A predictive analytics model for differentiating between transient ischemic attacks (TIA) and its mimics

Machine Learning-based Cerebral Venous Thrombosis Diagnosis with Clinical Data

OBJECTIVES

Cerebral Venous Thrombosis (CVT) poses diagnostic challenges due to the variability in disease course and symptoms. The prognosis of CVT relies on early diagnosis. Our study focuses on developing a machine learning-based screening algorithm using clinical data from a large neurology referral center in southern Iran.

METHODS

The Iran Cerebral Venous Thrombosis Registry (ICVTR code: 9001013381) provided data on 382 CVT cases from Namazi Hospital. The control group comprised of adult headache patients without CVT as confirmed by neuroimaging and was retrospectively selected from those admitted to the same hospital. We collected 60 clinical and demographic features for model development and validation. Our modeling pipeline involved imputing missing values and evaluating four machine learning algorithms: generalized linear model, random forest, support vector machine, and extreme gradient boosting.

RESULTS

A total of 314 CVT cases and 575 controls were included. The highest AUROC was reached when imputation was used to estimate missing values for all the variables, combined with the support vector machine model (AUROC=0.910, Recall=0.73, Precision=0.88). The best recall was achieved also by the support vector machine model when only variables with less than 50% missing rate were included (AUROC=0.887, Recall=0.77, Precision=0.86). The random forest model yielded the best precision by using variables with less than 50% missing rate (AUROC=0.882, Recall=0.61, Precision=0.94).

CONCLUSION

The application of machine learning techniques using clinical data showed promising results in accurately diagnosing CVT within our study population. This approach offers a valuable complementary assistive tool or an alternative to resource-intensive imaging methods.

Machine Learning-Based Prediction of Stroke in Emergency Departments

Background

Stroke misdiagnosis, associated with poor outcomes, is estimated to occur in 9% of all stroke patients.

Objectives

We hypothesized that machine learning (ML) could assist in the diagnosis of ischemic stroke in emergency departments (EDs).

Design

The study was conducted and reported according to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis guidelines. We performed model development and prospective temporal validation, using data from pre- and post-COVID periods; we also performed a case study on a small cohort of previously misdiagnosed stroke patients.

Methods

We used structured and unstructured electronic health records (EHRs) of 56,452 patient encounters from 13 hospitals in Pennsylvania, from September 2003 to January 2021. ML pipelines, including natural language processing, were created using pre-event clinical data and provider notes in the EDs.

Results

Using pre-event information, our model's area under the receiver operating characteristics curve (AUROC) ranged from 0.88 to 0.92 with a similar range accuracy (0.87-0.90). Using provider notes, we identified five models that reached a balanced performance in terms of AUROC, sensitivity, and specificity. Model AUROC ranged from 0.93 to 0.99. Model sensitivity and specificity reached 0.90 and 0.99, respectively. Four of the top five performing models were based on the post-COVID provider notes; however, no performance difference between models tested on pre- and post-COVID was observed.

Conclusion

This study leveraged pre-event and at-encounter level EHR for stroke prediction. The results indicate that available clinical information can be used for building EHR-based stroke prediction models and ED stroke alert systems.

Transient Ischemic Attack Outpatient Clinic: Past Journey and Future Adventure

A transient ischemic attack (TIA), a constellation of temporary neurological symptoms, precedes stroke in one-fifth of patients. Thus far, many clinical models have been introduced to optimize the quality, time to treatment, and cost of acute TIA care, either in an inpatient or outpatient setting. In this article, we aim to review the characteristics and outcomes of outpatient TIA clinics across the globe. In addition, we discussed the main challenges for outpatient management of TIA, including triage and diagnosis, and the system dynamics of the clinics. We further reviewed the potential developments in TIA care, such as telemedicine, predictive analytics, personalized medicine, and advanced imaging.

