Development and validation of a comorbidity scoring system for patients with cirrhosis

Abstract

Background & aims: At least 40% of patients with cirrhosis have comorbidities that increase mortality. We developed a cirrhosis-specific comorbidity scoring system (CirCom) to help determine how these comorbidities affect mortality and compared it with the generic Charlson Comorbidity Index.

Methods: We used data from nationwide health care registries to identify Danish citizens diagnosed with cirrhosis in 1999-2008 (n = 12,976). They were followed through 2010 and characterized by 34 comorbidities. We used Cox regression to assign severity weights to comorbidities with an adjusted mortality hazard ratio (HR) ≥ 1.20. Each patient's CirCom score was based on, at most, 2 of these comorbidities. Performance was measured with Harrell's C statistic and the Net Reclassification Index (NRI) and results were compared with those obtained using the Charlson Index (based on 17 comorbidities). Findings were validated in 2 separate cohorts of patients with alcohol-related cirrhosis or chronic hepatitis C.

Results: The CirCom score included chronic obstructive pulmonary disease, acute myocardial infarction, peripheral arterial disease, epilepsy, substance abuse, heart failure, nonmetastatic cancer, metastatic cancer, and chronic kidney disease; 24.2% of patients had 1 or more of these, and mortality correlated with the CirCom score. Patients' CirCom score correlated with their Charlson Comorbidity Index (Kendall's τ = 0.57; P < .0001). Compared with the Charlson Index, the CirCom score increased Harrell's C statistic by 0.6% (95% confidence interval: 0.3%-0.8%). The NRI for the CirCom score was 5.2% (95% confidence interval: 3.7%-6.9%), and the NRI for the Charlson Index was 3.6% (95% confidence interval: 2.3%-5.0%). Similar results were obtained from the validation cohorts.

Conclusions: We developed a scoring system to predict mortality among patients with cirrhosis based on 9 comorbidities. This system had higher C statistic and NRI values than the Charlson Comorbidity Index, and is easier to use. It could therefore be a preferred method to predict death or survival of patients and for use in epidemiologic studies.

Keywords: CI; End-Stage Liver Disease; HR; MELD; Model for End-Stage Liver Disease; NRI; Net Reclassification Index; Outcome; Prediction Model; Prognostic Factors; confidence interval; hazard ratio.