Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges

Abstract

Risk prediction plays an important role in clinical cardiology research. Traditionally, most risk models have been based on regression models. While useful and robust, these statistical methods are limited to using a small number of predictors which operate in the same way on everyone, and uniformly throughout their range. The purpose of this review is to illustrate the use of machine-learning methods for development of risk prediction models. Typically presented as black box approaches, most machine-learning methods are aimed at solving particular challenges that arise in data analysis that are not well addressed by typical regression approaches. To illustrate these challenges, as well as how different methods can address them, we consider trying to predicting mortality after diagnosis of acute myocardial infarction. We use data derived from our institution's electronic health record and abstract data on 13 regularly measured laboratory markers. We walk through different challenges that arise in modelling these data and then introduce different machine-learning approaches. Finally, we discuss general issues in the application of machine-learning methods including tuning parameters, loss functions, variable importance, and missing data. Overall, this review serves as an introduction for those working on risk modelling to approach the diffuse field of machine learning.

Keywords: Electronic health records; Personalized medicine; Precision medicine; Risk prediction.

Figure 1

One perspective on the intersection of statistical modelling (blue) and machine-learning (green) goals. The figure highlights that while the processes differ the overarching goals are often the same.

Figure 2

Observed non-linearities in the predicted probabilities of death for calcium (A) and haemoglobin (B) on the logistic (i.e. linear) and probability (C/D) scales. Both show a sharp decreasing relationship with death at lower values, which levels off at more moderate values. Models were estimated using cubic splines and were adjusted for age, sex, and race. 95% confidence bands are shown.

Figure 3

Classification and regression trees (A) for predicting acute myocardial infarction. The first split is on minimum CO₂. Splits on different sides of the tree (creatinine on left, sodium on right) indicate potential interactions. Interaction plot (B) between CO₂ and creatinine highlights the differential relationship.

Figure 4

High-level overview of process of applying machine-learning routines to data.