*2.4. Covariates*

The panel used a process of consensus to decide on the pertinent covariates to include in the model, and the measures that should be used in each trial to ascertain the variables/covariates of interest. Decisions were made based on the literature and standard practice. Demographic covariates considered across the protocols included age, gender, and race/ethnicity.

Given the correlation between depression and suicidality, the prevalence of baseline depression was also assessed by creating a binary indicator of depression using the instrument that had been part of each trial's procedures. Depression measures included the 9-item Patient Health Questionnaire (PHQ-9 [26]: CTN0067 [38,46], CTN0068 [39,47]), the 18-item Brief Symptom Inventory (BSI-18 [48]: CTN0049 [33,41], CTN0064 [37,45]), the Addiction Severity Index Lite (ASI Lite [49];CTN0037 [32,40], CTN0051 [34,42]), the Medical and Psychiatric History (CTN0054 [36,44]), and the Hospital Anxiety and Depression Scale (HADS [50]: CTN0053 [35,43]). Because lifetime heroin use [51], recent alcohol and benzodiazepine use [52], and past psychiatric history [53], increase risk of overdose, these factors were also included as covariates in the model. Each of these covariates was assessed by creating binary variables of each distinct instrument or question across the 8 trials. The assessment of lifetime use of heroin included the Addiction Severity Index Lite (ASI Lite [49]: 0037 [32,40], 0049 [33,41], 0051 [34,42], 0064 [37,45], 0067 [38,46]) and the Alcohol and Substance History (0054 [36,44], 0068 [39,47]). One trial, CTN 0054, did not assess lifetime use of heroin. Recent alcohol and benzodiazepine use was determined by creating

a binary indicator using each trial's instrument to assess this variable. The assessment of alcohol and benzodiazepine use was determined by the ASI Lite (0037 [32,40], 0049 [33,41], 0051 [34,42], 0064 [37,45], 0067 [38,46]), and the DSM-5 checklist [54] (CTN0054 [36,44], CTN0053 [35,43], CTN0068 [39,47]). Psychiatric history exclusive of depression was determined as a binary indicator, using each trial's instrument to assess this variable. The assessment of psychiatric history included the Medical History Form (0051 [34,42], 0054 [36,44], 0067 [38,46], 0068 [39,47]), ASI Lite (CTN 0037 [32,40]), Additional Psychiatric Diagnosis Form (CTN0064 [37,45]), and the Mini International Neuropsychiatric Interview, version 6.0 (MINI 6.0 [55]: 0053 [35,43]). For participants in CTN0049, psychiatric diagnosis was considered via two instruments. First, the team reviewed the Initial Hospital Admission form, and included patients as having a history of psychiatric diagnosis if the primary diagnosis during admission and/or any comorbid diagnoses included terms or conditions such as "suicidal ideation", "psychosis", "schizophrenia", "bipolar disorder", "PTSD", "hallucinations", "mood disorder" and "altered mental state". Second, if CTN0049 participants reported that they saw a professional for the primary purpose of getting help for psychological or emotional issues in the past 6 months (Service Utilization Detail Form [33,41]), they were also included as having a history of a mental health diagnosis.

Because these trials were diverse in the study treatments, settings, targeted substance use disorders and specific populations, we included trial as a covariate to account for this variability. Treatment arm (experimental or control) was also included to account for the difference in treatments within trials.

#### *2.5. Analytic Plan*

Descriptive statistics, including mean and standard deviation for continuous variables and frequencies and proportions for categorical variables, were calculated across trials and among participants with and without suicidality and with and without overdose events. A multivariate logistic regression, using a generalized estimating equation, analyzed continuous CHRT-SR score at baseline as a predictor of binary overdose event (present/not present), while controlling for covariates. Adjusted odds ratios and 95% confidence intervals were calculated. For all analyses, two-tailed *p*-values less than 0.05 were considered statistically significant. All analyses were performed using SAS version 9.4 [56].
