Association of body mass index with postoperative pulmonary complications and potential explanatory pathways in paediatric scoliosis correction surgery: a multicentre cohort study

Background Postoperative pulmonary complications (PPCs) contribute substantially to morbidity after paediatric scoliosis correction surgery, but the relationship between preoperative body mass index (BMI) and PPCs remains uncertain. We evaluated this association and explored potential explanatory pathways and the incremental predictive value of BMI.Methods This retrospective cohort study included 1,165 patients from three Chinese tertiary hospitals. Associations between BMI and the composite PPC outcome were evaluated using multivariable regression. An orthogonalized causal forest estimated the average partial effect of continuous BMI on PPC probability, and variation in model-derived partial-effect predictions across sex and preoperative Cobb-angle strata was examined descriptively. Exploratory pathway analyses considered the Prognostic Nutritional Index (PNI), intraoperative blood loss, and fusion levels. At the completion of surgery, the incremental predictive value of BMI was assessed using an elastic-net model developed in centres 1 and 3 and externally validated in centre 2.Results PPCs occurred in 321 patients (27.55%). In the primary preoperative-adjusted model, each 1-kg/m2 higher BMI was associated with lower odds of PPCs (aOR, 0.87; 95% CI, 0.83–0.91; p < 0.001). A secondary exploratory perioperative-adjusted model yielded a similar estimate (aOR, 0.85; 95% CI, 0.81–0.89; p < 0.001). Compared with normal weight, thinness was associated with higher odds of PPCs (aOR, 4.26; 95% CI, 2.74–6.67; p < 0.001), whereas overweight was associated with lower odds (aOR, 0.57; 95% CI, 0.34–0.91; p = 0.022). The causal forest yielded an inverse average partial-effect estimate for BMI (−0.031; 95% CI, −0.037 to −0.025; p < 0.001). Exploratory pathway analyses suggested small indirect components through PNI, blood loss, and fusion levels. Adding BMI increased the external-validation AUC from 0.658 to 0.703 (p < 0.001), representing a modest improvement in discrimination.Conclusions Higher preoperative BMI was associated with lower odds of PPCs, with a concordant inverse causal-forest estimate under its identification assumptions. Low BMI or thinness may serve as a preoperative risk marker, while the examined pathways accounted for only small proportions of the association. At the end-of-surgery prediction time point, adding BMI provided modest incremental predictive value, although the model’s external performance was modest and further prospective validation is required.