Skip to content

Performance of prediction models for opportunities for improvement in trauma care across patient cohorts

Notifications You must be signed in to change notification settings

noacs-io/ml-performance-cohorts

Repository files navigation

Performance of prediction models for opportunities for improvement in trauma care among women compared to men and across clinical trauma cohorts

This repository documents our research project on the performance of prediction models for opportunities for improvement in trauma care among women compared to men and across clinical trauma cohorts.

Summary

Introduction

Trauma, the clinical entity composed of physical injury and the body’s associated response, is a leading cause of mortality and morbidity worldwide. A cornerstone of trauma quality improvement programs is multidisciplinary mortality and morbidity review to identify opportunities for improvement and implement corrective actions. Examples of opportunities for improvement may include lack of resources and management errors. We have developed machine learning based prediction models that outperform the conventional methods to predict opportunities for improvement. It is important that these models have similar performance in different subgroups.

Aim

To assess the difference in performance of prediction models for opportunities of improvement in trauma care among women compared to men and across clinical trauma cohorts.

Methods

This is a registry based cohort study using data from the trauma registry and trauma care quality database at the Karolinska University Hospital in Solna. The trauma registry includes about 12000 patients treated between 2012 and 2022. The trauma care quality database is a subset of the trauma registry and includes about 6000 patients selected for review between 2014 and 2022. The outcome is opportunities for improvement, as identified by the multidisciplinary review board and recorded in the trauma care quality database. This project will link the two databases and assess the performance of our developed machine learning based prediction models in women, men, and and different clinical trauma cohorts respectively. The perfomance in terms of discrimination and calibration will then be compared between men and women and between the different trauma cohorts. A 5% significance level and 95% confidence levels will be used.

About

Performance of prediction models for opportunities for improvement in trauma care across patient cohorts

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published