StrokeClassifier: ischemic stroke etiology classification by ensemble consensus modeling using electronic health records.

Determining acute ischemic stroke (AIS) etiology is fundamental to secondary stroke prevention efforts but can be diagnostically challenging. We trained and validated an automated classification tool, StrokeClassifier, using electronic health record (EHR) text from 2039 non-cryptogenic AIS patients at 2 academic hospitals to predict the 4-level outcome of stroke etiology adjudicated by agreement of at least 2 board-certified vascular neurologists' review of the EHR. StrokeClassifier is an ensemble consensus meta-model of 9 machine learning classifiers applied to features extracted from discharge summary texts by natural language processing. StrokeClassifier was externally validated in 406 discharge summaries from the MIMIC-III dataset reviewed by a vascular neurologist to ascertain stroke etiology. Compared with vascular neurologists' diagnoses, StrokeClassifier achieved the mean cross-validated accuracy of 0.74 and weighted F1 of 0.74 for multi-class classification. In MIMIC-III, its accuracy and weighted F1 were 0.70 and 0.71, respectively. In binary classification, the two metrics ranged from 0.77 to 0.96. The top 5 features contributing to stroke etiology prediction were atrial fibrillation, age, middle cerebral artery occlusion, internal carotid artery occlusion, and frontal stroke location. We designed a certainty heuristic to grade the confidence of StrokeClassifier's diagnosis as non-cryptogenic by the degree of consensus among the 9 classifiers and applied it to 788 cryptogenic patients, reducing cryptogenic diagnoses from 25.2% to 7.2%. StrokeClassifier is a validated artificial intelligence tool that rivals the performance of vascular neurologists in classifying ischemic stroke etiology. With further training, StrokeClassifier may have downstream applications including its use as a clinical decision support system.

StrokeClassifier: Ischemic Stroke Etiology Classification by Ensemble Consensus Modeling Using Electronic Health Records.

Determining the etiology of an acute ischemic stroke (AIS) is fundamental to secondary stroke prevention efforts but can be diagnostically challenging. We trained and validated an automated classification machine intelligence tool, StrokeClassifier, using electronic health record (EHR) text data from 2,039 non-cryptogenic AIS patients at 2 academic hospitals to predict the 4-level outcome of stroke etiology determined by agreement of at least 2 board-certified vascular neurologists' review of the stroke hospitalization EHR. StrokeClassifier is an ensemble consensus meta-model of 9 machine learning classifiers applied to features extracted from discharge summary texts by natural language processing. StrokeClassifier was externally validated in 406 discharge summaries from the MIMIC-III dataset reviewed by a vascular neurologist to ascertain stroke etiology. Compared with stroke etiologies adjudicated by vascular neurologists, StrokeClassifier achieved the mean cross-validated accuracy of 0.74 (±0.01) and weighted F1 of 0.74 (±0.01). In the MIMIC-III cohort, the accuracy and weighted F1 of StrokeClassifier were 0.70 and 0.71, respectively. SHapley Additive exPlanation analysis elucidated that the top 5 features contributing to stroke etiology prediction were atrial fibrillation, age, middle cerebral artery occlusion, internal carotid artery occlusion, and frontal stroke location. We then designed a certainty heuristic to deem a StrokeClassifier diagnosis as confidently non-cryptogenic by the degree of consensus among the 9 classifiers, and applied it to 788 cryptogenic patients. This reduced the percentage of the cryptogenic strokes from 25.2% to 7.2% of all ischemic strokes. StrokeClassifier is a validated artificial intelligence tool that rivals the performance of vascular neurologists in classifying ischemic stroke etiology for individual patients. With further training, StrokeClassifier may have downstream applications including its use as a clinical decision support system.

Stroke metrics during the first year of the COVID-19 pandemic, a tale of two comprehensive stroke centers.

Although a decrease in stroke admissions during the SARS-CoV-2 pandemic has been observed, detailed analyses of the evolution of stroke metrics during the pandemic are lacking. We analyzed changes in stroke presentation, in-hospital systems-of-care, and treatment time metrics at two representative Comprehensive Stroke Centers (CSCs) during the first year of Coronavirus disease 2019 pandemic. From January 2018 to May 2021, data from stroke presentations to two CSCs were obtained. The study duration was split into: period 0 (prepandemic), period 1 (Wave 1), period 2 (Lull), and period 3 (Wave 2). Acute stroke therapies rates and workflow times were compared among pandemic and prepandemic periods. Analyses were adjusted for age, sex, comorbidities, and pre-morbid care needs. There was a significant decrease in monthly hospital presentations of stroke during Wave 1. Both centers reported declines in reperfusion therapies during Wave 1, slowly catching up but never to pre pandemic numbers, and dropping again in Wave 2. Both CSCs experienced in-hospital workflow delays during Waves 1 and 2, and even during the Lull period. Our results highlight the need for proactive strategies to reduce barriers to workflow and hospital avoidance for stroke patients during crisis periods.

