Multitask learning and benchmarking with clinical time series data
This work provides standardized benchmarks for researchers in healthcare machine learning, addressing a key bottleneck in measuring progress, though it is incremental in nature.
The authors tackled the lack of public benchmarks in machine learning for healthcare by proposing four clinical prediction tasks using MIMIC-III data, and they established strong baselines with linear and neural models, achieving competitive performance across tasks such as mortality risk and length-of-stay forecasting.
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.