Forecasting new home oxygen requirements after lung resection: A predictive model.

OBJECTIVE

There are currently no reliable means for accurately predicting which patients will require discharge with supplemental oxygen following lung resection. To address this problem, we endeavored to develop and validate a prediction model and risk nomogram for new home oxygen requirement after lung resection.

METHODS

We performed a retrospective cohort study using an institutional Society of Thoracic Surgeons database. Adult patients who underwent lung resection for any indication from 2013 to 2023 were included. We excluded patients with preoperative home oxygen and missing pulmonary function tests. Bivariate analyses compared characteristics of patients who required home oxygen postdischarge with those who did not. Multivariable logistic regression models were used to identify risk factors for home oxygen, including the creation of a risk nomogram developed and validated via bootstrapping.

RESULTS

Of 1875 patients, 6.4% had a new home oxygen requirement. Overall, a majority of patients (72%) were older than age 60 years, 54% were women, and 82% were White. Most patients underwent resection for lung cancer (69%). The most commonly performed procedure was a lobectomy (50%), and a minimally invasive approach was utilized in 84%. On multivariable analysis, patients with a new home oxygen requirement were more likely to be obese (adjusted odds ratio [aOR], 2.62; 95% CI, 1.67-4.11), be current smokers (aOR, 2.52; 95% CI, 1.58-3.96), have a preoperative diffusing capacity for carbon monoxide <50% of predicted (aOR, 2.88; 95% CI, 1.65-4.91), and have a preoperative forced expiratory volume in 1 second <50% of predicted (aOR, 2.14; 95% CI, 1.07-4.07).

CONCLUSIONS

In this large retrospective study, we identified 4 primary factors associated with increased odds of home oxygen requirement after lung resection: obesity, current tobacco use, reduced forced expiratory volume in 1 second, and reduced diffusion capacity of carbon monoxide. Prediction modeling may aid in preoperative risk assessment and targeting prehabilitation deployment.