Background: Delirium following cardiac surgery is common, morbid, and costly, but may be prevented with risk stratification and targeted intervention. Preoperative protein signatures may identify patients at increased risk for worse postoperative outcomes, including delirium. In this study, we aimed to identify plasma protein biomarkers and develop a predictive model for postoperative delirium in older patients undergoing cardiac surgery, while also uncovering possible pathophysiological mechanisms. Methods: SOMAscan analysis of 1,305 proteins in the plasma from 57 older adults undergoing cardiac surgery requiring cardiopulmonary bypass was conducted to define delirium-specific protein signatures at baseline (PREOP) and postoperative day 2 (POD2). Selected proteins were validated in 115 patients using the ELLA multiplex immunoassay platform. Proteins were combined with clinical and demographic variables to build multivariable models that estimate the risk of postoperative delirium and bring light to the underlying pathophysiology. Results: A total of 666 proteins from SOMAscan analysis were found altered between PREOP and POD2 (Benjamini-Hochberg (BH)-p<0.01). Using these results and findings from other studies, 12 biomarker candidates (Tukey's fold change (|tFC|)>1.4) were selected for ELLA multiplex validation. Eight proteins were significantly altered at PREOP, and seven proteins at POD2 (p<0.05), in patients who developed postoperative delirium compared to non-delirious patients. Statistical analyses of model fit resulted in selecting a combination of age, sex, and different 3-protein biomarker panels highly correlated with delirium at PREOP (angiopoietin-2, ANGPT2; C-C motif chemokine 5, CCL5; and metalloproteinase inhibitor 1, TIMP1; AUC=0.829) and POD2 (lipocalin-2, LCN2; neurofilament light chain, NFL; and CCL5; AUC=0.845). The delirium-associated proteins identified as biomarker candidates are involved with inflammation, glial dysfunction, vascularization, and hemostasis, highlighting the multifactorial pathophysiology of delirium. Conclusion: Our study proposes two models of postoperative delirium that include a combination of older age, female sex, and altered levels of proteins both preoperatively and postoperatively. Our results support the identification of patients at higher risk of developing postoperative delirium after cardiac surgery and provide insights on the underlying pathophysiology. ClinicalTrials.gov (NCT02546765).