Dataset Information

Novel machine learning models to predict pneumonia events in supratentorial intracerebral hemorrhage populations: An analysis of the Risa-MIS-ICH study.

ABSTRACT:

Background

Stroke-associated pneumonia (SAP) contributes to high mortality rates in spontaneous intracerebral hemorrhage (sICH) populations. Accurate prediction and early intervention of SAP are associated with prognosis. None of the previously developed predictive scoring systems are widely accepted. We aimed to derive and validate novel supervised machine learning (ML) models to predict SAP events in supratentorial sICH populations.

Methods

The data of eligible supratentorial sICH individuals were extracted from the Risa-MIS-ICH database and split into training, internal validation, and external validation datasets. The primary outcome was SAP during hospitalization. Univariate and multivariate analyses were used for variable filtering, and logistic regression (LR), Gaussian naïve Bayes (GNB), random forest (RF), K-nearest neighbor (KNN), support vector machine (SVM), extreme gradient boosting (XGB), and ensemble soft voting model (ESVM) were adopted for ML model derivations. The accuracy, sensitivity, specificity, and area under the curve (AUC) were adopted to evaluate the predictive value of each model with internal/cross-/external validations.

Results

A total of 468 individuals with sICH were included in this work. Six independent variables [nasogastric feeding, airway support, unconscious onset, surgery for external ventricular drainage (EVD), larger sICH volume, and intensive care unit (ICU) stay] for SAP were identified and selected for ML prediction model derivations and validations. The internal and cross-validations revealed the superior and robust performance of the GNB model with the highest AUC value (0.861, 95% CI: 0.793-0.930), while the LR model had the highest AUC value (0.867, 95% CI: 0.812-0.923) in external validation. The ESVM method combining the other six methods had moderate but robust abilities in both cross-validation and external validation and achieved an AUC of 0.843 (95% CI: 0.784-0.902) in external validation.

Conclusion

The ML models could effectively predict SAP in sICH populations, and our novel ensemble model demonstrated reliable robust performance outcomes despite the populational and algorithmic differences. This attempt indicated that ML application may benefit in the early identification of SAP.

SUBMITTER: Zheng Y

PROVIDER: S-EPMC9452786 | biostudies-literature | 2022

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Novel machine learning models to predict pneumonia events in supratentorial intracerebral hemorrhage populations: An analysis of the <i>Risa-MIS-ICH</i> study.

Zheng Yan Y Lin Yuan-Xiang YX He Qiu Q Zhuo Ling-Yun LY Huang Wei W Gao Zhu-Yu ZY Chen Ren-Long RL Zhao Ming-Pei MP Xie Ze-Feng ZF Ma Ke K Fang Wen-Hua WH Wang Deng-Liang DL Chen Jian-Cai JC Kang De-Zhi DZ Lin Fu-Xin FX

Frontiers in neurology 20220825

<h4>Background</h4>Stroke-associated pneumonia (SAP) contributes to high mortality rates in spontaneous intracerebral hemorrhage (sICH) populations. Accurate prediction and early intervention of SAP are associated with prognosis. None of the previously developed predictive scoring systems are widely accepted. We aimed to derive and validate novel supervised machine learning (ML) models to predict SAP events in supratentorial sICH populations.<h4>Methods</h4>The data of eligible supratentorial sI ...[more]

