Development and validation of a nomogram prediction model for perioperative delirium in older patients with osteoporotic fractures based on LASSO regression and multiple parameters

Background Perioperative delirium (POD) is a severe postoperative complication in older patients with osteoporotic fractures. This study aimed to establish a nomogram prediction model incorporating multisystem stress, inflammation and frailty-related biomarkers for risk prediction. Methods This retrospective cohort study included 176 older patients (≥65 years) with osteoporotic fractures. Potential predictors, including the rate-pressure product (RPP), systemic inflammation response index (SIRI) and prognostic nutritional index (PNI), were analysed. Least absolute shrinkage and selection operator regression was utilised for feature selection to construct a multivariable logistic regression nomogram model, which was internally validated via bootstrap resampling. Results Least absolute shrinkage and selection operator regression and logistic regression identified six independent predictors of POD: age, preoperative dementia (cognitive dysfunction), elevated admission RPP, elevated SIRI, decreased PNI and prolonged postoperative intensive care unit (ICU) stay. These indicators serve as predictive markers rather than confirmed causal factors for POD. The nomogram model presented favourable discriminative ability (area under the curve >0.85). The calibration curve showed favourable consistency between the predicted probability of POD and the actual incidence. Decision curve analysis indicated that the model might yield a potential net clinical benefit. Conclusion A prediction model integrating RPP, SIRI and PNI showed favourable performance in assessing POD risk among older patients with osteoporotic fractures, which may provide a reference for clinicians to conduct individualised pre-emptive interventions. However, interpretation of model performance should be conducted with caution given the inherent limitations of this study.