Utilization of Telestroke Prior to and Following the COVID-19 Pandemic.

For over two decades, telestroke has been utilized as a means for improving acute access to a stroke specialist when this expertise is otherwise unavailable. During this time, telestroke use has increased and improvements in care metrics have been widely reported. Several telestroke model variations are utilized; each has different workflow implications. A successful telestroke system should include adequate protocols and training, equipment, documentation system, and tracking of quality metrics. Upfront costs of needed technology and devices, credentialing hurdles, and limited reimbursement are all reported barriers to the utilization of telestroke. Emphasis on safety measures during the COVID-19 pandemic resulted in the dramatic upscaling of telehealth utilization, although overall stroke volumes declined in many areas in the early phases of the pandemic. Going forward, continued reduction in cost of required devices and broadband connections, increased use of automated and advanced analytical software, and a universal licensing and credentialing system are needed to continue the expansion of telestroke use.

Cohort design and natural language processing to reduce bias in electronic health records research.

Electronic health record (EHR) datasets are statistically powerful but are subject to ascertainment bias and missingness. Using the Mass General Brigham multi-institutional EHR, we approximated a community-based cohort by sampling patients receiving longitudinal primary care between 2001-2018 (Community Care Cohort Project [C3PO], n = 520,868). We utilized natural language processing (NLP) to recover vital signs from unstructured notes. We assessed the validity of C3PO by deploying established risk models for myocardial infarction/stroke and atrial fibrillation. We then compared C3PO to Convenience Samples including all individuals from the same EHR with complete data, but without a longitudinal primary care requirement. NLP reduced the missingness of vital signs by 31%. NLP-recovered vital signs were highly correlated with values derived from structured fields (Pearson r range 0.95-0.99). Atrial fibrillation and myocardial infarction/stroke incidence were lower and risk models were better calibrated in C3PO as opposed to the Convenience Samples (calibration error range for myocardial infarction/stroke: 0.012-0.030 in C3PO vs. 0.028-0.046 in Convenience Samples; calibration error for atrial fibrillation 0.028 in C3PO vs. 0.036 in Convenience Samples). Sampling patients receiving regular primary care and using NLP to recover missing data may reduce bias and maximize generalizability of EHR research.

Sex-specific differences in presentations and determinants of outcomes after endovascular thrombectomy for large vessel occlusion stroke.

INTRODUCTION: Sex-specific differences in ischemic stroke outcomes are prevalent. We sought to investigate sex differences in the determinants of reperfusion and functional outcomes after endovascular thrombectomy (EVT) for emergent large vessel occlusion ischemic stroke (ELVO). METHODS: Patients presenting to a single referral center with an anterior circulation ELVO that underwent EVT from 2011 to 2019 were included in this retrospective analysis. Sex differences in history, presentation, adequate reperfusion (TICI 2b-3), and 90-day good outcome [delta modified Rankin Scale (mRS) ≤ 2 from pre-stroke] were examined. Multivariable logistic regression analyses were performed to assess sex-specific associations with outcomes. RESULTS: Three hundred and eighty-one consecutive ELVO patients were identified. Women (N = 193) were older (75 vs 64 years, p < 0.0001), had more pre-stroke disability (17% vs 9%, p = 0.032), more atrial fibrillation (41% vs 30%, p = 0.033), but less carotid atherosclerosis (8% vs 16%, p = 0.027). Rates of TICI 2b-3 and good outcome were similar between sexes. Carotid atherosclerosis (OR 0.315, 95% CI 0.130, 0.762) and dissection (OR 0.124, 95% CI 0.027, 0.569) independently decreased the odds of TICI 2b-3 among men but not women. Older age, more severe stroke, and not achieving TICI 2b-3 independently decreased the odds of good outcome among both sexes, while prior stroke (OR 0.258, 95% CI 0.083, 0.797) and hemorrhagic transformation (OR 0.111, 0.021, 0.592) were determinants exclusive to men. CONCLUSION: In a real-world analysis of ELVO stroke patients treated with EVT, we found that despite advanced age and more pre-stroke disability, women have comparable reperfusion rates and functional outcomes compared to men. Sex-specific determinants of reperfusion and functional outcome were identified that require further study.

Re-CHARGE-AF: Recalibration of the CHARGE-AF Model for Atrial Fibrillation Risk Prediction in Patients With Acute Stroke.

