Prospective external validation of the automated PIPRA multivariable prediction model for postoperative delirium on real-world data from a consecutive cohort of non-cardiac surgery inpatients.

Abstract

Objectives: Postoperative delirium (POD) is a common complication in surgical patients over 60, increasing morbidity, mortality and hospital stays. While international guidelines recommend risk screening, resource constraints limit implementation. This study externally validated the Pre-Interventional Preventive Risk Assessment (PIPRA) algorithm, a CE-certified tool for identifying high-risk patients to enable targeted prevention.

Methods: A prospective validation study was conducted at a 335-bed Swiss hospital as part of a quality improvement initiative. Data from 866 patients aged ≥60 undergoing non-cardiac, non-intracranial surgery (May-June 2023) were analysed. The PIPRA model's performance was assessed on discrimination (Area Under the Receiver Operating Characteristic Curve (AUROC)) and calibration.

Results: POD occurred in 11.5% (n=100) of patients. The PIPRA model showed good discrimination (AUROC=0.77, 95% CI: 0.72 to 0.82) and generally accurate calibration, though slightly overpredicting risk in high-risk patients. POD was associated with higher mortality, prolonged intensive care unit (ICU)/hospital stays and increased nursing care needs. The model effectively stratified patients for targeted interventions.

Discussion: The PIPRA algorithm demonstrated robust performance in a real-world setting, affirming its utility for POD risk prediction. The study highlighted the model's applicability across diverse clinical environments, despite differences in patient populations and screening protocols.

Conclusions: The PIPRA algorithm is a reliable tool for identifying surgical patients at risk of POD, supporting early intervention strategies to improve patient outcomes. Its integration into clinical workflows may enhance POD prevention efforts and optimise resource allocation in perioperative care.