A Point-based Mortality Prediction System for Older Adults with Diabetes

Area of Science:

• Gerontology
• Diabetology
• Biostatistics

Background:

• Established mortality prediction models exist for the general diabetic population.
• A specific, validated model for elderly individuals with diabetes is currently lacking.
• Accurate risk stratification is crucial for managing diabetes in older adults.

Purpose of the Study:

• To develop and validate a mortality prediction model tailored for elderly patients with diabetes (≥65 years).
• To identify significant predictors of five-year mortality in this demographic.
• To create an accessible tool for assessing mortality risk in older diabetic individuals.

Main Methods:

• Utilized data from Taiwan's nationwide adult health screening program (2007-2010) for model construction.
• Employed 10-fold cross-validation for internal validation.
• Performed external validation using the MJ health screening database (2004-2007).
• Applied multivariable Cox proportional hazards models to predict five-year mortality.

Main Results:

• The final model included predictors such as age, gender, BMI, fasting glucose, blood pressure, and various biochemical markers (e.g., cholesterol, hemoglobin, albumin).
• The model demonstrated good predictive performance with Harrell's C values of 0.737 (development), 0.746 (internal validation), and 0.685 (external validation).
• A total of 220,832 older diabetic subjects were analyzed, with a 10.5% five-year mortality rate.

Conclusions:

• An easy-to-use, point-based mortality prediction model for elderly individuals with diabetes was successfully developed.
• The model provides accurate five-year mortality risk assessment for older adults with diabetes.
• This tool can support clinical decision-making and proactive healthcare management for geriatric diabetic patients.