To investigate how the locus of control in the aggregation mechanism affects clinical utility and treatment decisions in AI-assisted antidepressant selection.
Approach:
Study Design: A proof-of-concept, within-subjects clinician pilot study was conducted with 22 physicians comparing three weighting schemes for antidepressant selection.
Weighting Schemes: The three schemes evaluated were: Implicit Weighting (raw probabilities), Static Expert-Derived Weighting (fixed expert weights), and Dynamic Clinician-Determined Weighting (clinician-defined weights).
This scheme led to data-informed revisions of initial antidepressant choices in 33.3% of cases.
The positive effects were observed among both psychiatrists and primary care physicians.
Interpretation:
Effective integration of AI into psychiatric practice requires flexible decision support systems that maintain clinical agency while incorporating data-driven predictions.
Limitations:
The study involved a small sample size of 22 physicians.
The findings may not be generalizable beyond the study's specific context.
Conclusion:
Dynamic weighting may better reflect the nuanced and individualized nature of mental health care.
by Akiva Kleinerman, David Benrimoh, Amit Yaniv-Rosenfeld, Grace Golden, Myriam Tanguay-Sela, Howard C. Margolese, Teddy Lazebnik, Ben H. Amit, Hadar Samuel, Ariel Rosenfeld
