Thursday, May 30, 2019
Pacific A (Hyatt Regency San Francisco)
* noted as presenting author
Opiate use among youth, in particular, American Indian (AI) youth is rising which has resulted in a large number of accidental overdoses and deaths. There is limited research on risk and protective factors for current opiate use among youth. In order to develop effective prevention strategies, we need a more comprehensive understanding of factors predictive of opiate use. An exploratory approach for identifying risk and protective factors from big datasets is machine learning. An advantage of machine learning is the identification of previously unknown predictive variables from large survey data. The present study is an application of machine learning to the Our Youth Our Future epidemiological survey of youth (both AI and non-AI) living on or near reservations (N = 6360) to determine salient risk and protective factors for past 30-day opiate use. After analyzing over 300 variables using classification tree analysis, we identified 6 protective factors and 6 risk factors. The most important predictor was recent cocaine use, with past 30-day cocaine use portending risk for past 30-day opiate use. The other risk factors identified by the model were ever having used narcotics other than heroin, having ever used concentrated THC, reporting feeling lonely, and identifying as Native American. Importantly, the highest risk was conferred for those reporting recent cocaine use, having ever tried a narcotic other than heroin, and identifying as Native American. Protective factors included, never having used a narcotic other than heroin, not having used concentrated THC, having fewer friends asking you to use illegal drugs, and attending more religious services. This model explained 64% of the variance in 30-day opiate use in the sample, and on average 35% of the variance across 1000 bootstrapped samples. Taken together, this model identifies known predictors of 30-day opiate use, e.g., recent substance use, as well as unknown predictors including being AI, social isolation, and peer encouragement for use. Notably, recent cocaine use was a more important predictor of recent opiate use than lifetime opiate use. Based on these results, prevention efforts aimed at reducing opiate use should target substance use broadly, as well as social and spiritual factors.