Adapting machine learning techniques to censored time-to-event health record data: A general-purpose approach using inverse probability of censoring weighting

Abstract

Models for predicting the probability of experiencing various health outcomes or adverse events over a certain time frame (e.g., having a heart attack in the next 5 years) based on individual patient characteristics are important tools for managing patient care. Electronic health data (EHD) are appealing sources of training data because they provide access to large amounts of rich individual-level data from present-day patient populations. However, because EHD are derived by extracting information from administrative and clinical databases, some fraction of subjects will not be under observation for the entire time frame over which one wants to make predictions; this loss to follow-up is often due to disenrollment from the health system. For subjects without complete follow-up, whether or not they experienced the adverse event is unknown, and in statistical terms the event time is said to be right-censored. Most machine learning approaches to the problem have been relatively ad hoc; for example, common approaches for handling observations in which the event status is unknown include 1) discarding those observations, 2) treating them as non-events, 3) splitting those observations into two observations: one where the event occurs and one where the event does not. In this paper, we present a general-purpose approach to account for right-censored outcomes using inverse probability of censoring weighting (IPCW). We illustrate how IPCW can easily be incorporated into a number of existing machine learning algorithms used to mine big health care data including Bayesian networks, k-nearest neighbors, decision trees, and generalized additive models. We then show that our approach leads to better calibrated predictions than the three ad hoc approaches when applied to predicting the 5-year risk of experiencing a cardiovascular adverse event, using EHD from a large U.S. Midwestern healthcare system.

Modeling the Factors Associated with Mortality in Patients with Breast Cancer: A Machine Learning Approach

“…This approach can be used with many machine learning methods. 11 Interestingly, in our data, the GAM using IPCW approach had the best performance which had the highest accuracy, sensitivity and negative predictive value among the investigated machine learning methods. So…”

A Systematic Review on Healthcare Analytics: Application and Theoretical Perspective of Data Mining

“…Elimination and exclusion of the censored data create bias in prediction results. To address the censorship of the data in their study on CVD event risk prediction after time, two studies [ 43 , 44 ] used Inverse Probability Censoring Weighting (IPCW). IPCW is a pre-processing step used to calculate the weights on data which are later classified using Bayesian Network.…”

Evaluating the effect of right-censored endpoint transformation for dimensionality reduction of radiomic features of oropharyngeal cancer patients

“…We used inverse probability of censoring weighting (IPCW) to extend machine learning methods for survival analysis [9].…”

“…It is well-known that machine learning approaches non-cognizant to right-censoring lead to poorly calibrated risk prediction as they fail to accurately implement risk over the entire follow-up period (but discrimination is relatively unaffected) [9]. However, it is unclear what the effect of data pre-preprocessing approaches to handle right censoring have on algorithms for feature selection.…”

“…2. Assessment of the reduced feature set upon subsequent predictive model specification, using a library of established and novel risk prediction approaches (Cox proportional hazard, random forest, random survival forest, logistic regression, logistic-elastic net) which have been adapted to right-censored outcomes using inverse probability of censoring weights [9].…”