Journal of Biometrics & Biostatistics

People with type II diabetes are having more chances to develop diabetic retinopathy which is generally viewed as a multi-factorial disease. Identifying the risk of any disease, is very important for health care planning and creating score cards for identifying the risk of any disease is pervasive in medical diagnostics. This involves statistical techniques using parameter estimation of multivariable models such as linear regression, logistic regression or Cox proportional hazards regression. Geographic and/or disease specific methods for risk score estimation provide a scope to develop and evaluate new possible statistical methods for risk score analysis. This work explores the weighted scoring procedure through logistic regressions to develop two methods using Wald statistic and maximum regression coefficient by precluding the selection of protective risk factors. Further, to avoid the numerical errors due to interim rounding of digits in any computations the study includes a standard method that avoids such rounding of digits. Three widely applicable methods for score estimation of different diseases have also been considered for comparative study. All these methods are applied to Sankara Nethralaya Diabetic Retinopathy Epidemiology and Molecular Genetics Study III, a cross-sectional study to estimate the prevalence and risk factors for diabetic retinopathy in rural south India and then validated by comparing with methods used in Australian Type 2 Diabetes Risk Assessment Tool. Results have indicated that the new methods are more suitable in score estimation for diabetic retinopathy by considering the statistical property of the methods.