Predicting non-surgical treatment failure in patients with spontaneous pneumothorax-the Base-C score: a retrospective...

The probability of failure for non-surgical treatment of spontaneous pneumothorax (SP) is relatively high. Early identification of patients at risk of non-surgical treatment failure, with timely intervention, can lead to improved outcomes. This study aimed to develop a model to identify patients at high risk of non-surgical treatment failure, thereby personalizing treatment strategies and improving patient outcomes.

We conducted a retrospective cohort study using data from two large comprehensive hospitals. Demographic data, comorbidities, imaging findings and clinical outcomes were collected. The primary outcome was treatment failure, defined as the need for surgical intervention or the failure of non-surgical management to resolve the pneumothorax within 8 weeks. We employed multivariable logistic regression to identify risk factors and established the Base-C score based on the regression coefficients, which was validated externally.

In the present study, the derivation cohort consisted of 1,378 patients, and the validation cohort comprised 506 patients. The Base-C score included pulmonary bullae, age, smoking, emphysema and the degree of lung collapse. The Base-C score demonstrated excellent discriminative ability in predicting SP treatment failure, with an area under the curve (AUC) of 0.836 (95% confidence interval (CI) [0.813-0.860]) in the derivation cohort and 0.829 (95% CI [0.791-0.866]) in the validation cohort. The Hosmer-Lemeshow (H-L) goodness-of-fit test and calibration curves indicated good calibration in the derivation and validation cohorts. In the clinical application context, Decision Curve Analysis (DCA) demonstrates that the Base-C score generates a positive net benefit for outcomes when the threshold probability is greater than 15%.

The Base-C score showed promise as an effective tool for identifying the risk of non-surgical treatment failure in patients with SP. It may aid in guiding clinical management strategies and has the potential to improve patient outcomes, though this requires prospective verification.