Heroin overdose risk profiles following long‐term heroin dependence: Findings from the 18‐ to 20‐year follow‐up of the Australian treatment outcome study

AIMS: To identify heroin overdose and polydrug risk profiles of Australian adults with long-term heroin dependence and assess if psychosocial, treatment, physical and mental health-related covariates were associated with class membership. DESIGN: The 18-20-year follow-up of the Australian Treatment Outcome Study (ATOS). SETTING: Sydney, New South Wales, Australia. PARTICIPANTS: 615 people with heroin dependence were recruited in 2001-2002, of whom 401 completed the follow-up interview between 2019 and 2021 and were included in this study (mean age: 47.3 years, standard deviation: 7.4; 66% male). MEASUREMENTS: Latent class analysis used indicator variables of heroin use, overdose history and past-month use of alcohol, benzodiazepines, other opioids, cannabis, amphetamines and cocaine. Odds ratios were estimated for covariates associated with class membership using multinomial logistic regression. Covariates included: age, sex, employment, crime, lifetime incarceration, unstable housing, age of first heroin use under 18 years, use of opioid maintenance therapy, naloxone history, trauma, post-traumatic stress disorder diagnosis, Short Form Health Survey-12 scores representing severely poor physical and mental health and depression. All latent class indicator variables and covariates were based on the latest 2019-2021 responses, except for three covariates (age of first heroin use, antisocial personality disorder and borderline personality disorder) where responses from 2001to 2002 were retained. RESULTS: Three latent classes were identified: high risk (7%), moderate risk (44%) and low risk (49%) of heroin overdose. The 3-class model had a Log Likelihood of -2133.31 and Entropy of 0.61. The 3-class model had the lowest sample size adjusted-Bayesian information of 601.02 and the bootstrapped likelihood ratio test suggested the 3-class model fit better than the 2-class model (P < 0.05). The high-risk class was defined by high probabilities of a past-year overdose, more than 10 lifetime overdoses, and 44-74% probability of past-month use of heroin, amphetamines, benzodiazepines and other opioids. Individuals in the moderate- and low-risk class were unlikely to have a past-year overdose. The moderate risk class had 30-50% probability of past-month use of heroin, amphetamines, benzodiazepines and other opioids, while the low-risk class had 6-17% probability of past-month use of these substances. Past-month crime, past-month depression and lifetime incarceration were associated with moderate- and high-risk class membership. Current methadone treatment and lifetime trauma were associated with moderate-risk class membership. SUMMARY: The Australian Treatment Outcome Study cohort with long-term heroin dependence can be classified into high, moderate and low risk of heroin overdose based on overdose history and polydrug use. People with higher rates of heroin, cannabis, amphetamines, benzodiazepines and other opioid use are more likely to belong to the high- and moderate-risk class.