Glycemic Variability and Clinical Outcomes in ICU Patients with Ischemic Stroke: A Multi-Database Retrospective Cohort Study

Glycemic Variability and Clinical Outcomes in ICU Patients with Ischemic Stroke: A Multi-Database Retrospective Cohort Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Glycemic Variability and Clinical Outcomes in ICU Patients with Ischemic Stroke: A Multi-Database Retrospective Cohort Study Yong Xia, Yanbin Chen, Liu Yang, Jingyuan Kang, Qiao Huang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7686095/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 28 Mar, 2026 Read the published version in European Journal of Medical Research → Version 1 posted 11 You are reading this latest preprint version Abstract Background Glycemic variability (GV) is an emerging marker of metabolic stress in critically ill patients, but its prognostic value in ICU patients with ischemic stroke remains unclear. Methods This retrospective cohort study used MIMIC-Ⅳ and eICU databases to assess associations between GV and outcomes in adult ischemic stroke patients. GV metrics—including mean blood glucose (MBG), standard deviation (SD), coefficient of variation (CV), and glucose range—were calculated from all ICU glucose readings. The primary outcome was in-hospital mortality; secondary was ICU length of stay (LOS). Multivariable regression and XGBoost machine learning were used, with external validation. Results A total of 418 patients were included (334 survivors, 84 non-survivors). In univariate analyses, non-survivors had significantly lower SD ( P = 0.007), CV ( P = 0.036), and glucose range ( P = 0.038) than survivors, while MBG did not differ significantly ( P = 0.132). However, no GV metric remained independently associated with mortality or ICU LOS after multivariable adjustment ( P > 0.05). XGBoost models showed moderate predictive performance (AUC = 0.657), with GV metrics contributing moderate feature importance. Subgroup analyses indicated a possible protective association between higher CV and mortality in older or non-hypertensive patients. Findings were consistent across databases ( P > 0.6). Conclusions In ICU patients with ischemic stroke, GV was linked to outcomes in unadjusted analyses but was not an independent predictor after adjustment. GV may aid in risk stratification, though traditional clinical variables remain more predictive. glycemic variability ischemic stroke intensive care unit in-hospital mortality blood glucose machine learning MIMIC eICU Figures Figure 1 Figure 2 1 Introduction Ischemic stroke is one of the leading causes of mortality and long-term disability worldwide, accounting for approximately 76% of all stroke cases and affecting over 12 million people annually(1, 2). Among critically ill patients, stroke represents a substantial clinical burden: intensive care unit (ICU) admission is required for up to 15–20% of hospitalized stroke cases, particularly those with severe neurological deficits, respiratory compromise, or multi-organ dysfunction(3, 4) In such high-acuity settings, metabolic dysregulation—particularly disturbances in blood glucose levels—is frequently observed and may exert a considerable impact on clinical outcomes (5, 6). Traditionally, research on glucose in acute stroke has focused on hyperglycemia at admission, which has been consistently associated with poor neurological recovery and higher mortality(7, 8). However, growing evidence highlights glycemic variability (GV)—the degree of glucose fluctuation over time—as a potentially superior prognostic indicator compared to static glucose levels (9, 10). GV reflects the complex interplay of stress-induced hyperglycemia, insulin resistance, and neuroendocrine dysregulation, all of which are amplified in critical care environments(11, 12). Despite these insights, prior investigations into GV have faced notable limitations, including small sample sizes, single-center designs, and limited control for confounding factors. With the increasing availability of large-scale, de-identified ICU databases—such as the Medical Information Mart for Intensive Care Ⅳ (MIMIC-Ⅳ) and the eICU Collaborative Research Database—researchers are now positioned to conduct multicenter, high-resolution studies with enhanced external validity(13, 14). These databases contain high-resolution, time-stamped glucose records, along with detailed patient characteristics, comorbidities, treatment interventions, and outcomes. In this study, we leveraged data from both MIMIC-IV and eICU to investigate the relationship between GV and clinical outcomes in ICU patients with ischemic stroke. Specifically, we aimed to: (1) characterize the distribution and clinical correlates of GV in this population; (2) examine the association between GV and key outcomes, including ICU length of stay and in-hospital mortality; and (3) evaluate the predictive utility of GV using both conventional regression and machine learning models. By providing a more nuanced understanding of glucose dynamics in critically ill stroke patients, this work may contribute to refining individualized glucose management strategies in the neurocritical care setting. 2 Methods 2.1 Data Sources This study utilized data from two publicly available critical care databases: the MIMIC-Ⅳ and the eICU Collaborative Research Database. MIMIC-Ⅳ was developed by the Massachusetts Institute of Technology in collaboration with the Beth Israel Deaconess Medical Center and includes detailed electronic health records of ICU patients from 2008 to 2019 at a single academic center. The eICU database, developed by Philips Healthcare, contains ICU data from over 200 hospitals across the United States between 2014 and 2015. Both databases include de-identified patient-level data in accordance with HIPAA regulations and provide extensive information on demographics, vital signs, laboratory results, glucose measurements, comorbidities, and clinical outcomes. Access to both datasets was granted to credentialed researchers who completed CITI training and data use agreements. As this study used publicly available de-identified data, it was exempt from institutional review board (IRB) approval and informed consent requirements. 2.2 Study Population We included adult patients (aged ≥ 18 years) who were admitted to the ICU with a primary diagnosis of ischemic stroke. Ischemic stroke was identified using International Classification of Diseases (ICD) codes: ICD-9 codes 434.x and ICD-10 codes I63.x. Eligible patients were required to have at least two or more blood glucose measurements recorded during their ICU stay to allow calculation of GV. Patients were excluded if they met any of the following criteria: (1) ICU length of stay less than 24 hours; (2) missing key variables such as glucose data, in-hospital mortality status, or ICU admission/discharge times; (3) presence of comorbid conditions known to severely affect glucose metabolism (e.g., diabetic ketoacidosis, hyperosmolar hyperglycemic state, or acute pancreatitis). If a patient had multiple ICU admissions, only the first ICU stay meeting all inclusion criteria was considered for analysis. Data preprocessing, including patient identification and eligibility filtering, was conducted separately for each database (MIMIC-Ⅳ and eICU) to ensure consistent cohort construction across sources. 2.3 Variables and Definitions 2.3.1 Exposure Variables GV during the ICU stay was assessed using the following four metrics: (1) Mean Blood Glucose (MBG): the arithmetic mean of all recorded glucose values. (2) Standard Deviation (SD): reflects the absolute variation in glucose values. (3) Coefficient of Variation (CV): calculated as (SD / MBG) × 100%, indicating relative variability. (4) Range: the difference between the maximum and minimum glucose values. 2.3.2 Outcome Variables (1) In-hospital mortality: defined as death occurring during the hospitalization associated with the ICU stay, recorded as a binary variable (0 = survived, 1 = died). (2) ICU length of stay (LOS): defined as the number of days between ICU admission and ICU discharge, treated as a continuous variable. 2.3.3 Covariates (1) Demographics : age, sex, and body mass index (BMI). (2) Comorbidities : including diabetes mellitus, hypertension, and atrial fibrillation, identified via ICD codes. (3) Neurological status : Glasgow Coma Scale (GCS) score at ICU admission. (4) ICU characteristics : ICU type (e.g., medical, surgical, neurological) and initial glucose level at ICU entry. All variables were harmonized across databases using consistent definitions and coding. Continuous variables were standardized when appropriate, and categorical variables were converted into binary or dummy-coded formats for inclusion in regression and machine learning models. Neurological status was assessed using the GCS score recorded at ICU admission. While specific imaging-based measures of infarct volume or NIH Stroke Scale (NIHSS) were not available in the databases, GCS was used as a surrogate indicator of baseline neurological function. GCS has been widely applied in ICU outcome research as a proxy for brain injury severity. 2.4 Data Extraction and Processing Clinical data were extracted separately from the MIMIC-Ⅳ and eICU databases using structured query language (SQL). For each patient, demographic information, diagnostic codes, blood glucose measurements, ICU admission and discharge times, comorbidities, and outcome variables were retrieved. (1) Patient Identification: ICU patients with a primary diagnosis of ischemic stroke were identified based on ICD-9 codes (434.x) and ICD-10 codes (I63.x). Only the first ICU admission was considered for patients with multiple eligible stays. Inclusion and exclusion criteria were applied to filter the final study cohort as described in Section 2.2. (2) Blood Glucose Data Handling: All available blood glucose values (including bedside point-of-care and laboratory glucose measurements) during the ICU stay were extracted. Values were cleaned to remove physiologically implausible readings (e.g., 600 mg/dL), and only patients with two or more valid glucose records were included for glycemic variability analysis. (3) Variable Harmonization and Transformation: Variables were standardized across the two databases by mapping equivalent fields and ensuring consistent units, formats, and definitions. Continuous variables (e.g., age, glucose metrics) were normalized as needed, and categorical variables (e.g., sex, ICU type, comorbidities) were encoded as binary indicators. (4) Missing Data Management: Patients with missing primary outcome variables (in-hospital mortality or ICU discharge time) were excluded. For covariates with low missingness (e.g., BMI, GCS), median imputation was used. Variables with extensive missing data were excluded from adjusted models. All preprocessing procedures were performed using Python (pandas, numpy), and were conducted separately within each database before merging results for comparative or pooled analyses. 2.5 Statistical Analysis All statistical analyses were conducted using R and Python. Continuous variables were presented as mean ± standard deviation or median [IQR], and categorical variables as counts and percentages. Group comparisons were performed using t-tests, Mann–Whitney U tests, or chi-squared tests, as appropriate. Logistic regression was used to evaluate the association between GV metrics and in-hospital mortality, and linear regression was applied for ICU length of stay, adjusting for age, sex, BMI, comorbidities, GCS, and ICU type. GV metrics were also divided into quartiles for trend analysis. Machine learning models, including XGBoost, were used to predict mortality, with model performance assessed by AUC and feature importance. A two-sided P < 0.05 was considered statistically significant. 