Serum glucose and potassium ratio as a predictive factor for prognosis of acute intracerebral hemorrhage

Effect of stress-induced hyperglycemia on long-term mortality in non-diabetic patients with acute type A aortic dissection: a retrospective analysis

BACKGROUND

Acute Type A Aortic Dissection (AAAD) is one of the most life-threatening diseases, often associated with transient hyperglycemia induced by acute physiological stress. The impact of stress-induced hyperglycemia on the prognosis of ST-segment elevation myocardial infarction has been reported. However, the relationship between stress-induced hyperglycemia and the prognosis of AAAD patients remains uncertain.

METHODS

The clinical data of 456 patients with acute type A aortic dissection were retrospectively reviewed. Patients were divided into two groups based on their admission blood glucose. Cox model regression analysis was performed to assess the relationship between stress-induced hyperglycemia and the 30-day and 1-year mortality rates of these patients.

RESULTS

Among the 456 patients, 149 cases (32.7%) had AAAD combined with stress-induced hyperglycemia (SIH). The results of the multifactor regression analysis of the Cox model indicated that hyperglycemia (RR = 1.505, 95% CI: 1.046-2.165, p = 0.028), aortic coarctation involving renal arteries (RR = 3.330, 95% CI: 2.237-4.957, p < 0.001), aortic coarctation involving superior mesenteric arteries (RR = 1.611, 95% CI: 1.056-2.455, p = 0.027), and aortic coarctation involving iliac arteries (RR = 2.034, 95% CI: 1.364-3.035, p = 0.001) were independent influences on 1-year postoperative mortality in AAAD patients.

CONCLUSION

The current findings indicate that stress-induced hyperglycemia measured on admission is strongly associated with 1-year mortality in patients with AAAD. Furthermore, stress-induced hyperglycemia may be related to the severity of the condition in patients with AAAD.

A prospective cohort study regarding usability of serum RIPK3 as a biomarker in relation to early hematoma growth and neurological outcomes after acute intracerebral hemorrhage

OBJECTIVE

The receptor-interacting protein kinase 3 (RIPK3) is a pivotal component for triggering necroptosis. We intended to investigate predictive effects of serum RIPK3 levels on early hematoma growth (EHG) and poor neurological outcome after acute intracerebral hemorrhage (ICH).

METHODS

In this prospective cohort study, 183 ICH patients and 100 controls were enrolled for measuring serum RIPK3 levels. National Institutes of Health Stroke Scale (NIHSS) and hematoma volume were recorded as the severity indicators. EHG and poststroke 6-month unfavorable outcome (modified Rankin Scale scores of 3-6) were registered as the two prognostic parameters. Multivariate analyses were implemented to discern relevance of serum RIPK3 to ICH severity and prognosis.

RESULTS

Serum RIPK3 levels of patients, which were dramatically higher than those of controls, were independently related to NIHSS scores, hematoma volume, EHG, 6-month mRS scores and unfavorable outcome. Risks of EHG and unfavorable outcome were linearly pertinent to and efficiently discriminated by RIPK3 levels under restricted cubic spline and receiver operating characteristic curve respectively. RIPK3 levels nonsignificantly interacted with age, gender, hypertension, etc. Predictive ability of RIPK3 levels resembled those of NIHSS scores and hematoma volume. The prediction models, in which serum RIPK3, NIHSS scores and hematoma volume were integrated, were visually displayed via nomograms. The models' predictive capabilities substantially surpassed that of serum RIPK3, NIHSS scores and hematoma volumes alone. The models kept stable under calibration curve.

CONCLUSION

A profound increase of serum RIPK3 levels after ICH is tightly relevant to severity, EHG and poor neurological outcomes, assuming that serum RIPK3 may emerge as a valuable prognostic predictor of ICH.

Systemic immune-inflammation index and serum glucose-potassium ratio predict poor prognosis in patients with spontaneous cerebral hemorrhage: An observational study

