Modeling drug detection and diagnosis with the 'drug evaluation and classification program'

Area of Science:

• Forensic Science
• Toxicology
• Computational Science

Background:

• Drug impairment poses significant public safety risks.
• Accurate identification of impairing substances is crucial for law enforcement.
• Existing methods for drug impairment detection have limitations.

Purpose of the Study:

• To develop and evaluate formal models and algorithms for detecting drug impairment.
• To identify specific impairing drug types using data from Drug Evaluation and Classification (DEC) investigations.
• To compare the accuracy of formal models against trained police officers in identifying drug impairment.

Main Methods:

• Utilized data from Drug Evaluation and Classification (DEC) investigations, including vital signs and observable signs/symptoms.
• Developed formal models and algorithms based on police officer-collected data.
• Employed logistic regression to analyze the accuracy of the formal models.

Main Results:

• Formal models achieved >60% sensitivity and >90% specificity for cannabis, alprazolam, and amphetamine impairment.
• Sensitivity for codeine impairment was low (~20%) despite high specificity (~90%).
• Formal models significantly outperformed trained officers in identifying cannabis, alprazolam, and amphetamine impairments.

Conclusions:

• The formal model demonstrates superior accuracy in identifying common drug impairments compared to trained officers.
• Joint application of DEC procedures and the formal model enhances drug detection and identification capabilities.
• Further refinement is needed to improve the detection of specific drug impairments, such as codeine.