A systematic review identifies valid comorbidity indices derived from administrative health data

Area of Science:

• Health Services Research
• Biostatistics
• Epidemiology

Background:

• Comorbidity indices are crucial for assessing patient complexity using administrative health data.
• Validating these indices is essential for accurate outcome prediction in diverse populations.

Purpose of the Study:

• To systematically review and compare the construct validity of comorbidity indices developed or validated using administrative health data.
• To evaluate the predictive ability of different comorbidity indices for various health outcomes.

Main Methods:

• Comprehensive literature search of MEDLINE and EMBASE databases up to September 2012.
• Inclusion of 76 articles after title and abstract screening by two independent reviewers.
• Assessment of predictive validity using c-statistic for dichotomous and R(2) for continuous outcomes.

Main Results:

• Two main categories of indices were identified: diagnosis-based (e.g., Elixhauser, Charlson) and medication-based (e.g., Chronic Disease Score).
• Predictive performance varied widely (C statistic 0.69 to >0.80) based on index, outcome, and population.
• Diagnosis-based indices showed higher accuracy for mortality prediction, while medication-based indices better predicted healthcare utilization.

Conclusions:

• Multiple valid comorbidity indices derived from administrative data exist.
• The choice of index should align with available data, study population, and the specific outcome of interest.