| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Ann Thorac Surg 2004;77:1239-1240
© 2004 The Society of Thoracic Surgeons
Department of Cardiovascular Thoracic Surgery, Albert Einstein College of Medicine, Montefiore Medical Center, 3400 Bainbridge Ave, 5th Floor, Suite 5B, Bronx, NY 10467, USA
e-mail: jgold{at}montefiore.org
Nilsson and colleagues describe a sophisticated statistical comparison of two risk algorithms for predicting outcome in patients undergoing coronary bypass surgery. The EuroSCORE and The Society of Thoracic Surgeons National Database risk stratification systems are described, compared, and reviewed. In a series of nearly 4500 single institution isolated coronary bypass surgery patients, the Hosmer-Lemeshop goodness of fit test was utilized. The clinical results were excellent, and good predictive accuracy was determined for both models. The authors calculated that the area under the receiver operating characteristic curve (ROC) was larger for the Euroscore than for the STS National Database, thus concluding significantly enhanced power of the EUROscore to predict 30-day mortality.
While this is an important type of comparative study, moving us closer to global database analysis, it underscores many of the challenges associated with attempting to do so. In particular, the fine differences in the data fields, differences in variable definitions, and a difference in the vintage of the statistical algorithms as they evolve over time, given the older age of the STS algorithm, all come into play in this type of comparison. It also raises questions regarding the management of unidentified noncardiac risk factors, management of incomplete data fields, surgical procedure selection criteria, and the differences inherent in single institution and multiple institution comparison studies, all of which might influence predicted cardiac surgical outcomes and therefore, the results of the comparison of these two highly regarded models. Indeed, the mere process of comparing the receiver operating characteristic curves (ROCs) in a statistically valid fashion to propose superiority of one model or the other is highly controversial.
These analysis and comparison challenges are by no means unique to the EUROscore and the STS National Database [1]. Similar difficulties arise when almost any comparison of the accuracy and predictive power of large data bases are attempted. These issues have been resolved to some extent previously by either eliminating noncomparable features of the two systems compared from the analysis, ignoring the noncomparable features including them in the analysis, or working to standardize as many of the data fields, definitions and analysis systems as possible.
Although the circumference, diameter and surface area of our planet are reportedly fixed, we in clinical medicine would all agree that the planet is definitely shrinking, particularly in our highly defined field of medicine. International scientific collaboration and sharing of clinical expertise have been the long term goals of virtually all of our respected leaders and institutions of higher learning. The time is right to set about the task of analyzing the strengths of the major data collection systems across the globe, and developing consensus on as many standards as possible. While not an easy task, with much history and personal investment at risk, the long term results will truly take us where we wish to go.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
