| Background:Aneurysmal subarachnoid hemorrhage(a SAH)is a disease with high morbidity and mortality.On average,1/3 of a SAH patients died,and 1/5 of the surviving patients had severe dysfunction.Moreover,a SAH affects younger people than other strokes.In recent years,a SAH has been dramatically improved in the diagnosis and treatment,but the overall treatment effect is still not satisfactory,and further research is needed to improve the treatment effect.The study of prognostic factors and predictive models of a SAH is the leading research direction of aneurysms.Accurate evaluation of prognostic factors and accurate prediction models of a SAH can provide important references for the choice of treatment methods.We designed a large retrospective cohort study to explore four prognostic factors and constructed prognosis prediction models.There are reasons for selecting the prognostic factors to study in the research.Inflammatory/immunologic reactions markedly influence outcomes and predict the clinical course in patients with stroke.Previous studies have revealed the potential of systemic inflammatory markers,including neutrophils,platelets,lymphocytes,and their combination ratios.Recent evidence has revealed that neutrophil-lymphocyte ratio offers predictive potential for mobility and mortality in patients with a SAH;however,those studies were limited by single-center evaluations with little generalizability,small sample sizes resulting in insufficient statistical power to capture true differences in inflammatory markers,and failure to adjust for major confounders.Importantly,these studies did not address the relationships between systemic inflammatory markers and infectious complications.Multiple studies have revealed that infectious complications are common and deadly in patients with a SAH,which highlights the urgent need to develop biomarkers that can help identify patients who are at increased risk for infectious complications.As the most numerous types of leukocytes,neutrophils play a major role in inflammation.Elevated neutrophil counts in stroke have been associated with poor outcomes.However,it remains unclear whether neutrophil count can predict outcomes in patients with a SAH.Chronic liver disease(CLD)is a significant risk factor for increased morbidity and mortality in acutely ill patients,including those with major trauma,or those undergoing major surgery.The number of patients hospitalized for CLD has increased significantly in the past decade,and such patients have an even higher mortality than patient with other chronic diseases such as congestive heart failure and chronic obstructive pulmonary disease.The mortality rates in patients presenting with an acute deterioration of CLD range from 36%to 85%.Some observational studies have shown an association between cirrhosis and worse outcomes in patients with intracranial hemorrhage.For neurosurgeons,CLD and its impact on bleeding risks is a significant clinical concern.There are a paucity of studies that investigate the association between CLD and clinical outcomes in patients with a SAH.Hyperglycemia is a common phenomenon in the acute phase after aneurysmal subarachnoid hemorrhage and is strongly and independently associated with mortality in many diseases,such as acute ischemic stroke and intracerebral hemorrhage.However,previous studies regarding the association in a SAH have shown varied and conflicting results These studies are limited by small sample sizes and single-center evaluations with little generalizability.Most studies to date are based on variable diagnostic criteria,inclusion of other types of subarachnoid hemorrhage and failure to adjust for major confounders.Moreover,none of the previous studies have specifically investigated continuous associations between mortality and glucose values.Hyponatremia is the most common electrolyte disorder encountered in clinical medicine.Hyponatremia has been associated with increased mortality and morbidity in various diseases such as heart failure,cirrhosis,chronic kidney disease.These associations were also seen in patients in the neurocritical care unit,including ischemic stroke,intracerebral hemorrhage,or traumatic brain injury.Potential mechanisms that may explain the association between hyponatremia and mortality in neurocritical ill include the development of cerebral edema,seizures,and delayed cerebral infarctions.In patients with a SAH,10-40%of patients experience hyponatremia,which is associated with the severity of hemorrhage and related complications such as delayed cerebral ischemia.However,no study to date has established the association between hyponatremia and mortality in patients with a SAH.Previous exploratory studies were limited by the small number of patients with a SAH and thus these studies were hampered by a lack of statistical power.Moreover,some of them included other types of SAH and did not adjust for confounders.In recent years,machine learning has made important progress in medicine.Some studies show that machine learning is better than traditional statistical models(such as linear logistics regression)and can significantly improve the ability of clinical prediction.At present,there is few research on the application of machine learning to predict the outcome of a SAH.Objective:The general purpose of this study was to verify whether the four important but neglected prognostic factors were associated with in-hospital mortality of a SAH and whether the prediction effect of the machine learning model was better than that of the traditional model.The specific purpose of each chapter is detailed below.The first chapter was aimed to evaluate the prognostic value of neutrophils count on admission in patients with a SAH:1)firstly,whether neutrophils count was independent prognostic factors of death and infection in patients with a SAH;2)Then compare the predictive ability of several inflammatory markers and screen the best markers.The second chapter was aimed to evaluate the predictive significance of the Model for End-Stage Liver Disease(MELD)score for death in patients with CLD complicated with a SAH:1)whether CLD was a prognostic factor for death in patients with a SAH;2)whether MELD score is an independent predictor of mortality in patients with both CLD and a SAH.The third chapter was aimed to explore whether admission hyperglycemia is related to the prognosis of patients with a SAH:1)whether hyperglycemia on admission was associated with in-hospital death;2)exploring the correlation between the severity of hyperglycemia and in-hospital death.The fourth chapter was aimed to explore the correlation between hyponatremia on