A new nomogram for individualized prediction of the probability of hemorrhagic transformation after intravenous thrombolysis for ischemic stroke patients
BMC neurology. 2020;20(1):426
BACKGROUND A reliable scoring tool to detect the risk of intracerebral hemorrhage (ICH) after intravenous thrombolysis for ischemic stroke is warranted. The present study was designed to develop and validate a new nomogram for individualized prediction of the probability of hemorrhagic transformation (HT) in patients treated with intravenous (IV) recombinant tissue plasminogen activator (rt-PA). METHODS We enrolled patients who suffered from acute ischemic stroke (AIS) with IV rt-PA treatment in our emergency green channel between August 2016 and July 2018. The main outcome was defined as any type of intracerebral hemorrhage according to the European Cooperative Acute Stroke Study II (ECASS II). All patients were randomly divided into two cohorts: the primary cohort and the validation cohort. On the basis of multivariate logistic model, the predictive nomogram was generated. The performance of the nomogram was evaluated by Harrell's concordance index (C-index) and calibration plot. RESULTS A total of 194 patients with complete data were enrolled, of whom 131 comprised the primary cohort and 63 comprised the validation cohort, with HT rate 12.2, 9.5% respectively. The score of chronic disease scale (CDS), the global burden of cerebral small vascular disease (CSVD), National Institutes of Health Stroke Scale (NIHSS) score ≥ 13, and onset-to-treatment time (OTT) ≥ 180 were detected important determinants of ICH and included to construct the nomogram. The nomogram derived from the primary cohort for HT had C- Statistics of 0.9562 and the calibration plot revealed generally fit in predicting the risk of HT. Furthermore, we made a comparison between our new nomogram and several other risk-assessed scales for HT with receiver operating characteristic (ROC) curve analysis, and the results showed the nomogram model gave an area under curve of 0.9562 (95%CI, 0.9221-0.9904, P < 0.01) greater than HAT (Hemorrhage After Thrombolysis), SEDAN (blood Sugar, Early infarct and hyper Dense cerebral artery sign on non-contrast computed tomography, Age, and NIHSS) and SPAN-100 (Stroke Prognostication using Age and NIHSS) scores. CONCLUSIONS This proposed nomogram based on the score of CDS, the global burden of CSVD, NIHSS score ≥ 13, and OTT ≥ 180 gives rise to a more accurate and more comprehensive prediction for HT in patients with ischemic stroke receiving IV rt-PA treatment.
Machine Learning to Predict Outcomes in Patients with Acute Gastrointestinal Bleeding: A Systematic Review
Digestive diseases and sciences. 2019
Risk stratification of patients with gastrointestinal bleeding (GIB) is recommended, but current risk assessment tools have variable performance. Machine learning (ML) has promise to improve risk assessment. We performed a systematic review to evaluate studies utilizing ML techniques for GIB. Bibliographic databases and conference abstracts were searched for studies with a population of overt GIB that used an ML algorithm with outcomes of mortality, rebleeding, hemostatic intervention, and/or hospital stay. Two independent reviewers screened titles and abstracts, reviewed full-text studies, and extracted data from included studies. Risk of bias was assessed with an adapted Quality in Prognosis Studies tool. Area under receiver operating characteristic curves (AUCs) were the primary assessment of performance with AUC ≥ 0.80 predefined as an acceptable threshold of good performance. Fourteen studies with 30 assessments of ML models met inclusion criteria. No study had low risk of bias. Median AUC reported in validation datasets for predefined outcomes of mortality, intervention, or rebleeding was 0.84 (range 0.40-0.98). AUCs were higher with artificial neural networks (median 0.93, range 0.78-0.98) than other ML models (0.81, range 0.40-0.92). ML performed better than clinical risk scores (Glasgow-Blatchford, Rockall, Child-Pugh, MELD) for mortality in upper GIB. Limitations include heterogeneity of ML models, inconsistent comparisons of ML models with clinical risk scores, and high risk of bias. ML generally provided good-excellent prognostic performance in patients with GIB, and artificial neural networks tended to outperform other ML models. ML was better than clinical risk scores for mortality in upper GIB.