Multilevel Multivariate Modelling of The Effect of Gender and Patient Co-morbidity on Spherocylindrical Refractive Outcome Following Cataract Surgery

Purpose: To investigate effects of co-morbidities on refractive outcomes following cataract surgery. Design: Case series database study Methods: Study population: patients on UK national ophthalmic cataract database. Variables examined included gender, age, diabetic retinopathy, glaucoma, high myopia, inherited eye disease, optic nerve disease, uveitis, pseudoexfoliation, vitreous opacities, retinal pathology, cataract type, previous surgery and posterior capsular rupture. A multivariate normal cross-classified model was fitted to the refractive outcome using Markov Chain Monte Carlo (MCMC) methods with diffuse priors to approximate maximum likelihood estimation. A MCMC chain was generated with a burn-in of 5000 iterations and a monitoring chain of 50,000 iterations. Main Outcome Measure Difference between postoperative and expected refractive outcome Results: 490,987 cataract operations were performed on 351,864 patients by 2567 surgeons. Myopic and astigmatic errors were associated with female gender, glaucoma, previous trabeculectomy or vitrectomy, high myopia and increasing age. Hyperopic and astigmatic errors were associated with diabetic retinopathy, pseudoexfoliation and male sex. A large proportion of the unexplained variance in EERO was attributed to unobserved eye-level factors such as biometry measurements and IOL prediction formulae. Conclusions: There was minimal variation between surgeons. Co-morbidities and gender accounted for small but significant differences significant differences in expected outcomes.