admission and the prognosis of patients with a SAH:1)whether hyponatremia was associated with in-hospital death;2)whether there was a dose-response relationship between severity of hyponatremia and poor prognosis.The fifth chapter was aimed to verify 1)whether the four prognostic factors in the above four chapters can simultaneously enter the final logistic regression model and to explore whether they are independently related to the mortality;2)whether these four variables can improve the prediction ability;3)whether the use of machine learning algorithm,compared with traditional statistical methods,can improve the prediction ability.Methods:1.Association of neutrophil Counts and MortalityIn a multicenter observational study of patients with a SAH patients,the counts of neutrophil,platelet,and lymphocyte were collected on admission.Patients were stratified based on neutrophil counts with propensity score matching to minimize confounding.We calculated the adjusted odds ratios(OR)with 95%confidence intervals(CI)for the primary outcome of in-hospital mortality and hospital-acquired infections.2.Association of Chronic Liver Disease and MortalityIn this retrospective cohort study,we analyzed consecutive a SAH patients admitted to the West China Hospital between 2009 and 2019.Patients were determined to have CLD based on their clinical diagnosis(e.g.,chronic hepatitis,steatosis,cirrhosis).The primary outcome was in-hospital all-cause mortality.Multivariable logistic regression and propensity score matching were performed to obtain the adjusted odds ratios(ORs)with 95%CI.MELD was calculated at clinical presentation.3.Association of Hyperglycemia and MortalityIn a multicenter observational study of patients with a SAH patients,we defined normal glycemia,mild hyperglycemia,moderate hyperglycemia,and severe hyperglycemia as blood glucose at admission 4.00-6.09 mmol/L,6.10–7.80 mmol/L,7.81–10.00 mmol/L,and>10.00 mmol/L,respectively We performed a propensity score matching to obtain the adjusted odds ratios(OR)with 95%confidence intervals(CI).The primary outcome was in-hospital mortality.4.Association of Hyponatremia and MortalityThis observational study included all consecutive a SAH patients admitted to the West China Hospital from 2009 to 2019.Propensity score matching was performed to obtain the adjusted odds ratios(ORs)with 95%confidence intervals(CI).The primary outcome was in-hospital mortality.5.Machine Learning–Based Model for Prediction of MortalityDifferent machine learning models for the prediction of in-hospital mortality were trained on a cohort of patients with aneurysmal subarachnoid hemorrhage(split into a training cohort[70%]and internal validation cohort[30%]).A total of 41clinical features were used to inform the models.We assessed model performance according to area under the receiver operating characteristic curve.Results:1.Association of neutrophil Counts and MortalityA total of 6041 patients were included in this study.Propensity score matching analyses indicated that compared with the lower neutrophil counts,higher neutrophil counts were associated with increased risk of in-hospital mortality(OR 1.53,95%CI1.14-2.06)and hospital-acquired infections(OR 1.61,95%CI 1.38-1.79).Moreover,out of all the inflammatory factors studied,neutrophil counts demonstrated the highest correlation with in-hospital mortality and hospital-acquired infections.2.Association of Chronic Liver Disease and MortalityThis study included 6228 cases of a SAH,489(7.9%)of whom also had CLD.After matching,319 patients were included in CLD group and 1276 in non-CLD groups.CLD was associated with increased mortality in patients with a SAH compared with non-CLD(OR 2.04,95%CI 1.43-2.92).Similarly,patients with CLD had significantly higher odds of repeat hemorrhage,pneumonia,acute renal failure,and longer length of hospital stay.In a SAH patients with CLD,a high MELD score was still associated with increased odds of mortality.3.Association of Hyperglycemia and MortalityOf 6771 patients with a SAH,hyperglycemia at admission was observed in 70.9%.Propensity scores matching analyses indicated that compared with normal glycemia,the odds of in-hospital mortality were slight lower in patients with mild hyperglycemia(OR 0.89,95%CI 0.56-1.40),significantly higher in patients with moderate hyperglycemia(OR 1.90,95%CI 1.20-3.01),and in patients with severe hyperglycemia(OR 3.45,95%CI 2.15-5.53;_P趋势<0.001).,Similar trends were evident for poor functional outcome and major disability.Hyperglycemia was associated with increased risk of hospital-acquired infections(OR 1.46,95%CI 1.29-1.66)and re-bleeding(OR 1.58,95%CI 1.06-2.35).4.Association of Hyponatremia and MortalityOf 4336 patients with a SAH,825(19.0%)had hyponatremia at admission.Propensity score matching resulted in 792:1584 patients matched pairs with standardized mean differences in patients’characteristics within 10%.Hyponatremia at admission was associated with increased in-hospital mortality(OR 1.76,95%CI1.24–2.49)and longer length of hospital stay(P<0.001).Similarly,patients with hyponatremia had significantly higher odds of delayed ischemic neurological deficits(OR 1.30,95%CI 1.04-1.68).5.Machine Learning–Based Model for Prediction of MortalityA total of 6228 patients with a SAH were included in this study.The area under the curve for random forest was)和0.93(0.89-0.97),which was significantly higher than that of the logistic regression 0.85(0.81-0.89);P=0.007).Conclusion:1 Among a SAH patients,high neutrophil counts at admission were associated with increased mortality and hospital-acquired infections.The neutrophil count is a simple,useful marker with prognostic value in a SAH patients.2 Among a SAH patients,CLD was associated with increased mortality compared to non-CLD.Among a SAH patients with CLD,a higher MELD score was associated with increased odds of mortality.3 Among a SAH patients,hyperglycemia at admission was independently associated with increased in-hospital mortality and hospital-acquired infections.4 Among a SAH patients,hyponatremia at admission is associated with increased mortality.Hyponatremia also represents a significant risk factor for delayed neurological ischemic deficits.5 In the logistic regression model,the four newly found prognostic factors,hyponatremia,hyperglycemia,neutrophil counts,chronic liver disease)were associated with in-hospital mortality,and adding these can improve the prediction ability of in-hospital mortality after a SAH;Compared with traditional statistical methods,machine learning algorithm improved the prediction ability of in-hospital mortality in patients with a SAH. |
PDF Full Text Request