3 Results 3.1 Baseline Characteristics A total of 418 ICU patients with ischemic stroke were included in the analysis, of whom 334 survived and 84 died during hospitalization. Baseline characteristics between survivors and non-survivors were compared (Table 1 ). Significant differences were observed in age ( P = 0.012) and GCS score ( P < 0.001), with non-survivors tending to be older and having lower levels of consciousness. Other variables, including sex, BMI, diabetes, hypertension, atrial fibrillation, and ICU type, showed no statistically significant differences between groups ( P > 0.05 for all). Table 1 Baseline Characteristics of ICU Patients with Ischemic Stroke Variable Total (n = 418) Survivors (n = 334) Non-survivors (n = 84) t / χ 2 P Age (years) 65.25 ± 14.44 64.37 ± 14.07 68.77 ± 15.40 -2.518 0.012 Gender [n (%)] 1.321 0.250 Male 215 (51.44%) 177 (52.99%) 38 (45.24%) Female 203 (48.56%) 157 (47.01%) 46 (54.76%) BMI (kg/m 2 ) 26.00 ± 4.05 26.10 ± 4.07 25.58 ± 3.96 1.046 0.296 GCS score (points) 9.68 ± 2.84 9.97 ± 2.73 8.51 ± 2.96 4.299 < 0.001 Diabetes [n (%)] 129 (30.86%) 102 (30.54%) 27 (32.14%) 0.023 0.879 Hypertension [n (%)] 174 (41.63%) 137 (41.02%) 37 (44.05%) 0.144 0.704 Atrial fibrillation [n (%)] 103 (24.64%) 89 (26.65%) 14 (16.67%) 3.083 0.079 ICU type [n (%)] 1.064 0.587 Medical ICU 142 (33.97%) 112 (33.53%) 30 (35.71%) Neuro ICU 139 (33.25%) 115 (34.43%) 24 (28.57%) Surgical ICU 137 (32.78%) 107 (32.04%) 30 (35.71%) 3.2 Distribution of Glycemic Variability Metrics Glycemic variability metrics were compared between survivors and non-survivors (Table 2 ). The SD, CV, and glucose range were all significantly lower in non-survivors compared to survivors ( P = 0.007, P = 0.036, and P = 0.038, respectively). In contrast, MBG did not differ significantly between the two groups ( P = 0.132). Table 2 Distribution of Glycemic Variability Metrics by Clinical Outcome Variable Total (n = 418) Survivors (n = 334) Non-survivors (n = 84) t / χ 2 P Mean Blood Glucose (MBG, mg/dL) 142.15 ± 10.90 142.55 ± 10.78 140.54 ± 11.29 1.511 0.132 Standard Deviation (SD, mg/dL) 28.88 ± 6.64 29.32 ± 6.75 27.14 ± 5.91 2.711 0.007 Coefficient of Variation (CV, %) 20.47 ± 5.11 20.73 ± 5.23 19.43 ± 4.48 2.099 0.036 Glucose Range (mg/dL) 90.59 ± 22.96 91.75 ± 23.30 85.96 ± 21.04 2.077 0.038 3.3 Univariate Associations To further explore the association between glycemic variability and clinical outcomes, patients were stratified into quartiles based on their CV values (Table 3 ). A decreasing trend in in-hospital mortality was observed across increasing CV quartiles, although this trend did not reach statistical significance ( P = 0.463). Median ICU length of stay appeared similar across quartiles, with no consistent trend observed. Figure 1 presents a scatter plot of CV and ICU length of stay. No significant linear correlation was found between CV and ICU stay duration ( r = 0.07, P = 0.147), suggesting limited direct association in the univariate context. Table 3 Quartile Analysis of CV and Its Association With Clinical Outcomes CV Quartile CV Range (%) In-hospital Mortality (%) ICU Length of Stay (median [IQR]) Trend Statistic (Z) P Q1 (Lowest) 4.95–16.97 24.76 4.2 [2.4–6.5] Q2 16.98–20.11 21.15 3.7 [2.2–5.5] Q3 20.15–23.38 17.31 4.6 [2.5–6.7] Q4 (Highest) 23.42–40.24 17.14 4.6 [2.7–7.4] 2.570 0.463 Scatter plot showing the relationship between glycemic CV and ICU length of stay. 3.4 Multivariable Regression Analysis Multivariable logistic regression was performed to assess the independent association between glycemic variability metrics and in-hospital mortality (Table 4 ). After adjusting for demographic and clinical covariates, GCS score ( P < 0.001) and age ( P = 0.005) remained significantly associated with mortality. However, none of the glycemic variability indicators—including CV, SD, MBG, and range—showed a statistically significant association with in-hospital death in the adjusted model ( P > 0.05 for all). In the linear regression model predicting ICU length of stay (Table 5 ), most glycemic variability metrics were not significantly associated with the outcome. The only variable reaching statistical significance was hypertension, which was independently associated with a shorter ICU stay ( P = 0.010). Glycemic metrics, including CV, SD, and MBG, did not show significant associations with ICU length of stay in the adjusted model ( P > 0.05). Table 4 Logistic Regression: Predictors of In-hospital Mortality Variable OR (95% CI) Wald χ 2 P Age 1.03 (1.01–1.05) 2.78 0.005 Gender 0.69 (0.41–1.16) -1.39 0.164 BMI 0.97 (0.91–1.03) -1.02 0.309 GCS score 0.81 (0.74–0.89) -4.33 < 0.001 Diabetes 1.03 (0.59–1.80) 0.11 0.912 Hypertension 1.11 (0.66–1.87) 0.39 0.697 Atrial fibrillation 0.63 (0.32–1.21) -1.39 0.165 MBG 0.95 (0.86–1.04) -1.17 0.243 CV 0.82 (0.44–1.55) -0.60 0.550 SD 1.02 (0.64–1.63) 0.09 0.925 Range 1.02 (0.99–1.05) 1.12 0.264 Table 5 Multivariable Linear Regression: Predictors of ICU Length of Stay Variable β Coefficient (95% CI) Standard Error t P Age -0.00 (-0.02–0.02) 0.01 -0.113 0.910 Gender 0.47 (-0.16–1.10) 0.32 1.449 0.148 BMI 0.08 (-0.00–0.15) 0.04 1.910 0.057 GCS score -0.01 (-0.13–0.10) 0.06 -0.242 0.809 Diabetes 0.32 (-0.36–0.99) 0.35 0.910 0.363 Hypertension -0.84 (-1.47–-0.20) 0.33 -2.571 0.010 Atrial fibrillation 0.45 (-0.28–1.18) 0.37 1.206 0.229 MBG -0.03 (-0.14–0.08) 0.06 -0.487 0.626 CV -0.05 (-0.75–0.64) 0.36 -0.150 0.881 SD 0.09 (-0.43–0.60) 0.26 0.327 0.744 Range -0.01 (-0.04–0.02) 0.02 -0.554 0.580 3.5 Machine Learning Model Performance To further assess the predictive value of glycemic variability and clinical features, an XGBoost classification model was developed to predict in-hospital mortality. As shown in Table 6 , the model achieved an area under the ROC curve (AUC) of 0.657 and an accuracy of 75.4%. However, sensitivity (recall) was relatively low at 16.0%, with precision at 28.6% and an F1 score of 90.1%. Calibration assessed by the Hosmer–Lemeshow test indicated no significant miscalibration ( P = 0.144), and the Brier score was 0.196. Figure 2 A displays the ROC curve, demonstrating moderate discriminative ability. Figure 2 B presents the SHAP-based feature importance, with atrial fibrillation, BMI, GCS score, and age ranking highest in predictive contribution. Glycemic variability metrics such as CV and SD had moderate importance, indicating they provide complementary predictive value alongside major clinical factors. Table 6 Performance Metrics of the XGBoost Model for Predicting In-hospital Mortality Metric Value Area Under the Curve (AUC) 0.657 Accuracy (%) 75.40 Precision (%) 28.57 Recall (Sensitivity, %) 16.00 Specificity (%) 20.51 F1 Score (%) 90.10 AUCPR 0.313 Brier Score 0.196 Calibration P (Hosmer–Lemeshow) 0.144 Panel A shows the ROC curve of the XGBoost model predicting in-hospital mortality among ICU patients with ischemic stroke, with an area under the curve (AUC) of 0.657, indicating moderate discriminative ability. Panel B presents the SHAP (Shapley Additive Explanations) summary plot of feature importance, ranking variables based on their contribution to the model’s predictions. Clinical factors such as atrial fibrillation, BMI, GCS score, and age showed the highest importance, while glycemic variability metrics such as CV and SD contributed moderate predictive value. 3.6 Subgroup and Sensitivity Analyses Subgroup analyses were performed to evaluate whether the association between glycemic CV and in-hospital mortality differed across key clinical characteristics (Table 7 , Fig. 6). Among patients without hypertension, CV was significantly associated with reduced mortality risk (OR = 0.91, 95% CI: 0.85–0.98, P = 0.013), whereas this association was not observed in patients with hypertension ( P = 0.790). In the age-stratified analysis, the protective association was more evident in patients aged ≥ 65 years (OR = 0.93, 95% CI: 0.87–1.00, P = 0.038), but not in those 0.05 for all), suggesting the effect of CV on mortality was generally consistent across these subgroups. Table 7 Subgroup Analysis of the Association Between CV and In-hospital Mortality Subgroup n OR (95% CI) P Gender Male 215 0.94 (0.88–1.01) 0.105 Female 203 0.96 (0.90–1.02) 0.210 Diabetes status Diabetic 129 0.96 (0.88–1.04) 0.316 Non-diabetic 289 0.94 (0.89–1.00) 0.068 GCS score Low (≤ 8) 152 0.95 (0.88–1.02) 0.179 High (> 8) 266 0.95 (0.88–1.02) 0.135 Age group < 65 years 209 0.97 (0.90–1.04) 0.374 ≥ 65 years 209 0.93 (0.87–1.00) 0.038 Hypertension Yes 174 0.99 (0.92–1.06) 0.790 No 244 0.91 (0.85–0.98) 0.013 3.7 Model Stability and External Validation To assess the robustness and generalizability of the findings, key models were independently applied to both the MIMIC-Ⅳ and eICU databases. As shown in Table 8 , the association between CV and in-hospital mortality was consistent across datasets, with an OR of 1.05 (95% CI: 0.60–1.85) in MIMIC-Ⅳ and 0.77 (95% CI: 0.34–1.75) in eICU. The difference was not statistically significant ( P = 0.863, Z-test). Similarly, XGBoost models trained on each dataset demonstrated comparable performance, yielding AUCs of 0.624 (MIMIC-Ⅳ) and 0.650 (eICU), with no significant difference detected by Delong’s test ( P = 0.623). These results indicate stable model behavior and suggest that the predictive value of glycemic variability is relatively consistent across independent ICU cohorts. Table 8 Model Comparison Between MIMIC-Ⅳ and eICU Databases Model Type MIMIC-Ⅳ Result eICU Result Difference Test P Logistic Regression (CV, OR) 1.05 (0.60–1.85) 0.77 (0.34–1.75) Z-test (OR) 0.863 XGBoost AUC 0.624 0.650 Delong test 0.623 4 Discussion In this large-scale, multi-database retrospective cohort study, we investigated the relationship between GV and clinical outcomes among ICU patients with ischemic stroke. Using data from the MIMIC-IV and eICU databases, we evaluated multiple GV metrics and assessed their association with in-hospital mortality and ICU LOS. While survivors demonstrated significantly higher GV values—such as SD, CV, and glucose range—in unadjusted analyses, none of the GV measures remained independently associated with mortality or ICU LOS following adjustment for demographic and clinical covariates. Machine learning models confirmed that GV had only moderate predictive importance, and external validation showed comparable model performance across datasets. The univariate analysis revealed that non-survivors had lower GV, as reflected by significantly reduced SD ( P = 0.007), CV ( P = 0.036), and glucose range ( P = 0.038), while MBG did not differ significantly ( P = 0.132). Stratification by CV quartiles suggested a potential inverse relationship between GV and mortality; however, this trend did not reach statistical significance ( P = 0.463). Median ICU length of stay appeared similar across quartiles, with no consistent trend observed. Figure 2 presents a scatter plot of CV and ICU length of stay. No significant linear correlation was found between CV and ICU stay duration (r = 0.07, P = 0.147), suggesting limited direct association in the univariate context. Several prior studies have reported a strong association between elevated GV and poor outcomes in critically ill patients. Krinsley et al.(9) conducted a retrospective analysis of 4,084 ICU patients and found that higher GV (measured by CV) was independently associated with increased mortality among non-diabetic patients, even after adjusting for illness severity and excluding those with hypoglycemia ( P < 0.0001). However, no such association was observed among diabetic patients. For example, in non-diabetic patients with a mean glucose level of 70–99 mg/dL, mortality rose dramatically from 10.2% with CV < 15% to 58.3% with CV ≥ 50%. These findings highlight the prognostic relevance of GV particularly in non-diabetic ICU patients. Compared with the study by Zhu et al.(15), which utilized the MIMIC-IV database to analyze the prognostic value of GV using MAGE in over 13,000 ICU patients, our findings show partial consistency. Zhu et al. reported a strong association between higher MAGE levels and increased ICU, in-hospital, and 28-day mortality, particularly in non-diabetic patients (HR up to 3.59). In our study, although GV metrics such as CV and SD were not independently associated with in-hospital mortality in the fully adjusted models ( P > 0.05), subgroup analyses revealed potential associations in specific populations, such as older adults and those without hypertension. While different GV metrics were used—MAGE in Zhu et al. versus CV and SD in our study—both investigations underscore the importance of glycemic fluctuations as a prognostic indicator in critical illness. Variations in predictive performance may be attributed to differences in sample size, patient composition, and statistical approaches. Qi et al.