This study is a retrospective study based on hospital medical records. The database was established by the Department of Neurology at Yangzhou University Affiliated Hospital. Patients with Acute minor cerebral infarction diagnosed and treated at Yangzhou University Affiliated Hospital from September 2023 to October 2024 were selected. Each enrolled patient underwent a structured telephone follow-up assessment (mRS) precisely at 90 ± 7 days after their individual stroke onset date. Thus, follow-up assessments were conducted between December 2023 and January 2025. All clinical data collection and database locking were completed in 2025. Inclusion criteria: (1) Adults aged 18–90 years, regardless of gender; (2) mRS score ≤ 2 upon admission, patients with first-time cerebral infarction or patients with a history of cerebral infarction who have new-onset cerebral infarction; (3) Completion of polysomnography. Exclusion criteria: (1) Exclusion of patients with Restless Legs Syndrome (RLS), Rapid Eye Movement Sleep Behavior Disorder (RBD), etc.; (2) Exclusion of patients who have received interventions for PLMS (such as dopaminergic drugs, iron supplements, etc.) or rehabilitation treatment plans after stroke; (3) Acute or non-acute cerebral infarction with NHISS ≥ 5, history of acute cerebrovascular accident, brain tumor, head injury, or spinal cord injury; (4) Severe limb paralysis unable to cooperate, severe and unstable conditions (congestive heart failure, r0065spiratory failure, liver failure, or end-stage renal disease) or comorbidities with other central nervous system diseases (Parkinson’s disease, dementia, or mental disorders); (5) Patients with dementia or severe cognitive impairment; (6) Patients with incomplete or missing information on certain scales. This study was approved by the Ethics Committee of Yangzhou University (Ethics 2023-YKL09), and all patients provided informed consent.

Collect basic information about patients from the medical record system, including gender, age, BMI, medical history (hypertension, diabetes, coronary heart disease, etc.), smoking and drinking history. At the same time, collect laboratory biochemical parameters, including white blood cell count, C-reactive protein, lipid levels, HCY, creatinine, VB12, glycosylated hemoglobin, and the NHISS score of patients upon admission.

The included patients have all undergone monitoring using the SOMNOmedics V6 polysomnography device, which includes the following neurophysiological indicators: six-channel electroencephalogram, two-channel electrooculogram, two-channel electromyogram of the jaw, electrocardiogram, nasal airflow, snoring microphone, chest and abdominal movement, blood oxygen saturation, and leg movement events. According to The World Association of Sleep Medicine (WASM) standards (18), leg movement events are manually analyzed by three professional polysomnographic technologists. If the interpretation results differ, the three technologists will discuss and decide together. Leg movement events are represented by the periodic limb movement index (PLMI), which indicates the frequency of periodic leg movements occurring per hour, with each movement lasting 0.5 to 1.0 s and an interval of 5.0 to 90.0 s. The occurrence of three or more movements is considered one instance of periodic limb movement.

All patients underwent neuroimaging examinations in the radiology department of our hospital, and MR images were obtained using a 3.0T MR scanner. The locations of the infarct lesions were collected, including the cortex, basal ganglia, centrum semiovale, corona radiata, thalamus, cerebellum, and brainstem.

The baseline National Institutes of Health Stroke Scale (NIHSS) score at the time of admission was collected to assess the degree of baseline neurological deficit.

The modified Rankin Scale (mRS) was assessed via structured telephone interview at a target timepoint of 90 ± 7 days post-stroke onset. To ensure rating consistency and minimize inter-rater variability, all participating neurologists completed a certified online mRS training and certification module prior to study commencement. Furthermore, any cases with ambiguous functional status during the telephone interview were independently adjudicated by a third senior neurologist blinded to the initial assessment. Actual follow-up intervals (defined as days from stroke onset to mRS assessment) were recorded for all patients. As shown in Supplementary Table 1, the median follow-up time was 91 days (IQR: 90–94 days). This narrow window minimizes differential misclassification bias. The 3-month follow-up was selected based on established evidence that spontaneous neurological recovery plateaus by 90 days in minor cerebral infarction patients, with minimal functional gains thereafter. This is consistent with the AHA/ASA guideline recommendation that assessing outcomes at this critical time point captures the greatest potential for recovery before chronic disability has stabilized (19). The mRS Scale is a widely used and clinically validated measure of global disability and functional outcomes after stroke. It assesses the degree of dependence in daily activities on a scale ranging from 0 (no symptoms) to 6 (death). A score of 0–2 indicates functional independence (good outcome), while a score of 3–5 indicates varying degrees of disability requiring assistance, and 6 indicates death (20). Based on the scoring results, follow-up participants were grouped into a good prognosis group (0–2 points) and a poor prognosis group (3–6 points). All scores were completed by more than two experienced senior neurologists.

Statistical analysis was performed using SPSS 26.0 (IBM Corp., Armonk, United States). All data were first subjected to normality and homogeneity of variance tests. Normally distributed numerical data were expressed as mean ± standard deviation (SD), while skewed numerical data were expressed as median and interquartile range (IQR). Intergroup comparisons were made using independent samples t-test or rank-sum test. Categorical data were expressed as percentages, and intergroup comparisons were conducted using chi-square tests. The correlation between sleep-related scales and polysomnography with prognosis was analyzed using Spearman correlation analysis. Benjamini–Hochberg false discovery rate (FDR) correction was applied in the correlation analysis to control for type I error inflation. All p-values reported by Spearman correlation analysis are FDR-corrected q-values. Further, indicators related to prognosis (data with p < 0.05 in correlation analysis) along with age, hypertension, cholesterol, and brain infarction in the corona radiata were included in a binary logistic regression analysis to assess factors affecting the prognosis of patients with acute mild stroke. All statistical tests were two-sided, showing odds ratios (OR) with 95% confidence intervals (CI), and p < 0.05 was considered statistically significant.

