An evaluation of machine-learning methods for predicting pneumonia mortality

• Published in: Artificial intelligence in medicine Vol. 9; no. 2; pp. 107 - 138
• Main Authors: Cooper, Gregory F., Aliferis, Constantin F., Ambrosino, Richard, Aronis, John, Buchanan, Bruce G., Caruana, Richard, Fine, Michael J., Glymour, Clark, Gordon, Geoffrey, Hanusa, Barbara H., Janosky, Janine E., Meek, Christopher, Mitchell, Tom, Richardson, Thomas, Spirtes, Peter
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.02.1997
• Subjects: Artificial Intelligence; Bayes Theorem; Clinical databases; Computer applications; Computer-based prediction; Databases, Factual; Evaluation Studies as Topic; Expert Systems; Hospitalization; Humans; Logistic Models; Machine learning; Medicine; Neural Networks (Computer); Pneumonia; Pneumonia - mortality; Predictive Value of Tests; Regression Analysis; Sample Size; United States - epidemiology; United States; Clinical databases; Pneumonia; Computer-based prediction; Machine learning
• ISSN: 0933-3657; 1873-2860
• DOI: 10.1016/S0933-3657(96)00367-3

This paper describes the application of eight statistical and machine-learning methods to derive computer models for predicting mortality of hospital patients with pneumonia from their findings at initial presentation. The eight models were each constructed based on 9847 patient cases and they were each evaluated on 4352 additional cases. The primary evaluation metric was the error in predicted survival as a function of the fraction of patients predicted to survive. This metric is useful in assessing a model's potential to assist a clinician in deciding whether to treat a given patient in the hospital or at home. We examined the error rates of the models when predicting that a given fraction of patients will survive. We examined survival fractions between 0.1 and 0.6. Over this range, each model's predictive error rate was within 1% of the error rate of every other model. When predicting that approximately 30% of the patients will survive, all the models have an error rate of less than 1.5%. The models are distinguished more by the number of variables and parameters that they contain than by their error rates; these differences suggest which models may be the most amenable to future implementation as paper-based guidelines.