(16) analyzed ICU patients with traumatic brain injury (TBI) using the MIMIC-IV database and found that higher glycemic variability, measured by coefficient of variation (CV), was significantly associated with worse neurological outcomes and increased in-hospital mortality (HR = 1.74, P = 0.003). Their findings align with our observation that elevated GV may negatively impact prognosis. In contrast, our study did not find an independent association between GV and in-hospital mortality in ischemic stroke patients. This discrepancy may be attributable to differences in study populations, as our cohort was restricted to stroke patients—a group in whom neurological status, such as GCS score, may exert a more dominant influence on prognosis than glycemic metrics. Furthermore, our study adjusted for a wider range of confounding factors, and the inclusion criteria excluded patients with major glucose-altering comorbidities, possibly attenuating the GV effect. GV is considered a surrogate marker of physiological stress and metabolic instability. Fluctuations in blood glucose may trigger oxidative stress, endothelial dysfunction, and systemic inflammation, all of which are detrimental in critically ill patients(7, 17). However, in ischemic stroke, such effects may be masked by factors more directly related to cerebral injury—such as neurological severity and cerebral perfusion status. The paradoxical finding that lower GV was associated with higher mortality may reflect an impaired metabolic adaptive response in severely ill patients (18, 19). The observation that GV appeared protective in non-hypertensive and older patients raises the possibility that moderate glucose fluctuations could serve as indicators of residual metabolic flexibility in selected populations(20, 21). These nuances underscore the need for personalized glycemic targets and the limitations of "one-size-fits-all" glycemic management strategies in stroke ICUs. Our XGBoost model achieved an AUC of 0.657 for mortality prediction, with traditional predictors (GCS, age, atrial fibrillation) ranking highest in feature importance. GV metrics such as CV and SD contributed moderately, indicating they offer complementary—but not central—value in outcome prediction. This finding is consistent with previous studies showing that GV enhances, but does not replace, conventional risk stratification methods in ICU patients(22, 23). To test model generalizability, we applied both regression and machine learning models to the eICU database. Neither the logistic regression ORs nor the AUCs from XGBoost differed significantly between datasets ( P > 0.6), supporting the stability of GV-related findings across geographically and temporally distinct ICU populations. This external validation strengthens the reproducibility of our results and reinforces the utility of combining multiple databases for critical care research. Several limitations of this study should be acknowledged. First, the retrospective design limits the ability to draw causal inferences between glycemic variability and patient outcomes. Second, although we adjusted for basic demographic and clinical variables, some important confounders—such as stroke subtype, neuroimaging findings, and neurological scores like the NIHSS—were not available in the databases. We used the GCS score at ICU admission as a surrogate for baseline neurological status, which may not fully capture the extent of brain injury. Third, data on critical interventions, including thrombolytic therapy (e.g., rtPA) and endovascular procedures, were incomplete or inconsistently recorded, limiting our ability to assess treatment-related differences. Fourth, ICU-related complications such as aspiration pneumonia, mechanical ventilation, ventilator-associated pneumonia, and sepsis may have influenced both glucose variability and outcomes. While some of these were identifiable via diagnostic codes, the potential for residual confounding remains. Finally, decisions regarding withdrawal of life-sustaining treatment—an important factor affecting in-hospital mortality—were not systematically documented in either dataset. Prospective studies with protocolized glucose monitoring and treatment data are needed to validate the role of GV in ischemic stroke outcomes. Future work could explore real-time GV monitoring as a clinical decision tool, investigate the interplay between GV and neuroimaging findings, and determine whether GV-targeted interventions can improve recovery in specific stroke subgroups. 5 Conclusion In this multi-database cohort study of ICU patients with ischemic stroke, we found that glycemic variability, as measured by SD, CV, and glucose range, was significantly higher in survivors than in non-survivors in univariate analysis. However, these metrics were not independently associated with in-hospital mortality or ICU length of stay after multivariable adjustment. Machine learning models identified glycemic variability as a moderately important feature in mortality prediction, and external validation across the MIMIC-Ⅳ and eICU databases demonstrated stable and reproducible model performance. These findings suggest that while glycemic variability reflects aspects of metabolic stress in critically ill stroke patients, it may not serve as a strong independent prognostic factor when traditional clinical variables are accounted for. Future prospective studies are needed to further clarify the role of glycemic dynamics in guiding personalized glucose management strategies in neurocritical care. Declarations Acknowledgements Not applicable. Authors’ contributions Yong Xia and Qiao Huang participated in the conceptualization of protocol, quality assessment, data extraction, formal analysis, methodology, writing-original draft, writing-review and editing, and approving the final draft. Yanbin Chen participated in the formal analysis, Methodology, writing-review and editing,Liu Yang and Jingyuan Kang participated in Supervision and Visualization. All authors read and approved the manuscript. Funding This work was supported by Hunan Provincial Natural Science Foundation Project(2025JJ70588) Availability of data and materials The data sets used and/or analyzed during the current systematic review and meta-analysis are presented within the manuscript Ethics approval and consent to participate Not applicable. Because all reviewed studies provided appropriate information on obtaining ethical approval in accordance with the Declaration of Helsinki. Consent for publication Not applicable. Because all reviewed studies provided appropriate information regarding participant consent. Competing interests The authors declare that they have no competing interests to report. References Golubnitschaja O, Potuznik P, Polivka J, Jr., Pesta M, Kaverina O, Pieper CC, et al. Ischemic stroke of unclear aetiology: a case-by-case analysis and call for a multi-professional predictive, preventive and personalised approach. Epma j. 2022;13(4):535-45. Zhang FL, Guo ZN, Wu YH, Liu HY, Luo Y, Sun MS, et al. Prevalence of stroke and associated risk factors: a population based cross sectional study from northeast China. BMJ Open. 2017;7(9):e015758. Cai W, Xu J, Wu X, Chen Z, Zeng L, Song X, et al. Association between triglyceride-glucose index and all-cause mortality in critically ill patients with ischemic stroke: analysis of the MIMIC-IV database. Cardiovasc Diabetol. 2023;22(1):138. Li XD, Li MM. A novel nomogram to predict mortality in patients with stroke: a survival analysis based on the MIMIC-III clinical database. BMC Med Inform Decis Mak. 2022;22(1):92. Li S, Wang Y, Zhu X, Zheng H, Ni J, Li H, et al. Lipid on stroke in intracranial artery atherosclerotic stenosis: a mediation role of glucose. Front Endocrinol (Lausanne). 2024;15:1322114. Vasa D, Rossitto CP, Ezzat B, Bazil M, Schuldt B, Johnson B, et al. Stress hyperglycemia is associated with longer ICU length of stay after endoscopic intracerebral hemorrhage evacuation. J Stroke Cerebrovasc Dis. 2024;33(10):107911. Said AME, Sayed Hussien N, Negm MA. Dynamic of Glycemic State as a Prognostic Marker in Acute Ischemic Stroke: A Prospective Controlled Cohort Observational Study. QJM: An International Journal of Medicine. 2023;116(Supplement_1). Ma Z, Li S, Lin X. Body mass index, blood glucose, and mortality in patients with ischemic stroke in the intensive care unit: A retrospective cohort study. Front Neurosci. 2022;16:946397. Krinsley JS. Glycemic variability and mortality in critically ill patients: the impact of diabetes. J Diabetes Sci Technol. 2009;3(6):1292-301. Laredo C, Renú A, Llull L, Tudela R, López-Rueda A, Urra X, et al. Elevated glucose is associated with hemorrhagic transformation after mechanical thrombectomy in acute ischemic stroke patients with severe pretreatment hypoperfusion. Sci Rep. 2020;10(1):10588. Cai W, Li Y, Guo K, Wu X, Chen C, Lin X. Association of glycemic variability with death and severe consciousness disturbance among critically ill patients with cerebrovascular disease: analysis of the MIMIC-IV database. Cardiovasc Diabetol. 2023;22(1):315. Su Y, Fan W, Liu Y, Hong K. Glycemic variability and in-hospital death of critically ill patients and the role of ventricular arrhythmias. Cardiovasc Diabetol. 2023;22(1):134. Wu WT, Li YJ, Feng AZ, Li L, Huang T, Xu AD, et al. Data mining in clinical big data: the frequently used databases, steps, and methodological models. Mil Med Res. 2021;8(1):44. Kim Y, Kim M, Kim Y, Choi M. Using nursing data for machine learning-based prediction modeling in intensive care units: A scoping review. Int J Nurs Stud. 2025;169:105133. Zhu Q, Zong Q, Guo S, Ye H, Ma Z, Zhang R, et al. Mean amplitude of glycemic excursion and mortality in critically ill patients: A retrospective analysis using the MIMIC-IV database. Diabetes Obes Metab. 2025;27(7):3831-9. Qi L, Geng X, Feng R, Wu S, Fu T, Li N, et al. Association of glycemic variability and prognosis in patients with traumatic brain injury: A retrospective study from the MIMIC-IV database. Diabetes Res Clin Pract. 2024;217:111869. Hou Y, Wu X, Shi Y, Xu X, Zhang Y, Jiang L, et al. METS-IR as an important predictor of neurological impairment severity in patients with severe cerebral infarction: a multicenter study based on the Chinese population. Front Neurol. 2024;15:1450825. Pensato U, Bosshart S, Stebner A, Rohr A, Kleinig TJ, Gupta R, et al. Effect of Hemoglobin and Blood Glucose Levels on CT Perfusion Ischemic Core Estimation: A Post Hoc Analysis of the ESCAPE-NA1 Trial. Neurology. 2024;103(10):e209939. Lu Z, Tao G, Sun X, Zhang Y, Jiang M, Liu Y, et al. Association of Blood Glucose Level and Glycemic Variability With Mortality in Sepsis Patients During ICU Hospitalization. Front Public Health. 2022;10:857368. Shi X, Yang S, Guo C, Sun W, Song J, Fan S, et al. Impact of stress hyperglycemia on outcomes in patients with large ischemic stroke. J Neurointerv Surg. 2025. Emgin Ö, Yavuz M, Sahin A, Günes M, Eser M, Yavuz T, et al. The Association Between Glycemic Variability and Mortality in Critically Ill Patients: A Multicenter Prospective Observational Study. Journal of Clinical Medicine. 2024;13(22):6939. Guo Y, Qiu Y, Xue T, Zhou Y, Yan P, Liu S, et al. Association between glycemic variability and short-term mortality in patients with acute kidney injury: a retrospective cohort study of the MIMIC-IV database. Sci Rep. 2024;14(1):5945. Lin L, Liang Z. Association Between Glycemic Variability and All-Cause Mortality in Patients with Acute Pancreatitis in the Intensive Care Unit: A Retrospective Analysis. Dig Dis Sci. 2025;70(6):2194-203. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 28 Mar, 2026 Read the published version in European Journal of Medical Research → Version 1 posted Editorial decision: Revision requested 10 Feb, 2026 Reviews received at journal 10 Feb, 2026 Reviews received at journal 30 Jan, 2026 Reviewers agreed at journal 26 Jan, 2026 Reviews received at journal 20 Jan, 2026 Reviewers agreed at journal 20 Jan, 2026 Reviewers agreed at journal 10 Jan, 2026 Reviewers invited by journal 24 Nov, 2025 Editor assigned by journal 25 Sep, 2025 Submission checks completed at journal 24 Sep, 2025 First submitted to journal 22 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7686095","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":550031847,"identity":"b6948ac8-8644-4053-98fa-736cc1817bc7","order_by":0,"name":"Yong Xia","email":"","orcid":"","institution":"Affiliated Hospital of Xiangnan University","correspondingAuthor":false,"prefix":"","firstName":"Yong","middleName":"","lastName":"Xia","suffix":""},{"id":550031848,"identity":"31474337-5a2e-439a-9077-3a66c1fb17b3","order_by":1,"name":"Yanbin Chen","email":"","orcid":"","institution":"Affiliated Hospital of Xiangnan University","correspondingAuthor":false,"prefix":"","firstName":"Yanbin","middleName":"","lastName":"Chen","suffix":""},{"id":550031849,"identity":"df75fde4-00ae-43cf-b592-de9697d139a6","order_by":2,"name":"Liu Yang","email":"","orcid":"","institution":"Affiliated Hospital of Xiangnan University","correspondingAuthor":false,"prefix":"","firstName":"Liu","middleName":"","lastName":"Yang","suffix":""},{"id":550031850,"identity":"96a261f7-0ed0-478c-9358-da46d3aee9c3","order_by":3,"name":"Jingyuan Kang","email":"","orcid":"","institution":"Affiliated Hospital of Xiangnan University","correspondingAuthor":false,"prefix":"","firstName":"Jingyuan","middleName":"","lastName":"Kang","suffix":""},{"id":550031851,"identity":"a9c92a8d-2d3d-4909-a69b-7256df6cb980","order_by":4,"name":"Qiao Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYBACNobzjx984LHh4WdvPkCcFj7GM2yGM2TSZCR7jiUQp0WO+QyDNI/NYRuDGzkGRDqM7ewBwxk5aTxALR9vvGGwk9NtIKSF51zCgw9nbHgkz7zdbDmHIdnY7AAhLRIHDAxn9qTx8B3P3SbNw3AgcRtBLfIPDKR5/x0GKs55RqQWhjMG0jw8h3kETuSwEavlWJrhDJ40HmAgG1vOMSDCL/INhw+DotIeGJUPb7ypsJMjqAUFSPAQGTXIWkjVMQpGwSgYBSMCAABbHUTuSzt66gAAAABJRU5ErkJggg==","orcid":"","institution":"Xiangnan