Data‐driven discovery of probable Alzheimer's disease and related dementia subphenotypes using electronic health records

Abstract

Introduction: We sought to assess longitudinal electronic health records (EHRs) using machine learning (ML) methods to computationally derive probable Alzheimer's Disease (AD) and related dementia subphenotypes. Methods: A retrospective analysis of EHR data from a cohort of 7587 patients seen at a large, multi-specialty urban academic medical center in New York was conducted. Subphenotypes were derived using hierarchical clustering from 792 probable AD patients (cases) who had received at least one diagnosis of AD using their clinical data. The other 6795 patients, labeled as controls, were matched on age and gender with the cases and randomly selected in the ratio of 9:1. Prediction models with multiple ML algorithms were trained on this cohort using 5-fold cross-validation. XGBoost was used to rank the variable importance. Results: Four subphenotypes were computationally derived. Subphenotype A (n = 273; 28.2%) had more patients with cardiovascular diseases; subphenotype B (n = 221; 27.9%) had more patients with mental health illnesses, such as depression and anxiety; patients in subphenotype C (n = 183; 23.1%) were overall older (mean (SD) age, 79.5 (5.4) years) and had the most comorbidities including diabetes, cardiovascular diseases, and mental health disorders; and subphenotype D (n = 115; 14.5%) included patients who took antidementia drugs and had sensory problems, such as deafness and hearing impairment. The 0-year prediction model for AD risk achieved an area under the receiver operating curve (AUC) of 0.764

Identifying and evaluating clinical subtypes of Alzheimer’s disease in care electronic health records using unsupervised machine learning

“…We observed a recurring cluster enriched for mental health disorders in three out of four clustering approaches. These findings highlight a clinically distinct cluster of AD patients that has been found in previous research [ 9 ] and can be a target for clinical intervention and further research. Future research should examine the best way to pick a cluster method and evaluate the results.…”

“…They had a faster rate of progression—roughly 3 times as fast as the other clusters, which may be driven by the earlier onset of disease. A cluster with high mental health issues was also found in a previous study subtyping AD and EHR [ 9 ], and a subtype of early onset mostly female patients was found in another [ 10 ]. The latter study also suggested that in early onset cases prodromal signs of cognitive decline can be misdiagnosed as depression.…”

“…A CVD cluster was found in three out of four of the methods employed, a finding seen in one other study subtyping AD using EHR [ 9 ]. Interestingly, patients with hypertension did not form part of the CVD cluster, with the two clusters having differing outcomes, suggesting that having hypertension or hypertension treatment results in a different AD profile than that of other CVD patients [ 60 ].…”

“…Other studies have used EHR to identify subtypes of AD. Xu et al [ 9 ] used data from the multi-specialty urban academic medical center and hierarchical clustering and found 4 subtypes: one with patients who had higher proportions of cardiovascular disease, another subtype with higher prevalence of mental health conditions, a third subtype of multimorbid patients with later onset and a final subtype of patients who took anti-dementia drugs [ 9 ]. A second study used a representation learning model and hierarchical clustering on Mount Sinai Health System data and found 3 subtypes: an early onset mostly female cluster, a late onset cluster with mild neuropsychiatric symptoms and cerebrovascular disease and finally a cluster with mild to moderate dementia symptoms [ 10 ].…”

The Impact of Disease Registries on Advancing Knowledge and Understanding of Dementia Globally

“…NACC data investigation showed that sleep disturbance, depression as well as APOE genotype are associated with subsequent diagnosis of AD dementia (Burke et al, 2016). Overall, mental health disorders are among the strong predictors of AD diagnosis, so enhancing individuals' mental health might decrease the risk of AD developing later on (Xu J. et al, 2020). Further exploration of registries data would help identifying more risk factors and how they increase the risk of dementia and hence, determining the most appropriate strategies to avoid them.…”