Authors:
Hynek Kružík
1
;
Jiří Vomlel
2
;
Václav Kratochvíl
2
;
Petr Tůma
1
and
Petr Somol
2
Affiliations:
1
GNOMON Healthcare Solutions s.r.o., Czech Republic
;
2
Institute of Information Theory and Automation, Czech Republic
Keyword(s):
Data mining, Machine learning, Artificial intelligence, Logistic regression, Predictive model, Acute myocardial infarction.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Business Analytics
;
Clinical Problems and Applications
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
Decision Support Systems
;
Enterprise Information Systems
;
Health Information Systems
;
Information Systems Analysis and Specification
;
Knowledge Management
;
Ontologies and the Semantic Web
;
Pattern Recognition and Machine Learning
;
Sensor Networks
;
Signal Processing
;
Society, e-Business and e-Government
;
Soft Computing
;
Web Information Systems and Technologies
Abstract:
We propose and investigate a prediction model of inpatient mortality for patients with myocardial infarction. The model is based on complex clinical data from a hospital information system used in the Czech Republic. The prediction of the outcome is an important risk-adjustment factor for objective measurement of the quality of healthcare; thus it is a very important factor in healthcare quality assessment. For our experiments we studied hospital mortality in acute myocardial infarction, because: (1) this indicator is reliably detectable from available data; (2) treatment of acute myocardial infarction has a significant socio-economic impact; and (3) the prediction of mortality based on admission findings is the subject of many research papers and thus, we have a good benchmark for our experimental results. We considered only variables that convey information about the patient at the time of admission. We selected 21 out of 637 variables and used them as predictors in logistic regre
ssion to form a prediction model for hospital mortality. The achieved prediction accuracy was 85% and the size of the area under the ROC curve was 0.802. The results are based on a relatively small data sample of 486 patient records. Our future work will aim at increasing the accuracy by using a larger data set.
(More)