Measurement bias for age, sex, and years of education in selected RDoC tasks

Measurement bias - systematic errors that lead to inaccuracies in assessing latent constructs - threatens to invalidate analysis and interpretation of ourcome scores collected in the context of clinical trials, longtiudinal research, and other types of neurocognitive studies; however, there has been little effort to examine this issue in tasks recommended by the National Institute of Health's Research Domain Criteria (RDoC) for neuropsychological research.

We aimed to evaluate measurement bias for sex, age, years of education, educational discouragement, underrepresented racial/ethnic identity, and household income in seven RDoC-recommended tasks. We recruited a sample of 320 individuals balanced for race, age, sex, and income. All participants completed a demographic survey and battery of neurocognitive tasks. We used Multiple Indicators Multiple Causes models to assess bias related to these factors.

Evidence of negative bias associated with older age was found for the Continuous Performance Task (CPT), and evidence of positive bias with older age was found for the Flanker task. Evidence of positive bias associated with sex, favoring women, was found for the Self-Ordered Pointing Task (SOPT). Additionally, evidence of negative bias associated with years of education was found for the CPT. No evidence of bias was found for the Sternberg task, Probabilistic Learning Task, or Effort-Expenditure for Rewards Task (EEfRT). Overall, we did not find evidence of bias related to income, race, or educational discouragement.

Certain tasks in this battery may be biased due to age, sex, and education, suggesting that researchers should be cautious when using these particular tasks as part of a battery in neuropsychological research with diverse populations.