PCV55 - Cost-Effectiveness of Different Diagnostic Strategies For Prediction of Length Of Stay In Hospital After Cardiac Surgery

Abstract

Cardiac surgery is sensitive area of surgery which is connected with great risk of prolonged and unplanned hospital stay caused by surgery-related complications. This study was aimed to investigate the cost and effectiveness of the different calculators for prediction of length of stay (LOS) in hospital after the cardiac surgery. We have used logistic regression analysis (LR) and artificial neural network (ANN) in order to get the best classification system capable to predict LOS. The both classification systems included pre and post-operative risk factors. Data of the costs included costs of visiting, ICU wards, ordinary wards, examinations, radiology, disposals, operation, anesthesia, specific medical materials for operations, blood transfusion, laboratory analyses and medication and were collected at the Department for Cardiac Surgery, Military Medical Academy, Belgrade, Serbia. For each calculator we calculated probabilities of correct patient classification, true and false positive and negative values and combined it with associated cost. If model detected patients with high probability for prolonged LOS it will result in better risk patient’s treatment, less unwanted post-operative complication and with lower costs. High false positive error should be connected with false insight about overall good patients’ condition, unnecessary error equipment utilization and with higher costs. Effectiveness of LR model expressed as percent of correctly classify patients was 76% and for ANN model was 86%. Associated costs for patient treatment based on LR and ANN model were 1902.62 euro and 1895.00 euro, respectively. Cost for patients without application of calculators was 1940.00 euro. ANN calculator has been perceived to be cost-effective strategy, since it enables additional effectiveness, along with lower costs compared to LR. ANN calculator was selected as dominant strategy. Both calculators improved patient’s outcomes and reduced costs in comparison to strategy without LOS probability calculation.