Artificial Intelligence in Acute Ischemic Stroke Subtypes According to Toast Classification: A Comprehensive Narrative Review

The correct recognition of the etiology of ischemic stroke (IS) allows tempestive interventions in therapy with the aim of treating the cause and preventing a new cerebral ischemic event. Nevertheless, the identification of the cause is often challenging and is based on clinical features and data obtained by imaging techniques and other diagnostic exams. TOAST classification system describes the different etiologies of ischemic stroke and includes five subtypes: LAAS (large-artery atherosclerosis), CEI (cardio embolism), SVD (small vessel disease), ODE (stroke of other determined etiology), and UDE (stroke of undetermined etiology). AI models, providing computational methodologies for quantitative and objective evaluations, seem to increase the sensitivity of main IS causes, such as tomographic diagnosis of carotid stenosis, electrocardiographic recognition of atrial fibrillation, and identification of small vessel disease in magnetic resonance images. The aim of this review is to provide overall knowledge about the most effective AI models used in the differential diagnosis of ischemic stroke etiology according to the TOAST classification. According to our results, AI has proven to be a useful tool for identifying predictive factors capable of subtyping acute stroke patients in large heterogeneous populations and, in particular, clarifying the etiology of UDE IS especially detecting cardioembolic sources.

Description of neurological mimics presented to the neurology service of a small animal referral hospital

BACKGROUND

Clinicians observe that cats and dogs referred to neurology services often do not have an underlying neurological disorder. There has been no analysis of the frequency or categorisation of these neurological mimics.

METHODS

Retrospective study of 520 cases was carried out. Data on signalment, presenting clinical signs, neurological examination findings and final diagnosis were collected. Final diagnoses were classified as primary neurological, non-neurological in origin but with neurological clinical manifestation, completely non-neurological (neurological mimics) or undiagnosed. Presenting clinical signs and neurological examination results were compared between neurological mimics and primary neurological cases using Chi-square or Fischer exact test. Relative risk (RR) was calculated for significant associations.

RESULTS

A total of 74% were primary neurological conditions, 8% neurological mimics, 3% non-neurological with neurological manifestation and 15% undiagnosed. An animal referred for lameness was approximately five times more likely to be diagnosed as a neurological mimic than as a primary neurological disorder (RR = 5.42, p < 0.001). Cases with a normal neurological examination were approximately 15 times more likely to be a neurological mimic (RR = 14.97, p < 0.001).

CONCLUSION

Thorough examination with consideration of alternative diagnoses is important when a neurological condition is suspected in an animal that presents with lameness or normal neurological examination.

Risk of Subsequent Stroke Among Patients Receiving Outpatient vs Inpatient Care for Transient Ischemic Attack: A Systematic Review and Meta-analysis

IMPORTANCE

Transient ischemic attack (TIA) often indicates a high risk of subsequent cerebral ischemic events. Timely preventive measures improve the outcome.

OBJECTIVE

To estimate and compare the risk of subsequent ischemic stroke among patients with TIA or minor ischemic stroke (mIS) by care setting.

DATA SOURCES

MEDLINE, Web of Science, Scopus, Embase, International Clinical Trials Registry Platform, ClinicalTrials.gov, Trip Medical Database, CINAHL, and all Evidence-Based Medicine review series were searched from the inception of each database until October 1, 2020.

STUDY SELECTION

Studies evaluating the occurrence of ischemic stroke after TIA or mIS were included. Cohorts without data on evaluation time for reporting subsequent stroke, with retrospective diagnosis of the index event after stroke occurrence, and with a report of outcomes that were not limited to patients with TIA or mIS were excluded. Two authors independently screened the titles and abstracts and provided the list of candidate studies for full-text review; discrepancies and disagreements in all steps of the review were addressed by input from a third reviewer.

DATA EXTRACTION AND SYNTHESIS

The study was prepared and reported following the Preferred Reporting Items for Systematic Reviews and Meta-analyses, Meta-analysis of Observational Studies in Epidemiology, Methodological Expectations of Cochrane Intervention Reviews, and Enhancing the Quality and Transparency of Health Research guidelines. The Risk of Bias in Nonrandomized Studies-of Exposures (ROBINS-E) tool was used for critical appraisal of cohorts, and funnel plots, Begg-Mazumdar rank correlation, Kendall τ2, and the Egger bias test were used for evaluating the publication bias. All meta-analyses were conducted under random-effects models.

MAIN OUTCOMES AND MEASURES

Risk of subsequent ischemic stroke among patients with TIA or mIS who received care at rapid-access TIA or neurology clinics, inpatient units, emergency departments (EDs), and unspecified or multiple settings within 4 evaluation intervals (ie, 2, 7, 30, and 90 days).

