Using topological data analysis and pseudo time series to infer temporal phenotypes from electronic health records.

Arianna Dagliati,Riccardo Bellazzi,Niels Peek,Allan Tucker,John H Holmes,Nophar Geifman,Seyed Erfan Sajjadi,Lucia Sacchi

doi:10.1016/j.artmed.2020.101930

Abstract

Temporal phenotyping enables clinicians to better understand observable characteristics of a disease as it progresses. Modelling disease progression that captures interactions between phenotypes is inherently challenging. Temporal models that capture change in disease over time can identify the key features that characterize disease subtypes that underpin these trajectories. These models will enable clinicians to identify early warning signs of progression in specific sub-types and therefore to make informed decisions tailored to individual patients. In this paper, we explore two approaches to building temporal phenotypes based on the topology of data: topological data analysis and pseudo time-series. Using type 2 diabetes data, we show that the topological data analysis approach is able to identify disease trajectories and that pseudo time-series can infer a state space model characterized by transitions between hidden states that represent distinct temporal phenotypes. Both approaches highlight lipid profiles as key factors in distinguishing the phenotypes.

Highlights

Electronic temporal phenotyping is the identification of clinically meaningful event sequences from patient data that have been collected over time
We propose the joint use of pseudo time-series with Topological Data Analysis (TDA) in order to illustrate the temporal characteristics of disease progression, so that disease trajectories can be constructed from the data using the topological model as a guide
We focus on microvascular complications of type 2 diabetes mellitus (T2DM) and explore both TDA and pseudo time-series (PTS) for building different trajectories from health record data in order to better understand the temporal phenotypes that can identify different sub-phenotypes of T2DM

Summary

Introduction

Electronic temporal phenotyping is the identification of clinically meaningful event sequences from patient data that have been collected over time. The identification of temporal phenotypes that are specific to subgroups of patients can assist researchers in identifying useful cohorts and could be used to generate hypotheses for precision medicine research. What is more, they help experts to better understand the disease in question and how it progresses over time, while ensuring that existing guidelines and care plans are appropriate. Unlike most previous research that is based on extracting phenotypes from longitudinal electronic health records, we are interested in the construction of temporal phenotypes based on the overall structure of data (that is not necessarily longitudinal) and the identification of realistic trajectories through this structure in time

Methods

Results

Conclusion