Counting co-occurring diseases to predict mortality is as accurate as multimorbidity indices: an external validation...

Area of Science:

• Gerontology and Public Health
• Epidemiology
• Biostatistics

Background:

• A systematic review identified seven multimorbidity indices for mortality prediction.
• However, a direct comparison of their predictive performance was lacking.
• This study aimed to externally validate these indices and compare them to simple disease counting.

Purpose of the Study:

• To compare the performance of seven established multimorbidity indices against simple disease counts for predicting all-cause mortality.
• To determine the most practical and reliable method for assessing prognosis in community-dwelling older adults with multiple conditions.

Main Methods:

• External validation of seven multimorbidity indices using data from the prospective Rotterdam Study (n=14,926).
• Comparison of index performance against four benchmark models, including age/sex, disease counts, and individual diseases.
• Assessment of discriminative ability (C-statistic) and accuracy (Brier score), and calibration.

Main Results:

• Multimorbidity indices and disease count models showed nearly identical discriminative performance (max C-statistic difference: 0.06).
• Both were superior to age-and-sex-only models.
• Prediction accuracy was similar across models (max Brier score improvement: 4%), but calibration was poor for indices with short follow-up times.

Conclusions:

• Counting co-occurring diseases is as accurate as using complex multimorbidity indices for predicting mortality in community-dwelling older adults.
• Disease counting offers a more practical and reliable method for patient prognosis.
• These findings support simpler approaches for managing multimorbidity in primary care settings.