Mining the risk factors for stroke occurrence and dietary protective factors based on the NHANES database: Analysis using SHAP.

BACKGROUND

Stroke has a high morbidity and mortality rate and its incidence is rising. This study aimed to identify risk factors for stroke, develop an accurate prediction model and scoring system, thereby providing novel primary prevention strategies for stroke and post-stroke depression.

METHODS

Data from 49,491 participants in the National Health and Nutrition Examination Survey (NHANES) (1999-2016) were analyzed. A training cohort of 3972 individuals was created. Logistic regression was applied for univariate and multivariate analyses to identify risk factors. SHAP values analyzed the impact of these factors, leading to the development of a prediction model and scoring scale. Model performance was validated using DCA, ROC, and calibration curves.

RESULTS

A prediction model with an AUC of 82 % was developed, alongside a scoring scale with an AUC of 79 %. Key risk factors were identified as "age(OR = 1.061, 95%CI:1.055-1.066, P<0.001)", "income level(OR= 0.812, 95%CI:0.777-0.85, P<0.001)", "cholesterol(OR = 0.845, 95%CI:0.793-0.901, P<0.001)", "albumin(OR = 0.614, 95%CI:0.497-0.757, P<0.001)", "lymphocyte percentage(OR= 1.045, 95%CI:1-1.091, P=0.048)", "alkaline phosphatase(OR= 1.005, 95%CI:1.002-1.002, P<0.001)", " Glucose(OR= 1.059, 95%CI:1.026-1.093, P<0.001)", "triglycerides(OR = 1.061, 95%CI:1.004-1.123, P = 0.037)", and "neutrophil percentage(OR= 1.058, 95%CI:1.015-1.102, P=0.008)".

CONCLUSION

In this study, a simple scoring scale was successfully developed based on the research findings. The scale demonstrates excellent clinical usability and operational convenience, providing a novel self-screening tool for the occurrence of stroke and post-stroke depression.