RESUMO
RATIONALE: As the prevalence of multimorbidity increases, understanding the impact of isolated comorbidities in people COPD becomes increasingly challenging. A simplified model of common comorbidity patterns may improve outcome prediction and allow targeted therapy. OBJECTIVES: To assess whether comorbidity phenotypes derived from routinely collected clinical data in people with COPD show differences in risk of hospitalisation and mortality. METHODS: Twelve clinical measures related to common comorbidities were collected during annual reviews for people with advanced COPD and k-means cluster analysis performed. Cox proportional hazards with adjustment for covariates was used to determine hospitalisation and mortality risk between clusters. MEASUREMENTS AND MAIN RESULTS: In 203 participants (age 66 ± 9 years, 60 % male, FEV1%predicted 31 ± 10 %) no comorbidity in isolation was predictive of worse admission or mortality risk. Four clusters were described: cluster A (cardiometabolic and anaemia), cluster B (malnourished and low mood), cluster C (obese, metabolic and mood disturbance) and cluster D (less comorbid). FEV1%predicted did not significantly differ between clusters. Mortality risk was higher in cluster A (HR 3.73 [95%CI 1.09-12.82] p = 0.036) and B (HR 3.91 [95%CI 1.17-13.14] p = 0.027) compared to cluster D. Time to admission was highest in cluster A (HR 2.01 [95%CI 1.11-3.63] p = 0.020). Cluster C was not associated with increased risk of mortality or hospitalisation. CONCLUSIONS: Despite presence of advanced COPD, we report striking differences in prognosis for both mortality and hospital admissions for different co-morbidity phenotypes. Objectively assessing the multi-system nature of COPD could lead to improved prognostication and targeted therapy for patients.