University","correspondingAuthor":true,"prefix":"","firstName":"Qiao","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2025-09-23 01:08:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7686095/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7686095/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s40001-026-04325-9","type":"published","date":"2026-03-28T16:10:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":97140911,"identity":"12b7219d-165d-4e41-9d8d-cde3b685e82c","added_by":"auto","created_at":"2025-12-01 10:05:56","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":174087,"visible":true,"origin":"","legend":"","description":"","filename":"GlycemicVariabilityandClinicalOutcomesinICUPatientswithIschemicStroke9.22.docx","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/0b2be8d829c5dbcda50302e6.docx"},{"id":97122941,"identity":"b7e08633-d070-4dd6-b55d-8197f982ab06","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6863,"visible":true,"origin":"","legend":"","description":"","filename":"b6ad848736294d6b90665f35096537dd.json","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/84e15a8d14ae6b506070809c.json"},{"id":97122940,"identity":"4af252cc-12da-439a-a0aa-8bbf2a495521","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82131,"visible":true,"origin":"","legend":"","description":"","filename":"b6ad848736294d6b90665f35096537dd1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/6b866ade3a2f310101960f18.xml"},{"id":97141308,"identity":"fbce8814-8c14-4e25-adea-25a21dabb993","added_by":"auto","created_at":"2025-12-01 10:06:33","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55961,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/52360b4ad6976c5a8b62ce03.png"},{"id":97122946,"identity":"982952ea-379c-4b34-afdc-f41025b53234","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29244,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/056dfaab997a27445c577936.png"},{"id":97122938,"identity":"27aa74e3-c7ab-4da3-b11f-24755ab8b4c1","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17916,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/9efedf4e6b4b85e414f6c60e.png"},{"id":97142967,"identity":"3efafdb2-e9ab-4d21-83b3-575cbab03794","added_by":"auto","created_at":"2025-12-01 10:08:09","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10434,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/6f539650955f7ab9f9cfe48e.png"},{"id":97122944,"identity":"eea6f5df-207d-4beb-830e-6a4663425bbc","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82383,"visible":true,"origin":"","legend":"","description":"","filename":"b6ad848736294d6b90665f35096537dd1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/a7342f194ec34c219ee4a3e8.xml"},{"id":97122947,"identity":"d93e8be2-6671-464f-b765-344bf6de6179","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"html","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91404,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/ca0f5973b38f4c811e1e41ef.html"},{"id":97122937,"identity":"4edbee33-8f2c-401e-860c-bf58626d98a4","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":113139,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eScatter Plot of CV and ICU Length of Stay\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eScatter plot showing the relationship between glycemic CV and ICU length of stay.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/d5ce46fc40b80e2cd517649c.png"},{"id":97122936,"identity":"2d5c841c-e776-4f3a-a95a-4f49aa81eb4a","added_by":"auto","created_at":"2025-12-01 08:03:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":59625,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePerformance Evaluation of the XGBoost Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePanel A shows the ROC curve of the XGBoost model predicting in-hospital mortality among ICU patients with ischemic stroke, with an area under the curve (AUC) of 0.657, indicating moderate discriminative ability. Panel B presents the SHAP (Shapley Additive Explanations) summary plot of feature importance, ranking variables based on their contribution to the model’s predictions. Clinical factors such as atrial fibrillation, BMI, GCS score, and age showed the highest importance, while glycemic variability metrics such as CV and SD contributed moderate predictive value.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/ee6da29a1efc8e1eff179fd7.png"},{"id":105755889,"identity":"0408a6c2-44a5-4eb6-8cd3-5e1757c8d923","added_by":"auto","created_at":"2026-03-30 16:32:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1158876,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7686095/v1/4af4d1cd-1537-4d3d-8cf5-644d3477f3b8.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Glycemic Variability and Clinical Outcomes in ICU Patients with Ischemic Stroke: A Multi-Database Retrospective Cohort Study","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eIschemic stroke is one of the leading causes of mortality and long-term disability worldwide, accounting for approximately 76% of all stroke cases and affecting over 12\u0026nbsp;million people annually(1, 2). Among critically ill patients, stroke represents a substantial clinical burden: intensive care unit (ICU) admission is required for up to 15\u0026ndash;20% of hospitalized stroke cases, particularly those with severe neurological deficits, respiratory compromise, or multi-organ dysfunction(3, 4) In such high-acuity settings, metabolic dysregulation\u0026mdash;particularly disturbances in blood glucose levels\u0026mdash;is frequently observed and may exert a considerable impact on clinical outcomes (5, 6).\u003c/p\u003e\u003cp\u003eTraditionally, research on glucose in acute stroke has focused on hyperglycemia at admission, which has been consistently associated with poor neurological recovery and higher mortality(7, 8). However, growing evidence highlights glycemic variability (GV)\u0026mdash;the degree of glucose fluctuation over time\u0026mdash;as a potentially superior prognostic indicator compared to static glucose levels (9, 10). GV reflects the complex interplay of stress-induced hyperglycemia, insulin resistance, and neuroendocrine dysregulation, all of which are amplified in critical care environments(11, 12).\u003c/p\u003e\u003cp\u003eDespite these insights, prior investigations into GV have faced notable limitations, including small sample sizes, single-center designs, and limited control for confounding factors. With the increasing availability of large-scale, de-identified ICU databases\u0026mdash;such as the Medical Information Mart for Intensive Care Ⅳ (MIMIC-Ⅳ) and the eICU Collaborative Research Database\u0026mdash;researchers are now positioned to conduct multicenter, high-resolution studies with enhanced external validity(13, 14). These databases contain high-resolution, time-stamped glucose records, along with detailed patient characteristics, comorbidities, treatment interventions, and outcomes.\u003c/p\u003e\u003cp\u003eIn this study, we leveraged data from both MIMIC-IV and eICU to investigate the relationship between GV and clinical outcomes in ICU patients with ischemic stroke. Specifically, we aimed to: (1) characterize the distribution and clinical correlates of GV in this population; (2) examine the association between GV and key outcomes, including ICU length of stay and in-hospital mortality; and (3) evaluate the predictive utility of GV using both conventional regression and machine learning models. By providing a more nuanced understanding of glucose dynamics in critically ill stroke patients, this work may contribute to refining individualized glucose management strategies in the neurocritical care setting.\u003c/p\u003e"},{"header":"2 Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Data Sources\u003c/h2\u003e\u003cp\u003eThis study utilized data from two publicly available critical care databases: the MIMIC-Ⅳ and the eICU Collaborative Research Database. MIMIC-Ⅳ was developed by the Massachusetts Institute of Technology in collaboration with the Beth Israel Deaconess Medical Center and includes detailed electronic health records of ICU patients from 2008 to 2019 at a single academic center. The eICU database, developed by Philips Healthcare, contains ICU data from over 200 hospitals across the United States between 2014 and 2015. Both databases include de-identified patient-level data in accordance with HIPAA regulations and provide extensive information on demographics, vital signs, laboratory results, glucose measurements, comorbidities, and clinical outcomes.\u003c/p\u003e\u003cp\u003eAccess to both datasets was granted to credentialed researchers who completed CITI training and data use agreements. As this study used publicly available de-identified data, it was exempt from institutional review board (IRB) approval and informed consent requirements.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Study Population\u003c/h2\u003e\u003cp\u003eWe included adult patients (aged\u0026thinsp;\u0026ge;\u0026thinsp;18 years) who were admitted to the ICU with a primary diagnosis of ischemic stroke. Ischemic stroke was identified using International Classification of Diseases (ICD) codes: ICD-9 codes 434.x and ICD-10 codes I63.x. Eligible patients were required to have at least two or more blood glucose measurements recorded during their ICU stay to allow calculation of GV.\u003c/p\u003e\u003cp\u003ePatients were excluded if they met any of the following criteria: (1) ICU length of stay less than 24 hours; (2) missing key variables such as glucose data, in-hospital mortality status, or ICU admission/discharge times; (3) presence of comorbid conditions known to severely affect glucose metabolism (e.g., diabetic ketoacidosis, hyperosmolar hyperglycemic state, or acute pancreatitis).\u003c/p\u003e\u003cp\u003eIf a patient had multiple ICU admissions, only the first ICU stay meeting all inclusion criteria was considered for analysis. Data preprocessing, including patient identification and eligibility filtering, was conducted separately for each database (MIMIC-Ⅳ and eICU) to ensure consistent cohort construction across sources.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Variables and Definitions\u003c/h2\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.3.1 Exposure Variables\u003c/h2\u003e\u003cp\u003eGV during the ICU stay was assessed using the following four metrics:\u003c/p\u003e\u003cp\u003e(1) Mean Blood Glucose (MBG): the arithmetic mean of all recorded glucose values.\u003c/p\u003e\u003cp\u003e(2) Standard Deviation (SD): reflects the absolute variation in glucose values.\u003c/p\u003e\u003cp\u003e(3) Coefficient of Variation (CV): calculated as (SD / MBG) \u0026times; 100%, indicating relative variability.\u003c/p\u003e\u003cp\u003e(4) Range: the difference between the maximum and minimum glucose values.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.3.2 Outcome Variables\u003c/h2\u003e\u003cp\u003e(1) In-hospital mortality: defined as death occurring during the hospitalization associated with the ICU stay, recorded as a binary variable (0\u0026thinsp;=\u0026thinsp;survived, 1\u0026thinsp;=\u0026thinsp;died).\u003c/p\u003e\u003cp\u003e(2) ICU length of stay (LOS): defined as the number of days between ICU admission and ICU discharge, treated as a continuous variable.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003e2.3.3 Covariates\u003c/h2\u003e\u003cp\u003e(1) \u003cb\u003eDemographics\u003c/b\u003e: age, sex, and body mass index (BMI).\u003c/p\u003e\u003cp\u003e(2) \u003cb\u003eComorbidities\u003c/b\u003e: including diabetes mellitus, hypertension, and atrial fibrillation, identified via ICD codes.\u003c/p\u003e\u003cp\u003e(3) \u003cb\u003eNeurological status\u003c/b\u003e: Glasgow Coma Scale (GCS) score at ICU admission.\u003c/p\u003e\u003cp\u003e(4) \u003cb\u003eICU characteristics\u003c/b\u003e: ICU type (e.g., medical, surgical, neurological) and initial glucose level at ICU entry.\u003c/p\u003e\u003cp\u003eAll variables were harmonized across databases using consistent definitions and coding. Continuous variables were standardized when appropriate, and categorical variables were converted into binary or dummy-coded formats for inclusion in regression and machine learning models. Neurological status was assessed using the GCS score recorded at ICU admission. While specific imaging-based measures of infarct volume or NIH Stroke Scale (NIHSS) were not available in the databases, GCS was used as a surrogate indicator of baseline neurological function. GCS has been widely applied in ICU outcome research as a proxy for brain injury severity.