Abstract: BACKGROUND: Currently, constructing a risk predictive model for post-stroke cognitive impairment mostly depends on logistic regression, with relatively few studies incorporating Lasso regression for variable selection to address collinearity and overfitting.
OBJECTIVE: To explore the factors associated with post-stroke cognitive impairment following hypertensive intracerebral hemorrhage and to construct a nomogram prediction model using LASSO regression, followed by model validation.
METHODS: A total of 260 intracerebral hemorrhage patients admitted to the Neurology Emergency Department of the Affiliated Hospital of Guizhou Medical University from August 2022 to August 2024 were initially selected, of whom 127 were classified into the post-stroke cognitive impairment group and 133 into the post-stroke non-cognitive impairment group. Feature selection was optimized using Lasso-logistic regression, and all cohorts were randomly divided into a training set (182 cases) and a validation set (78 cases) in a 7:3 ratio using R Studio software. A risk prediction nomogram model was constructed based on independent risk factors identified from the training set. The model’s discriminative ability was evaluated using the receiver operating characteristic curve, calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test and calibration curve, and clinical benefits were evaluated using a decision analysis curve.
RESULTS AND CONCLUSION: (1) Lasso-logistic regression analysis identified the following independent risk factors for post-stroke cognitive impairment after hypertensive intracerebral hemorrhage: age [odds ratio (OR)=1.112, 95% confidence interval (CI)=1.068-1.157, P=0.000), hematoma diameter (OR=2.021, 95% CI=1.025-3.983, P=0.042), intraventricular rupture (OR=2.398, 95% CI=1.149-5.006, P=0.020), surgery (OR=2.542, 95% CI=1.278-5.056, P=0.008), and serum creatinine levels (OR=1.017, 95% CI=1.004-1.031, P=0.010). (2) A nomogram prediction model was constructed accordingly. The receiver operating characteristic curve analysis revealed an area under the curve for the training and validation sets to be 0.826 (95% CI=0.765-0.885) and 0.795 (95% CI=0.693-0.898), respectively. (3) The Hosmer-Lemeshow goodness-of-fit test and calibration curve analysis showed a good fit of the nomogram model, with a χ2 value of 12.710 and a P-value of 0.122 (P > 0.05) for the training set (χ2=12.170, P=0.122 > 0.05) and the validation set (χ2=4.328, P=0.826 > 0.05). (4) The clinical decision curve demonstrated considerable clinical net benefit of the model. In conclusion, the nomogram model based on predictive factors such as age, hematoma diameter > 3 cm, intraventricular rupture, surgery, and serum creatinine levels has a significant predictive value for cognitive impairment within 3 months after hypertensive intracerebral hemorrhage.

Key words: hypertensive intracerebral hemorrhage, cognitive impairment, novel inflammatory markers, nomogram prediction model, risk factors, lasso regression