Recent studies have shown systemic inflammatory response, serum glucose, and serum potassium are associated with poor prognosis in spontaneous intracerebral hemorrhage (SICH). This retrospective study aimed to investigate the association of systemic immune-inflammatory index (SII) and serum glucose-potassium ratio (GPR) with the severity of disease and the poor prognosis of patients with SICH at 3 months after hospital discharge. We reviewed the clinical data of 105 patients with SICH, assessed the extent of their disease using Glasgow Coma Scale score, National Institutes of Health Stroke Scale (NIHSS) score, and hematoma volume, and categorized them into a good prognosis group (0-3 scores) and a poor prognosis group (4-6 scores) based on their mRS scores at 3 months after hospital discharge. Demographic characteristics, clinical, laboratory, and imaging data at admission were compared between the 2 groups, bivariate correlations were analyzed using Spearman's correlation coefficients, multivariate logistic regression analysis was used to determine the independent risk factors for poor prognosis of patients with SICH, and finally, SII, GPR, and platelet/lymphocyte ratio (PLR) were examined using the subject's work characteristics (ROC) curve, lymphocyte/monocyte ratio (LMR), and neutrophil/lymphocyte ratio (NLR) for their predictive efficacy for poor prognosis. Patients in the poor prognosis group had significantly higher SII and serum GPR than those in the good prognosis group, and Spearman analysis showed that SII and serum GPR were significantly correlated with the admission Glasgow Coma Scale score as well as the NIHSS score and that SII and GPR increased with the increase in mRS score. Multivariate logistic regression analysis showed that admission NIHSS score, hematoma volume SII, GPR, NLR, and PLR were independently associated with poor patient prognosis. Analysis of the subjects' work characteristic curves showed that the areas under the SII, GPR, NLR, PLR, LMR, and coSII-GPR curves were 0.838, 0.837, 0.825, 0.718, 0.616, and 0.883. SII and GRP were significantly associated with disease severity and short-term prognosis in SICH patients 3 months after discharge, and SII and GPR had better predictive value compared with NLR, PLR, and LMR. In addition, coSII-GPR, a joint indicator based on SII and GPR, can improve the predictive accuracy of poor prognosis 3 months after discharge in patients with SICH.

Association between admission serum potassium concentration and the island sign on cranial CT in HICH patients: a cross-sectional study

Objective

This study aimed to explore the correlation between serum potassium (K+) concentration upon admission and the presence of the Island Sign (IS) in cranial CT scans of patients diagnosed with Hypertensive Intracerebral Hemorrhage (HICH), including the potential presence of a non-linear relationship.

Methods

This investigation constituted a single-center cross-sectional study. We systematically gathered comprehensive general clinical characteristics, biological indicators, and imaging data from a cohort of 330 patients diagnosed with HICH. These patients received treatment within the neurosurgery department of Chongqing Emergency Medical Center during the period spanning from July 1, 2018, to July 7, 2023. Our primary objective was to scrutinize the potential connection between serum K+ concentration upon admission and the presence of the IS observed in cranial CT scans. To meticulously address this inquiry, we employed logistic regression modeling, thereby meticulously evaluating the correlation aforementioned. Moreover, in order to delve deeper into the intricacies of the relationship, we extended our analysis by employing a smoothed curve-fitting model to meticulously authenticate the potential non-linear interrelation between these two critical variables.

Results

In this investigation, a total of 330 patients diagnosed with HICH were ultimately enrolled, exhibiting an average age of 58.4 ± 13.1 years, comprising 238 (72.1%) males and 92 (27.9%) females. Among these participants, 118 individuals (35.7%) presented with the IS upon admission cranial CT scans, while 212 patients (64.3%) did not exhibit this characteristic. Upon comprehensive multifactorial adjustments, a non-linear association was uncovered between serum K+ concentration and the presence of IS. Notably, an inflection point was identified at approximately 3.54 mmol/L for serum K+ concentration. Prior to the patient's serum K+ concentration reaching around 3.54 mmol/L upon admission, a discernible trend was observed-every 0.1 mmol/L increment in serum K+ concentration was associated with an 8% decrease in the incidence of IS (OR: 0.914, 95% CI: 0.849-0.983, p = 0.015).

Conclusion

The findings of this study underscore a negative association between reduced serum K+ concentration upon admission and the occurrence of the IS on cranial CT scans among patients diagnosed with hypertensive cerebral hemorrhage. Furthermore, this negative correlation appears to manifest within the realm of a non-linear relationship. This study elucidates the potential significance of serum K+ concentration levels among patients with HICH, highlighting the role they play. Moreover, the maintenance of a physiological equilibrium in serum K+ concentrations emerges as a conceivable protective factor for individuals within the stroke population.

The utility of serum glucose potassium ratio as a predictive factor for haemorrhagic transformation, stroke recurrence, and mortality among ischemic stroke patients

Background and Objective

Glucose-Potassium Ratio (GPR) has emerged as a biomarker in several pathophysiological conditions. However, the association between GPR and long-term outcomes in stroke patients has not been investigated. Our study evaluated the applicability of baseline GPR as a predictive prognostic tool for clinical outcomes in ischemic stroke patients.

