Multivariate normative comparisons for neuropsychological assessment by a multilevel factor structure or multiple...

Area of Science:

• Neuropsychology
• Clinical Psychology
• Biostatistics

Background:

• Neuropsychological tests aid in diagnosing cognitive impairment by comparing patient scores to normative data.
• Multivariate normative comparisons, analyzing entire score profiles, enhance clinical decision-making accuracy.
• A key challenge is the unavailability of comprehensive multivariate normative datasets.

Purpose of the Study:

• To propose a method for creating multivariate normative datasets by aggregating data from existing studies.
• To address the issue of missing data in aggregated datasets.
• To evaluate statistical approaches for handling missing data in this context.

Main Methods:

• Aggregating healthy control group data from multiple neuropsychological studies.
• Employing statistical techniques, specifically multiple imputation and factor modeling, to manage missing data.
• Conducting simulation studies to compare the performance of imputation methods.

Main Results:

• Factor modeling demonstrated superior performance over multiple imputation for handling missing data in aggregated neuropsychological datasets.
• The effectiveness of factor modeling is contingent upon adequate specification of the factor model.
• This approach facilitates the creation of robust multivariate normative datasets.

Conclusions:

• Factor modeling offers a viable solution for constructing multivariate normative datasets from heterogeneous sources.
• The proposed method enables routine use of multivariate normative comparisons in clinical practice.
• This will ultimately lead to more precise and reliable clinical decision-making in neuropsychology.