RESULTS

The analysis included 226 683 patients from 71 articles recruited between 1981 and 2018; 5636 patients received care at TIA clinics (mean [SD] age, 65.7 [3.9] years; 2291 of 4513 [50.8%] men), 130 139 as inpatients (mean [SD] age, 78.3 [4.0] years; 49 458 of 128 745 [38.4%] men), 3605 at EDs (mean [SD] age, 68.9 [3.9] years; 1596 of 3046 [52.4%] men), and 87 303 patients received care in an unspecified setting (mean [SD] age, 70.8 [3.8] years, 43 495 of 87 303 [49.8%] men). Among the patients who were treated at a TIA clinic, the risk of subsequent stroke following a TIA or mIS was 0.3% (95% CI, 0.0%-1.2%) within 2 days, 1.0% (95% CI, 0.3%-2.0%) within 7 days, 1.3% (95% CI, 0.4%-2.6%) within 30 days, and 2.1% (95% CI, 1.4%-2.8%) within 90 days. Among the patients who were treated as inpatients, the risk of subsequent stroke was to 0.5% (95% CI, 0.1%-1.1%) within 2 days, 1.2% (95% CI, 0.4%-2.2%) within 7 days, 1.6% (95% CI, 0.6%-3.1%) within 30 days, and 2.8% (95% CI, 2.1%-3.5%) within 90 days. The risk of stroke among patients treated at TIA clinics was not significantly different from those hospitalized. Compared with the inpatient cohort, TIA clinic patients were younger and had had lower ABCD2 (age, blood pressure, clinical features, duration of TIA, diabetes) scores (inpatients with ABCD2 score >3, 1101 of 1806 [61.0%]; TIA clinic patients with ABCD2 score >3, 1933 of 3703 [52.2%]).

CONCLUSIONS AND RELEVANCE

In this systematic review and meta-analysis, the risk of subsequent stroke among patients who were evaluated in a TIA clinic was not higher than those hospitalized. Patients who received treatment in EDs without further follow-up had a higher risk of subsequent stroke. These findings suggest that TIA clinics can be an effective component of the TIA care component pathway.

Artificial Intelligence: A Shifting Paradigm in Cardio-Cerebrovascular Medicine

The future of healthcare is an organic blend of technology, innovation, and human connection. As artificial intelligence (AI) is gradually becoming a go-to technology in healthcare to improve efficiency and outcomes, we must understand our limitations. We should realize that our goal is not only to provide faster and more efficient care, but also to deliver an integrated solution to ensure that the care is fair and not biased to a group of sub-population. In this context, the field of cardio-cerebrovascular diseases, which encompasses a wide range of conditions-from heart failure to stroke-has made some advances to provide assistive tools to care providers. This article aimed to provide an overall thematic review of recent development focusing on various AI applications in cardio-cerebrovascular diseases to identify gaps and potential areas of improvement. If well designed, technological engines have the potential to improve healthcare access and equitability while reducing overall costs, diagnostic errors, and disparity in a system that affects patients and providers and strives for efficiency.

Adherence to anticoagulant guideline for atrial fibrillation: A large care gap among stroke patients in a rural population

OBJECTIVE

This study aimed to investigate the prevalence and factors associated with oral anticoagulant undertreatment of atrial fibrillation (AF) among a cohort of rural patients with stroke outcomes and examine how undertreatment may influence a patient's one-year survival after stroke.

METHODS

This retrospective cohort study examined ischemic stroke patients with pre-stroke AF diagnosis from September 2003 to May 2019 and divided them into proper treatment and undertreatment group. Analysis included chi-square test, variance analysis, Kruskal-Wallis test, logistic regression, Kaplan-Meier estimator, and Cox proportional-hazards model.

RESULTS

Out of 1062 ischemic stroke patients with a pre-stroke AF diagnosis, 1015 patients had a CHA₂DS₂-VASc score ≥2, and 532 (52.4%) of those were undertreated. Median time from AF diagnosis to index stroke was significantly lower among undertreated patients (1.9 years vs. 3.6 years, p < 0.001). Other thromboembolism, excluding stroke, TIA, and myocardial infarction (OR 0.41, p < 0.001), the number of encounters per year (OR 0.90, p < 0.001), and the median time between AF diagnosis and stroke event (OR 0.86, p < 0.001) were negatively associated with undertreatment. Kaplan-Meier estimator showed no statistical difference in the one-year survival probability between groups (log-rank test, p = 0.29), while the Cox-Hazard model showed that age (HR 1.05, p < 0.001) and history of congestive heart failure (HR 1.88, p < 0.001) increased the risk of mortality.