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Data Extraction and Processing\u003c/h2\u003e\u003cp\u003eClinical data were extracted separately from the MIMIC-Ⅳ and eICU databases using structured query language (SQL). For each patient, demographic information, diagnostic codes, blood glucose measurements, ICU admission and discharge times, comorbidities, and outcome variables were retrieved.\u003c/p\u003e\u003cp\u003e(1) Patient Identification: ICU patients with a primary diagnosis of ischemic stroke were identified based on ICD-9 codes (434.x) and ICD-10 codes (I63.x). Only the first ICU admission was considered for patients with multiple eligible stays. Inclusion and exclusion criteria were applied to filter the final study cohort as described in Section 2.2.\u003c/p\u003e\u003cp\u003e(2) Blood Glucose Data Handling: All available blood glucose values (including bedside point-of-care and laboratory glucose measurements) during the ICU stay were extracted. Values were cleaned to remove physiologically implausible readings (e.g., \u0026lt;\u0026thinsp;20 mg/dL or \u0026gt;\u0026thinsp;600 mg/dL), and only patients with two or more valid glucose records were included for glycemic variability analysis.\u003c/p\u003e\u003cp\u003e(3) Variable Harmonization and Transformation: Variables were standardized across the two databases by mapping equivalent fields and ensuring consistent units, formats, and definitions. Continuous variables (e.g., age, glucose metrics) were normalized as needed, and categorical variables (e.g., sex, ICU type, comorbidities) were encoded as binary indicators.\u003c/p\u003e\u003cp\u003e(4) Missing Data Management: Patients with missing primary outcome variables (in-hospital mortality or ICU discharge time) were excluded. For covariates with low missingness (e.g., BMI, GCS), median imputation was used. Variables with extensive missing data were excluded from adjusted models.\u003c/p\u003e\u003cp\u003eAll preprocessing procedures were performed using Python (pandas, numpy), and were conducted separately within each database before merging results for comparative or pooled analyses.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Statistical Analysis\u003c/h2\u003e\u003cp\u003eAll statistical analyses were conducted using R and Python. Continuous variables were presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation or median [IQR], and categorical variables as counts and percentages. Group comparisons were performed using t-tests, Mann\u0026ndash;Whitney U tests, or chi-squared tests, as appropriate. Logistic regression was used to evaluate the association between GV metrics and in-hospital mortality, and linear regression was applied for ICU length of stay, adjusting for age, sex, BMI, comorbidities, GCS, and ICU type. GV metrics were also divided into quartiles for trend analysis. Machine learning models, including XGBoost, were used to predict mortality, with model performance assessed by AUC and feature importance. A two-sided \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Baseline Characteristics\u003c/h2\u003e\u003cp\u003eA total of 418 ICU patients with ischemic stroke were included in the analysis, of whom 334 survived and 84 died during hospitalization. Baseline characteristics between survivors and non-survivors were compared (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Significant differences were observed in age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.012) and GCS score (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with non-survivors tending to be older and having lower levels of consciousness. Other variables, including sex, BMI, diabetes, hypertension, atrial fibrillation, and ICU type, showed no statistically significant differences between groups (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05 for all).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eBaseline Characteristics of ICU Patients with Ischemic Stroke\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;418)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSurvivors (n\u0026thinsp;=\u0026thinsp;334)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNon-survivors (n\u0026thinsp;=\u0026thinsp;84)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003et\u003c/em\u003e/\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65.25\u0026thinsp;\u0026plusmn;\u0026thinsp;14.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e64.37\u0026thinsp;\u0026plusmn;\u0026thinsp;14.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e68.77\u0026thinsp;\u0026plusmn;\u0026thinsp;15.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e-2.518\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.012\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender [n (%)]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.321\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.250\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e215 (51.44%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e177 (52.99%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e38 (45.24%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e203 (48.56%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e157 (47.01%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e46 (54.76%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI (kg/m\u003csup\u003e2\u003c/sup\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e26.00\u0026thinsp;\u0026plusmn;\u0026thinsp;4.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26.10\u0026thinsp;\u0026plusmn;\u0026thinsp;4.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e25.58\u0026thinsp;\u0026plusmn;\u0026thinsp;3.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.046\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.296\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS score (points)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e9.68\u0026thinsp;\u0026plusmn;\u0026thinsp;2.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.97\u0026thinsp;\u0026plusmn;\u0026thinsp;2.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.51\u0026thinsp;\u0026plusmn;\u0026thinsp;2.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.299\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes [n (%)]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e129 (30.86%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e102 (30.54%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e27 (32.14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.879\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension [n (%)]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e174 (41.63%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e137 (41.02%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e37 (44.05%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.144\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.704\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAtrial fibrillation [n (%)]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e103 (24.64%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e89 (26.65%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e14 (16.67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.083\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.079\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eICU type [n (%)]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.064\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.587\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedical ICU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e142 (33.97%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e112 (33.53%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e30 (35.71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeuro ICU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e139 (33.25%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e115 (34.43%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e24 (28.57%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSurgical ICU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e137 (32.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e107 (32.04%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e30 (35.71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Distribution of Glycemic Variability Metrics\u003c/h2\u003e\u003cp\u003eGlycemic variability metrics were compared between survivors and non-survivors (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The SD, CV, and glucose range were all significantly lower in non-survivors compared to survivors (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.036, and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.038, respectively). In contrast, MBG did not differ significantly between the two groups (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.132).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDistribution of Glycemic Variability Metrics by Clinical Outcome\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\"\u0026plusmn;\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;418)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSurvivors (n\u0026thinsp;=\u0026thinsp;334)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNon-survivors (n\u0026thinsp;=\u0026thinsp;84)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003et\u003c/em\u003e/\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMean Blood Glucose (MBG, mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e\u003cp\u003e142.15\u0026thinsp;\u0026plusmn;\u0026thinsp;10.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e\u003cp\u003e142.55\u0026thinsp;\u0026plusmn;\u0026thinsp;10.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e\u003cp\u003e140.54\u0026thinsp;\u0026plusmn;\u0026thinsp;11.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.511\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.132\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStandard Deviation (SD, mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e\u003cp\u003e28.88\u0026thinsp;\u0026plusmn;\u0026thinsp;6.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e\u003cp\u003e29.32\u0026thinsp;\u0026plusmn;\u0026thinsp;6.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e\u003cp\u003e27.14\u0026thinsp;\u0026plusmn;\u0026thinsp;5.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.711\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCoefficient of Variation (CV, %)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e\u003cp\u003e20.47\u0026thinsp;\u0026plusmn;\u0026thinsp;5.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e\u003cp\u003e20.73\u0026thinsp;\u0026plusmn;\u0026thinsp;5.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e\u003cp\u003e19.43\u0026thinsp;\u0026plusmn;\u0026thinsp;4.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.099\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.036\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlucose Range (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c2\"\u003e\u003cp\u003e90.59\u0026thinsp;\u0026plusmn;\u0026thinsp;22.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c3\"\u003e\u003cp\u003e91.75\u0026thinsp;\u0026plusmn;\u0026thinsp;23.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026plusmn;\" colname=\"c4\"\u003e\u003cp\u003e85.96\u0026thinsp;\u0026plusmn;\u0026thinsp;21.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.077\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.038\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Univariate Associations\u003c/h2\u003e\u003cp\u003eTo further explore the association between glycemic variability and clinical outcomes, patients were stratified into quartiles based on their CV values (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). A decreasing trend in in-hospital mortality was observed across increasing CV quartiles, although this trend did not reach statistical significance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.463). Median ICU length of stay appeared similar across quartiles, with no consistent trend observed. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents a scatter plot of CV and ICU length of stay. No significant linear correlation was found between CV and ICU stay duration (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.07, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.147), suggesting limited direct association in the univariate context.