Skip to content

criticaldatacourse/hst953-2018-team_h

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

2018 HST.953

Prediction Model for Hemodilution Effect among patients with Fluid Resuscitation

Abstract

Background

The administration of intravenous fluid (resuscitation) is one of the most common interventions in intensive care medicine. Decreased concentration of hemoglobin in intravascular plasma volume due to IV fluid infusion is called hemodilution. Clear guidelines do not exist for determination of the expected amount of hemodilution after fluid administration. The main goal of this study is to qualify and identify any correlation in the change in hemoglobin to the amount of fluid administered, and to evaluate also for any other factors that may contribute to hemodilution.

Methods

Using the MIMIC III database (2001 – 2012), a study cohort of patients from 18 – 80 years of age, who didn’t have a diagnosis of internal or external bleeding or any bleeding disorder, and were given normal saline within the first 24 hours of their ICU stay, was formed. Covariates like gender, weight, urine output over 6 hours, creatinine levels at admission, hemoglobin at baseline and after 24 hours, and total amount of fluid administered were collected. Six prediction models were built, viz., linear regression, stepwise linear regression (AIC), random forest, support vector machine, gradient boosting machine, neural network.

Results

Comparing mean square error of all six models, SVM with MSE of 1.02g/dl was finalized as best model. Model performance is better near the median change in hemoglobin among the cohort and has increasing error on both extremes.

Conclusion

Our study has proven that hemodilution exists in the critically ill cohort using normal saline over the first 24 hours, but there is either no simple model to predict the expected drop or we are limited by the data.

Project Repository

Analysis folder - 1. SQL queries and data extraction from MIMIC III dataset in folder named SQL - 2. Two .csv files containing extracted data (everything and final_weights) - 3. Analysis_original markdown file - prediction modelling using non transformed variables - 4. Analysis_transformed markdown file - prediction modelling using transformed variables figure_tables - tables and graphs in final report proposal - project proposal reference - pdf file for references final_report - Final report is submitted here.

About

hst953-2018-team_h created by GitHub Classroom

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •