ICU Patient Status Prediction Using Markov Chain Model

Intensive care medicine usually involves making quick decisions based on large amounts of information. In making medical decisions, ICU physicians generally rely on personal experience to make subjective evaluations. It is necessary to continuously monitor the parameters related to the admission and health of ICU patients, and it seems necessary to equip each intensive care unit with a special estimation system. This paper develops a Markov chain model to predict patient conditions in the ICU. The dataset we used for this model is pediatric congenital heart disease ICU patients. The state is determined by prior age, weight, CVP, blood pressure, and urine output. We propose a state-based transition probability matrix using these parameters. Experimental results show that deteriorated patients rarely go to an improvement state. This analysis will be helpful in significantly improving the quality of care in the ICU.