Replicating psychiatric ratings through multiple regression analysis: The Midtown Manhattan Restudy

Area of Science:

• Psychiatry
• Psychometrics
• Behavioral Science

Background:

• Assessing mental health relies on psychiatrists' subjective judgments.
• Predictive validity of these judgments for future outcomes is crucial.
• Exploring computational methods for objective assessment is needed.

Purpose of the Study:

• To determine if multiple regression can predict psychiatrists' mental health ratings.
• To assess the validity of these ratings as a measure of mental health over time.
• To examine the relationship between initial mental health ratings and subsequent behavioral outcomes.

Main Methods:

• Utilized multiple regression techniques to predict psychiatrists' judgments.
• Compared computer-derived ratings with original psychiatrist ratings.
• Analyzed the correlation between 1954 mental health ratings and later reported behaviors (e.g., hospitalization).

Main Results:

• Computer-derived ratings adequately substitute for psychiatrists' ratings, explaining 69% of the variance.
• Computer-derived ratings demonstrated similar behavior to original ratings concerning other variables.
• Neither psychiatrists' nor computer-derived ratings accurately predicted subsequent self-reported mental impairment behaviors.

Conclusions:

• Computational methods, specifically multiple regression, offer a viable alternative to psychiatrists' direct ratings for mental health assessment.
• Despite high concordance with original ratings, the predictive power of these mental health assessments for future behavioral issues remains limited.