Thomas Potter

BACKGROUND: Acute ischemic stroke (AIS) survivors are at high risk of major adverse cardiovascular events (MACE), highlighting the need to identify modifiable secondary prevention targets. We assessed whether postacute discharge destination is associated with 1-year MACE risk among AIS survivors. METHODS: In this retrospective cohort study, adult AIS survivors were identified from the state inpatient and emergency department databases of 5 US states (2016–2019) and categorized by discharge destination: home, inpatient rehabilitation facility (IRF), or skilled nursing facility (SNF). Other destinations were excluded. The primary outcome was 1-year MACE, defined as recurrent stroke, acute myocardial infarction, systemic embolism, or vascular death. Multivariable Cox models, adjusted for demographic and clinical factors, were used to evaluate the association between discharge disposition and MACE risk. Fine-Gray models were used for nonfatal secondary outcomes, and restricted mean survival time analyses were used to estimate absolute risk. Effect modification by age (P _interaction =0.014). Restricted mean survival time analyses indicated that IRF discharge was associated with 3.47 and 3.87 additional MACE-free days versus home and SNF discharge, respectively (both P <0.001). Adjusted number-needed-to-be-exposed were 115 (95% CI, 80–207) for IRF versus home and 87 (95% CI, 64–136) for IRF versus SNF. CONCLUSIONS: Postacute IRF discharge is associated with a lower 1-year MACE risk after AIS, highlighting the postacute care setting as a potentially modifiable system-level target for secondary prevention of long-term vascular outcomes.

BACKGROUND: Studies evaluating health system factors associated with major adverse cardiovascular events (MACE) among intracerebral hemorrhage (ICH) survivors are lacking. We evaluate differences in MACE incidence across postacute ICH care settings—inpatient rehabilitation facilities (IRF), home, or skilled nursing facilities (SNF). METHODS: Using data from Florida, New York, Maryland, Washington, and Georgia, we identified adult ICH survivors discharged to home, IRF, or SNF (April 2016–December 2018). Multivariable logistic models, adjusted for sociodemographic factors, treatment intensity, comorbidities, and frailty, estimated adjusted odds ratios (aORs) and 95% CI for the association between discharge disposition (IRF versus home; IRF versus SNF) and MACE (a composite of acute stroke, acute myocardial infarction, systemic embolism, and vascular death), recurrent ICH, acute ischemic stroke, acute myocardial infarction, vascular death, and all-cause mortality within 1 year. Cardiovascular outcomes were ascertained using International Classification of Diseases, Tenth Revision codes. We assessed interaction between age and discharge disposition, performing stratified analyses for patients P =0.047). In stratified analysis, IRF discharge (versus home) was only significantly associated with MACE in patients aged <65 years (aOR, 0.70 [CI, 0.54–0.92]), not in those aged ≥65 years (aOR, 0.94 [CI, 0.77–1.15]). Patients discharged to IRF had significantly lower odds of recurrent ICH (versus SNF: aOR, 0.60 [CI, 0.45–0.80]), vascular death (versus SNF: aOR, 0.70 [CI, 0.49–0.99]), and all-cause mortality (versus SNF: aOR, 0.63 [CI, 0.50–0.79]). CONCLUSIONS: IRF care (versus SNF and home) was associated with lower odds of MACE. Further research is needed to determine specific components of IRF care contributing to better outcomes.

