| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
a National Centre of Epidemiology, Surveillance and Health Promotion, Istituto Superiore di Sanità, Rome, Italy
b Department of Epidemiology, ASL RME, Rome, Italy
Received 23 October 2008; accepted 27 October 2008.
* Corresponding author. Address: National Centre for Epidemiology, Surveillance and Health Promotion, Istituto Superiore di Sanità, Via Giano della Bella, 34, I-00161 Rome, Italy. Tel.: +39 06 49904236; fax: +39 06 49904230. (Email: paola.derrigo{at}iss.it).
Key Words: Coronary artery bypass graft • Outcome • Risk adjustment • EuroSCORE
Some years have passed since the CABG Outcome Study [1] was published and we are still answering questions concerning the original methodology already discussed several times. This is the case of Ranucci's letter [2]. We hope this is the very last time.
As usual, we answer in detail:
Concerning the matter of model validation we would like to quote a passage from Iezzoni's book ‘Risk adjustment for measuring health care outcomes’ [3]: ‘... How cross-validation is done depends primarily on the size of the data set. If the database is sufficiently large, direct estimates of how the model will perform with new data can be obtained by performing the following steps. First, the data are randomly divided in half. Second, the model is developed on one half of the data and then validated on the other half ...’. We think that a database of 34,310 records can be considered ‘sufficiently large’ to justify the use of the suggested methodology. As the cross-validation procedure allows obtaining an ‘estimate of how the model will perform with new data’, we think the conclusion affirming the better performance of the Italian CABG Model (ICM) as compared to the EuroSCORE is not biased.
Moreover, although it could not be considered a real validation, in a recently published work [4] we applied the ICM to an Italian sub-population selected on the basis of the National Hospital Discharge Records (years 2002–2004). The aim was to compare hospitals’ performances obtained using the ICM with those derived from a model built on current administrative data. The results confirmed the ICM goodness of fit.
As a final remark, we would like to remind that EuroSCORE, as well as many other works addressing the same issue, used a similar methodology [5]. In fact, EuroSCORE was validated on an external population only some years after its development, application and publication.
Concerning the EuroSCORE logistic model recalibration, contrary to Ranucci's statement, we do not suggest to multiply by 0.4 the logistic EuroSCORE values but the number of the expected deaths. Actually, in this way, the number of the expected and observed deaths in estimated risk classes becomes closer and, as a consequence, the Hosmer–Lemeshow 
2 becomes not significant (p
= 0.092).
Finally, concerning the discrepancies in reporting numbers, we admit our fault. Actually, values in the abstract were wrongly reported. The right numbers are those reported in the text and in Table 4. We apologise for the slip.
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 |
