Authors (including presenting author) :
Yeung LY(1)(2), Li WC(1), Wan HYS(2), Fung SS(2), Leung KN(2), Wong LY(2)
Affiliation :
(1)School of Nursing, The University of Hong Kong (2)Department of Cardiothoracic Surgery, Queen Mary Hospital
Keyword 1: :
postoperative delirium prevention
Keyword 2: :
cardiac surgery
Keyword 3: :
latent profile analysis
Keyword 4: :
preoperative vulnerability
Keyword 5: :
cognitive frailty
Keyword 6: :
perioperative risk stratification
Introduction :
Postoperative delirium (POD) occurs in 15-55% of cardiac surgery patients and substantially increases mortality, prolongs hospitalization, and impairs long-term cognitive recovery. Conventional univariate risk models fail to capture the co-occurrence of multiple preoperative vulnerabilities. Whether distinct preoperative vulnerability phenotypes exist and differentially predict POD risk remains unexplored.
Objectives :
To identify preoperative vulnerability phenotypes using latent profile analysis (LPA) and examine their association with POD.
Methodology :
A prospective cohort study enrolled adults (≥18 years) undergoing elective coronary artery bypass grafting (CABG), valve repair/replacement, combined CABG and valve surgery, or great vessel replacement surgery at a university-affiliated hospital. Preoperative vulnerability was operationalized across five domains: (1) psychological distress (Patient Health Questionnaires-4, Perceived Stress Scale-10); (2) sleep dysregulation (Insomnia Severity Index); (3) neurocognitive reserve (Montreal Cognitive Assessment); (4) frailty (Edmonton Frailty Score, NYHA class); and (5) medical burden (Charlson Comorbidity Index). All measures were z-score standardized and entered into latent profile analysis. POD was assessed using the 4 ‘A’s Test (4AT), administered twice daily from postoperative days 0–5 by trained nurses. POD was defined as ≥1 positive event. Between-phenotype differences were examined using BCH-corrected distal outcome analysis with an overall Wald test (p< 0.05). Sensitivity analysis excluded antipsychotic users.
Result & Outcome :
A total of 159 participants were recruited (mean age 61.7±12.1 years; 65.4% male). Surgical procedures included valve surgery (46.5%), CABG (32.1%), great vessel replacement (11.3%), and combined surgery (10.1%). Overall, POD prevalence was 18.9%. A four-class solution demonstrated superior fit (entropy = 0.87): low vulnerability/resilient (58.1%, n ≈ 92), multimorbidity–cognitive vulnerability (21.0%, n ≈ 33), psychological–sleep vulnerability (15.7%, n ≈ 25), and severe psychiatric vulnerability (5.1%, n ≈ 8). In sensitivity analysis, four-class latent phenotype structure remained stable and demonstrated preserved pattern of POD risk across phenotypes. POD incidence differed significantly across phenotypes (Wald χ² = 11.335, p = 0.010). POD incidence was highest in the multimorbidity–cognitive vulnerability phenotype (47.4%, 95% CI 27.2–67.6%) and lowest in the low vulnerability/resilient phenotype (8.7%, 95% CI 1.4–16.0%). Compared with the low vulnerability/resilient phenotype, the multimorbidity–cognitive vulnerability phenotype showed 9-fold higher odds of POD (OR = 9.4; 95% CI: 2.5–37.0; p = 0.001). Four distinct preoperative vulnerability phenotypes were identified, with multimorbidity-cognitive vulnerability phenotype demonstrating 9-fold elevated POD risk. Phenotype-informed stratification may support targeted, multi-component delirium prevention aligned with dominant vulnerability pathways and facilitate more efficient perioperative resource allocation.