PMID: 36090880

Similar Datasets

Project description:Background and Purpose: Perihematomal edema (PHE) is associated with poor functional outcomes after intracerebral hemorrhage (ICH). Early identification of risk factors associated with PHE growth may allow for targeted therapeutic interventions. Methods: We used data contained in the risk stratification and minimally invasive surgery in acute intracerebral hemorrhage (Risa-MIS-ICH) patients: a prospective multicenter cohort study. Patients' clinical, laboratory, and radiological data within 24 h of admission were obtained from their medical records. The absolute increase in PHE volume from baseline to day 3 was defined as iPHE volume. Poor outcome was defined as modified Rankin Scale (mRS) of 4 to 6 at 90 days. Binary logistic regression was used to assess the relationship between iPHE volume and poor outcome. The receiver operating characteristic curve was used to find the best cutoff. Linear regression was used to identify variables associated with iPHE volume (ClinicalTrials.gov Identifier: NCT03862729). Results: One hundred ninety-seven patients were included in this study. iPHE volume was significantly associated with poor outcome [P = 0.003, odds ratio (OR) 1.049, 95% confidence interval (CI) 1.016-1.082] after adjustment for hematoma volume. The best cutoff point of iPHE volume was 7.98 mL with a specificity of 71.4% and a sensitivity of 47.5%. Diabetes mellitus (P = 0.043, β = 7.66 95% CI 0.26-15.07), black hole sign (P = 0.002, β = 18.93 95% CI 6.84-31.02), and initial ICH volume (P = 0.018, β = 0.20 95% CI 0.03-0.37) were significantly associated with iPHE volume. After adjusting for hematoma expansion, the black hole sign could still independently predict the increase of PHE (P < 0.001, β = 21.62 95% CI 10.10-33.15). Conclusions: An increase of PHE volume >7.98 mL from baseline to day 3 may lead to poor outcome. Patients with diabetes mellitus, black hole sign, and large initial hematoma volume result in more PHE growth, which should garner attention in the treatment.

Project description:ObjectiveThe effect of surgical treatment for supratentorial spontaneous intracerebral hemorrhage (ICH) and whether it is modified by key baseline characteristics and timing remains uncertain.MethodsWe performed a systematic review and meta-analysis of randomized controlled trials of surgical treatment of supratentorial spontaneous ICH aimed at clot removal. We searched MEDLINE, Embase, and Cochrane databases up to February 21, 2019. Primary outcome was good functional outcome at follow-up; secondary outcomes were death and serious adverse events. We analyzed all types of surgery combined and minimally invasive approaches separately. We pooled risk ratios with 95% confidence intervals and assessed the modifying effect of age, Glasgow Coma Scale, hematoma volume, and timing of surgery with meta-regression analysis.ResultsWe included 21 studies with 4,145 patients; 4 (19%) were of the highest quality. Risk ratio of good functional outcome after any type of surgery was 1.40 (95% confidence interval [CI] = 1.22-1.60, I2 = 46%, 20 studies), and after minimally invasive surgery it was 1.47 (95% CI = 1.26-1.72, I2 = 47%, 12 studies). For death, the risk ratio for any type of surgery was 0.77 (95% CI = 0.68-0.85, I2 = 23%, 21 studies), and for minimally invasive surgery it was 0.68 (95% CI = 0.56-0.83, I2 = 14%, 13 studies). Serious adverse events were reported infrequently. Surgery seemed more effective when performed sooner after symptom onset (p = 0.04, 12 studies). Age, Glasgow Coma Scale, and hematoma volume did not modify the effect of surgery.InterpretationSurgical treatment of supratentorial spontaneous ICH may be beneficial, in particular with minimally invasive procedures and when performed soon after symptom onset. Further well-designed randomized trials are needed to demonstrate whether (minimally invasive) surgery improves functional outcome after ICH and to determine the optimal time window of the treatment after symptom onset. ANN NEUROL 2020;88:239-250.

Project description:PurposeWe develop a new risk score to predict patients with stroke-associated pneumonia (SAP) who have an acute intracranial hemorrhage (ICH).MethodWe applied logistic regression to develop a new risk score called ICH-LR2S2. It was derived from examining a dataset of 70,540 ICH patients between 2015 and 2018 from the Chinese Stroke Center Alliance (CSCA). During the training of ICH-LR2S2, patients were randomly divided into two groups - 80% for the training set and 20% for model validation. A prospective test set was developed using 12,523 patients recruited in 2019. To further verify its effectiveness, we tested ICH-LR2S2 on an external dataset of 24,860 patients from the China National Stroke Registration Management System II (CNSR II). The performance of ICH-LR2S2 was measured by the area under the receiver operating characteristic curve (AUROC).ResultsThe incidence of SAP in the dataset was 25.52%. A 24-point ICH-LR2S2 was developed from independent predictors, including age, modified Rankin Scale, fasting blood glucose, National Institutes of Health Stroke Scale admission score, Glasgow Coma Scale score, C-reactive protein, dysphagia, Chronic Obstructive Pulmonary Disease, and current smoking. The results showed that ICH-LR2S2 achieved an AUC = 0.749 [95% CI 0.739-0.759], which outperforms the best baseline ICH-APS (AUC = 0.704) [95% CI 0.694-0.714]. Compared with the previous ICH risk scores, ICH-LR2S2 incorporates fasting blood glucose and C-reactive protein, improving its discriminative ability. Machine learning methods such as XGboost (AUC = 0.772) [95% CI 0.762-0.782] can further improve our prediction performance. It also performed well when further validated by the external independent cohort of patients (n = 24,860), ICH-LR2S2 AUC = 0.784 [95% CI 0.774-0.794].ConclusionICH-LR2S2 accurately distinguishes SAP patients based on easily available clinical features. It can help identify high-risk patients in the early stages of diseases.

