The choice between a mechanical or biological (bioprosthetic) valve is a consequential decision in heart valve surgery. Mechanical valves offer superior durability but require lifelong anticoagulation, carrying bleeding and thromboembolic risks. Bioprosthetic valves avoid long-term anticoagulation but are subject to structural valve deterioration, often necessitating reoperation, particularly in younger patients. Current guidelines (ACC/AHA, ESC/EACTS) provide age-based thresholds and risk-factor considerations to guide this choice, yet substantial unexplained variation in prosthesis selection persists across patients with similar clinical profiles. This research project applies multi-level (hierarchical) modelling to understand the sources of variation in valve prosthesis selection for both aortic and mitral valve replacement surgery. By fitting multi-level logistic regression models with random intercepts (and potentially random slopes) at surgeon and hospital levels, this project aims to quantify how much of the variation in mechanical-versus-biological valve choice is explained by patient characteristics (age, sex, comorbidities, life expectancy, anticoagulation risk, patient preference) as opposed to surgeon-level tendencies and hospital-level factors (volume, resources, institutional protocols). Measures such as the intraclass correlation coefficient (ICC) and median odds ratio (MOR) will be used to express the magnitude of surgeon and hospital effects in clinically interpretable terms. If a substantial proportion of prosthesis choice is driven by surgeon or hospital identity rather than patient characteristics, this points to unwarranted practice variation that could be addressed through standardised decision aids, shared decision-making tools, or targeted education. Conversely, confirming that patient factors dominate would support the appropriateness of current individualised decision-making. The findings may also inform quality improvement initiatives and help identify outlier surgeons or centres whose prosthesis selection patterns diverge markedly from evidence-based norms, prompting further investigation into outcomes and practice.

I may also undertake an additional project topic on pre-operative sizing of the valve annulus for valve replacement surgery. Undersizing risks patient-prosthesis mismatch, where the effective orifice area of the implanted valve is too small relative to the patient's body size, leading to elevated transvalvular gradients, impaired hemodynamics, and worse long-term outcomes including heart failure, reduced exercise tolerance, and higher mortality. Oversizing carries risks of annular injury, paravalvular leak, and conduction abnormalities. Current clinical practice typically estimates annulus size from preoperative CT using two-dimensional measurements taken in a single cross-sectional plane. However, the aortic and mitral annuli are non-planar, saddle-shaped structures, and 2D measurement systematically underestimates true annulus area and perimeter by failing to capture this three-dimensional geometry. By comparing 3D model-derived measurements against conventional 2D measurement, manual expert annotation, and surgical outcomes, the project aims to demonstrate that 3D estimation more accurately captures true annulus size, reduces the risk of undersizing, and improves prosthesis selection accuracy relative to current clinical practice.