Design and testing of a knowledge-based system for treatment of heart failure patients in a hospital system

The purpose of this study is to determine the best prediction of heart failure outcomes, resulting from two methods—standard epidemiologic analysis using logistic regression and knowledge discovery with supervised learning/data mining using decision tree, nearest neighbor and neural network classifiers. Heart failure was chosen for this study as it exhibits higher prevalence and cost of treatment than most other hospitalized diseases. Findings of this study should be useful for the design of quality improvement initiatives, as particular aspects of patient comorbidity and treatment are found to be associated with mortality. This is also a proof of concept study, considering the feasibility of emerging health informatics methods of data mining in conjunction with or in lieu of traditional logistic regression methods of prediction. Administrative data were obtained from eight hospitals in Iowa, affiliated with the Iowa Health System. These data included insurance claims and care process variables abstracted from all medical records for heart failure patients treated from July 2002 through July 2004. Mortality records originating with the Iowa Department of Public Health were merged with these data to ascertain 30 day mortality, where applicable. These data were pre-classified with Elixhauser comorbidities and Clinical Classification System algorithms, as a means of reducing the diagnoses and procedures into clinically meaningful classes which would be considered as predictors. Stronger predictors of mortality included: age, number of comorbidities and number of prior hospitalizations. Additional predictors included, lack of follow-up planning and comorbid conditions of renal failure, liver disease, arthritis/collagen vascular disease. The classification methods of logistic regression and alternative forms of data mining were evaluated statistically for sensitivity and specificity with ROC curves. Observations were made about their strength in identifying clinical and administrative opportunities for improving clinical care of heart failure patients.