Objective: To expand cognitive indices derived from PPMI cognitive battery (PPMI-CB) by testing hierarchical models and measurement invariance across pertinent groups.

Background: Parkinson’s can cause cognitive impairment, which should be expertly measured. Recent analyses have identified parsimonious and powerful indices of cognitive function. However, a single global score did not fit the PPMI-CB, which complicates the creation of a simple composite. This can sometimes happen with hierarchical constructs (a global factor subsuming subfactors). Additionally, it is unknown if these indices function similarly across groups, which is important for unbiased comparisons and diagnostic utility.

Method: Individuals were selected from PPMI at baseline, which included 4,546 individuals, who were mostly men (54.2%), older (M_age=65, SD=8.9), with higher education (M_years=16.3 SD= 3.4). Cognitive measures from the PPMI-CB included: Letter-Number Sequencing, Symbol Digit Modalities, Trail Making, Judgment of Line Orientation, Letter fluency, Category fluency, and Hopkins Verbal Learning. Hierarchical versions of a recently proposed three-factor model for the PPMI-CB were tested in higher order and bifactor form using confirmatory factor analysis. These used robust Satorra-Bentler estimation and confirmatory fit index, Tucker-Lewis index, standardized root mean square residual, and root mean square error of approximation to determine goodness of model fit. Change in CFI across increasing constraints for equal structure, loadings, and intercepts were used to evaluate measurement invariance across gender and cognitive stages.

Results: Although both hierarchical models fit well, the bifactor model had better fit than the higher order model (CFI=.99, TLI=.98, RMSEA=.03, SRMR=.02 vs CFI=.97, TLI=.96, RMSEA=.05, SRMR=.03). When evaluating nested constraints across groups, ΔCFIs were <.01, indicating the three-factor structure fit well across gender and cognitive impairment status with functionally equivalent structure, loadings, and means among these groups.

Conclusion: Previously proposed indices for the PPMI-CB are best conceptualized hierarchically, and a hierarchical three-factor solution is optimal across pertinent subgroups. Simple composites that do not consider the hierarchical, multifactorial nature of cognition are suboptimal. This can inform research and clinical practice addressing cognitive impairment in PD.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/optimal-and-invariant-hierarchical-cognitive-indices-in-the-parkinsons-progression-markers-initiative/