Early Prediction Model of Post-Stroke Cognitive Impairment Based on Routine Clinical Blood Biomarkers

Objective: To evaluate the predictive value of routine clinical blood biomarkers obtained early after admission for post-stroke cognitive impairment (PSCI) and to develop and validate an early prediction model. Methods: Consecutive patients with first-ever acute ischemic stroke admitted to Shanghai Fourth People’s Hospital between March and December 2024 were enrolled. Ten routine biomarkers measured within 24 h of admission were collected: C-reactive protein (CRP), thyroid-stimulating hormone (TSH), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), lymphocyte count (LYC), mean corpuscular hemoglobin concentration (MCHC), glycated hemoglobin (HbA1c), homocysteine (HCY), β-hydroxybutyrate (β-HB), and total cholesterol (TC). Cognitive function was assessed at 3 months using the Telephone Montreal Cognitive Assessment (T-MoCA); PSCI was defined as T-MoCA <19. Patients were randomly split (7:3) into a training set (n=92) and a validation set (n=39). Predictors were selected using LASSO regression, and a Cox proportional hazards model was built and visualized with a nomogram. Model performance was evaluated by ROC/AUC, C-index, decision curve analysis (DCA), and calibration. The study protocol complied with the Declaration of Helsinki was registered in the Chinese Clinical Trial Registry (ChiCTR2400082449). Results: A total of 131 patients were included (PSCI, n=65; non-PSCI, n=66). LASSO selected four candidate predictors: TSH, LYC, HCY, and TC. The model achieved an AUC of 0.69 (95% CI, 0.632–0.742) and a C-index of 0.687 in the training set, and an AUC of 0.58 (95% CI, 0.485–0.673) and a C-index of 0.579 in the validation set. DCA suggested a net clinical benefit across threshold probabilities of 0%–75% in the training set and 0%–56.25% in the validation set. The Hosmer–Lemeshow test indicated good calibration (P>0.05). Conclusion: TSH, LYC, HCY, and TC showed potential value for early PSCI prediction; however, the model based solely on routine blood biomarkers demonstrated limited discrimination. Larger multicenter cohorts and more specific biomarkers and/or multimodal features are warranted to improve early PSCI risk stratification.