On criterion weighting in the DSM-IV

Area of Science:

• Psychiatry
• Psychometrics
• Clinical Psychology

Background:

• Differential weighting of illness signs and symptoms is a recurring theme in psychiatric nosology.
• The Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV) provides criteria for diagnosing alcohol dependence.

Purpose of the Study:

• To compare the predictive efficiency of six alternately weighted algorithms for diagnosing alcohol dependence based on DSM-IV criteria.
• To explore an alternative strategy using factorially complex criteria to develop a novel diagnostic algorithm.

Main Methods:

• Six algorithms (statistical, unit, rational, random weighting) were applied to DSM-IV criteria in a sample of 365 participants.
• Predictive efficiency was assessed against concurrent validators and 6-month drinking outcomes.
• An experimental algorithm with 39 candidate criteria was developed and weighted differently.

Main Results:

• Comparable predictive efficiency was observed across all tested DSM-IV algorithms, attributed to high internal consistency.
• The experimental algorithm with complex criteria demonstrated low internal consistency but substantial surplus predictive power when differentially weighted.
• Rational weighting models for the experimental algorithm outperformed the random weight model.

Conclusions:

• A conflict exists in nosology between high internal consistency and the desire for differential weighting for prognosis.
• Algorithms with high internal consistency may not benefit from differential weighting.
• Diagnostic criteria with complex structures offer a potential approach to balance brevity, validity, and weighting performance.