Effects of longitudinal changes in Charlson comorbidity on prognostic survival model performance among newly...

Area of Science:

• Clinical Epidemiology
• Biostatistics

Background:

• Hypertension is a prevalent condition requiring accurate mortality risk assessment.
• Traditional prognostic models often use static comorbidity data.
• Dynamic changes in patient health status can impact long-term outcomes.

Purpose of the Study:

• To evaluate the impact of updated comorbidity information on predicting mortality in adults with newly diagnosed hypertension.
• To compare the prognostic performance of time-invariant versus time-varying covariate models for mortality prediction.

Main Methods:

• A cohort of adults with incident hypertension in Alberta, Canada was analyzed.
• Cox regression models were employed, comparing time-invariant covariates (TIC) with time-varying covariates (TVC) for Charlson comorbidities.
• Model fit was assessed using the Akaike Information Criterion (AIC).

Main Results:

• Time-varying covariate (TVC) models demonstrated superior prognostic performance compared to time-invariant covariate (TIC) models.
• The association between comorbidities and mortality was stronger in TVC models (e.g., myocardial infarction HR 1.20 vs. 1.07).
• The TVC model incorporating all 17 individual Charlson comorbidities exhibited the best fit and predictive accuracy.

Conclusions:

• Incorporating dynamic comorbidity data enhances the accuracy of patient health risk assessment.
• Time-varying covariate analysis provides a more precise prediction of mortality in hypertensive adults.
• Updated comorbidity information improves predictive model performance over traditional static assessments.