To estimate the prevalence of and risk factors associated with concurrent injection of multiple substances (co-injection) among a community-recruited cohort of people who inject drugs.

Cross-sectional study.

Melbourne, Australia.

A sample of 720 actively injecting participants from the Melbourne Injecting Drug User Cohort Study (33% female) was extracted.

We constructed two statistical models: a logistic regression model analysing correlates of co-injection of any substance combination in the past month and a multinomial logistic regression model analysing correlates of three mutually exclusive groups: heroin-diphenhydramine co-injection only, co-injection of other substances and no co-injection. Risk factors examined included drug use characteristics, demographic characteristics, health service use, hepatitis C status, injection risk behaviours and previous experience of non-fatal overdose.

One-third [n = 226, 31%; 95% confidence interval (CI): 28-34%] of participants reported co-injecting substances within the past month, with equal numbers of participants reporting injecting combinations of heroin-diphenhydramine (n = 121, 54%; 95% CI = 48-60%) and heroin-methamphetamine (n = 121, 54%; 95% CI = 48-60%). In logistic regression analyses, reporting co-injection of any substance combination was associated with male sex [adjusted odds ratio (aOR) = 1.80, 95% CI = 1.18-2.74, P = 0.006] and injecting daily or more frequently (aOR = 2.04, 95% CI = 1.31-3.18, P = 0.002). In multinomial logistic regression analyses, participants reporting heroin-diphenhydramine co-injection only were significantly more likely to report groin injecting [adjusted relative risk ratio (aRRR) = 6.16, 95% CI = 2.80-13.56, P < 0.001] and overdose (requiring an ambulance) in the past 12 months (aRRR = 2.81, 95% CI = 1.17-6.72, P = 0.021) compared with participants reporting no co-injection or co-injection of other substances.

A substantial proportion of people who inject drugs report co-injection of multiple substances, which is associated with a range of socio-demographic, drug use and health service use risk factors.