Project description:BackgroundIntraventricular hemorrhage (IVH) after intracerebral hemorrhage (ICH) is a strong independent predictor of poor outcomes. Although the location and volume of ICH are associated with IVH, our knowledge concerning the mechanism of IVH after ICH is still limited. This study aimed to investigate the relationship between hematoma morphology and IVH in patients with supratentorial deep ICH.MethodsWe retrospectively analyzed adult patients (aged ≥18 years) with spontaneous supratentorial deep ICH who underwent computed tomography (CT) within 48 h after ICH symptom onset in Peking University First Hospital between January 2017 and August 2022. We collected the clinical and imaging data of the patients and assessed hematoma morphology using several quantitative radiological parameters including hematoma volume, sphericity index, A/B ratio (A: the largest area of hematoma; B: the largest diameter 90° to A on the same slice), and our newly proposed largest diameter-midline angle (LMA). Multivariable logistic regression analysis was used to analyze the relationship between these parameters and the presence of IVH on the initial CT scan.ResultsAmong 114 patients with spontaneous supratentorial deep ICH, 41 (36.0%) had IVH. In patients with IVH, the sphericity index was lower than that in individuals without IVH, while the LMA was larger. Multivariate logistic regression analysis showed that sphericity index [0.1-unit odds ratio (OR) =0.252; 95% CI: 0.089-0.709; P=0.009] and the LMA (10-unit OR =1.281; 95% CI: 1.007-1.630; P=0.04) were independently associated with the presence of IVH in patients with supratentorial deep ICH. Univariate analyses showed that hematoma volume, A/B ratio, sphericity index, and the LMA were significantly associated with poor outcomes at discharge.ConclusionsTwo quantitative parameters of hematoma morphology, sphericity index and the LMA, were significantly associated with the presence of IVH in patients with supratentorial deep ICH. Further prospective studies with larger sample sizes are needed to validate our results.

Project description:ImportanceHematoma volume is the strongest predictor of outcome in intracerebral hemorrhage (ICH). Despite known differences in the underlying biology between deep and lobar ICHs, limited data are available on location specificity of factors reported to affect hematoma volume.ObjectiveTo evaluate whether determinants of ICH volume differ by topography, we sought to estimate location-specific effects for potential predictors of this radiological outcome.DesignProspective cohort study.SettingAcademic medical center.ParticipantsA total of 744 supratentorial primary ICH patients (388 deep and 356 lobar) aged older than 18 years admitted between January 1, 2000, and December 31, 2010.Main outcomes and measuresIntracerebral hemorrhage volume measured from the computed tomography scan obtained on presentation to the emergency department. Linear regression analysis, stratified by ICH location, was implemented to identify determinants of log-transformed ICH volume.ResultsMedian ICH volume was larger in lobar hemorrhages (39 mL; interquartile range, 16-75 mL) than in deep hemorrhages (13 mL; interquartile range, 5-40 mL; P < .001). In multivariable linear regression, independent predictors of deep ICH volume were intensity of anticoagulation (β = 0.32; standard error [SE] = 0.08; P < .001; test for trend across 4 categories of the international normalized ratio), history of coronary artery disease (β = 0.33; SE = 0.17; P = .05), male sex (β = 0.28; SE = 0.14; P = .05), and age (β = -0.02; SE = 0.01; P = .001). Independent predictors of lobar ICH volume were intensity of anticoagulation (β = 0.14; SE = 0.06; P = .02) and antiplatelet treatment (β = 0.27; SE = 0.13; P = .03).Conclusions and relevancePredictors of hematoma volume only partially overlap between deep and lobar ICHs. These findings suggest that the mechanisms that determine the extent of bleeding differ for deep and lobar ICHs. Further studies are needed to characterize the specific biological pathways that underlie the observed associations.