Detection of Cognitive Impairment From eSAGE Metadata Using Machine Learning.

OBJECTIVE: Using the metadata collected in the digital version of the Self-Administered Gerocognitive Examination (eSAGE), we aim to improve the prediction of mild cognitive impairment (MCI) and dementia (DM) by applying machine learning methods. PATIENTS AND METHODS: A total of 66 patients had a diagnosis of normal cognition (NC), MCI, or DM, and eSAGE scores and metadata were used. eSAGE scores and metadata were obtained. Each eSAGE question was scored and behavioral features (metadata) such as the time spent on each test page, drawing speed, and average stroke length were extracted for each patient. Logistic regression (LR) and gradient boosting models were trained using these features to detect cognitive impairment (CI). Performance was evaluated using 10-fold cross-validation, with accuracy, precision, recall, F1 score, and receiver operating characteristic area under the curve (AUC) score as evaluation metrics. RESULTS: LR with feature selection achieved an AUC of 89.51%, a recall of 87.56%, and an F1 of 85.07% using both behavioral and scoring. LR using scores and metadata also achieved an AUC of 84.00% in detecting MCI from NC, and an AUC of 98.12% in detecting DM from NC. Average stroke length was particularly useful for prediction and when combined with 4 other scoring features, LR achieved an even better AUC of 92.06% in detecting CI. The study shows that eSAGE scores and metadata are predictive of CI. CONCLUSIONS: eSAGE scores and metadata are predictive of CI. With machine learning methods, the metadata could be combined with scores to enable more accurate detection of CI.

Self-administered gerocognitive examination (SAGE) aids early detection of cognitive impairment at primary care provider visits.

Introduction: Current estimates indicate that up to 50-75% of dementia cases are undiagnosed at an early stage when treatments are most effective. Conducting robust accurate cognitive assessments can be time-consuming for providers and difficult to incorporate into a time-limited Primary Care Provider (PCP) visit. We wanted to compare PCP visits with and without using the self-administered SAGE to determine differences in identification rates of new cognitive disorders. Methods: Three hundred patients aged 65-89 without diagnosed cognitive disorders completing a non-acute office visit were enrolled (ClinicalTrials.gov identifier: NCT04063371). Two PCP offices conducted routine visits for 100 consecutive eligible patients each. One office used the SAGE in an additional 100 subjects and asked available informants about cognitive changes over the previous year. Chart reviews were conducted 60 days later. One-way analysis of variance and Fisher exact tests were used to compare the groups and outcomes. Results: When SAGE was utilized, the PCP documented the detection of new cognitive conditions/concerns six times (9% versus 1.5%) as often (p = 0.003). The detection rate was nearly 4-fold for those with cognitively impaired SAGE scores (p = 0.034). Patients having impaired SAGE score and informant concerns were 15-fold as likely to have new cognitive conditions/concerns documented (p = 0.0007). Among providers using SAGE, 86% would recommend SAGE to colleagues. Discussion: SAGE was easily incorporated into PCP visits and significantly increased identification of new cognitive conditions/concerns leading to new diagnoses, treatment, or management changes. The detection rate increased 15-fold for those with impaired SAGE scores combined with informant reports.