Eur J Cardiothorac Surg 2009;35:379-380. doi:10.1016/j.ejcts.2008.10.031
Copyright © 2009, European Association for Cardio-thoracic Surgery. Published by Elsevier. All rights reserved.
Italian hospital mortality risk model vs additive and logistic EuroSCORE in coronary operations
Marco Ranucci*
Department of Cardiothoracic-vascular Anesthesia and Intensive Care, IRCCS Policlinico S.Donato, Via Morandi 30, 20097 San Donato Milanese, Milan, Italy
Received 29 August 2008;
accepted 27 October 2008.
* Corresponding author. Tel.: +39 02 52774320; fax: +39 02 55602262. (Email: cardioanestesia{at}virgilio.it).
Key Words: Cardiac surgery • Risk assessment • Calibration
D’ Errigo and co-workers [1] recently published an interesting article, where they propose an Italian CABG Model (ICM) for assessing the mortality risk in coronary patients and evaluate its performance compared to the additive EuroSCORE and the logistic EuroSCORE. They concluded that the ICM had a better performance than both the additive and logistic EuroSCORE. This last model demonstrated a poor calibration.
The authors are to be congratulated for addressing this relevant topic. However, I have some concerns about the statistical methodology of their study.
- (a) A database of 34,130 isolated coronary operations was used to create and validate their model. The authors randomly split the population into a first series of 17,231 used to build the model, and a validation series of 17,079 patients.
A validation series is generally created at different sites or in different time-periods. Parsonnet and co-workers [2] applied an external validation using data from other hospitals; Higgins and co-workers [3] used a validation series in their own hospital, based on data collected after the first series; two transfusion risk models used again two consecutive series in the same hospital [4] or an external validation in different hospitals [5].The reasons for this are quite obvious: testing a score under different clinical conditions.
D’Errigo and co-workers randomly attributed the patients to the test or the validation series. Following a correct randomization, it is expected that each hospital contributed to the test or the validation series with the same amount of patients collected in the same period of time. Therefore, the authors performed something similar to an internal validation for the ICM, compared to an external validation for the additive and logistic EuroSCORE. This methodological bias may justify the better performance of the ICM.
- (b) There are some discrepancies in reporting numbers. In the abstract, it is stated that ‘the logistic EuroSCORE shows a significant Hosmer-Lemeshow test (
, p < 0.0001)’. In the text, the same significant 
2 value is attributed not to the Hosmer-Lemeshow test within the EuroSCORE model, but to the difference between the receiver operating characteristic (ROC) areas of the ICM vs the logistic EuroSCORE model. In Table 4, the Hosmer-Lemeshow 2 value for the logistic EuroSCORE model is reported as 798.756. It is frankly difficult to understand and interpret these figures.
- (c) The authors suggest an adjustment for the logistic EuroSCORE, by simply applying a 0.4 multiplier. After this, they claim for non-significant Hosmer-Lemeshow values, although they do not report them. Applying a fixed adjustment to the individual logistic EuroSCORE values simply produces a downward displacement of the logistic regression curve (same β value, constant = 0.4 times unadjusted constant). The Hosmer-Lemeshow 2 remains the same, as well as the ROC area. The adjusted model may appear more accurate, but actually has the same calibration and discrimination power of the unadjusted model.
I think that the conclusions of the authors are not supported by enough evidence, and that only an external validation process may determine the actual performance of the ICM.
References
- D’Errigo P, Seccareccia F, Rosato S, Manno V, Badoni G, Fusco CA, the Research Group of the Italian CABG Outcome Project Eur J Cardiothorac Surg 2008;33:325-333.[Abstract/Free Full Text]
- Parsonnet V, Dean D, Bernstein AD. A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989;79(Suppl. 1):3-12.
- Higgins TL, Estafanous FG, Loop FD, Beck GJ, Blum JM, Paranandi L. Stratification of morbidity and mortality outcome by preoperative risk factors in coronary artery bypass patients. JAMA 1992;267:2344-2348.[Abstract/Free Full Text]
- Karkouti K, O’Farrell R, Yau TM, Beattie S, for the Reducing Bleeding in Cardiac Surgery (RBC) Research Group Prediction of massive blood transfusion in cardiac surgery. Can J Anesth 2006;53:781-794.[Medline]
- Alghamdi AA, Davis A, Brister S, Corey P, Logan A. Development and validation of Transfusion Risk Understanding Scoring Tool (TRUST) to stratify cardiac surgery patients according to their blood transfusion needs. Transfusion 2006;46:1120-1129.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
P. D'Errigo, F. Seccareccia, D. Fusco, and C. A. Perucci
Reply to Ranucci
Eur. J. Cardiothorac. Surg.,
February 1, 2009;
35(2):
380 - 381.
[Full Text]
[PDF]
|
|
|
