Using machine learning to evaluate treatment effects in multiple-group interrupted time series analysis.

Area of Science:

• Epidemiology
• Biostatistics
• Machine Learning

Background:

• Interrupted time series analysis (ITSA) is a robust method for evaluating interventions.
• Internal validity in ITSA is enhanced by using comparable control groups.
• Evaluating multiple-group ITSA often requires advanced analytical techniques.

Purpose of the Study:

• Introduce a novel machine learning approach, optimal discriminant analysis (ODA), for multiple-group ITSA.
• Evaluate the effectiveness of California's Proposition 99 using ODA and compare it to traditional ITSA regression.
• Assess the generalizability and advantages of ODA in intervention effect evaluation.

Main Methods:

• Applied optimal discriminant analysis (ODA) to evaluate treatment effects in a multiple-group ITSA.
• Utilized California's Proposition 99 (cigarette sales reduction) as a case study, comparing California (treated) to Montana (control).
• Contrasted ODA results with interrupted time series regression (ITSAREG).

Main Results:

• Both ODA and ITSAREG found comparable pre-intervention time series for California and Montana.
• Both methods indicated statistically significant reductions in California's cigarette sales post-intervention (P < 0.0001).
• ODA demonstrated high classification accuracy (91.67% sensitivity), maintaining 75.00% after leave-one-out cross-validation.

Conclusions:

• Optimal discriminant analysis (ODA) provides results comparable to traditional ITSAREG, increasing confidence in intervention effect findings.
• ODA offers advantages over conventional methods, including handling skewed data and model-free P-value derivation.
• ODA's features, such as threshold identification and cross-validation, make it a potentially superior approach for multiple-group ITSA.