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eQuartile Analysis of CV and Its Association With Clinical Outcomes\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCV Quartile\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCV Range (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eIn-hospital Mortality (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eICU Length of Stay (median [IQR])\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eTrend Statistic (Z)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQ1 (Lowest)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e4.95\u0026ndash;16.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e24.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.2 [2.4\u0026ndash;6.5]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQ2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e16.98\u0026ndash;20.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e21.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.7 [2.2\u0026ndash;5.5]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQ3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e20.15\u0026ndash;23.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e17.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.6 [2.5\u0026ndash;6.7]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQ4 (Highest)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e23.42\u0026ndash;40.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e17.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.6 [2.7\u0026ndash;7.4]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.570\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.463\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eScatter plot showing the relationship between glycemic CV and ICU length of stay.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Multivariable Regression Analysis\u003c/h2\u003e\u003cp\u003eMultivariable logistic regression was performed to assess the independent association between glycemic variability metrics and in-hospital mortality (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). After adjusting for demographic and clinical covariates, GCS score (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and age (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005) remained significantly associated with mortality. However, none of the glycemic variability indicators\u0026mdash;including CV, SD, MBG, and range\u0026mdash;showed a statistically significant association with in-hospital death in the adjusted model (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05 for all).\u003c/p\u003e\u003cp\u003eIn the linear regression model predicting ICU length of stay (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e), most glycemic variability metrics were not significantly associated with the outcome. The only variable reaching statistical significance was hypertension, which was independently associated with a shorter ICU stay (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.010). Glycemic metrics, including CV, SD, and MBG, did not show significant associations with ICU length of stay in the adjusted model (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eLogistic Regression: Predictors of In-hospital Mortality\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOR (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWald \u003cem\u003eχ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.03 (1.01\u0026ndash;1.05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.005\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.69 (0.41\u0026ndash;1.16)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-1.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.164\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.97 (0.91\u0026ndash;1.03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.309\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.81 (0.74\u0026ndash;0.89)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-4.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.03 (0.59\u0026ndash;1.80)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.912\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.11 (0.66\u0026ndash;1.87)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.697\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAtrial fibrillation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.63 (0.32\u0026ndash;1.21)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-1.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.165\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMBG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.95 (0.86\u0026ndash;1.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-1.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.243\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.82 (0.44\u0026ndash;1.55)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.550\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.02 (0.64\u0026ndash;1.63)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.925\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRange\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.02 (0.99\u0026ndash;1.05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.264\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eMultivariable Linear Regression: Predictors of ICU Length of Stay\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eβ Coefficient (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStandard Error\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003et\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.00 (-0.02\u0026ndash;0.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.113\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.910\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e0.47 (-0.16\u0026ndash;1.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.449\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.148\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e0.08 (-0.00\u0026ndash;0.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.057\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.01 (-0.13\u0026ndash;0.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.242\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.809\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e0.32 (-0.36\u0026ndash;0.99)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.363\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.84 (-1.47\u0026ndash;-0.20)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-2.571\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.010\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAtrial fibrillation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e0.45 (-0.28\u0026ndash;1.18)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.206\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.229\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMBG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.03 (-0.14\u0026ndash;0.08)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.487\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.626\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.05 (-0.75\u0026ndash;0.64)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.150\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.881\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e0.09 (-0.43\u0026ndash;0.60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.327\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.744\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRange\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c2\"\u003e\u003cp\u003e-0.01 (-0.04\u0026ndash;0.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.554\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.580\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Machine Learning Model Performance\u003c/h2\u003e\u003cp\u003eTo further assess the predictive value of glycemic variability and clinical features, an XGBoost classification model was developed to predict in-hospital mortality. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, the model achieved an area under the ROC curve (AUC) of 0.657 and an accuracy of 75.4%. However, sensitivity (recall) was relatively low at 16.0%, with precision at 28.6% and an F1 score of 90.1%. Calibration assessed by the Hosmer\u0026ndash;Lemeshow test indicated no significant miscalibration (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.144), and the Brier score was 0.196.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA displays the ROC curve, demonstrating moderate discriminative ability. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB presents the SHAP-based feature importance, with atrial fibrillation, BMI, GCS score, and age ranking highest in predictive contribution. Glycemic variability metrics such as CV and SD had moderate importance, indicating they provide complementary predictive value alongside major clinical factors.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance Metrics of the XGBoost Model for Predicting In-hospital Mortality\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetric\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eArea Under the Curve (AUC)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.657\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAccuracy (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e75.40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePrecision (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e28.57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRecall (Sensitivity, %)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e16.00\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpecificity (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e20.51\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eF1 Score (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e90.10\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAUCPR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.313\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBrier Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.196\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCalibration \u003cem\u003eP\u003c/em\u003e (Hosmer\u0026ndash;Lemeshow)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.144\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003ePanel A shows the ROC curve of the XGBoost model predicting in-hospital mortality among ICU patients with ischemic stroke, with an area under the curve (AUC) of 0.657, indicating moderate discriminative ability. Panel B presents the SHAP (Shapley Additive Explanations) summary plot of feature importance, ranking variables based on their contribution to the model\u0026rsquo;s predictions. Clinical factors such as atrial fibrillation, BMI, GCS score, and age showed the highest importance, while glycemic variability metrics such as CV and SD contributed moderate predictive value.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Subgroup and Sensitivity Analyses\u003c/h2\u003e\u003cp\u003eSubgroup analyses were performed to evaluate whether the association between glycemic CV and in-hospital mortality differed across key clinical characteristics (Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, Fig.\u0026nbsp;6). Among patients without hypertension, CV was significantly associated with reduced mortality risk (OR\u0026thinsp;=\u0026thinsp;0.91, 95% CI: 0.85\u0026ndash;0.98, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013), whereas this association was not observed in patients with hypertension (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.790). In the age-stratified analysis, the protective association was more evident in patients aged\u0026thinsp;\u0026ge;\u0026thinsp;65 years (OR\u0026thinsp;=\u0026thinsp;0.93, 95% CI: 0.87\u0026ndash;1.00, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.038), but not in those\u0026thinsp;\u0026lt;\u0026thinsp;65 years. No statistically significant interactions were observed by sex, diabetes status, or GCS score (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05 for all), suggesting the effect of CV on mortality was generally consistent across these subgroups.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSubgroup Analysis of the Association Between CV and In-hospital Mortality\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSubgroup\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003en\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOR (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGender\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e215\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.94 (0.88\u0026ndash;1.01)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.105\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e203\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.96 (0.90\u0026ndash;1.