Objective To explore the correlation between the geriatric nutritional risk index （GNRI） and unplanned readmission （readmission） in elderly stroke patients. Methods The clinical data of 307 elderly stroke patients admitted to the Yancheng Third People's Hospital， Jiangsu Province from June 2019 to May 2022 were retrospectively analyzed. According to the median GNRI of all patients， they were divided into a normal GNRI group （GNRI≥92） and a low GNRI group （GNRI＜92）. The propensity score matching （PSM） was used to obtain two sets of samples with 13 variables balanced， such as age and albumin at admission. Logistic regression was used to analyze the impact of GNRI on readmission before and after PSM. The predictive value of GNRI on readmission was evaluated using the area under the receiver operating characteristic （ROC） curve （AUC） alone and after adjusting for other covariates. Results Among 307 elderly stroke patients， the GNRI was （92.62±25.30）， and 43.32%（133/307）patients had normal GNRI. After PSM， a total of 126 pairs were successfully matched. There was no statistical difference between the two groups in age， body mass index，diabetes， hospital stay， surgical treatment ratio， albumin and other clinical data （P ＞ 0.05）. After PSM， the readmission rate of the low GNRI group was 28.57% （36/126）， which was higher than the 11.90% （15/126） of the normal GNRI group， and the difference was statistically significant （χ2 =10.841， P=0.001）. Logistic regression analysis after PSM showed that in Model 1， which only included GNRI， the readmission risk of patients with low GNRI increased by 2.232 times compared to those with normal GNRI， and the difference was statistically significant ［OR=3.232，95%CI（1.701，6.138），P=0.001］. In Model 2， after adjusting for age and gender， the readmission risk of patients with low GNRI increased by 2.206 times compared to those with normal GNRI， and the difference was statistically significant ［OR=3.206，95%CI（1.674，6.138），P=0.001］. In Model 3， after adjusting for all variables such as age and gender， the readmission risk of patients with low GNRI increased by 2.052 times compared to those with normal GNRI， with a statistically significant difference ［OR=3.052， 95%CI （1.469， 6.339）， P=0.003］. ROC curve analysis showed that the predicted AUC for readmission of Model 1 was 0.821 ［95%CI （0.755， 0.886）］. After adjusting for covariates using Model 2 and Model 3， the predicted AUC for readmission were 0.828 ［95%CI （0.770， 0.886）］ and 0.847 ［95%CI （0.797， 0.898）］， respectively. The AUC of Model 3 was higher than that of Model 1 and Model 2， and the difference was statistically significant （Z=3.036， 2.457； P=0.002， 0.014）. Conclusions GNRI can effectively predict the readmission of elderly stroke patients and has an important warning role， which may guide medical and nursing staff in nutritional management of elderly stroke patients.