Use of Pharmacy Database to Investigate Patterns of Physician Practice as Related to Patient Outcomes

Abstract

Background: It has been observed that physicians vary considerably in their dose prescribing in the cardiac catheterization laboratory. Objective: To investigate the relationship between heparin dose and patient outcome and demonstrate the effectiveness of data mining in investigating variation in physician practice. Methods: The pharmacy database containing information about heparin dose prior to angioplasty was merged with a clinical database containing information about patient weight and length of hospital stay. The data were analyzed using data mining and kernel density estimation. Results: Cardiologists practice differing methods for administering heparin doses. This variability has impact on patient length-of-stay. Patients receiving heparin doses greater than 80 mg/kg stayed in the catheterization lab longer than patients receiving less (p < 0.001). This was independent of the number of vessels treated during angioplasty. Conclusions: Length of stay related to angioplasty is significantly related to physician decisions concerning the heparin dose prescribed. Data mining is an effective tool that can be used to exploit the pharmacy database for research into physician practice.