CONCLUSIONS

More than half of our rural stroke patients with a pre-index AF diagnosis were not on guideline-recommended treatment. The study highlights a large care gap and an opportunity to improve AF management.

Prediction of Long-Term Stroke Recurrence Using Machine Learning Models

Background: The long-term risk of recurrent ischemic stroke, estimated to be between 17% and 30%, cannot be reliably assessed at an individual level. Our goal was to study whether machine-learning can be trained to predict stroke recurrence and identify key clinical variables and assess whether performance metrics can be optimized. Methods: We used patient-level data from electronic health records, six interpretable algorithms (Logistic Regression, Extreme Gradient Boosting, Gradient Boosting Machine, Random Forest, Support Vector Machine, Decision Tree), four feature selection strategies, five prediction windows, and two sampling strategies to develop 288 models for up to 5-year stroke recurrence prediction. We further identified important clinical features and different optimization strategies. Results: We included 2091 ischemic stroke patients. Model area under the receiver operating characteristic (AUROC) curve was stable for prediction windows of 1, 2, 3, 4, and 5 years, with the highest score for the 1-year (0.79) and the lowest score for the 5-year prediction window (0.69). A total of 21 (7%) models reached an AUROC above 0.73 while 110 (38%) models reached an AUROC greater than 0.7. Among the 53 features analyzed, age, body mass index, and laboratory-based features (such as high-density lipoprotein, hemoglobin A1c, and creatinine) had the highest overall importance scores. The balance between specificity and sensitivity improved through sampling strategies. Conclusion: All of the selected six algorithms could be trained to predict the long-term stroke recurrence and laboratory-based variables were highly associated with stroke recurrence. The latter could be targeted for personalized interventions. Model performance metrics could be optimized, and models can be implemented in the same healthcare system as intelligent decision support for targeted intervention.

Age-Dependent Differences in the Rate and Symptoms of TIA Mimics in Patients Presenting With a Suspected TIA to a Neurological Emergency Room

Background: Transient ischemic attack (TIA) needs further diagnostic evaluation to prevent future ischemic stroke. However, prophylaxis can be harmful in elderly if the diagnosis is wrong. We aimed at characterizing differences in TIA mimics in younger and older patients to enhance diagnostic accuracy in elderly patients. Methods: In a dedicated neurological emergency room (nER) of a tertiary care University hospital, patients with transient neurological symptoms suspicious of TIA (<24 h) were retrospectively analyzed regarding their final diagnoses and their symptoms. These parameters were compared between patients aged 18-70 and >70 years using descriptive, univariable, and multivariable statistics. Results: From November 2018 until August 2019, 386 consecutive patients were included. 271 (70%) had cardiovascular risk factors and all patients received cerebral imaging, mostly CT [376 (97%)]. There was no difference in the rate of diagnosed TIA between the age groups [85 (46%) vs. 58 (39%); p = 0.213].TIA mimics in the elderly were more often internal medicine diseases [35 (19%) vs. 7 (5%); p < 0.001] and epileptic seizures [48 (26%) vs. 24 (16%); p = 0.032] but less often migraine [2 (1%) vs. 20 (13%); p < 0.001]. The most frequent symptoms in all patients were aphasia and dysarthria [107 (28%) and 92 (24%)]. Sensory impairments were less frequent in elderly patients [23 (11%) vs. 54 (30%); p < 0.001]. Impaired consciousness and orientation were independent predictors for TIA mimics (p < 0.001) whereas facial palsy (p < 0.001) motor weakness (p < 0.001), dysarthria (p = 0.022) and sensory impairment (p < 0.001) were independent predictors of TIA. Conclusion: TIA mimics in elderly patients are more likely to be internal medicine diseases and epilepsy compared to younger patients. Excluding internal medicine diseases seems to be important in elderly patients. Facial palsy, motor weakness, dysarthria and sensory impairment are associated with TIA.