Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 198 - Innovations in Patient-Focused Clinical Trials
Type: Contributed
Date/Time: Monday, August 8, 2022 : 2:00 PM to 3:50 PM
Sponsor: Biopharmaceutical Section
Abstract #322965
Title: A Comparison of K-Means and Consensus Clustering Algorithms in the Analysis of Wearable and Biosensor Data
Author(s): Vanja Vlajnic* and Steve Simske
Companies: Bayer and Colorado State University
Keywords: k-means clustering; consensus clustering; biosensors; clinical trials; heart failure; wearables

Wearable and biosensor data have become more common within the clinical trial space as they provide information complementary to data collected as part of traditional clinical trials. However, the most appropriate analytical approaches for examining baseline clusters of patients is not clear. In this case study, wearable and biosensor data from a recent heart failure trial were analyzed using both a k-means and consensus clustering approach. For k-means clustering, the R package "NbClust" was utilized. Here, 30 indices were used to identify the appropriate value of K as 3. For consensus clustering, the R package "M3C" was utilized and, specifically, a partitioning around medoids (PAM) clustering algorithm with Euclidean distance and an entropy objective function were used. The PAM method identified the appropriate value of K as 4. The clusters were examined from each algorithm across their baseline characteristics, medical history, and activity duration and intensity. Unique clinical phenotypes were identified and the utility of these approaches to identify possible clusters of heterogeneous patient populations is discussed.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2022 program