Mortality risk assessment for ICU patients using logistic regression

Prediction of outcome for patients in Intensive Care Unit (ICU) is of great interest since early 1980s. Various techniques had been proposed to evade this issue. Using Physionet/CinC Challenge 2012 data set we have identified maximum, mean and minimum as potential features extracted from the parameters measured during patients stay of 48hrs at ICU to accurately predict in-hospital mortality risk. The study was done with adult patients who were admitted for a wide variety of reasons to Coronary Care Unit, Cardiac Surgery Recovery Unit, Medical ICU, Surgical ICU. The proposed risk prediction model used a logistic regression technique for assessing the probability of mortality based on the selected features. The technique shows significant accuracy on test data set-c with final event 1 score: 0.45128, event 2 score: 45.0101 and ranked within top 10 for both the events.