Risk score may predict death in patients with type 2 diabetes

A risk score using commonly measured clinical and biochemical variables may help predict mortality in patients with type 2 diabetes mellitus.

Researchers from Hong Kong conducted a prospective cohort study of 7,583 patients with type 2 diabetes enrolled in the Hong Kong Diabetes Registry since 1995. The data was censored on July 30, 2005 or after six years of follow up. Linear-transforming formulas were created using a restricted cubic spline model and the researchers randomly assigned the data to a training or test dataset.

During a median follow-up of 5.51 years, 619 patients died—a mortality rate of 18.69 per 1,000 person-years. Numerous variables, including age, sex and cancer history, were used to create a risk score for all-cause mortality.

Predicted and observed death rates in the test dataset were similar (P>.70), according to the study. The researchers used the risk score to rank cause-specific deaths and found that for genitourinary death, the area under the receiver operating characteristic curve was 0.95; for circulatory death, 0.85; for respiratory death, 0.85; and for neoplasm death, 0.71.

The researchers wrote that additional substantiation is needed before using the risk score in a clinical setting. – by Stacey L. Adams

Arch Intern Med. 2008;168:451-457.