Could the probability of surgical indication be determined after first episode of primary spontaneous pneumothorax?

To develop a risk score model for primary spontaneous pneumothorax surgery (prolonged air leak or ipsilateral recurrence). The model was internally validated for risk estimation.

We analyzed 453 patients with primary spontaneous pneumothorax between 2014 and 2018. Patients were randomly assigned a 2:1 ratio to the development dataset (n = 302, study cohort) or the internal validation dataset (n = 151, validation cohort). The final outcomes of patients with primary spontaneous pneumothorax, the presence or absence of surgical indications, were tracked. Multivariable logistic regression models were prepared to estimate the probability of surgical indication and a scoring model was created. It was internally validated using the validation cohort. Calibration was ascertained using the Hosmer-Lemeshow method and Brier score.

The surgery indication rate was 47.8% (n = 217) (prolonged air leak, n = 130; ipsilateral recurrence, n = 87). There were no demographic or radiological differences between the validation and the study cohorts. Logistic regression analysis showed that the presence of bullae or blebs (p < 0.001, odds ratio = 3.340, 95%CI = 1.753-6.363) and pneumothorax volume (p < 0.001, odds ratio = 1.033, 95%CI = 1.019-1.048) were independent risk factors for surgical indication. The scoring model significantly predicted surgical indications (area under the curve, AUC = 0.768, 95%CI = 0.714-0.821, p < 0.001). Our model showed acceptable discrimination with an AUC > 0.75 in the validation set (AUC = 0.777, 95%CI = 0.702-0.852, p < 0.001) and had an adequate calibration (Hosmer-Lemeshow test p = 0.249, Brier score = 0.25).

The internally validated primary spontaneous pneumothorax scoring model was a good predictor of the need for surgery in patients with primary spontaneous pneumothorax. Prospective external validation studies with larger patient cohorts are required.