Background: Although stroke is well recognized as a critical disease, treatment options are often limited. Inpatient stroke encounters carry critical information regarding the mechanisms of stroke and patient outcomes; however, these data are typically formatted to support administrative functions instead of research. To support improvements in the care of patients with stroke, a substantive research data platform is needed. Objective: To advance a stroke-oriented learning health care system, we sought to establish a comprehensive research repository of stroke data using the Houston Methodist electronic health record (EHR) system. Methods: Dedicated processes were developed to import EHR data of patients with primary acute ischemic stroke, intracerebral hemorrhage (ICH), transient ischemic attack, and subarachnoid hemorrhage under a review board–approved protocol. Relevant patients were identified from discharge diagnosis codes and assigned registry patient identification numbers. For identified patients, extract, transform, and load processes imported EHR data of primary cerebrovascular disease admissions and available data from any previous or subsequent admissions. Data were loaded into patient-focused SQL objects to enable cross-sectional and longitudinal analyses. Primary data domains (admission details, comorbidities, laboratory data, medications, imaging data, and discharge characteristics) were loaded into separate relational tables unified by patient and encounter identification numbers. Computed tomography, magnetic resonance, and angiography images were retrieved. Imaging data from patients with ICH were assessed for hemorrhage characteristics and cerebral small vessel disease markers. Patient information needed to interface with other local and national databases was retained. Prospective patient outreach was established, with patients contacted via telephone to assess functional outcomes 30, 90, 180, and 365 days after discharge. Dashboards were constructed to provide investigators with data summaries to support access. Results: The Registry of Neurological Endpoint Assessments among Patients with Ischemic and Hemorrhagic Stroke (REINAH) database was constructed as a series of relational category-specific SQL objects. Encounter summaries and dashboards were constructed to draw from these objects, providing visual data summaries for investigators seeking to build studies based on REINAH data. As of June 2022, the database contains 18,061 total patients, including 1809 (10.02%) with ICH, 13,444 (74.43%) with acute ischemic stroke, 1221 (6.76%) with subarachnoid hemorrhage, and 3165 (17.52%) with transient ischemic attack. Depending on the cohort, imaging data from computed tomography are available for 85.83% (1048/1221) to 98.4% (1780/1809) of patients, with magnetic resonance imaging available for 27.85% (340/1221) to 85.54% (11,500/13,444) of patients. Outcome assessment has successfully contacted 56.1% (240/428) of patients after ICH, with 71.3% (171/240) of responders providing consent for assessment. Responders reported a median modified Rankin Scale score of 3 at 90 days after discharge. Conclusions: A highly curated and clinically focused research platform for stroke data will establish a foundation for future research that may fundamentally improve poststroke patient care and outcomes.

Background: Nationally representative data demonstrating the impact of the COVID-19 pandemic on hemorrhagic stroke outcomes are lacking. Methods: In this pooled cross-sectional analysis, we used the National Inpatient Sample (2016-2020) to identify adults (>=18 years) with primary intracerebral hemorrhage (ICH) or subarachnoid hemorrhage (SAH). We fit segmented logistic regression models to evaluate the differences in the rates of in-hospital outcomes (in-hospital mortality, home discharge, and receiving neurosurgical procedures) between the pre-pandemic (January 2016-February 2020) and pandemic periods (March 2020-December 2020). We used multivariable logistic regression models to evaluate the differences in mortality between patients admitted from April to December 2020, with and without COVID-19, and those admitted during a similar period in 2019. Stratified analyses were conducted among patients residing in low and high-income zip codes and among patients with extreme loss of function (E-LoF) and those with minor to major loss of function (MM-LoF). Results: Overall, 309,965 ICH patients (mean age [SD]: 68[14.8], 47% female, 56% low-income) and 112,210 SAH patients (mean age [SD]: 60.2[15.4], 62% female, 55% low-income) were analyzed. Pre-pandemic, ICH mortality was decreasing by ≈ 1 % per month (adjusted odds ratio, 95% confidence interval: 0.99, 0.99-1.00). However, during the pandemic, the overall ICH mortality rate increased by ≈ 2% per month (1.02, 1.00-1.02) and ≈ 4% per month among low-income patients (1.04, 1.01-1.07). However, there was no change in trend among high-income ICH patients during the pandemic (1.00, 0.97-1.03). Patients with comorbid COVID-19 in 2020 had significantly higher odds of mortality compared to the 2019 comparison cohort, overall (ICH: 1.83, 1.33-2.51; SAH: 2.76, 1.68-4.54), and among patients with MM-LoF (ICH: 2.15, 1.12-4.16; SAH: 5.77, 1.57-21.17). However, patients with E-LoF and comorbid COVID-19 had similar mortality rates with the 2019 cohort. Conclusion: Sustained efforts are needed to address socioeconomic disparities in healthcare access, quality, and outcomes during public health emergencies.

Background: Neuroimaging is the gold-standard diagnostic modality for all patients suspected of stroke. However, the unstructured nature of imaging reports remains a major challenge to extracting useful information from electronic health records systems. Despite the increasing adoption of natural language processing (NLP) for radiology reports, information extraction for many stroke imaging features has not been systematically evaluated. Objective: In this study, we propose an NLP pipeline, which adopts the state-of-the-art ClinicalBERT model with domain-specific pretraining and task-oriented fine-tuning to extract 13 stroke features from head computed tomography imaging notes. Methods: We used the model to generate structured data sets with information on the presence or absence of common stroke features for 24,924 patients with strokes. We compared the survival characteristics of patients with and without features of severe stroke (eg, midline shift, perihematomal edema, or mass effect) using the Kaplan-Meier curve and log-rank tests. Results: Pretrained on 82,073 head computed tomography notes with 13.7 million words and fine-tuned on 200 annotated notes, our HeadCT_BERT model achieved an average area under receiver operating characteristic curve of 0.9831, F1-score of 0.8683, and accuracy of 97%. Among patients with acute ischemic stroke, admissions with any severe stroke feature in initial imaging notes were associated with a lower probability of survival (P<.001). Conclusions: Our proposed NLP pipeline achieved high performance and has the potential to improve medical research and patient safety.

