The findings, published in Mayo Clinic Proceedings: Digital Health, reported 89 percent accuracy for women and 86 percent for men in predicting Type 2 diabetes from acoustic voice features.