Gaming in risk-adjusted mortality rates: Effect of misclassification of risk factors in the benchmarking of cardiac surgery risk-adjusted mortality rates

详细信息    查看全文

• 作者：Sabrina Siregar ; MD^a ; ^b ; ^{s.siregar@umcutrecht.nl} ; Rolf H.H. Groenwold ; MD ; PhD^b ; Michel I.M. Versteegh ; MD^c ; Luc Noyez ; MD ; PhD^d ; Willem Jan P.P. ter Burg ; MSc^e ; Michiel L. Bots ; MD ; PhD^b ; Yolanda van der Graaf ; MD ; PhD^b ; Lex A. van Herwerden ; MD ; PhD^a
• 关键词：CI ; confidence interval ; EuroSCORE ; European system for cardiac operative risk evaluation ; LVF ; left ventricular function
• 刊名：The Journal of Thoracic and Cardiovascular Surgery
• 出版年：2013
• 出版时间：March, 2013
• 年：2013
• 卷：145
• 期：3
• 页码：781-789
• 全文大小：694 K

文摘

| Figures/TablesFigures/Tables | ReferencesReferences

Objective

Upcoding or undercoding of risk factors could affect the benchmarking of risk-adjusted mortality rates. The aim was to investigate the effect of misclassification of risk factors on the benchmarking of mortality rates after cardiac surgery.

Methods

A prospective cohort was used comprising all adult cardiac surgery patients in all 16 cardiothoracic centers in The Netherlands from January 1, 2007, to December 31, 2009. A random effects model, including the logistic European system for cardiac operative risk evaluation (EuroSCORE) was used to benchmark the in-hospital mortality rates. We simulated upcoding and undercoding of 5 selected variables in the patients from 1 center. These patients were selected randomly (nondifferential misclassification) or by the EuroSCORE (differential misclassification).

Results

In the random patients, substantial misclassification was required to affect benchmarking: a 1.8-fold increase in prevalence of the 4 risk factors changed an underperforming center into an average performing one. Upcoding of 1 variable required even more. When patients with the greatest EuroSCORE were upcoded (ie, differential misclassification), a 1.1-fold increase was sufficient: moderate left ventricular function from 14.2 % to 15.7 % , poor left ventricular function from 8.4 % to 9.3 % , recent myocardial infarction from 7.9 % to 8.6 % , and extracardiac arteriopathy from 9.0 % to 9.8 % .

Conclusions

Benchmarking using risk-adjusted mortality rates can be manipulated by misclassification of the EuroSCORE risk factors. Misclassification of random patients or of single variables will have little effect. However, limited upcoding of multiple risk factors in high-risk patients can greatly influence benchmarking. To minimize ¡°gaming,¡± the prevalence of all risk factors should be carefully monitored.