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.210\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes status\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e129\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.96 (0.88\u0026ndash;1.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.316\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNon-diabetic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e289\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.94 (0.89\u0026ndash;1.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.068\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGCS score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLow (\u0026le;\u0026thinsp;8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e152\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.95 (0.88\u0026ndash;1.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.179\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHigh (\u0026gt;\u0026thinsp;8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e266\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.95 (0.88\u0026ndash;1.02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.135\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge group\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;65 years\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e209\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.97 (0.90\u0026ndash;1.04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.374\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026ge;\u0026thinsp;65 years\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e209\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.93 (0.87\u0026ndash;1.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.038\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e174\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.99 (0.92\u0026ndash;1.06)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.790\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e244\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.91 (0.85\u0026ndash;0.98)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.013\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e3.7 Model Stability and External Validation\u003c/h2\u003e\u003cp\u003eTo assess the robustness and generalizability of the findings, key models were independently applied to both the MIMIC-Ⅳ and eICU databases. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, the association between CV and in-hospital mortality was consistent across datasets, with an OR of 1.05 (95% CI: 0.60\u0026ndash;1.85) in MIMIC-Ⅳ and 0.77 (95% CI: 0.34\u0026ndash;1.75) in eICU. The difference was not statistically significant (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.863, Z-test).\u003c/p\u003e\u003cp\u003eSimilarly, XGBoost models trained on each dataset demonstrated comparable performance, yielding AUCs of 0.624 (MIMIC-Ⅳ) and 0.650 (eICU), with no significant difference detected by Delong\u0026rsquo;s test (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.623). These results indicate stable model behavior and suggest that the predictive value of glycemic variability is relatively consistent across independent ICU cohorts.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eModel Comparison Between MIMIC-Ⅳ and eICU Databases\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMIMIC-Ⅳ Result\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eeICU Result\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDifference Test\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic Regression (CV, OR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.05 (0.60\u0026ndash;1.85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.77 (0.34\u0026ndash;1.75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eZ-test (OR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.863\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost AUC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.624\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDelong test\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.623\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eIn this large-scale, multi-database retrospective cohort study, we investigated the relationship between GV and clinical outcomes among ICU patients with ischemic stroke. Using data from the MIMIC-IV and eICU databases, we evaluated multiple GV metrics and assessed their association with in-hospital mortality and ICU LOS. While survivors demonstrated significantly higher GV values\u0026mdash;such as SD, CV, and glucose range\u0026mdash;in unadjusted analyses, none of the GV measures remained independently associated with mortality or ICU LOS following adjustment for demographic and clinical covariates. Machine learning models confirmed that GV had only moderate predictive importance, and external validation showed comparable model performance across datasets.\u003c/p\u003e\u003cp\u003eThe univariate analysis revealed that non-survivors had lower GV, as reflected by significantly reduced SD (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007), CV (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.036), and glucose range (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.038), while MBG did not differ significantly (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.132). Stratification by CV quartiles suggested a potential inverse relationship between GV and mortality; however, this trend did not reach statistical significance (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.463). Median ICU length of stay appeared similar across quartiles, with no consistent trend observed. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents a scatter plot of CV and ICU length of stay. No significant linear correlation was found between CV and ICU stay duration (r\u0026thinsp;=\u0026thinsp;0.07, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.147), suggesting limited direct association in the univariate context.\u003c/p\u003e\u003cp\u003eSeveral prior studies have reported a strong association between elevated GV and poor outcomes in critically ill patients. Krinsley et al.(9) conducted a retrospective analysis of 4,084 ICU patients and found that higher GV (measured by CV) was independently associated with increased mortality among non-diabetic patients, even after adjusting for illness severity and excluding those with hypoglycemia (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). However, no such association was observed among diabetic patients. For example, in non-diabetic patients with a mean glucose level of 70\u0026ndash;99 mg/dL, mortality rose dramatically from 10.2% with CV\u0026thinsp;\u0026lt;\u0026thinsp;15% to 58.3% with CV\u0026thinsp;\u0026ge;\u0026thinsp;50%. These findings highlight the prognostic relevance of GV particularly in non-diabetic ICU patients. Compared with the study by Zhu et al.(15), which utilized the MIMIC-IV database to analyze the prognostic value of GV using MAGE in over 13,000 ICU patients, our findings show partial consistency. Zhu et al. reported a strong association between higher MAGE levels and increased ICU, in-hospital, and 28-day mortality, particularly in non-diabetic patients (HR up to 3.59). In our study, although GV metrics such as CV and SD were not independently associated with in-hospital mortality in the fully adjusted models (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05), subgroup analyses revealed potential associations in specific populations, such as older adults and those without hypertension. While different GV metrics were used\u0026mdash;MAGE in Zhu et al. versus CV and SD in our study\u0026mdash;both investigations underscore the importance of glycemic fluctuations as a prognostic indicator in critical illness. Variations in predictive performance may be attributed to differences in sample size, patient composition, and statistical approaches. Qi et al.(16) analyzed ICU patients with traumatic brain injury (TBI) using the MIMIC-IV database and found that higher glycemic variability, measured by coefficient of variation (CV), was significantly associated with worse neurological outcomes and increased in-hospital mortality (HR\u0026thinsp;=\u0026thinsp;1.74, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003). Their findings align with our observation that elevated GV may negatively impact prognosis. In contrast, our study did not find an independent association between GV and in-hospital mortality in ischemic stroke patients. This discrepancy may be attributable to differences in study populations, as our cohort was restricted to stroke patients\u0026mdash;a group in whom neurological status, such as GCS score, may exert a more dominant influence on prognosis than glycemic metrics. Furthermore, our study adjusted for a wider range of confounding factors, and the inclusion criteria excluded patients with major glucose-altering comorbidities, possibly attenuating the GV effect.\u003c/p\u003e\u003cp\u003eGV is considered a surrogate marker of physiological stress and metabolic instability. Fluctuations in blood glucose may trigger oxidative stress, endothelial dysfunction, and systemic inflammation, all of which are detrimental in critically ill patients(7, 17). However, in ischemic stroke, such effects may be masked by factors more directly related to cerebral injury\u0026mdash;such as neurological severity and cerebral perfusion status. The paradoxical finding that lower GV was associated with higher mortality may reflect an impaired metabolic adaptive response in severely ill patients (18, 19).\u003c/p\u003e\u003cp\u003eThe observation that GV appeared protective in non-hypertensive and older patients raises the possibility that moderate glucose fluctuations could serve as indicators of residual metabolic flexibility in selected populations(20, 21). These nuances underscore the need for personalized glycemic targets and the limitations of \"one-size-fits-all\" glycemic management strategies in stroke ICUs.\u003c/p\u003e\u003cp\u003eOur XGBoost model achieved an AUC of 0.657 for mortality prediction, with traditional predictors (GCS, age, atrial fibrillation) ranking highest in feature importance. GV metrics such as CV and SD contributed moderately, indicating they offer complementary\u0026mdash;but not central\u0026mdash;value in outcome prediction. This finding is consistent with previous studies showing that GV enhances, but does not replace, conventional risk stratification methods in ICU patients(22, 23).\u003c/p\u003e\u003cp\u003eTo test model generalizability, we applied both regression and machine learning models to the eICU database. Neither the logistic regression ORs nor the AUCs from XGBoost differed significantly between datasets (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.6), supporting the stability of GV-related findings across geographically and temporally distinct ICU populations. This external validation strengthens the reproducibility of our results and reinforces the utility of combining multiple databases for critical care research.\u003c/p\u003e\u003cp\u003eSeveral limitations of this study should be acknowledged. First, the retrospective design limits the ability to draw causal inferences between glycemic variability and patient outcomes. Second, although we adjusted for basic demographic and clinical variables, some important confounders\u0026mdash;such as stroke subtype, neuroimaging findings, and neurological scores like the NIHSS\u0026mdash;were not available in the databases. We used the GCS score at ICU admission as a surrogate for baseline neurological status, which may not fully capture the extent of brain injury. Third, data on critical interventions, including thrombolytic therapy (e.g., rtPA) and endovascular procedures, were incomplete or inconsistently recorded, limiting our ability to assess treatment-related differences. Fourth, ICU-related complications such as aspiration pneumonia, mechanical ventilation, ventilator-associated pneumonia, and sepsis may have influenced both glucose variability and outcomes. While some of these were identifiable via diagnostic codes, the potential for residual confounding remains. Finally, decisions regarding withdrawal of life-sustaining treatment\u0026mdash;an important factor affecting in-hospital mortality\u0026mdash;were not systematically documented in either dataset.\u003c/p\u003e\u003cp\u003eProspective studies with protocolized glucose monitoring and treatment data are needed to validate the role of GV in ischemic stroke outcomes. Future work could explore real-time GV monitoring as a clinical decision tool, investigate the interplay between GV and neuroimaging findings, and determine whether GV-targeted interventions can improve recovery in specific stroke subgroups.