factors or as a direct consequence of the disorder itself [65–68]. For instance, patients who subsequently develop AUD may exhibit preexisting impulsivity, which has been associated with death by suicide [69, 70]. Moreover, comorbid AUD, as a central nervous system depressant, may intensify opioid-induced respiratory depression, potentially resulting in hypoxic respiratory failure and increasing the risk of fatal overdose [34]. A previous study found that comorbid AUD increased the risk of fatal overdose by 2.57 times in patients with OUD [71]. Furthermore, patients with comorbid AUD often demonstrate poorer performance in working memory and inhibitory control [72]. Poor inhibitory control correlated with a 6.6% elevated risk of death, and this correlation became more significant with advancing age among homeless individuals or those experiencing housing instability [73]. Additionally, patients with AUD are susceptible to liver disease, as total alcohol consumption correlates with cirrhosis and may accelerate the progression of liver pathology, particularly in the context of hepatitis C virus infection [74]. A previous study identified that comorbid AUD confers a 7.26-fold increased risk of hepatic-related death among patients with OUD [71]. Although OAT is an effective intervention for OUD, it is not designed to treat other SUDs. Most research indicates that alcohol consumption remains stable or may even increase concomitantly with a decrease in opioid use during OAT [75, 76]. For instance, a systematic review reported that only 3 out of 15 studies observed a decrease in alcohol use in MMT [77]. Therefore, considering the elevated risk of death and the limited effectiveness of OAT for AUD, screening tools such as the Alcohol Use Disorders Identification Test (AUDIT) for initial screening are recommended, and phosphatidylethanol can contribute to more accurate detection of alcohol consumption [78]. Additional treatments, such as brief interventions and disulfiram, need to be examined and applied to support patients with AUD undergoing OAT [76, 79].

Individuals receiving buprenorphine solely appeared to exhibit a reduced risk of overdose compared to those on methadone alone, and a lower risk of overdose and self-harm than individuals who have had both methadone and buprenorphine medications at some point during OAT. These results aligned with current observations and knowledge. The pharmacological safety advantages of buprenorphine over methadone are widely recognized [80, 81]. Buprenorphine’s ceiling effect as a partial agonist decreases respiratory depression [82], and a study indicated that methadone carries a sixfold higher overdose risk than buprenorphine among the general population in England and Wales [80]. Additionally, in Finland, patients using methadone constitute a complex patient group with multiple risk factors for nonfatal overdose [42, 83]. These include more frequent and more severe concomitant illicit substance use, particularly benzodiazepines; a greater number of previous OAT periods; and higher rates of psychiatric comorbidities and hepatitis C virus infections compared with those receiving buprenorphine [83]. Our study also identified a higher prevalence of non-SUD psychiatric disorders and PDs among patients undergoing methadone treatment compared with those receiving buprenorphine, based on available information collected during OAT. In a clinical setting, buprenorphine is the most widely used medication in Finland [23, 81]; methadone may be more commonly used for those who are not responding to buprenorphine in OAT [84]. Thus, a record of both methadone and buprenorphine medications during OAT may also be an indicator of a suboptimal treatment response.

Our study demonstrates significant strengths. It presents findings related to death, overdose, and self-harm among patients with psychiatric comorbidities undergoing OAT, supported by an extended follow-up period. These findings contribute to the existing knowledge base for clinicians and researchers regarding this distinct population.

However, our study has some limitations. First, in our study, cohort entry coincided with the date of the initial F11.22 diagnosis. However, this date might be inaccurate because nationwide register-based diagnoses were not used. For example, an individual may have received OAT at other facilities before this date, which was not captured in our data. Second, the number of patients with overdose events and those with self-harm events identified through diagnostic codes may have been underestimated in this study. Such events could have been missed if patients sought care outside the studied units or if these events did not come to the attention of medical staff. For example, some patients may not have presented to services following these events. Third, as an observational study, it is inherently limited in establishing causality due to its nonexperimental design. Fourth, the number of deaths within the psychiatric comorbidity subgroups was small, particularly in the group with psychosis, which constrained the potential for further analysis. Fifth, the OAT medication type variable had missing data for some patients, and overdose and self-harm diagnoses might appear multiple times and be misclassified as outcomes. The effects of non-SUD psychiatric disorders and PDs were reduced in the corresponding sensitivity analyses; note that the results should be interpreted with caution. Sixth, this study was conducted in Finland, where buprenorphine was the primary medication used in OAT, potentially limiting the generalizability of the findings.

In conclusion, patients recorded as having comorbid psychosis and other SUDs consistently exhibited a higher risk of overdose and self-harm in OAT, and those with both comorbidities were particularly vulnerable to self-harm and overdose. Additionally, comorbid AUD was associated with an increased risk of death. The research underscores the importance of screening and detecting psychosis and other SUDs (especially AUD) in patients receiving OAT to reduce the risks of overdose, self-harm, and death. Moreover, it is crucial to evaluate and appropriately incorporate pharmacological and psychosocial interventions, tailored to address both comorbidities in OAT.

Statement of Ethics

Finnish law does not require an Ethics Committee statement for registry studies. According to Finnish law, written informed consent is not required for registry studies.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was funded by the Finnish Ministry of Education and Culture, and the governmental subsidiaries for research (VTR grant from the Wellbeing Services County of South-West Finland). The funders had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

Y.L.: formal analysis, validation, and writing – original draft, review, and editing; A.D.: methodology, validation, supervision, and writing – review and editing; M.H.: writing – review and editing; L.S.: data curation and writing – review and editing; S.N.: conceptualization, methodology, supervision, and writing – review and editing. All authors approved the final manuscript.

Funding Statement

This study was funded by the Finnish Ministry of Education and Culture, and the governmental subsidiaries for research (VTR grant from the Wellbeing Services County of South-West Finland). The funders had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to Yangyang Li or the corresponding author.