Generalizable Predictions of ICU Readmissions & Deaths Using Machine Learning
Intensive Care Unit (ICU) readmissions cause serious problems for hospitals, patients, and insurers. Readmissions have been shown to have longer length of stay, higher healthcare costs for both patient and insurer, as well as increased mortality rates. While many approaches have been outlined in recent research, many avenues remain unexplored as the availability of big data in the healthcare field continues to expand. Our capstone group seeks to use various machine learning techniques to create a model that can accurately predict which ICU patients are at risk of having negative outcomes (including readmission and death). This model supports existing research in the field by:
a) Focusing on building interpretable models in order to help providers understand why a patient was flagged for risk of readmission or death
b) Seeking to understand the transferability of different approaches using cross validation from two different datasets.
See the full description of our project and results on our website: https://natasha-flowers.github.io/icu_outcomes_capstone/