\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eIn this multi-database cohort study of ICU patients with ischemic stroke, we found that glycemic variability, as measured by SD, CV, and glucose range, was significantly higher in survivors than in non-survivors in univariate analysis. However, these metrics were not independently associated with in-hospital mortality or ICU length of stay after multivariable adjustment. Machine learning models identified glycemic variability as a moderately important feature in mortality prediction, and external validation across the MIMIC-Ⅳ and eICU databases demonstrated stable and reproducible model performance.\u003c/p\u003e\u003cp\u003eThese findings suggest that while glycemic variability reflects aspects of metabolic stress in critically ill stroke patients, it may not serve as a strong independent prognostic factor when traditional clinical variables are accounted for. Future prospective studies are needed to further clarify the role of glycemic dynamics in guiding personalized glucose management strategies in neurocritical care.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eAuthors\u0026rsquo; contributions\u003c/p\u003e\n\u003cp\u003eYong Xia and Qiao Huang participated in the conceptualization of protocol, quality assessment, data extraction, formal analysis, methodology, writing-original draft, writing-review and editing, and approving the final draft. Yanbin Chen participated in the formal analysis, Methodology, writing-review and editing,Liu Yang and Jingyuan Kang participated in Supervision and Visualization. All authors read and approved the manuscript.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis work was supported by Hunan Provincial Natural Science Foundation Project(2025JJ70588)\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe data sets used and/or analyzed during the current systematic review and meta-analysis are presented within the manuscript\u003c/p\u003e\n\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable. Because all reviewed studies provided appropriate information on obtaining ethical approval in accordance with the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNot applicable. Because all reviewed studies provided appropriate information regarding participant consent.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests to report.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGolubnitschaja O, Potuznik P, Polivka J, Jr., Pesta M, Kaverina O, Pieper CC, et al. Ischemic stroke of unclear aetiology: a case-by-case analysis and call for a multi-professional predictive, preventive and personalised approach. Epma j. 2022;13(4):535-45.\u003c/li\u003e\n\u003cli\u003eZhang FL, Guo ZN, Wu YH, Liu HY, Luo Y, Sun MS, et al. Prevalence of stroke and associated risk factors: a population based cross sectional study from northeast China. BMJ Open. 2017;7(9):e015758.\u003c/li\u003e\n\u003cli\u003eCai W, Xu J, Wu X, Chen Z, Zeng L, Song X, et al. Association between triglyceride-glucose index and all-cause mortality in critically ill patients with ischemic stroke: analysis of the MIMIC-IV database. Cardiovasc Diabetol. 2023;22(1):138.\u003c/li\u003e\n\u003cli\u003eLi XD, Li MM. A novel nomogram to predict mortality in patients with stroke: a survival analysis based on the MIMIC-III clinical database. BMC Med Inform Decis Mak. 2022;22(1):92.\u003c/li\u003e\n\u003cli\u003eLi S, Wang Y, Zhu X, Zheng H, Ni J, Li H, et al. Lipid on stroke in intracranial artery atherosclerotic stenosis: a mediation role of glucose. Front Endocrinol (Lausanne). 2024;15:1322114.\u003c/li\u003e\n\u003cli\u003eVasa D, Rossitto CP, Ezzat B, Bazil M, Schuldt B, Johnson B, et al. Stress hyperglycemia is associated with longer ICU length of stay after endoscopic intracerebral hemorrhage evacuation. J Stroke Cerebrovasc Dis. 2024;33(10):107911.\u003c/li\u003e\n\u003cli\u003eSaid AME, Sayed Hussien N, Negm MA. Dynamic of Glycemic State as a Prognostic Marker in Acute Ischemic Stroke: A Prospective Controlled Cohort Observational Study. QJM: An International Journal of Medicine. 2023;116(Supplement_1).\u003c/li\u003e\n\u003cli\u003eMa Z, Li S, Lin X. Body mass index, blood glucose, and mortality in patients with ischemic stroke in the intensive care unit: A retrospective cohort study. Front Neurosci. 2022;16:946397.\u003c/li\u003e\n\u003cli\u003eKrinsley JS. Glycemic variability and mortality in critically ill patients: the impact of diabetes. J Diabetes Sci Technol. 2009;3(6):1292-301.\u003c/li\u003e\n\u003cli\u003eLaredo C, Ren\u0026uacute; A, Llull L, Tudela R, L\u0026oacute;pez-Rueda A, Urra X, et al. Elevated glucose is associated with hemorrhagic transformation after mechanical thrombectomy in acute ischemic stroke patients with severe pretreatment hypoperfusion. Sci Rep. 2020;10(1):10588.\u003c/li\u003e\n\u003cli\u003eCai W, Li Y, Guo K, Wu X, Chen C, Lin X. Association of glycemic variability with death and severe consciousness disturbance among critically ill patients with cerebrovascular disease: analysis of the MIMIC-IV database. Cardiovasc Diabetol. 2023;22(1):315.\u003c/li\u003e\n\u003cli\u003eSu Y, Fan W, Liu Y, Hong K. Glycemic variability and in-hospital death of critically ill patients and the role of ventricular arrhythmias. Cardiovasc Diabetol. 2023;22(1):134.\u003c/li\u003e\n\u003cli\u003eWu WT, Li YJ, Feng AZ, Li L, Huang T, Xu AD, et al. Data mining in clinical big data: the frequently used databases, steps, and methodological models. Mil Med Res. 2021;8(1):44.\u003c/li\u003e\n\u003cli\u003eKim Y, Kim M, Kim Y, Choi M. Using nursing data for machine learning-based prediction modeling in intensive care units: A scoping review. Int J Nurs Stud. 2025;169:105133.\u003c/li\u003e\n\u003cli\u003eZhu Q, Zong Q, Guo S, Ye H, Ma Z, Zhang R, et al. Mean amplitude of glycemic excursion and mortality in critically ill patients: A retrospective analysis using the MIMIC-IV database. Diabetes Obes Metab. 2025;27(7):3831-9.\u003c/li\u003e\n\u003cli\u003eQi L, Geng X, Feng R, Wu S, Fu T, Li N, et al. Association of glycemic variability and prognosis in patients with traumatic brain injury: A retrospective study from the MIMIC-IV database. Diabetes Res Clin Pract. 2024;217:111869.\u003c/li\u003e\n\u003cli\u003eHou Y, Wu X, Shi Y, Xu X, Zhang Y, Jiang L, et al. METS-IR as an important predictor of neurological impairment severity in patients with severe cerebral infarction: a multicenter study based on the Chinese population. Front Neurol. 2024;15:1450825.\u003c/li\u003e\n\u003cli\u003ePensato U, Bosshart S, Stebner A, Rohr A, Kleinig TJ, Gupta R, et al. Effect of Hemoglobin and Blood Glucose Levels on CT Perfusion Ischemic Core Estimation: A Post Hoc Analysis of the ESCAPE-NA1 Trial. Neurology. 2024;103(10):e209939.\u003c/li\u003e\n\u003cli\u003eLu Z, Tao G, Sun X, Zhang Y, Jiang M, Liu Y, et al. Association of Blood Glucose Level and Glycemic Variability With Mortality in Sepsis Patients During ICU Hospitalization. Front Public Health. 2022;10:857368.\u003c/li\u003e\n\u003cli\u003eShi X, Yang S, Guo C, Sun W, Song J, Fan S, et al. Impact of stress hyperglycemia on outcomes in patients with large ischemic stroke. J Neurointerv Surg. 2025.\u003c/li\u003e\n\u003cli\u003eEmgin \u0026Ouml;, Yavuz M, Sahin A, G\u0026uuml;nes M, Eser M, Yavuz T, et al. The Association Between Glycemic Variability and Mortality in Critically Ill Patients: A Multicenter Prospective Observational Study. Journal of Clinical Medicine. 2024;13(22):6939.\u003c/li\u003e\n\u003cli\u003eGuo Y, Qiu Y, Xue T, Zhou Y, Yan P, Liu S, et al. Association between glycemic variability and short-term mortality in patients with acute kidney injury: a retrospective cohort study of the MIMIC-IV database. Sci Rep. 2024;14(1):5945.\u003c/li\u003e\n\u003cli\u003eLin L, Liang Z. Association Between Glycemic Variability and All-Cause Mortality in Patients with Acute Pancreatitis in the Intensive Care Unit: A Retrospective Analysis. Dig Dis Sci. 2025;70(6):2194-203.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"glycemic variability, ischemic stroke, intensive care unit, in-hospital mortality, blood glucose, machine learning, MIMIC, eICU","lastPublishedDoi":"10.21203/rs.3.rs-7686095/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7686095/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eGlycemic variability (GV) is an emerging marker of metabolic stress in critically ill patients, but its prognostic value in ICU patients with ischemic stroke remains unclear.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eThis retrospective cohort study used MIMIC-Ⅳ and eICU databases to assess associations between GV and outcomes in adult ischemic stroke patients. GV metrics\u0026mdash;including mean blood glucose (MBG), standard deviation (SD), coefficient of variation (CV), and glucose range\u0026mdash;were calculated from all ICU glucose readings. The primary outcome was in-hospital mortality; secondary was ICU length of stay (LOS). Multivariable regression and XGBoost machine learning were used, with external validation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eA total of 418 patients were included (334 survivors, 84 non-survivors). In univariate analyses, non-survivors had significantly lower SD (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007), CV (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.036), and glucose range (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.038) than survivors, while MBG did not differ significantly (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.132). However, no GV metric remained independently associated with mortality or ICU LOS after multivariable adjustment (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05). XGBoost models showed moderate predictive performance (AUC\u0026thinsp;=\u0026thinsp;0.657), with GV metrics contributing moderate feature importance. Subgroup analyses indicated a possible protective association between higher CV and mortality in older or non-hypertensive patients. Findings were consistent across databases (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.6).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eIn ICU patients with ischemic stroke, GV was linked to outcomes in unadjusted analyses but was not an independent predictor after adjustment. GV may aid in risk stratification, though traditional clinical variables remain more predictive.\u003c/p\u003e","manuscriptTitle":"Glycemic Variability and Clinical Outcomes in ICU Patients with Ischemic Stroke: A Multi-Database Retrospective Cohort Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-01 08:03:17","doi":"10.21203/rs.3.rs-7686095/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-10T12:02:16+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-10T10:18:05+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-30T21:11:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"75181019885479650261177249290637883307","date":"2026-01-26T17:43:51+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-20T17:00:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"230416640489895190380611506186792840679","date":"2026-01-20T15:34:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"39175754215560372619113931505373513066","date":"2026-01-10T23:07:17+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-24T11:47:24+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-25T15:30:40+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-24T10:50:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"European Journal of Medical Research","date":"2025-09-22T14:52:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"european-journal-of-medical-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ejmr","sideBox":"Learn more about [European Journal of Medical Research](http://eurjmedres.biomedcentral.com)","snPcode":"40001","submissionUrl":"https://submission.nature.com/new-submission/40001/3","title":"European Journal of Medical Research","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d8291622-0cd6-4ffb-a8f9-5ae863d2b48a","owner":[],"postedDate":"December 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-30T16:26:35+00:00","versionOfRecord":{"articleIdentity":"rs-7686095","link":"https://doi.org/10.1186/s40001-026-04325-9","journal":{"identity":"european-journal-of-medical-research","isVorOnly":false,"title":"European Journal of Medical Research"},"publishedOn":"2026-03-28 16:10:04","publishedOnDateReadable":"March 28th, 2026"},"versionCreatedAt":"2025-12-01 08:03:17","video":"","vorDoi":"10.1186/s40001-026-04325-9","vorDoiUrl":"https://doi.org/10.1186/s40001-026-04325-9","workflowStages":[]},"version":"v1","identity":"rs-7686095","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7686095","identity":"rs-7686095","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}