Could a 2-year mortality prediction model have prevented deaths from respiratory failure: a single UK centre experience

Introduction: People with cystic fibrosis (PwCF) have unpredictable clinical courses, making timing and listing for lung transplantation (LTx) difficult. FEV1 <30% predicted remains the gold standard as the referral threshold but is a blunt tool. Stanojevic et al (ERJ 2019; 54; 1900224) created a 2-year mortality prediction model, with the aim of identifying high-risk patients (defined as risk of mortality >20% over 2 years). Aim: To establish whether this model would have identified patients who died without LTx as high risk of mortality at time of referral or death (whichever came first). Method: Retrospective application of the mortality prediction model to PwCF who were transplanted (n=32) or died with a respiratory-related illness without LTx (n=36) 2015-2019. The model was applied at time of decision to refer, decision not to refer or death. Results: In the transplanted group, all PwCF had a risk of death >20% within 2 years at time of referral (median risk of death 58%, range 30%-91%). In the group that died (n=36), 58% (n=21) had a risk of death <20% within 2 years, (median risk of death 17%, range 7%-42%). Of these, 71% (n=15) had started transplant discussions or had already had a transplant decision made. Conclusion: Most PwCF die from respiratory failure and many have acute unpredictable declines, which prediction models do not have the sensitivity to discern. Although this model identified many patients appropriately for LTx, it was not sensitive enough to identify 58% of the patients who died, many of whom the clinical team had recognised as being at high risk of mortality. Better prediction models are needed given the variable clinical course.