Background: We evaluate nationwide trends and urban–rural disparities in case fatality (in‐hospital mortality) and discharge dispositions among patients with primary intracerebral hemorrhage (ICH). Methods and Results: In this repeated cross‐sectional study, we identified adult patients (≥18 years of age) with primary ICH from the National Inpatient Sample (2004–2018). Using a series of survey design Poisson regression models, with hospital location–time interaction, we report the adjusted risk ratio (aRR), 95% CI, and average marginal effect (AME) for factors associated with ICH case fatality and discharge dispositions. We performed a stratified analysis of each model among patients with extreme loss of function and minor to major loss of function. We identified 908 557 primary ICH hospitalizations (overall mean age [SD], 69.0 [15.0] years; 445 301 [49.0%] women; 49 884 [5.5%] rural ICH hospitalizations). The crude ICH case fatality rate was 25.3% (urban hospitals: 24.9%, rural hospitals:32.5%). Urban (versus rural) hospital patients had a lower likelihood of ICH case fatality (aRR, 0.86 [95% CI, 0.83–0.89]). ICH case fatality is declining over time; however, it is declining faster in urban hospitals (AME, −0.049 [95% CI, −0.051 to −0.047]) compared with rural hospitals (AME, −0.034 [95% CI, −0.040 to −0.027]). Conversely, home discharge is increasing significantly among urban hospitals (AME, 0.011 [95% CI, 0.008–0.014]) but not significantly changing in rural hospitals (AME, −0.001 [95% CI, −0.010 to 0.007]). Among patients with extreme loss of function, hospital location was not significantly associated with ICH case fatality or home discharge. Conclusions: Improving access to neurocritical care resources, particularly in resource‐limited communities, may reduce the ICH outcomes disparity gap.

BACKGROUND Neuroimaging is the gold-standard diagnostic modality for all patients suspected of stroke. However, the unstructured nature of imaging reports remains a major challenge to extracting useful information from electronic health records systems. Despite the increasing adoption of natural language processing (NLP) for radiology reports, information extraction for many stroke imaging features has not been systematically evaluated. OBJECTIVE In this study, we propose an NLP pipeline, which adopts the state-of-the-art ClinicalBERT model with domain-specific pretraining and task-oriented fine-tuning to extract 13 stroke features from head computed tomography imaging notes. METHODS We used the model to generate structured data sets with information on the presence or absence of common stroke features for 24,924 patients with strokes. We compared the survival characteristics of patients with and without features of severe stroke (eg, midline shift, perihematomal edema, or mass effect) using the Kaplan-Meier curve and log-rank tests. RESULTS Pretrained on 82,073 head computed tomography notes with 13.7 million words and fine-tuned on 200 annotated notes, our HeadCT_BERT model achieved an average area under receiver operating characteristic curve of 0.9831, F1-score of 0.8683, and accuracy of 97%. Among patients with acute ischemic stroke, admissions with any severe stroke feature in initial imaging notes were associated with a lower probability of survival (P<.001). CONCLUSIONS Our proposed NLP pipeline achieved high performance and has the potential to improve medical research and patient safety.

