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.
Highlights
In the United States alone, each year over 30 million patients visit hospitals1, 83% of which use an electronic health record (EHR) system[2]. This trove of digital clinical data presents a significant opportunity for data mining and machine learning researchers to solve pressing health care problems, such as early triage and risk assessment, prediction of physiologic decompensation, identification of high cost patients, and characterization of complex, multi-system diseases[3,4,5,6,7]
In this paper we proposed four standardized benchmarks for machine learning researchers interested in clinical data problems, including in-hospital mortality, decompensation, length-of-stay, and phenotype prediction
Our benchmark data set is similar to other MIMIC-III patient cohorts described in machine learning publications but makes use of a larger number of patients and is immediately accessible to other researchers who wish to replicate our experiments or build upon our work
Summary
In the United States alone, each year over 30 million patients visit hospitals1, 83% of which use an electronic health record (EHR) system[2] This trove of digital clinical data presents a significant opportunity for data mining and machine learning researchers to solve pressing health care problems, such as early triage and risk assessment, prediction of physiologic decompensation, identification of high cost patients, and characterization of complex, multi-system diseases[3,4,5,6,7]. The main challenge is the absence of widely accepted benchmarks to evaluate competing models. Such benchmarks accelerate progress in machine learning by focusing the community and facilitating reproducibility and competition. Public benchmarks lower the barrier to entry by enabling new researchers to start without having to negotiate data access or recruit expert collaborators
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
