Analyzing the Impact of Diagnostic Measurements and Medical Variables in Predicting the Onset of Diabetes Mellitus in a Subpopulation of the Pima

The question of what factors can predict the outcome of diabetes mellitus and what are the associated risks among different groups was explored in this analysis. The initial study group consisted of 768 recorded observations of female Pima Indians greater than 21 years old located Arizona who were followed for 5 years of development of diabetes mellituspost initial clinical examination. The methods used in this analysis involved elements of categorical data analysis, regression analysis, and epidemiology. Namely, the main predictivemodel followed logistic regression techniques and the risk characterization involved techniques for strengths of association. The results were mainly consistent with previous research over the same topic but there were some differences as well. For example, the AUROC value reported by the original article by Smith was 0.76 whereas this AUROC was 0.86. Furthermore, the main binary logistic regression model for this research showcased the5 variables which were the number of pregnancies, plasma glucose concentration, body mass index, pedigree function of diabetes genetic traits, and the age of individuals as being significant, important, and sufficient in predicting diabetes. Also, the risk characterization as shown by the relative risk and odds ratios showed higher risk and odds of diabetes for the increasing and different levels of the variables of plasma glucose concentration, body mass index, pedigree function, and age with the opposite occurring for pregnancies. Furthermore, the assumptions of our final model as evaluated by the residual diagnostics and other tests on the predictor variables were not violated which helped support these findings. Thus, while the variables which were chosen were interesting and some might not have been originally expected such as the variable of serum insulin level not being included in the final model, they provided great clarity into understanding how to predict and characterize the risk of diabetes. These findings can help highlight what variables to focus on for further research and for assessing community needs from narrowed criteria in big healthcare data. This is useful to efficiently gather benchmark information to understand Pima Indian or Native American groups in the United States to target interventions towards at risk groups.