A risk model to predict an unplanned admission to the intensive care unit following lung resection

Abstract

Abstract OBJECTIVES The goal of this study was to develop a risk-adjusting model to stratify the risk of an unplanned admission to the intensive care unit (following lung resection). METHODS We performed a retrospective analysis of 3123 patients undergoing anatomical lung resections (2014–2019) in 2 centres. A risk score was developed by testing several variables for a possible association with a subsequent ICU admission using stepwise logistic regression analyses, validated by the bootstrap resampling technique. Variables associated with ICU admission were assigned weighted scores based on their regression coefficients. These scores were summed for each patient to generate the ICU risk score, and patients were grouped into risk classes. RESULTS A total of 103 patients (3.3%) required an unplanned admission to the ICU after the operation. The average ICU stay was 17.6 days. The following variables remained significantly associated with ICU admission following logistic regression: male gender (P = 0.004), body mass index <18.5 (P = 0.002), predicted postoperative forced expiratory volume in 1 s < 60% (P = 0.004), predicted postoperative carbon monoxide lung diffusion capacity <50% (P = 0.013), open access (P = 0.004) and pneumonectomy (P = 0.041). All variables were weighted 1 point except body mass index <18.5 (2 points). The final ICU risk score ranged from 0 to 7 points. Patients were grouped into 6 risk classes showing an incremental unplanned ICU admission rate: class A (score 0), 0.7%; class B (score 1), 1.7%; class C (score 2), 3%; class D (score 3), 7.1%; class E (score 4), 12%; and class F (score > 4), 13% (P < 0.001). CONCLUSIONS This risk score may assist in reliably planning the response to a sudden increase in the demand of critical care resources.