Methods

The multicenter retrospective cohort study included acute-subacute adult ischemic stroke patients who had their baseline serum GPR levels measured. Eligible patients were categorized into two sub-cohorts based on the baseline GPR levels (<1.67 vs. ≥ 1.67). The primary outcome was the incidence of 30-day hemorrhagic transformation, while stroke recurrence, and all-cause mortality within twelve months, were considered secondary.

Results

Among 4083 patients screened, 1047 were included in the current study. In comparison with GPR < 1.67 group, patients with ≥ 1.67 GPR had a significantly higher ratio of all-cause mortality within twelve months (aHR 2.07 [95 % CI 1.21-3.75] p = 0.01), and higher ratio of 30-day hemorrhagic transformation but failed to reach the statistical significance (aHR 1.60 [95 % CI 0.95-2.79], p = 0.08).

Conclusion

Overall, baseline GPR serum is an independent predictor of all-cause mortality within twelve months in patients with acute and subacute ischemic stroke. Further clinical studies are necessary to validate these findings.

Usability of serum AIM2 as a predictive biomarker of stroke-associated pneumonia and poor prognosis after acute supratentorial intracerebral hemorrhage: A prospective longitudinal cohort study

Background

Absent in melanoma 2 (AIM2) is implicated in inflammatory processes. We measured serum AIM2 with intent to unveil its predictive significance for stroke-associated pneumonia (SAP) and functional prognosis following acute intracerebral hemorrhage (ICH).

Methods

In this prospective cohort study, serum AIM2 concentrations of 163 ICH patients were gauged upon admission and 57 of them also consented for measurements at days 1, 3, 5, 7, 10 and 14. Coupled with 57 individuals without health conditions, dynamic change of serum AIM2 levels were uncovered. National Institutes of Health Stroke Scale (NIHSS) scores and hematoma volume were identified as the dual indicators of severity. Poststroke six-month modified Rankin Scale (mRS) scores ranging from 3 to 6 indicated an unfavorable outcome. SAP was observed during the first seven days after ICH. Sequential univariate and multivariate analyses were performed to discern predictors of SAP and adverse prognosis.

Results

The serum levels of AIM2 in patients exhibited a marked elevation upon admission, reaching peak levels on the third and fifth days, and remained notably elevated until day 14 compared to those of the control group. Serum AIM2 levels showed independent correlations with both NIHSS scores and the volume of hematoma. Additionally, AIM2 concentrations were independently associated with a poor prognosis and SAP at the six-month mark. Within the framework of restricted cubic spline analysis, serum AIM2 concentrations exhibited a linear correlation with the likelihood of developing SAP and experiencing a poor prognosis. In the context of receiver operating characteristic (ROC) curve analysis, serum AIM2 concentrations effectively differentiated risks of SAP and poor prognosis. By employing segmented analysis, serum AIM2 concentrations showed negligible interactions with several traditional variables, such as age, gender, smoking habits, alcohol consumption, and more. The integrated model incorporating serum AIM2, NIHSS scores, and the volume of hematoma was depicted by employing a nomogram and demonstrated strong predictive performance for poor prognosis or SAP across various evaluation metrics, including ROC curve analysis, calibration curve analysis, and decision curve analysis.

Conclusion

Serum AIM2 levels show a marked increase shortly after intracerebral hemorrhage (ICH), which may accurately reflect stroke severity, and effectively predict SAP and poor neurological outcomes, and therefore serum AIM2 stands out as an encouraging predictive indicator for ICH.

Study of certain easily available biochemical markers in prognostication in severe traumatic brain injury requiring surgery

Background

This study aimed to identify easily available prognostic factors in severe traumatic brain injury (TBI) patients undergoing craniotomy.

Methods

We retrospectively analyzed the clinical characteristics (age, sex, Glasgow coma scale score, cause of TBI, and oral antithrombotic drug use), laboratory parameters (hemoglobin, sodium, C-reactive protein, D-dimer, activated partial thromboplastin time, prothrombin time-international normalized ratio, and glucose-potassium [GP] ratio), and neuroradiological findings of 132 patients who underwent craniotomy for severe TBI in our hospital between January 2015 and December 2021. The patients were divided into two groups: Those with fatal clinical outcomes and those with non-fatal clinical outcomes, and compared between the two groups.

