Predicting cognitive resiliency in older women using machine learning: Epidemiology / Risk and protective factors in MCI and dementia. (7th December 2020)

• Record Type: Journal Article
• Authors: Casanova, Ramon; Gaussoin, Sarah A; Wallace, Robert B; Baker, Laura D; Sachs, Bonnie C; Justice, Jamie; Chen, Jiu‐Chiuan; Manson, JoAnn E; Henderson, Victor; Whitsel, Eric A; Hayden, Kathleen M; Rapp, Stephen R
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 10
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.041137
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Identification of factors that help preserve cognitive function in late life could facilitate interventions that can improve the lives of millions of people as they age. However, the large number of potential factors associated with cognitive resiliency pose an analytical challenge. Machine learning can be used to discover complex functional relationships in high dimensional data. We used data from the Women's Health Initiative Memory Study (WHIMS) and Random Forests (RF) methods to investigate predictors of cognitive resiliency in 80+ year old women from a set of demographic, biological, behavioral, social, and psychological variables. Method: We included WHIMS women who were at least 80 years old as of 01/01/18, and had at least one cognitive assessment following their 80 th birthday. We created two groups of participants: a cognitively resilient group and an impaired group of participants. The cognitively resilient group (N=205) included participants: (a) not previously adjudicated to have mild cognitive impairment (MCI) or probable dementia; (b) had a score > 39 on the most recent administration of the TICSm and (c) the average of scores in previous visits>=39. The cognitively impaired group (N=176) included participants: (a) adjudicated with probable dementia within the first 14 years of follow‐up or; (b) who had at least two classifications of MCI within the first 14 years of follow‐up and (c) whose last TICS score was<31. RF classification was used to identify discriminative predictors of these two groups based on 67 baseline variables available in WHIMS. Analyses used the randomForestSRC R library. The minimal depth index (MDI) was used to determine predictors' relevance. Result: Discrimination between the resilient and impaired groups based on area under the curve was 0.80(95% CI‐0.76‐0.85). Figure 1 shows that according to the MDI values the most relevant predictors were age, self‐reported forgetfulness, physical function, optimism, hemoglobin, glucose, sleep disturbances, education, systolic pressure and depression. Conclusion: We investigated discrimination of cognitively resilient older women from women who develop cognitive impairment using machine learning applied to a large set of predictors. The potential contributors to cognitive resilience include demographic, psychological, physical, metabolic and vascular factors suggesting multiple influences contributing to cognitive resiliency.