Accounting for practice effects improves stability of MCI diagnosis and uncovers new impairments

Abstract

AbstractBackgroundWe previously showed that failure to account for practice effects (PEs) delays detection of conversion from unimpaired to mild cognitive impairment (MCI) and reduces diagnostic accuracy based on biomarker positivity data. Even among those who decline, PE can mask steeper declines by inflating cognitive scores. Within MCI samples, PEs may increase reversion rates and impede detection of new impairments. Here, within a baseline MCI sample, we evaluated how PEs affect prevalence, reversion rates, and progression from single‐domain to multi‐domain MCI at follow‐up.MethodWe examined 347 baseline Alzheimer’s Disease Neuroimaging Initiative MCI participants (mean age=73.2; SD=7.4). We used Jak/Bondi diagnostic criteria (6 tests). We also calculated PEs on a cognitive composite (ADAS‐cog 11) and a cognitive screener (MMSE). We identified participants who were demographically matched to returning subjects at their 1‐year follow‐up. The only difference was that one group was tested once and the other had completed prior testing. We subtracted PEs—based on comparing these groups—from follow‐up scores and recalculated MCI diagnoses.ResultPEs were absent or small for 5 tests (Cohen’s d=.00‐.06) and medium for 1 (d=.50). There was a PE on the MMSE (Cohen’s d =.16), but none was found on the ADAS‐cog 11. Accounting for PEs increased MCI prevalence at follow‐up by 6% (252 vs 268), and reduced reversion rate by 15% (95 vs 79). PEs also masked progression from single‐domain to multi‐domain MCI: 24 vs 36 (+50.0% after accounting for PEs).ConclusionAlthough PEs were small, accounting for them in baseline MCI participants after 1‐year, increased MCI prevalence and progression from single‐domain to multi‐domain, and decreased reversion rates. PEs may be masked within a composite measures, but were present on a short cognitive screener. Failure to consider PEs reduces accuracy regarding conclusions about MCI stability/change and weakens our ability to determine when impairment spreads to new domains.