Results

The patients comprised 79 (59.8%) male and 53 (40.2%) female patients. Their mean age was 67 ± 17 years (range, 16-94 years). Computed tomography revealed acute subdural hematoma in 108 (81.8%) patients, acute epidural hematoma in 31 (23.5%), traumatic brain contusion in 39 (29.5%), and traumatic subarachnoid hemorrhage in 62 (47.0%). All 132 patients underwent craniotomy, and 41 eventually died. There were significant differences in the D-dimer, GP ratio, and optic nerve sheath diameter between the groups (all P < 0.01). Multivariate logistic regression analysis showed elevated GP ratio and D-dimer were associated with the death group (P < 0.01, P < 0.01, respectively). A GP ratio of >42 was the optimal cutoff value for the prediction of a fatal outcome of TBI (sensitivity, 85.4%; specificity, 51.1%).

Conclusion

The GP ratio and D-dimer were significantly associated with poor outcomes of TBI. A GP ratio of >42 could be a predictor of a fatal outcome of TBI.

The blood glucose-potassium ratio at admission predicts in-hospital mortality in patients with acute type A aortic dissection

Acute type A aortic dissection (ATAAD) is a serious cardiovascular emergency with high risk and mortality after surgery. Recent studies have shown that serum glucose-potassium ratio (GPR) is associated with the prognosis of cerebrovascular diseases. The purpose of this study was to investigate the relationship between GPR and in-hospital mortality in patients with ATAAD. From June 2019 to August 2021, we retrospectively analyzed the clinical data of 272 patients who underwent ATAAD surgery. According to the median value of GPR (1.74), the patients were divided into two groups. Univariate and multivariate logistic regression analysis were used to determine the risk factors of in-hospital mortality after ATAAD. In-hospital death was significantly more common in the high GPR group (> 1.74) (24.4% vs 13.9%; P = 0.027). The incidence of renal dysfunction in the low GPR group was significantly higher than that in the high GPR group (26.3% vs 14.8%: P = 0.019). After controlling for potential confounding variables and adjusting for multivariate logistic regression analysis, the results showed a high GPR (> 1.74) (AOR 4.70, 95% confidence interval (CI) 2.13-10.40; P =  < 0.001), lactic acid (AOR 1.14, 95% CI 1.03-1.26; P = 0.009), smokers (AOR 2.45, 95% CI 1.18-15.07; P = 0.039), mechanical ventilation (AOR 9.47, 95% CI 4.00-22.38; P =  < 0.001) was independent risk factor for in-hospital mortality in ATAAD patients, albumin (AOR 0.90, 95% CI 0.83-0.98; P = 0.014) was a protective factor for in-hospital prognosis. High GPR is a good predictor of in-hospital mortality after ATAAD surgery.

Prognostic and predictive significance of serum soluble scavenger receptor A in acute primary basal ganglia hemorrhage: A prospective cohort study

BACKGROUND

Scavenger receptor A (SRA) can regulate immune response and is involved in pathophysiological processes of acute brain injury. We analyzed the prognostic role of serum soluble SRA in intracerebral hemorrhage (ICH).

METHODS

In this prospective cohort study of 110 healthy controls and 110 patients with acute basal ganglia hemorrhage, serum soluble SRA concentrations were detected. Univariate analyses, followed by multivariate logistic regression analyses, were utilized to explore the relationship between serum soluble SRA concentrations and early neurologic deterioration (END) plus post-stroke 3-month poor prognosis (modified Rankin Scale scores of 3-6).

