Title: Preoperative and postoperative factors affecting 30-day mortality of patients undergoing mitral valve repair or replacement in Singapore

Biography:

Jonathan Kia-Sheng Phua is currently a Medical student at Yong Loo Lin School of Medicine, National University of Singapore.

Abstract

Aim: To use novel quantitative methods to identify the preoperative and postoperative factors for predicting 30-day mortality for multiracial Asian patients undergoing mitral valve repair or replacement.

Methods: The de-identified data of 197 patients who had undergone mitral valve repair or replacement from 2009 to 2015 were analyzed with generalized structural equation model (gSEM). The chi-square automated interaction detector (CHAID) algorithm was applied for variable-selection. Analyzed with Stata MP version 14, all statistical tests were carried out with 95% confidence intervals.

Results: Female gender, BMI>29 kg/m², the need for emergency surgery and prolonged mechanical ventilation were identified as significant predictors of 30-day mortality. Emergency surgery was also significant in explaining other postoperative outcomes such as unplanned reoperation, prolonged mechanical ventilation and prolonged intensive care. Patients who had valve replacement were more likely to suffer from postoperative complications. The final model was deemed to be valid externally (AUC: 0.99; 95% C.I.: 0.98-1.00).

Conclusion: The results were congruent with traditional negative prognosticators for mitral valve surgery. The novel quantitative techniques are able to provide more comprehensive analysis than the conventional methods. With its multi-way splitting, the CHAID could identify meaningful cut-offs of the quantitative predictors objectively and jointly. On the other hand, gSEM is able to analyze the multiple outcomes in a single setting that could result in higher precision. A predictor’s effect could also be divided into direct and indirect, thus providing more information about its overall effect on the outcomes. We have validated the use of these new statistical methods for use in future studies on identifying risk factors for medical conditions.