Coding algorithms for defining Charlson and Elixhauser co-morbidities in Read-coded databases

Area of Science:

• Health Services Research
• Clinical Informatics
• Epidemiology

Background:

• Comorbidity indices like Charlson Comorbidity Index (CCI) and Elixhauser Method (EM) are vital for risk adjustment in healthcare research.
• Primary care databases often use Read codes, necessitating specific code lists for these indices.
• This study addresses the need for validated Read code lists for CCI and EM in UK primary care.

Purpose of the Study:

• To develop comprehensive lists of Read codes for the Charlson Comorbidity Index (CCI) and Elixhauser Method (EM).
• To compare the predictive performance of CCI and EM for mortality in hip fracture patients.
• To evaluate the impact of look-back periods on the predictive accuracy of these comorbidity measures.

Main Methods:

• Two researchers independently screened over 111,000 Read codes to establish CCI and EM comorbidity categories.
• Hip fracture patients and matched controls were identified from the Clinical Practice Research Datalink (CPRD) primary care database.
• Logistic regression models and Receiver Operating Characteristic (ROC) curves were used to assess predictive properties for 30-day and 365-day mortality.

Main Results:

• 5832 CCI and 7156 EM comorbidity codes were identified.
• The Elixhauser Method (EM) demonstrated superior prediction of 30-day mortality (AUROC 0.744) compared to the Charlson Comorbidity Index (CCI) (AUROC 0.686).
• Extending the comorbidity look-back period beyond one year yielded only modest improvements in predictive value for both indices.

Conclusions:

• The generated Read code lists enable researchers to calculate CCI and EM in Read coded primary care databases.
• The Elixhauser Method (EM) is recommended over the Charlson Comorbidity Index (CCI) for risk adjustment in this population.
• Utilizing comorbidity data beyond a one-year look-back period offers limited additional predictive benefit.