BACKGROUND Although stroke is well recognized as a critical disease, treatment options are often limited. Inpatient stroke encounters carry critical information regarding the mechanisms of stroke and patient outcomes; however, these data are typically formatted to support administrative functions instead of research. To support improvements in the care of patients with stroke, a substantive research data platform is needed. OBJECTIVE To advance a stroke-oriented learning health care system, we sought to establish a comprehensive research repository of stroke data using the Houston Methodist electronic health record (EHR) system. METHODS Dedicated processes were developed to import EHR data of patients with primary acute ischemic stroke, intracerebral hemorrhage (ICH), transient ischemic attack, and subarachnoid hemorrhage under a review board–approved protocol. Relevant patients were identified from discharge diagnosis codes and assigned registry patient identification numbers. For identified patients, extract, transform, and load processes imported EHR data of primary cerebrovascular disease admissions and available data from any previous or subsequent admissions. Data were loaded into patient-focused SQL objects to enable cross-sectional and longitudinal analyses. Primary data domains (admission details, comorbidities, laboratory data, medications, imaging data, and discharge characteristics) were loaded into separate relational tables unified by patient and encounter identification numbers. Computed tomography, magnetic resonance, and angiography images were retrieved. Imaging data from patients with ICH were assessed for hemorrhage characteristics and cerebral small vessel disease markers. Patient information needed to interface with other local and national databases was retained. Prospective patient outreach was established, with patients contacted via telephone to assess functional outcomes 30, 90, 180, and 365 days after discharge. Dashboards were constructed to provide investigators with data summaries to support access. RESULTS The Registry of Neurological Endpoint Assessments among Patients with Ischemic and Hemorrhagic Stroke (REINAH) database was constructed as a series of relational category-specific SQL objects. Encounter summaries and dashboards were constructed to draw from these objects, providing visual data summaries for investigators seeking to build studies based on REINAH data. As of June 2022, the database contains 18,061 total patients, including 1809 (10.02%) with ICH, 13,444 (74.43%) with acute ischemic stroke, 1221 (6.76%) with subarachnoid hemorrhage, and 3165 (17.52%) with transient ischemic attack. Depending on the cohort, imaging data from computed tomography are available for 85.83% (1048/1221) to 98.4% (1780/1809) of patients, with magnetic resonance imaging available for 27.85% (340/1221) to 85.54% (11,500/13,444) of patients. Outcome assessment has successfully contacted 56.1% (240/428) of patients after ICH, with 71.3% (171/240) of responders providing consent for assessment. Responders reported a median modified Rankin Scale score of 3 at 90 days after discharge. CONCLUSIONS A highly curated and clinically focused research platform for stroke data will establish a foundation for future research that may fundamentally improve poststroke patient care and outcomes.

Background: We report contemporary trends in nationwide incidence of intracerebral hemorrhage (ICH) across demographic and regional strata over a 15-year period. Methods: Utilizing the Nationwide Inpatient Sample (2004–2018) and US Census Bureau data, we calculated ICH incidence rates for age, race/ethnicity, sex, and hospital region sub-cohorts across 5 consecutive 3-year periods (2004–2006 to 2016–2018). We fit Poisson and log binomial regression models to evaluate demographic and regional differences in ICH incidence and trends in prevalence of hypertension and past/current anticoagulant use among hospitalized ICH patients. Results: Overall, the annual incidence rate (95% CI) of ICH per 100 000 was 23.15 (23.10–23.20). The 3-year incidence of ICH (per 100 000) increased from 62.79 in 2004 to 2006 to 78.86 in 2016 to 2018 (adjusted incidence rate ratio, CI: 1.11 [1.02–1.20]), coinciding with increased 3-year prevalence of hypertension and anticoagulant use among hospitalized ICH patients (adjusted risk ratio, CI: hypertension—1.16 [1.15–1.17]; anticoagulant use—2.30 [2.14–2.47]). We found a significant age-time interaction, whereby ICH incidence increased significantly faster among those aged 18 to 44 years (adjusted incidence rate ratio, CI: 1.10 [1.05–1.14]) and 45 to 64 years (adjusted incidence rate ratio, CI: 1.08 [1.03–1.13]), relative to those aged ≥75 years. Conclusions: Rising ICH incidence among young and middle-aged Americans warrants ICH prevention strategies targeting these economically productive age groups.

Recent technological advances have enabled researchers to collect large amounts of electroencephalography (EEG) signals in labeled and unlabeled datasets. It is expensive and time consuming to collect labeled EEG data for use in brain-computer interface (BCI) systems, however. In this paper, a novel active learning method is proposed to minimize the amount of labeled, subject-specific EEG data required for effective classifier training, by combining measures of uncertainty and representativeness within an extreme learning machine (ELM). Following this approach, an ELM classifier was first used to select a relatively large batch of unlabeled examples, whose uncertainty was measured through the best-versus-second-best (BvSB) strategy. The diversity of each sample was then measured between the limited labeled training data and previously selected unlabeled samples, and similarity is measured among the previously selected samples. Finally, a tradeoff parameter is introduced to control the balance between informative and representative samples, and these samples are then used to construct a powerful ELM classifier. Extensive experiments were conducted using benchmark and multiclass motor imagery EEG datasets to evaluate the efficacy of the proposed method. Experimental results show that the performance of the new algorithm exceeds or matches those of several state-of-the-art active learning algorithms. It is thereby shown that the proposed method improves classifier performance and reduces the need for training samples in BCI applications.