Story memory in Alzheimer’s disease research: Concordance between logical memory and craft story 21 in a demographically diverse aging cohort

Abstract Background Episodic memory loss, the hallmark symptom of Alzheimer’s Disease (AD), is frequently quantified by performance on story memory tasks. The National Alzheimer’s Coordinating Center (NACC) Uniform Data Set (UDS) previously included Logical Memory (LM) as a story memory measure due to its widespread use in the field. Version 3 of the UDS shifted towards nonproprietary neuropsychological tests, hence the Craft Story 21 (CS) was introduced to replace LM. The concordance between LM and CS is poorly understood in demographically diverse older adults, possibly jeopardizing the integrity of longitudinal data in aging cohorts. Method Einstein Aging Study participants (subsample, n = 300; Mage = 76.6; Myears education = 14.9; 66.3% women; 43.3% Non‐Hispanic White) completed measures from the UDS test battery, including two story measures of immediate recall (LM1 and CS1) and delayed recall (LM2 CS2). Classification as cognitively normal (CN; n = 201) or mild cognitive impairment (MCI; n = 98) was based on Jak/Bondi criteria. Analyses included correlations, linear regression, and equipercentile equating at baseline to characterize the relationship between LM and CS. Multivariate linear mixed effects (LME) models (for immediate and delay) explored LM and CS performance across follow‐up ‘waves’ of longitudinal data, controlling for gender, ethnicity, education level, baseline depressive symptoms, and baseline age, and were further stratified by baseline cognitive status (CN vs MCI). Result LM1 and CS1 (r = 0.54) and LM2 and CS2 (r = 0.56) were moderately correlated, although the correlations were weaker in our sample than those previously reported by NACC (r = 0.73 and r = 0.77 respectively). Education, age, and depressive symptoms were significantly associated with story memory performance at baseline. Multivariate LME revealed significant interaction effects of follow‐up waves, depressive symptoms, and gender on story memory. These results were consistent regardless of baseline cognitive status. Conclusion While LM and CS are correlated they are not replicate measures. Our findings underscore the importance of taking into account key factors that may influence performance on story recall tests, including gender, depressive symptoms, and practice effects. Furthermore, the distinct content of the LM and CS stories may differentially impact learning and recall. These findings may aid in decision making about test selection in a given clinical or research sample.