Author + information
- James V. Freeman, MD, MPH⁎ (, )
- Yongfei Wang, MS,
- Jeptha P. Curtis, MD,
- Paul A. Heidenreich, MD, MS and
- Mark A. Hlatky, MD
- ↵⁎Stanford University School of Medicine, 300 Pasteur Drive, Falk Building, CVRC 5406, Stanford, California 94305-5405
We thank Dr. Navarese and colleagues for their interest in our paper (1). We completely agree that patient clustering by hospital should be accounted for in hierarchical logistic regression analysis. Our model did include a level for clustering by hospital, but we did not state so clearly in the Methods section. We appreciate the opportunity to clarify that the methods used in our study were appropriate.
Dr. Navarese and colleagues suggest that our analysis would have been better if we had analyzed hospital implantable converter-defibrillator (ICD) volume as a continuous variable rather than as quartiles. We agree that use of a continuous variable can be more sensitive in the case of a linear relationship, but point out that analysis of quartiles better accommodates the possibility of a nonlinear relationship without complex or arbitrary division of the data. Furthermore, the analysis of quartiles provides results that are far simpler to interpret than the approaches suggested by Dr. Navarese and colleagues. Because we found an essentially linear inverse relationship between annual ICD implantation volume and complications (Fig. 1), it is likely that this relationship would only have been more statistically significant if we had performed the analysis using hospital volume as a continuous variable.
We believe that the statistical methods used for our analysis were sound and support our conclusion that patients who have an ICD implanted at a high-volume hospital are less likely to have an adverse event associated with the procedure than patients who have an ICD implanted at a low-volume hospital.
- American College of Cardiology Foundation