Although mRS is ordinal, we selected Spearman correlation because: (1) It is the most widely used method for stroke prognosis studies(21); (2) Our primary predictors (PLMI) is continuous; (3) Simulation studies show minimal difference between Spearman and Kendall for n > 200 (22).

A total of 766 patients were included, with older age, a history of hypertension, and higher HCY levels associated with poorer prognosis (p = 0.003; p < 0.001; p = 0.047; Table 1). There were no statistically significant differences between the two groups in terms of gender, BMI, medical history (hypertension, diabetes, coronary heart disease, etc.), smoking and drinking history, and biochemical indicators (white blood cell count, C-reactive protein, lipid levels, creatinine, VB12, and glycated hemoglobin) (all p > 0.05).

Patients in the poor prognosis group had higher PSQI scores (p = 0.003, Table 2), while no significant abnormalities were observed in ESS, HADS, MoCA, and SCD-Q9 scores between the two groups (all p > 0.05). Table 3 presents a comparison of PSG results between the two groups, indicating that patients with poor prognosis had longer total sleep time (p = 0.014), lower sleep efficiency (p = 0.022), longer proportion of N1 sleep (p < 0.001), higher AHI (p = 0.020), and higher PLMS-related awakening index (p = 0.004) and PLMI (p = 0.001). No differences were found between the two groups for other indicators.

As shown in Table 4, patients with poor prognosis are more likely to have infarction located in the watershed area (p = 0.017). There are no statistically significant differences in other imaging locations between the two groups of patients (p > 0.05).

Figure 1 shows the correlation analysis between sleep-related scales and PSG results and the prognosis of patients. We have applied the Benjamini-Hochberg false discovery rate (FDR) correction to all Spearman correlation analyses. The results indicate that N1% (r_S = 0.117, p = 0.006), PLMSAI (r_S = 0.129, p = 0.006), and PLMI (r_S = 0.196, p = 0.006) are positively correlated with prognosis, while SE (r_S = −0.117, p = 0.006) is negatively correlated with prognosis.

Indicators with p < 0.05 from the correlation analysis, along with age, hypertension, HCY, and radiate crown area cerebral infarction, were included as covariates, with mRS grouping as the dependent variable (Table 5). The results indicated that patients with a history of hypertension (OR = 0.589, 95% CI = 0.401 to 0.863), higher HCY levels (OR = 1.037, 95% CI = 1.005 to 1.070), radiate crown area cerebral infarction (OR = 1.655, 95% CI = 1.150 to 2.382), longer N1% (OR = 1.032, 95% CI = 1.013 to 1.052), and higher PLMI (OR = 1.006, 95% CI = 1.001 to 1.010) are risk factors for the prognosis of Acute minor cerebral infarction patients at 3 months (p < 0.05). To enhance clinical interpretability, Figure 2 presents a forest plot of the adjusted odds ratios from the binary logistic regression. Although all five factors are risk factors for poor outcome, PLMI was a novel and modifiable target. Systematic sleep monitoring (e.g., polysomnography) in patients with minor acute stroke can identify high-risk groups for early PLMS intervention, which may improve long-term neurologic recovery.

To assess the robustness of our primary findings against potential variations in the precise timing of the 90-day assessment window. We performed a pre-specified sensitivity analysis. We restricted the analysis to patients assessed within a narrower, more stringent window of 85–95 days post-stroke (n = 734, excluding n = 32 outside this range). This sub-analysis confirmed that a higher PLMI remained significantly associated with poor prognosis (adjusted OR = 1.006, 95% CI = 1.001 to 1.011, p = 0.032; see Supplementary Table 2), supporting the stability of the association even when focusing on assessments conducted closest to the theoretical peak of recovery.

This study has some limitations. First, the single-center design with retrospective data collection may introduce regional/population selection bias and limit generalizability. Additionally, reliance on a single PSG session cannot account for night-to-night variability in sleep parameters, potentially affecting the robustness of sleep-related associations. Second, Regarding the issue of the ordinal nature of mRS. And the possibility of side-ranking in sleep-related measures such as PSQI, while Kendall’s tau-b is a suitable non-parametric correlation measure for ordinal data and less sensitive to ties, Spearman’s correlation was chosen primarily because of its wide use in stroke outcome studies and applicability to non-normally distributed variables. Crucially, the core findings relied on binary logistic regression, which essentially addressed the conventional question and confirmed that PLMI was an independent risk factor (OR = 1.006, p = 0.021). The consistency of the key correlations (e.g., PLMI-mRS: r_S = 0.196, p = 0.006) and regression results indicates that our conclusions are robust. Third, the follow-up period was relatively short, which may not adequately reflect the potential impact of PLMS on long-term recurrence rates, cognitive function deterioration, or mortality. Finally, the assessment of neurological outcome via telephone-administered mRS, while practical for large cohorts, may introduce subjectivity and inter-rater variability despite being conducted by experienced neurologists. Although we implemented standardized training and adjudication protocols, subtle functional deficits could be underdetected in remote assessments. Future prospective studies with multi-center designs, serial PSG, in-person assessments, and extended follow-up are warranted.