Researchers at Penn Medicine have developed a prediction tool to identify individuals at high risk of overdose from stimulants like cocaine and methamphetamine. The tool, detailed in JAMA Health Forum, addresses a critical gap: while overdose deaths involving stimulants account for 70% of all substance overdose deaths in Philadelphia and 60% nationwide, such deaths have received less attention than opioid-related fatalities.
The model was trained using de-identified Medicaid data covering nearly 71 million people and achieved extremely high accuracy, scoring above 9 out of 10 on statistical measures. Key risk factors identified include prior substance use diagnoses, previous overdoses, higher poverty levels, crowded housing conditions, and male sex.
The researchers envision the tool being used proactively in population health settings to direct resources such as cognitive behavioral therapy, naloxone provision, and incentive-based recovery programs to at-risk individuals. This approach treats stimulant use disorder as a chronic disease requiring proactive management rather than reactive or punitive responses. The transparent, open algorithm is designed to build trust among clinicians and public health officials, potentially enabling early intervention to prevent overdoses and save lives.
Covered by Penn Medicine Communications here: Overdose Prediction Tool for Cocaine, Other Stimulants Developed