# Multitask learning and benchmarking with clinical time series data

## TL;DR

This paper introduces four benchmark tasks for clinical prediction using EHR data from MIMIC-III, and evaluates various neural and linear models to advance machine learning in healthcare.

## Contribution

It provides publicly available benchmarks for clinical prediction and systematically assesses the impact of different modeling strategies on performance.

## Key findings

- Neural models outperform linear baselines on all tasks.
- Deep supervision and multitask training improve model accuracy.
- Architectural modifications tailored to data enhance performance.

## Abstract

Health care is one of the most exciting frontiers in data mining and machine learning. Successful adoption of electronic health records (EHRs) created an explosion in digital clinical data available for analysis, but progress in machine learning for healthcare research has been difficult to measure because of the absence of publicly available benchmark data sets. To address this problem, we propose four clinical prediction benchmarks using data derived from the publicly available Medical Information Mart for Intensive Care (MIMIC-III) database. These tasks cover a range of clinical problems including modeling risk of mortality, forecasting length of stay, detecting physiologic decline, and phenotype classification. We propose strong linear and neural baselines for all four tasks and evaluate the effect of deep supervision, multitask training and data-specific architectural modifications on the performance of neural models.