Patterns of substance use and mortality risk in a cohort of 'hard-to-reach' polysubstance users

Aims: To examine the mortality risk in a cohort of 'hard-to-reach' polysubstance users and its putative associations with substance use. Specifically, we estimated all-cause mortality risk as a function of individual substance use indicators, and then as a function of their complex, 'real-life' patterns as identified through latent class analysis (LCA).

Design: Prospective cohort study among street- and low-threshold service-recruited polysubstance users included between September and November 2013 and followed through the National Cause of Death Registry until 31 October 2015.

Setting: Seven Norwegian cities.

Participants: A total of 884 users of illegal opioids and/or stimulants. Of these, 357 were in opioid substitution treatment (OST) at the time of inclusion. Forty-four participants died during follow-up.

Measurements: Primary outcome: all-cause mortality risk. Unadjusted and adjusted Cox proportional hazard (PH) regression models (covariates: male, age, homelessness/shelter use, overdose experience, OST status, years of injecting, individual substance use indicators, city, use patterns). LCA models estimated separately for those in and those not in OST due to measurement invariance.

Findings: The crude mortality rate was 2.52 per 100 person-years. Standardized mortality ratio was 26.11 [95% confidence interval (CI) = 10.06-54.87] for women and 10.71 (95% CI = 6.39-16.81) for men. No single drug use indicator, such as 'heroin injection' or 'number of drugs used', was associated with the mortality risk. However, meaningful use patterns were identified; three OST and non-OST patterns each. The non-OST patterns 'polysubstance injectors' [hazard ratio (HR) = 3.45, 95% CI = 0.98-12.14] and 'low frequent injectors' (HR = 3.17, CI = 1.05-9.56) were associated significantly with the mortality risk even when adjusted for other known risk factors.

Conclusions: In a Norwegian prospective cohort study, 'hard-to-reach' polysubstance users had a more than 10 times higher mortality risk than the general population. Mortality risk was not a function of any single drug use indicator, but two distinct combinations of substances, frequencies and routes of administration were associated with the mortality risk.