Background Performance of existing atrial fibrillation (AF) risk prediction models in poststroke populations is unclear. We evaluated predictive utility of an AF risk model in patients with acute stroke and assessed performance of a fully refitted model. Methods and Results Within an academic hospital, we included patients aged 46 to 94 years discharged for acute ischemic stroke between 2003 and 2018. We estimated 5-year predicted probabilities of AF using the Cohorts for Heart and Aging Research in Genomic Epidemiology for Atrial Fibrillation (CHARGE-AF) model, by recalibrating CHARGE-AF to the baseline risk of the sample, and by fully refitting a Cox proportional hazards model to the stroke sample (Re-CHARGE-AF) model. We compared discrimination and calibration between models and used 200 bootstrap samples for optimism-adjusted measures. Among 551 patients with acute stroke, there were 70 incident AF events over 5 years (cumulative incidence, 15.2%; 95% CI, 10.6%-19.5%). Median predicted 5-year risk from CHARGE-AF was 4.8% (quartile 1-quartile 3, 2.0-12.6) and from Re-CHARGE-AF was 16.1% (quartile 1-quartile 3, 8.0-26.2). For CHARGE-AF, discrimination was moderate (C statistic, 0.64; 95% CI, 0.57-0.70) and calibration was poor, underestimating AF risk (Greenwood-Nam D'Agostino chi-square, P<0.001). Calibration with recalibrated baseline risk was also poor (Greenwood-Nam D'Agostino chi-square, P<0.001). Re-CHARGE-AF improved discrimination (P=0.001) compared with CHARGE-AF (C statistic, 0.74 [95% CI, 0.68-0.79]; optimism-adjusted, 0.70 [95% CI, 0.65-0.75]) and was well calibrated (Greenwood-Nam D'Agostino chi-square, P=0.97). Conclusions Covariates from an established AF risk model enable accurate estimation of AF risk in a poststroke population after recalibration. A fully refitted model was required to account for varying baseline AF hazard and strength of associations between covariates and incident AF.

Teleneurology-Enabled Determination of Death by Neurologic Criteria After Cardiac Arrest or Severe Neurologic Injury.

OBJECTIVE: To determine whether providing teleneurology (TN) consultations aiding in determination of death by neurologic criteria (DNC) to a bedside intensivist is feasible and whether timely access and expert input increase the quality of the DNC examination and identification of potential organ donors, we reviewed retrospective data related to outcomes of such consultations. METHODS: Between November 2017 and March 2019, TN consults were requested for sequential comatose patients in the intensive care unit (ICU). We recorded patients' demographic information, causes leading to coma or suspected DNC, and the results of TN consultations. We obtained data on the number of referrals to the organ bank and number of organ donors. RESULTS: Ninety-nine consults were performed with a median time from request to start of the consult of 20.2 minutes (interquartile range 5.4-65.3 minutes). Eighty consults were requested for determination of prognosis, whereas 19 consults were requested for supervision of the DNC examination. In 1 of 80 (1.2%) prognostication consults, the patient was determined by the neurologist to require assessment of DNC and was found to meet DNC criteria; determination of DNC occurred in 11 of the 19 (57.9%) consultations for a supervised DNC examination. In a comparison of the pre-TN (94 months) and post-TN (17 months) periods, there was 2.56-fold increase in the proportion of patients meeting DNC criteria who were medically suitable for donation (pre-TN 8.9% vs post-TN 21.1%, p = 0.02) and a 2.12-fold increase in the proportion of donors (pre-TN 6.14% vs post-TN 13.1%, p = 0.14). CONCLUSIONS: It is feasible to perform TN consultations for patients with severe neurologic damage and to allow expert supervision for DNC examination. Having a teleneurologist as part of the ICU assessment team helped differentiate severe neurologic deficits from DNC and was associated with increase in organ donation.

Acute ischemic stroke: improving access to intravenous tissue plasminogen activator.

INTRODUCTION: Since approval by the United States Food and Drug Administration in 1996, alteplase utilization rates for acute ischemic stroke have increased. Despite its efficacy for improving stroke outcomes, however, the majority of ischemic stroke patients still do not receive alteplase. To address this issue, different methods for improving access to alteplase have been tested with varying degrees of success. AREAS COVERED: This article gives an overview of the recent approaches pursued to improve access to alteplase for acute ischemic stroke patients. Utilization of stroke systems of care, quality metrics, and quality-improvement initiatives to improve alteplase treatment rates are discussed. The implementation of Telestroke networks to improve access and timely evaluation by a stroke specialist are also reviewed. Lastly, this review discusses the use of neuroimaging techniques to identify alteplase candidates in stroke of unknown symptom onset or beyond the 4.5-h treatment window. EXPERT COMMENTARY: Expanding access to alteplase therapy for acute ischemic stroke is a multi-faceted approach. Specific considerations based on region, population, and health-care resources should be considered for each strategy. Neuroimaging approaches to identify alteplase-eligible patients beyond the 4.5-h treatment window are a recent development in acute stroke care that holds promise for increasing alteplase treatment rates.