Using Recursive Logistic Regression to Develop a Patient-reported Outcome in Non-cystic Fibrosis Bronchiectasis

Many patient reported outcomes for non-cystic fibrosis bronchiectasis are time-consuming self-report surveys such as the St. George's Respiratory Questionnaire. Using data from the EMBRACE multi-center RCT, we use recursive logistic regression to develop a new outcome that only requires responses to three simple questions about pulmonary symptoms (sputum volume, sputum purulence, and dyspnoea). The low response burden means that the outcome can be calculated for each day of follow-up in real-time. The new outcome is constructed such that it is a good predictor of event-based exacerbation, a sustained worsening of condition requiring treatment with antibiotics, and validated against self-reported quality of life.