Causal Inference for Cardiology in the Age of Electronic Health Records

• Activity Number: 187
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): David Shilane*+
• Companies: Stanford University
• Keywords: causal inference ; observational studies ; cardiology ; biostatistics ; electronic medical records ; health research and policy

Abstract

The advent of electronic health records has greatly improved our ability to assess the quality, safety, and cost of medical treatments. In collaboration with a team of Stanford cardiologists and Kaiser Permanente's Division of Research, I have utilized a variety of electronic databases to provide statistical guidance in medical studies. Kaiser's cardiology records provide unprecedented patient information that integrates clinical, hospital, and pharmacy records with long--term followup. These registries provide larger sample sizes, richer covariate profiles, and greater population representation than randomized clinical trials at significantly reduced cost. However, because the data do not arise from designed experiments, causal inference methods are necessary to account for issues of treatment selection bias. I will present case studies that examine the safety of spironolactone treatments, provide a comparative cost--benefit analysis of imaging tests, and assess the factors impacting the prescription of, patient adherence to, and effectiveness of medications for ischemic heart disease. Within these studies, I will compare alternative methods for causal inference.