The relationship between periodic leg movements of sleep (PLMS) and cognitive function is controversial. This study aims to explore: (1) the effects of PLMS on cognitive function in middle-aged and elderly patients; and (2) the mediating role of rapid eye movement (REM) sleep duration and PLMS-related arousal between PLMS and cognitive performance. This retrospective study was conducted based on inpatient medical records. Middle-aged and elderly patients hospitalized in the Department of Neurology at the Affiliated Hospital of Yangzhou University from September 2023 to August 2024 were selected. Participants were those who underwent polysomnography (PSG) due to sleep-related complaints, such as daytime sleepiness, nocturnal snoring, insomnia, or frequent nighttime awakenings. The Montreal Cognitive Assessment (MoCA) was used to evaluate the cognitive function of patients. Participants were stratified into two groups based on MoCA scores (MoCA ≥ 26 and MoCA < 26) to investigate the differential effects of PLMS on cognitive function across distinct cognitive profiles. Multivariate linear regression models were used to evaluate the association between PLMS (periodic limb movement index, PLMI) and cognitive performance, adjusted for covariates including age, gender, education, smoking, total sleep time, and apnea-hypopnea index. Mediation analyses were performed to test whether REM sleep duration and PLMS-related arousal index (PLMAI) mediated the PLMS-cognition relationship. A total of 818 patients were included. Regression analysis showed that PLMI (β = - 0.033, 95% CI - 0.039 to - 0.026, p < 0.001) was negatively correlated with cognitive function. After adjusting for REM sleep duration and PLMAI, PLMI was still significantly associated with cognitive function (β = - 0.026, 95% CI - 0.034 to - 0.018, p < 0.001). PLMI was negatively correlated with REM sleep duration (β = - 0.157, 95% CI - 0.258 to - 0.056, p = 0.002) and positively correlated with PLMS-related arousal (β = 0.036, 95% CI 0.029-0.043, p < 0.001). The results of mediation analysis showed that REM sleep duration and PLMS-related arousal partially mediated the association between PLMS and cognitive function (indirect effect estimate = - 0.003, - 0.007; direct effect estimate = - 0.033). PLMS is associated with poorer cognitive function in middle-aged and elderly people, and REM sleep duration and PLMS-related arousal play a partial mediating role. Clinicians can utilize periodic limb movements during sleep (PLMS) as a potential early warning indicator of cognitive decline, enabling timely interventions to slow its progression. Additionally, improving REM sleep duration and suppressing PLMS-related arousals through pharmacological or non-pharmacological treatments may serve as critical strategies for preserving cognitive function.