Skip to content

NicolasHu11/Short-Long-Term-Mortality-Prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Short/Long Term Mortality Prediction

  • An initial evaluation of 8 machine learning models
  • Data Science and Health Final Project, Spring 2019
  • Group: Nicolas Hu, Zengtian Deng, Yuankai Qi, Marcelo Lerendegui¶

Overview and Background

Mortality prediction of ICU patients is a crucial task. It is important to accurately predict the mortality of a patient because it benefits both the patient and health care rpofessionals. Also, it helps to improve the health care quality, increase efficiency for hospitals and lower the increasing healthcare cost.

Machine learning models have been applied to many kinds of diseases such as heart failure, breast cancer, kidney disease for mortality prediction. Using the MIMIC-III dataset, we compared several machine learning models that predicts whether a patient will die in both short and long term, 30 days and 1 year respectively. We choose the data within first 24 hour after admission including patients vitals, lab results, risk scores, etc.

30-day mortality is an important feature which hospitals are graded on across the country. It's an indication of the health care quality provided by the hospitals. Thus prediction for 30-day mortality is crucial for hospitals. It helps physicians to decide which methods would be the best for given patients and perhaps provide palliative care.

1-year mortality is crucial for improving the quality of care over time. The patient and his/her physician are working cooperatively for health management. The risk for patients with chronic diseases is changing as time goes. Knowing the possible future scenarios is important to provide high quality, cost-effective medical care.

Presentation

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •