Data mining for censored time-to-event data: a Bayesian network model for predicting cardiovascular risk from electronic health record data

Abstract

Models for predicting the risk of cardiovascular events based on individual patient characteristics are important tools for managing patient care. Most current and commonly used risk prediction models have been built from carefully selected epidemiological cohorts. However, the homogeneity and limited size of such cohorts restricts the predictive power and generalizability of these risk models to other populations. Electronic health data (EHD) from large health care systems provide access to data on large, heterogeneous, and contemporaneous patient populations. The unique features and challenges of EHD, including missing risk factor information, non-linear relationships between risk factors and cardiovascular event outcomes, and differing effects from different patient subgroups, demand novel machine learning approaches to risk model development. In this paper, we present a machine learning approach based on Bayesian networks trained on EHD to predict the probability of having a cardiovascular event within five years. In such data, event status may be unknown for some individuals as the event time is right-censored due to disenrollment and incomplete follow-up. Since many traditional data mining methods are not well-suited for such data, we describe how to modify both modelling and assessment techniques to account for censored observation times. We show that our approach can lead to better predictive performance than the Cox proportional hazards model (i.e., a regression-based approach commonly used for censored, time-to-event data) or a Bayesian network with ad hoc approaches to right-censoring. Our techniques are motivated by and illustrated on data from a large U.S. Midwestern health care system.

A probabilistic topic model for clinical risk stratification from electronic health records

“…It must mention that nearly all studies along this line have been estimated using a small hand-picked subset of features from highly stratified patient cohorts. As a result, they merely account for a small number of predetermined risk factors and are fragmented where the conclusion only holds under wellcontrolled conditions [13].…”

“…With the widely adoption of electronic health record (EHR) in healthcare organizations, more advanced machine learning and data mining algorithms were introduced into risk stratification [6,13,19,20]. EHR typically contains a diverse set of information types, including patient demographics, symptoms, vital signs, laboratory tests and treatments, etc., which provides a comprehensive source for risk stratification.…”

“…EHR typically contains a diverse set of information types, including patient demographics, symptoms, vital signs, laboratory tests and treatments, etc., which provides a comprehensive source for risk stratification. Many machines learning techniques, such as decision trees [12], Bayesian network [13], and fuzzy inference system [16], have been proposed to explore the huge potentials of EHR data for risk stratification applications. For example, Karaolis et al, carried out data mining analysis using the C4.5 decision tree algorithm to assess the risk factors of coronary heart events [12].…”

“…These models are valuable and have been widely studied in clinical settings, however, they are fragmented where the conclusion only holds under wellcontrolled conditions [13].…”

“…In particular, heterogeneous clinical information for patients (e.g., the patient demographics, laboratory tests results, radiological examination reports, etc.) are recorded in EHRs, which suggests data-centric and hypothesis-free approach from which machine learning techniques can be utilized for risk stratification [13]. Considerable work has been done in this line in order to achieve automatic feature selection from a large volume of EHRs, and significantly improve the accuracy of risk stratification [6,13,21,23].…”