Handling Missing Phenotype Data with Random Forests for Diabetes Risk Prognosis