Modeling Missing Data in Clinical Time Series with RNNs

TL;DR

This paper presents a method using RNNs to handle missing data in clinical time series by leveraging missingness patterns as features, improving diagnosis prediction accuracy in pediatric ICU data.

Contribution

The study introduces a simple strategy for RNNs to utilize missing data patterns directly, outperforming traditional imputation methods in clinical time series classification.

Findings

RNNs using missingness indicators outperform imputation-based methods.

Missingness patterns alone can be highly predictive for certain diagnoses.

Treating missing data as features enhances predictive performance in clinical applications.

Abstract

We demonstrate a simple strategy to cope with missing data in sequential inputs, addressing the task of multilabel classification of diagnoses given clinical time series. Collected from the pediatric intensive care unit (PICU) at Children's Hospital Los Angeles, our data consists of multivariate time series of observations. The measurements are irregularly spaced, leading to missingness patterns in temporally discretized sequences. While these artifacts are typically handled by imputation, we achieve superior predictive performance by treating the artifacts as features. Unlike linear models, recurrent neural networks can realize this improvement using only simple binary indicators of missingness. For linear models, we show an alternative strategy to capture this signal. Training models on missingness patterns only, we show that for some diseases, what tests are run can be as predictive as the results themselves.