Battery length as a moderator of classification accuracy for cognitive impairment: a Monte Carlo simulation using actuarial and Bayesian methods.

Abstract

Objective: This Monte Carlo simulation study examined whether neuropsychological test battery length impacts cognitive impairment classification accuracy using actuarial and Bayesian interpretive frameworks. Three hypotheses were tested: (a) longer batteries yield higher false-positive rates due to cumulative probability effects; (b) longer batteries show superior classification accuracy; and (c) abbreviated batteries have higher false-negative rates.

Method: Four simulated batteries of varying length, domain coverage, and redundancy were applied to 10,000 virtual patient profiles. Sensitivity, specificity, predictive values, classification accuracy, and Bayesian post-test probabilities were evaluated across multiple prevalence conditions alongside base rates of low scores and resource demands.

Results: Hypothesis 1 was supported: longer batteries produced higher base rates of low scores in healthy controls. However, applying actuarial and Bayesian frameworks prevented this from translating into inflated false-positive rates. Hypotheses 2 and 3 were not supported: abbreviated batteries maintained classification accuracy, discrimination (AUC), false-negative rates, and predictive values comparable to longer batteries. Adding tests yielded diminishing returns, reaching a psychometric plateau at approximately eight tests.

Conclusions: Battery length does not meaningfully alter binary impairment classification when actuarial and Bayesian rules are applied. Brief core batteries maximize efficiency for global detection, while comprehensive batteries remain best suited for detailed cognitive profiling and differential diagnosis.