Characterization of Asthma Exacerbations in Primary Care Using Cluster Analysis

Abstract

Rationale. Patients with a history of asthma exacerbations are at a higher risk for future episodes of severe asthma exacerbations. Characterization of asthma phenotypes could help improve asthma management, including reducing exacerbations. Aim. The aim of this study is to identify distinctive patient characteristics associated with a history of asthma exacerbations using cluster analysis. Methods. We used data assessing asthma control from two cross-sectional surveys of adult and pediatric patients in the primary care setting. A supervised cluster analysis with recursive partitioning approach was applied to identify characteristics that maximized the differences across subgroups. Results. The sample comprised 2205 adults and 2435 children and adolescents with asthma. Key predictors were identified in seven adult clusters including visiting an asthma specialist, number of hours worked, and excessive use of rescue medication. The rate ratio (RR) for having an exacerbation was significantly higher (2.88; 95% confidence interval (CI), 2.46–3.36) in Cluster 7, with more female patients reporting severe disease, high body mass index, sinus infections, gastroesophageal reflux disease, skin allergies, and lower asthma control score. Features identified in the six pediatric clusters included visiting an asthma specialist, missed school days, race/ethnicity, and age. The RR for having an exacerbation was higher in Cluster 6 (2.36; 95% CI, 2.11–2.64), with patients reporting more severe disease, sinus and skin allergies, and lower asthma control score. Conclusions. Identification of specific risk factors can be enhanced by using supervised cluster analysis. This approach allows grouping of patients with unique characteristics to help identify patients at higher risk of exacerbations.

• asthma exacerbations
• cluster analysis
• primary care
• survey

Acknowledgments

The authors acknowledge Alicia Gilsenan, Ryan Ziemiecki, Abenah Vanderpuije, Xiaolei Zhou, and Christine Bui from RTI Health Solutions as well as Richard Stanford, William Lincourt, and Amanda Emmett from GlaxoSmithKline for their contribution to data collection for the primary studies. The authors are most appreciative of Drs Lynn Katz and Charlene Prazma for their helpful comments regarding the manuscript. Editorial support in the form of figure generation, grammatical editing, and referencing was provided by Dr Elaine F. Griffin at Evidence Scientific Solutions and was funded by GlaxoSmithKline.

Declaration of Interest

All authors were employees of GlaxoSmithKline at the time of the study. Dr Miller is currently an employee of UCB, Inc.