Derivation and Validation of a Risk Index to Predict All-Cause Mortality in Type 2 Diabetes Mellitus

Abstract

Background: Type 2 Diabetes Mellitus with severe complications have higher risk of mortality. An easy-to-use risk index to quantify the risk of mortality would help clinicians identify patients who might benefit from more intensive therapy. The study aims to develop a risk index to predict all-cause mortality for a cohort of T2DM patients seen at primary care clinics in Singapore.

Methods: In a retrospective cohort study, 28 patient-level variables were extracted from an automated clinical and administrative registry for T2DM who had at least 2 visits to the same National Healthcare Group Polyclinic in 2007. Demographic characteristics, inpatient and outpatient diagnoses, laboratory results and prescription were included. Mortality data were provided by the Ministry of Health. We used a split-sample design to derive and validate an index to predict the risk of death within 2 years of the index attendance. The c-statistic was used to assess model discrimination.

Results: Out of the 59,747 patients in the study, 2,977 (5%) patients died during the 2-year follow up. Age (A); diabetes-related complications (Diabetes Complication Severity Index) (C); and cancer history (C) were found to independently predict all-cause mortality (from which the mnemonic ACC was derived). The ACC risk index ranged from 0 to 20 with expected risk of mortality of 0.3% to 80.6%. The discriminatory accuracy of the ACC risk index for the validation data is excellent (c-statistic 0.83, 95% CI 0.82 - 0.84).

Conclusion: A simple risk index for all-cause mortality was successfully developed, and validated using routinely collected registry data. The risk index can be used to stratify T2 DM patients into varying risk of mortality. Further external validation of the risk index is needed before using it in a clinical setting.




J Endocrinol Metab. 2012;2(2):88-95
doi: https://doi.org/10.14740/jem67w