An Australian risk model for determining 30-day mortality following aortic valve replacement

Research output: Contribution to conferenceAbstractOtherpeer-review

Abstract

Background: Recent reports reveal that around 32% to 38% patients with severe Aortic Stenosis are not referred to Aortic Valve Replacement (AVR) due to factors such as old age, severe comorbidities and patient refusal. Preoperative risk associated with AVR can be ascertained through a variety of risk prediction models, none of which are specific to the Australian population. Aim: To identify risk factors associated with 30-day mortality following AVR in Australian patients, and to develop a multivariable logistic model for pre-operative risk prediction. Methods: Prospectively collected data from the Australasian Society for Cardiac and Thoracic Surgeons (ASCTS) database project was used. All AVR surgeries performed between 01 July 2001 and 30 June 2008 were included for analysis. Preoperative variables with a p-value of < 0.10 in chi-squared analysis were considered for multiple logistic regression analysis. Using bootstrap re-sampling technique, five plausible models were identified based on variables that were significant predictors of mortality. All models were validated internally using average receiver operating characteristic (ROC) curve and p-value of Hosmer Lemeshow (H-L) goodness-of-fit test via bootstrap n-fold (n=100) validation method on 70% of data. The Akaike Information Criterion (AIC) and prediction mean square error (MSE), the ROC and H-L p-value were used to select the final model (AVR-Score) from the five plausible models. Results: Between July 2001 and June 2008 a total of 3544 AVR procedures were performed, of which 147 (4.15%) reported a fatal outcome within 30-days. The final model, the AVR-Score (ROC:0.779; H-L p-value=0.042), comprised the following independent predictors of 30-day mortality following AVR: age, New York Heart Association class (NYHA), left main disease, infective endocarditis, cerebrovascular disease, renal dysfunction, previous cardiac surgery, arrhythmia, and estimated ejection fraction. Conclusions: We have identified 8 key predictors of early AVR mortality in Australian patients and developed a preoperative risk prediction model for 30-day mortality.
Original languageEnglish
Pages86-87
Number of pages2
Publication statusPublished - 6 Oct 2010
EventAustralasian Epidemiological Association Annual Scientific Meeting 2010: Translating research into practice - University of Sydney, Sydney, Australia
Duration: 29 Sep 20101 Oct 2010

Conference

ConferenceAustralasian Epidemiological Association Annual Scientific Meeting 2010
Abbreviated titleAEA 2010
CountryAustralia
CitySydney
Period29/09/101/10/10

Cite this

Ariyaratne, T. V., Billah, B., Yap, C., Dinh, D., & Reid, C. M. (2010). An Australian risk model for determining 30-day mortality following aortic valve replacement. 86-87. Abstract from Australasian Epidemiological Association Annual Scientific Meeting 2010, Sydney, Australia.
Ariyaratne, Thathya Venu ; Billah, Baki ; Yap, Chenghon ; Dinh, Diem ; Reid, Christopher M. / An Australian risk model for determining 30-day mortality following aortic valve replacement. Abstract from Australasian Epidemiological Association Annual Scientific Meeting 2010, Sydney, Australia.2 p.
@conference{10b50e2ac904437f854da8d2cba23830,
title = "An Australian risk model for determining 30-day mortality following aortic valve replacement",
abstract = "Background: Recent reports reveal that around 32{\%} to 38{\%} patients with severe Aortic Stenosis are not referred to Aortic Valve Replacement (AVR) due to factors such as old age, severe comorbidities and patient refusal. Preoperative risk associated with AVR can be ascertained through a variety of risk prediction models, none of which are specific to the Australian population. Aim: To identify risk factors associated with 30-day mortality following AVR in Australian patients, and to develop a multivariable logistic model for pre-operative risk prediction. Methods: Prospectively collected data from the Australasian Society for Cardiac and Thoracic Surgeons (ASCTS) database project was used. All AVR surgeries performed between 01 July 2001 and 30 June 2008 were included for analysis. Preoperative variables with a p-value of < 0.10 in chi-squared analysis were considered for multiple logistic regression analysis. Using bootstrap re-sampling technique, five plausible models were identified based on variables that were significant predictors of mortality. All models were validated internally using average receiver operating characteristic (ROC) curve and p-value of Hosmer Lemeshow (H-L) goodness-of-fit test via bootstrap n-fold (n=100) validation method on 70{\%} of data. The Akaike Information Criterion (AIC) and prediction mean square error (MSE), the ROC and H-L p-value were used to select the final model (AVR-Score) from the five plausible models. Results: Between July 2001 and June 2008 a total of 3544 AVR procedures were performed, of which 147 (4.15{\%}) reported a fatal outcome within 30-days. The final model, the AVR-Score (ROC:0.779; H-L p-value=0.042), comprised the following independent predictors of 30-day mortality following AVR: age, New York Heart Association class (NYHA), left main disease, infective endocarditis, cerebrovascular disease, renal dysfunction, previous cardiac surgery, arrhythmia, and estimated ejection fraction. Conclusions: We have identified 8 key predictors of early AVR mortality in Australian patients and developed a preoperative risk prediction model for 30-day mortality.",
author = "Ariyaratne, {Thathya Venu} and Baki Billah and Chenghon Yap and Diem Dinh and Reid, {Christopher M.}",
year = "2010",
month = "10",
day = "6",
language = "English",
pages = "86--87",
note = "Australasian Epidemiological Association Annual Scientific Meeting 2010 : Translating research into practice, AEA 2010 ; Conference date: 29-09-2010 Through 01-10-2010",

}

Ariyaratne, TV, Billah, B, Yap, C, Dinh, D & Reid, CM 2010, 'An Australian risk model for determining 30-day mortality following aortic valve replacement' Australasian Epidemiological Association Annual Scientific Meeting 2010, Sydney, Australia, 29/09/10 - 1/10/10, pp. 86-87.