RESULTS

Serum soluble SRA concentrations of patients were significantly higher than those of controls (median, 3.6 vs 0.9 ng/ml; P < 0.001). Serum soluble SRA concentrations of patients were independently correlated with hematoma volume (β, 0.201; 95 % confidence interval (CI), 0.093-0.309; P = 0.001), National Institutes of Health Stroke Scale (NIHSS) scores (β, 0.118; 95 % CI, 0.024-0.213; P = 0.024), and 3-month modified Rankin Scale scores (β, 0.148; 95 % CI, 0.063-0.232; P = 0.001). Serum soluble SRA concentrations independently predicted END and poor 3-month prognosis with odds ratio values of 1.394 (95 % CI, 1.024-1.899; P = 0.035) and 1.441 (95 % CI, 1.016-2.044; P = 0.040) respectively. Serum soluble SRA concentrations were efficiently predictive of the development of END (ROC AUC 0.746; 95 % CI, 0.631-0.861) and poor 3-month prognosis (AUC, 0.773; 95 % CI, 0.685-0.861). Serum soluble SRA concentrations significantly improved AUCs of NIHSS score and hematoma volume to 0.889 (95 % CI, 0.829-0.948; P = 0.035) and 0.873 (95 % CI, 0.811-0.936; P = 0.036) for prognostic prediction. The END predictive ability of serum sSRA concentrations combined with NIHSS score and ICH volume (AUC, 0.900; 95 % CI, 0.835-0.965) was significantly superior to those of NIHSS score (P = 0.020) and hematoma volume (P = 0.022). The prognostic predictive capability of serum sSRA concentrations combined with NIHSS score and ICH volume (AUC, 0.907; 95 % CI, 0.852-0.962) substantially exceeded those of NIHSS score (P = 0.009) and hematoma volume (P = 0.005).

CONCLUSIONS

Serum soluble SRA concentrations may reflect illness severity and neurologic function after ICH, indicating serum soluble SRA may serve as a promising prognostic biochemical marker of ICH.

Elevated serum nuclear factor erythroid 2-related factor 2 levels contribute to a poor prognosis after acute supratentorial intracerebral hemorrhage: A prospective cohort study

Objective

Nuclear factor erythroid 2-related factor 2 (Nrf2) is a key transcriptional factor for antioxidant response element-regulated genes. The purpose of this study was to assess the prognostic role of serum Nrf2 in intracerebral hemorrhage (ICH).

Materials and methods

In this prospective observational study, serum Nrf2 levels of 115 acute supratentorial ICH patients and 115 controls were gaged. Early neurologic deterioration (END) was defined as an increase of four or greater points in National Institutes of Health Stroke Scale (NIHSS) score or death at post-stroke 24 h. A poor outcome was referred to as the post-stroke 90-day modified Rankin scale (mRS) score of 3–6. END and a poor outcome were considered as the two prognostic parameters.

Results

As compared to controls, serum Nrf2 levels of patients were substantially elevated (P &lt; 0.001), with its levels increasing during the 6-h period immediately, peaking in 12–18 h, plateauing at 18–24 h, and decreasing gradually thereafter (P &lt; 0.05). Serum Nrf2 levels of patients were independently correlated with NIHSS score (t = 3.033; P = 0.003) and hematoma volume (t = 3.210; P = 0.002), independently predicted END (odds ratio 1.125; 95% confidence interval 1.027–1.232; P = 0.011) and poor outcome (odds ratio 1.217; 95% confidence interval 1.067–1.387; P = 0.013), as well as efficiently distinguished END (area under curve 0.771; 95% confidence interval 0.666–0.877; P &lt; 0.001) and poor outcome (area under curve 0.803; 95% confidence interval 0.725–0.882; P &lt; 0.001). Its predictive ability was equivalent to those of NIHSS score and hematoma volume (both P &gt; 0.05), and it also significantly improved their predictive abilities under receiver operating characteristic (ROC) curve (all P &lt; 0.05).

Conclusion

Elevated serum Nrf2 levels are closely correlated with severity, END, and 90-day poor outcome following ICH. Hence, Nrf2 may play an important role in acute brain injury after ICH, and serum Nrf2 may have the potential to serve as a prognostic biomarker of ICH.

Usability of serum annexin A7 as a biochemical marker of poor outcome and early neurological deterioration after acute primary intracerebral hemorrhage: A prospective cohort study

Objective

Annexin A7 (ANXA7), a calcium-dependent phospholipid-binding protein, may act to aggravate brain injury. This study aimed to assess the clinical utility of serum ANXA7 as a predictor of severity, early neurological deterioration (END), and prognosis after intracerebral hemorrhage (ICH).

Methods

A total of 126 ICH patients and 126 healthy controls were enrolled. Symptomatic severity was evaluated utilizing the National Institutes of Health Stroke Scale (NIHSS) score. The lesion volume of ICH was measured according to the ABC/2 method. END was referred to as an increase of 4 or greater points in the NIHSS score or death at post-stroke 24 h. The unfavorable functional outcome was a combination of death and major disability at post-stroke 90 days.

