Hospital Report Cards TM Bariatric Surgery Methodology 2007-2008 3
3
© Copyright 2007 Health Grades, Inc. All rights reserved.
May not be reprinted or reproduced without permission from Health Grades, Inc.
Developing HealthGrades Bariatric Surgery Ratings
Developing the HealthGrades Bariatric Surgery ratings involved four steps.
1.
First, the predicted value (predicted complications) was obtained using a logistic regression model
discussed in the next section.
2.
Second, the predicted value was compared with the actual or observed number of complications. Only
hospitals with at least 30 cases across three years of data and at least five cases in the most current year
were included.
3.
Third, a test was conducted to determine whether the difference between the predicted and actual values
was statistically significant. This test was performed to make sure that differences were very unlikely to be
caused by chance alone.
4.
Fourth, a star rating was assigned based upon the outcome of the statistical test.
The following rating system was applied to the data for all procedures and diagnoses:
Best--Actual performance was better than predicted and the difference
was statistically significant.
As Expected--Actual performance was not significantly different from
what was predicted.
Poor--Actual performance was worse than predicted and the difference
was statistically significant.
Statistical Models
Using the list of potential risk factors described above, we used logistic regression to determine to what extent each
one was correlated with the quality measure (complications). A risk factor stayed in the model if it had an odds ratio
greater than one (except clinically relevant procedures, cohort defining principal diagnoses, and some protective
factors as documented in the medical literature were allowed to have an odds ratio less than one) and was also
statistically significant (p<0.05).
Complications were not counted as risk factors as they were considered a result of care received during the
admission. Risk factors are those diagnoses that are the most highly correlated with the outcome studied
(complications). The most highly correlated risk factors are not necessarily those with the highest volume.
(See Appendix C for the Top Five Diagnosis Risk Factors.)
The statistical model was checked for validity and finalized. The final model was highly significant, with a C-statistic of
0.646. This model was then used to estimate the probability of a complication for each patient in the cohort. Patients
were then aggregated for each hospital to obtain the predicted number of complications for each hospital. Statistical
significance tests were performed to identify, by hospital, whether the actual and predicted rates were significantly
different.
