Maskininlärning för prediktion av dödlighet och hospitalisering hos hjärtsviktspatienter
(2025) EEML05 20251Department of Biomedical Engineering
- Abstract
- Heart failure is one of the most common causes of hospitalization in Sweden, and 30\% of all heart failure patients need unplanned readmissions within three months, creating a burden on the healthcare system. This project was conducted within the healthcare model "Hospital at Home", where there is currently no systematic follow-up method of heart failure patients. By monitoring those with the highest risk of readmission in real time, interventions could be made at the right time to reduce this risk. AI is considered to have great potential in this area. The aim of this project was to investigate the possibility of using machine learning to predict a patient’s risk for four different outcomes: death, death within one year, hospitalization,... (More)
- Heart failure is one of the most common causes of hospitalization in Sweden, and 30\% of all heart failure patients need unplanned readmissions within three months, creating a burden on the healthcare system. This project was conducted within the healthcare model "Hospital at Home", where there is currently no systematic follow-up method of heart failure patients. By monitoring those with the highest risk of readmission in real time, interventions could be made at the right time to reduce this risk. AI is considered to have great potential in this area. The aim of this project was to investigate the possibility of using machine learning to predict a patient’s risk for four different outcomes: death, death within one year, hospitalization, and hospitalization within 30 days.
To explore this, three different machine learning models were selected based on their explainability. These were logistic regression, support vector machine (SVM), and XGBoost. The models were trained using a retrospective database of 531 heart failure patients, with variables including demographics, cognitive factors, and clinical variables. Evaluation of the models was based on accuracy, AUC, and F1-score.
The results showed that logistic regression made the best predictions for death and hospitalization, while SVM performed best at predicting death within one year and hospitalization within 30 days. The models identified, among other factors, age and the result on the cognitive test MOCA as important variables for predicting these outcomes. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9205836
- author
- Österlund, Emilia LU and Andersson, Klara LU
- supervisor
-
- Martin Stridh LU
- Amra Jujic LU
- Marcus Ohlsson LU
- John Molvin LU
- Hannes Holm LU
- organization
- alternative title
- Machine Learning for Prediction of Mortality and Hospitalization among Heart Failure Patients
- course
- EEML05 20251
- year
- 2025
- type
- M2 - Bachelor Degree
- subject
- language
- Swedish
- id
- 9205836
- date added to LUP
- 2025-07-01 09:34:10
- date last changed
- 2025-07-01 09:34:10
@misc{9205836,
abstract = {{Heart failure is one of the most common causes of hospitalization in Sweden, and 30\% of all heart failure patients need unplanned readmissions within three months, creating a burden on the healthcare system. This project was conducted within the healthcare model "Hospital at Home", where there is currently no systematic follow-up method of heart failure patients. By monitoring those with the highest risk of readmission in real time, interventions could be made at the right time to reduce this risk. AI is considered to have great potential in this area. The aim of this project was to investigate the possibility of using machine learning to predict a patient’s risk for four different outcomes: death, death within one year, hospitalization, and hospitalization within 30 days.
To explore this, three different machine learning models were selected based on their explainability. These were logistic regression, support vector machine (SVM), and XGBoost. The models were trained using a retrospective database of 531 heart failure patients, with variables including demographics, cognitive factors, and clinical variables. Evaluation of the models was based on accuracy, AUC, and F1-score.
The results showed that logistic regression made the best predictions for death and hospitalization, while SVM performed best at predicting death within one year and hospitalization within 30 days. The models identified, among other factors, age and the result on the cognitive test MOCA as important variables for predicting these outcomes.}},
author = {{Österlund, Emilia and Andersson, Klara}},
language = {{swe}},
note = {{Student Paper}},
title = {{Maskininlärning för prediktion av dödlighet och hospitalisering hos hjärtsviktspatienter}},
year = {{2025}},
}