Artikel
External validation of six risk prediction models for patients undergoing elective open or endovascular repair of abdominal aortic aneurysm to predict immediate postoperative outcome
Suche in Medline nach
Autoren
Veröffentlicht: | 21. März 2014 |
---|
Gliederung
Text
Introduction: New risk prediction models have been developed taking into account the special circumstances for endovascular aneurysm repair (EVAR). The aim of this study is to determine whether these newly developed risk prediction models as well as older risk prediction models can be used to calculate 30-day mortality after elective open or endovascular abdominal aortic aneurysm repair.
Material and methods: Data of 707 patients that underwent elective open or endovascular AAA repair at a university vascular centre, from June 1998 to December 2010 were analysed using a prospective database. Examined scores were the Glasgow Aneurysm Score (GAS), the Revised Cardiac Risk Index (RCRI), the Combined prognostic Index (CPI), the Vascular Study Group of New England Cardiac Risk Index (VSG-CRI), the Medicare-Model and the Australian-Audit Score. 30-day mortality was used as an endpoint. The discriminative ability of all risk prediction models was studied by calculating the area under the curve after plotting the receiver-operator characteristic curves.
Results: Endovascular aneurysm repair (EVAR) was performed in 366 (51.8%) patients and open repair (OR) in 341 (48.2%) patients. The 30-day mortality rate for patients with EVAR was 1.6%, and for patients undergoing OR 1.5% (p=0.853).
The best performing score for patients undergoing OR was the CPI with an AUC of 0.780 (95% CI, 0.64-0.95; S.E. 0.08; P=0.024). The Medicare-Model with an AUC of 0.780 (95% CI, 0.64-0.93; S.E. 0.07; P=0.032), the VSG-CRI 0.773 (95% CI, 0.60-0.95; S.E. 0.09; P=0.061), the GAS 0.771 (95% CI, 0.59-0.95; S.E. 0.09; P=0.038), and the RCRI 0.763 (95% CI, 0.51-1.00; S.E. 0.13; P=0.044) also revealed good discriminative ability. The performance of the Australian-Audit Score was poor with an AUC of 0.671 (95% CI, 0.40-0.94; S.E. 0.14; P=0.188).
For patients undergoing EVAR the best performing score was the Medicare-Model with an AUC of 0.806 (95% CI, 0.55-1.00; S.E. 0.13; P=0.010). Also the GAS 0.741 (95% CI, 0.56-0.93; S.E. 0.09; P=0.043) and the VSG-CRI 0.715 (95% CI, 0.45-0.99; S.E. 0.14; P=0.140) were performing well. All the other Scores had an AUC less than 0.70 indicating poor discriminative ability for predicting postoperative death within 30 days.
Conclusion: The majority of scores are predictive for immediate postoperative outcome after open repair. For patients with EVAR, new scores like the Medicare-Model demonstrated good performance for predicting immediate postoperative outcome. Older scores should only applied with caution in this group of patients showing poor performance in this study.