Objective: To create a prediction model for stroke risk among middle-aged and elderly populations, so as to provide a basis for early identification of high-risk population for stroke. Methods: From October to December 2023, residents aged ≥45 years in Chongchuan District, Nantong City, Jiangsu Province were selected using a multi-stage stratified random sampling method. The demographic information, life behavior, and chronic disease data were collected through a questionnaire survey. The standardized prevalence of stroke was calculated using data from the seventh National Population Census. The subjects were randomly divided into the training set and the internal validation set according to the ratio of 8∶2. The basic demographic information, life behavior, and chronic diseases of residents aged ≥45 years in Rugao City were collected from July to August 2023 as the external validation set. Predictive factors were selected using multivariable logistic regression model, and a nomogram for stroke among residents aged ≥45 years was established. The prediction effect was evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC), calibration curve, and Hosmer-Lemeshow goodness of fit test. Results: A total of 6 290 residents aged ≥45 years were included, including 2 975 males (47.30%) and 3 315 females (52.70%). The average age was (61.90±10.20) years. The prevalence of stroke was 3.80%, and the standardized prevalence was 3.36%. The multivariable logistic regression showed that age, smoking, hypertension, and hyperlipidemia were predictors of stroke risk among residents aged ≥45 years, and the prediction model was ln[p/(1-p)]=-4.619+0.046×age+0.383×smoking+0.887×hypertension+0.678×hyperlipidemia. The AUC values of the training set, internal validation set, and external validation set were 0.748, 0.755, and 0.738, respectively. The consistency indexes were 0.748, 0.755, and 0.738, respectively. The Hosmer-Lemeshow goodness of fit test showed a good fitting effect (P>0.05). Conclusion The prediction model based on age, smoking, hypertension, and hyperlipidemia has good discrimination and calibration, and can be used to predict the risk of stroke among middle-aged and elderly populations aged ≥45 years.