An Australian risk model for determining 30-day mortality following aortic valve replacement. / Ariyaratne, Thathya Venu; Billah, Baki; Yap, Chenghon; Dinh, Diem; Reid, Christopher M.

2010. 86-87 Abstract from Australasian Epidemiological Association Annual Scientific Meeting 2010, Sydney, Australia.

Research output: Contribution to conferenceAbstractOtherpeer-review

TY - CONF

T1 - An Australian risk model for determining 30-day mortality following aortic valve replacement

AU - Ariyaratne, Thathya Venu

AU - Billah, Baki

AU - Yap, Chenghon

AU - Dinh, Diem

AU - Reid, Christopher M.

PY - 2010/10/6

Y1 - 2010/10/6

N2 - Background: Recent reports reveal that around 32% to 38% patients with severe Aortic Stenosis are not referred to Aortic Valve Replacement (AVR) due to factors such as old age, severe comorbidities and patient refusal. Preoperative risk associated with AVR can be ascertained through a variety of risk prediction models, none of which are specific to the Australian population. Aim: To identify risk factors associated with 30-day mortality following AVR in Australian patients, and to develop a multivariable logistic model for pre-operative risk prediction. Methods: Prospectively collected data from the Australasian Society for Cardiac and Thoracic Surgeons (ASCTS) database project was used. All AVR surgeries performed between 01 July 2001 and 30 June 2008 were included for analysis. Preoperative variables with a p-value of < 0.10 in chi-squared analysis were considered for multiple logistic regression analysis. Using bootstrap re-sampling technique, five plausible models were identified based on variables that were significant predictors of mortality. All models were validated internally using average receiver operating characteristic (ROC) curve and p-value of Hosmer Lemeshow (H-L) goodness-of-fit test via bootstrap n-fold (n=100) validation method on 70% of data. The Akaike Information Criterion (AIC) and prediction mean square error (MSE), the ROC and H-L p-value were used to select the final model (AVR-Score) from the five plausible models. Results: Between July 2001 and June 2008 a total of 3544 AVR procedures were performed, of which 147 (4.15%) reported a fatal outcome within 30-days. The final model, the AVR-Score (ROC:0.779; H-L p-value=0.042), comprised the following independent predictors of 30-day mortality following AVR: age, New York Heart Association class (NYHA), left main disease, infective endocarditis, cerebrovascular disease, renal dysfunction, previous cardiac surgery, arrhythmia, and estimated ejection fraction. Conclusions: We have identified 8 key predictors of early AVR mortality in Australian patients and developed a preoperative risk prediction model for 30-day mortality.

AB - Background: Recent reports reveal that around 32% to 38% patients with severe Aortic Stenosis are not referred to Aortic Valve Replacement (AVR) due to factors such as old age, severe comorbidities and patient refusal. Preoperative risk associated with AVR can be ascertained through a variety of risk prediction models, none of which are specific to the Australian population. Aim: To identify risk factors associated with 30-day mortality following AVR in Australian patients, and to develop a multivariable logistic model for pre-operative risk prediction. Methods: Prospectively collected data from the Australasian Society for Cardiac and Thoracic Surgeons (ASCTS) database project was used. All AVR surgeries performed between 01 July 2001 and 30 June 2008 were included for analysis. Preoperative variables with a p-value of < 0.10 in chi-squared analysis were considered for multiple logistic regression analysis. Using bootstrap re-sampling technique, five plausible models were identified based on variables that were significant predictors of mortality. All models were validated internally using average receiver operating characteristic (ROC) curve and p-value of Hosmer Lemeshow (H-L) goodness-of-fit test via bootstrap n-fold (n=100) validation method on 70% of data. The Akaike Information Criterion (AIC) and prediction mean square error (MSE), the ROC and H-L p-value were used to select the final model (AVR-Score) from the five plausible models. Results: Between July 2001 and June 2008 a total of 3544 AVR procedures were performed, of which 147 (4.15%) reported a fatal outcome within 30-days. The final model, the AVR-Score (ROC:0.779; H-L p-value=0.042), comprised the following independent predictors of 30-day mortality following AVR: age, New York Heart Association class (NYHA), left main disease, infective endocarditis, cerebrovascular disease, renal dysfunction, previous cardiac surgery, arrhythmia, and estimated ejection fraction. Conclusions: We have identified 8 key predictors of early AVR mortality in Australian patients and developed a preoperative risk prediction model for 30-day mortality.

UR - https://aea.asn.au/documents/ae/90-17-2-translating-research-into-practicel-aea-annual-scientific-meeting-2010/file

M3 - Abstract

SP - 86

EP - 87

ER -

Ariyaratne TV, Billah B, Yap C, Dinh D, Reid CM. An Australian risk model for determining 30-day mortality following aortic valve replacement. 2010. Abstract from Australasian Epidemiological Association Annual Scientific Meeting 2010, Sydney, Australia.