Applications of machine learning in addiction studies: A systematic review

Area of Science:

• Computational psychiatry
• Addiction research
• Data science in medicine

Background:

• Machine learning (ML) offers advanced analytical capabilities for complex datasets in healthcare.
• Addiction research traditionally relies on statistical methods, but ML presents new avenues for prediction and understanding.
• The integration of ML in addiction psychiatry is a growing area of interest.

Purpose of the Study:

• To systematically review the applications of machine learning methods in addiction research.
• To identify the types of addiction and ML techniques predominantly studied.
• To assess the trends in ML application within addiction psychiatry.

Main Methods:

• Systematic literature search across MEDLINE, Embase, and Cochrane Database of Systematic Reviews.
• Screening of 23 full-text articles, with 17 meeting inclusion criteria for the final review.
• Categorization of studies by addiction type (substance vs. non-substance) and ML methodology (supervised, unsupervised, reinforcement learning).

Main Results:

• The review included 17 studies, predominantly on substance addiction (82.4%), including smoking, alcohol, cocaine, and opioids.
• Non-substance addictions covered gambling and internet gaming (17.6%).
• Supervised learning was the most common ML approach (N=13), followed by unsupervised (N=2) and reinforcement learning (N=2).

Conclusions:

• Machine learning methods, particularly supervised learning, are increasingly utilized in addiction psychiatry.
• These techniques show promise for informing medical decisions in the diagnosis and treatment of addiction.
• Further research can explore the nuances of different ML algorithms across diverse addiction types.