| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2005;80:2119-2120
© 2005 The Society of Thoracic Surgeons
Research Department, Medical City Dallas Hospital, 7777 Forest Ln, Suite C-740, Dallas, TX75230
(Email: morley.herbert{at}hcahealthcare.com).
Current thinking is that the informed patient is a better consumer of medical services. In an effort to help the consumer make better decisions related to their care, many groups are offering advice on choosing the best doctors or hospitals. Some groups such as Healthgrades.com or the Texas Health Care Information Council develop algorithms that involve using data from administrative databases, then passing it through a black-box algorithm to generate their rating system.
Others use a more simplistic system tied to factors such as reputation and patient volume. For the US News and World Report magazine and USNews.com, volume contributes one third of the final score. Other groups such as the Leapfrog Group use only volume for their recommendations. Using a single univariate analysis, they have decided that the best hospitals have a volume of 450 cases per year in cardiac surgery.
A number of recent publications have challenged the conclusion that using volume alone is a useful measure for determining the best hospitals. This article by Welke and colleagues [1] suggests that using only volume to select the best hospitals (based on mortality rate) is only marginally better than randomly selecting hospitals.
Using the Centers for Medicare Services Medicare Provider and Analysis Review (MEDPAR) part A public files, the authors created a dataset of nearly 1 million patients who underwent coronary artery bypass grafting. They proceeded to look at the relationship between volumes and outcomes (defined as mortality). After dividing the data into groups based on hospital volume using quintiles, deciles, high and low volume (based on Leapfrog criteria), and volume as a continuous variable, they analyzed the mortality rates measured for each group. Although there was a rough correlation of declining 30-day mortality with increasing volume within any group, the range of values nearly totally overlapped the adjacent groups. Further, the "c" statistic, which measures the ability of the model to discriminate accurately, was within the range of 0.51 to 0.53, essentially no better than randomly selecting hospitals (randomness has the value c = 0.50). This poor predictability held even looking at 7, 14, 30, 90, and 365-day mortality rates. By comparison, other more complex multivariate models produce a "c" statistic of 0.72 to 0.81.
As the authors point out, they are not alone in providing evidence of a poor discriminatory ability of volume alone. Peterson's analysis using the The Society of Thoracic Surgeons database and Rathore's study of the National Inpatient Sample showed similarly poor results using volume.
The authors note that the range of mortality rates among hospitals with equal volumes is large. I would add that there is often a significant range in mortality rates among surgeons at any particular hospital. Thus if it were to be useful, picking a hospital based on volume alone would have to be expanded to using individual surgeon mortality rates, and then correcting them for the predicted risk of their caseload. Looking only at gross volume for a facility fails to do any of this.
Advising the public to use raw case volume to select the best hospital to have coronary artery bypass graft surgery may be misleading, and volume numbers should probably be reserved for use in multivariate analyses.
 |
References |
|---|
 Top 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 |