Results

Serum ANXA7 levels were significantly higher in patients than in controls (median, 46.5 vs. 9.7 ng/ml; P &lt; 0.001). Serum ANXA7 levels were independently correlated with NIHSS score [beta: 0.821; 95% confidence interval (CI): 0.106–1.514; variance inflation factor: 5.180; t = 2.573; P = 0.014] and hematoma volume (beta: 0.794; 95% CI: 0.418–1.173; variance inflation factor: 5.281; t = 2.781; P = 0.007). Serum ANXA7 levels were significantly elevated with increase in modified Rankin scale scores (P &lt; 0.001). Also, serum ANXA7, which was identified as a categorical variable, independently predicted END and an unfavorable outcome with odds ratio values of 3.958 (95% CI: 1.290–12.143; P = 0.016) and 2.755 (95% CI: 1.051–7.220; P = 0.039), respectively. Moreover, serum ANXA7 levels efficiently differentiated END (area under the curve: 0.781; 95% CI: 0.698–0.849) and an unfavorable outcome (area under the curve: 0.776; 95% CI: 0.693–0.846).

Conclusion

Serum ANXA7 may represent a useful blood-derived biomarker for assessing the severity, END, and prognosis of ICH.

Review of the Pathophysiologic and Clinical Aspects of Hypokalemia in Children and Young Adults: an Update

This article examines the regulatory function of the skeletal muscle, renal, and adrenergic systems in potassium homeostasis. The pathophysiologic bases of hypokalemia, systematic approach for an early diagnosis, and therapeutic strategy to avert life-threatening complications are highlighted. By promoting skeletal muscle uptake, intense physical exercise (post), severe trauma, and several toxins produce profound hypokalemia. Hypovolemia due to renal and extra-renal fluid losses and ineffective circulation activate secondary aldosteronism causing urinary potassium wasting. In addition to hypokalemic alkalosis, primary aldosteronism causes low-renin hypertension. Non-aldosterone mineralocorticoid activation leading to low-renin and low-aldosterone hypertension occurs in Liddle's syndrome and apparent mineralocorticoid excess. Although there is enzymatic inhibition of cortisol synthesis in congenital adrenal hyperplasia, precursors of aldosterone produce low-renin hypokalemic hypertension. In addition to the glucocorticoid effect, hypercortisolism activates mineralocorticoid receptors in Cushing's syndrome. Genetic mutations involving furosemide-sensitive Na+-K+-2Cl- co-transporters and thiazide-sensitive Na+-Cl- transporters result in (non-hypertensive) salt-wasting nephropathy. Proximal and distal renal tubular acidosis is associated with hypokalemia. Eating disorders causing hypokalemia include bulimia, laxative abuse, and diuretic misuse. Low urinary potassium (<15 mmol/day) and/or low urinary chloride (<20 mol/L) suggest a gastrointestinal pathology. Co-morbidity of hypokalemia with chronic pulmonary and cardiovascular diseases may increase the fatality rate.

Enhancing Robustness of Machine Learning Integration With Routine Laboratory Blood Tests to Predict Inpatient Mortality After Intracerebral Hemorrhage

Objective: The accurate evaluation of outcomes at a personalized level in patients with intracerebral hemorrhage (ICH) is critical clinical implications. This study aims to evaluate how machine learning integrates with routine laboratory tests and electronic health records (EHRs) data to predict inpatient mortality after ICH.

Methods: In this machine learning-based prognostic study, we included 1,835 consecutive patients with acute ICH between October 2010 and December 2018. The model building process incorporated five pre-implant ICH score variables (clinical features) and 13 out of 59 available routine laboratory parameters. We assessed model performance according to a range of learning metrics, such as the mean area under the receiver operating characteristic curve [AUROC]. We also used the Shapley additive explanation algorithm to explain the prediction model.

Results: Machine learning models using laboratory data achieved AUROCs of 0.71–0.82 in a split-by-year development/testing scheme. The non-linear eXtreme Gradient Boosting model yielded the highest prediction accuracy. In the held-out validation set of development cohort, the predictive model using comprehensive clinical and laboratory parameters outperformed those using clinical alone in predicting in-hospital mortality (AUROC [95% bootstrap confidence interval], 0.899 [0.897–0.901] vs. 0.875 [0.872–0.877]; P <0.001), with over 81% accuracy, sensitivity, and specificity. We observed similar performance in the testing set.

Conclusions: Machine learning integrated with routine laboratory tests and EHRs could significantly promote the accuracy of inpatient ICH mortality prediction. This multidimensional composite prediction strategy might become an intelligent assistive prediction for ICH risk reclassification and offer an example for precision medicine.