Predicting Patient Survival in Intensive Care Units (ICUs)
[Summary & Contributions] | [Relevant Publications]
Summary and Contributions
In the US and Europe, score systems for mortality prediction are widely used in intensive care units (ICUs) in hospitals, both to aid in decisions regarding patient care, and to assess the quality of ICUs. However, when used in developing countries, these same score systems fail to make accurate predictions due to differences in demographics. Following a request from the Critical Care Division at St. John’s Medical College and Hospital in Bangalore to help develop new score systems that would make more accurate predictions in the Indian context, we designed a machine learning algorithm to automatically learn a mortality prediction score system of the same form that is already familiar to ICU doctors, but whose parameters are fit to the training data provided. The resulting models, tested on both Indian patient data provided by St. John’s Hospital and other publicly available data, yield significantly better predictions than existing standard models.
Relevant Publications
- Aadirupa Saha, Chandrahas Dewangan, Harikrishna Narasimhan, Sriram Sampath, Shivani Agarwal.
Learning score systems for patient mortality prediction in intensive care units via orthogonal matching pursuit.
In Proceedings of the 13th International Conference on Machine Learning and Applications (ICMLA), 2014.
[pdf]
Shivani Agarwal 