Early prediction of in-hospital mortality of patients with hemoptysis: an approach to defining severe hemoptysis.

Background: The severity of hemoptysis is usually assessed on the amount of blood expectorated, although no threshold has been agreed upon. Respiratory or hemodynamic failures are additional severity criteria but occur in few cases.

Objectives: Early identification of the in-hospital mortality determinants might be helpful to best characterize severe hemoptysis.

Methods: This is a retrospective cohort study of consecutive patients with hemoptysis admitted to the ICU of a teaching hospital during a 14-year period. The model for early prediction of in-hospital mortality was developed on a derivation sample (67% of patients) using multiple logistic regression. Calibration and discrimination of the model were tested using the remaining validation sample. A scoring system was developed for clinical use.

Results: The in-hospital mortality of the 1,087 patients (age 54 years, 71% male) was 6.5% (95% CI 5-8). Chronic alcoholism, cancer or aspergillosis, pulmonary artery involvement, infiltrates involving two quadrants or more on the admission radiograph, and mechanical ventilation at referral predicted independently mortality. The model showed good concordance between predicted and observed probabilities of death and good discrimination (receiver operating characteristic curve area 0.87; 95% CI 0.82-0.92). The model-based score (chronic alcoholism, pulmonary artery involvement, and radiographic patterns, 1 point each; cancer, aspergillosis, and mechanical ventilation, 2 points each) predicted the probability of death as follows: score 0, 1%; score 1, 2%; score 2, 6%; score 3, 16%; score 4, 34%; score 5, 58%; score 6, 79%, and score 7, 91%.

Conclusions: Our results provide useful information about the short-term prognosis of patients with hemoptysis, which could help design therapeutic approaches and management plans according to the risk of in-hospital mortality.