A Clinical Score for Predicting Atrial Fibrillation in Patients with Cryptogenic Stroke or Transient Ischemic Attack

Prospective multicentric validation of a novel prediction model for paroxysmal atrial fibrillation

Background

The early recognition of paroxysmal atrial fibrillation (pAF) is a major clinical challenge for preventing thromboembolic events. In this prospective and multicentric study we evaluated prediction scores for the presence of pAF, calculated from non-invasive medical history and echocardiographic parameters, in patients with unknown AF status.

Methods

The 12-parameter score with parameters age, LA diameter, aortic root diameter, LV,ESD, TDI Aʹ, heart frequency, sleep apnea, hyperlipidemia, type II diabetes, smoker, ß-blocker, catheter ablation, and the 4-parameter score with parameters age, LA diameter, aortic root diameter and TDI A’ were tested. Presence of pAF was verified by continuous electrocardiogram (ECG) monitoring for up to 21 days in 305 patients.

Results

The 12-parameter score correctly predicted pAF in all 34 patients, in which pAF was newly detected by ECG monitoring. The 12- and 4-parameter scores showed sensitivities of 100% and 82% (95%-CI 65%, 93%), specificities of 75% (95%-CI 70%, 80%) and 67% (95%-CI 61%, 73%), and areas under the receiver operating characteristic (ROC) curves of 0.84 (95%-CI 0.80, 0.88) and 0.81 (95%-CI 0.74, 0.87). Furthermore, properties of AF episodes and durations of ECG monitoring necessary to detect pAF were analysed.

Conclusions

The prediction scores adequately detected pAF using variables readily available during routine cardiac assessment and echocardiography. The model scores, denoted as ECHO-AF scores, represent simple, highly sensitive and non-invasive tools for detecting pAF that can be easily implemented in the clinical practice and might serve as screening test to initiate further diagnostic investigations for validating the presence of pAF.

Graphic abstract

Prospective validation of a novel prediction model for paroxysmal atrial fibrillation based on echocardiography and medical history parameters by long-term Holter ECG

Electronic supplementary material

The online version of this article (10.1007/s00392-020-01773-z) contains supplementary material, which is available to authorized users.

Implantable Cardiac Monitoring in the Secondary Prevention of Cryptogenic Stroke

OBJECTIVE

In this study, we sought to evaluate the impact of implantable cardiac monitoring (ICM) in the prevention of stroke recurrence after a cryptogenic ischemic stroke or transient ischemic attack (TIA).

METHODS

We evaluated consecutive patients with cryptogenic ischemic stroke or TIA admitted in a comprehensive stroke center during an 8-year period. We compared the baseline characteristics and outcomes between patients receiving conventional cardiac monitoring with repeated 24-hour Holter-monitoring during the first 5 years in the outpatient setting and those receiving continuous cardiac monitoring with ICM during the last 3 years. Associations on the outcomes of interest were further assessed in multivariable regression models adjusting for potential confounders.

RESULTS

We identified a total of 373 patients receiving conventional cardiac monitoring and 123 patients receiving ICM. Paroxysmal atrial fibrillation (PAF) detection was higher in the ICM cohort compared to the conventional cardiac monitoring cohort (21.1% vs 7.5%, p < 0.001). ICM was independently associated with an increased likelihood of PAF detection during follow-up (hazard ratio [HR] = 1.94, 95% confidence interval [CI] = 1.16-3.24) in multivariable analyses. Patients receiving ICM were also found to have significantly higher rates of anticoagulation initiation (18.7% vs 6.4%, p < 0.001) and lower risk of stroke recurrence (4.1% vs 11.8%, p = 0.013). ICM was independently associated with a lower risk of stroke recurrence during follow-up (HR = 0.32, 95% CI = 0.11-0.90) in multivariable analyses.

INTERPRETATION

ICM appears to be independently associated with a higher likelihood of PAF detection and anticoagulation initiation after a cryptogenic ischemic stroke or TIA. ICM was also independently related to lower risk of stroke recurrence in our cryptogenic stroke / TIA cohort. ANN NEUROL 2020;88:946-955.

Detecting undiagnosed atrial fibrillation in UK primary care: Validation of a machine learning prediction algorithm in a retrospective cohort study

AIMS

To evaluate the ability of a machine learning algorithm to identify patients at high risk of atrial fibrillation in primary care.

METHODS

A retrospective cohort study was undertaken using the DISCOVER registry to validate an algorithm developed using a Clinical Practice Research Datalink (CPRD) dataset. The validation dataset included primary care patients in London, England aged ≥30 years from 1 January 2006 to 31 December 2013, without a diagnosis of atrial fibrillation in the prior 5 years. Algorithm performance metrics were sensitivity, specificity, positive predictive value, negative predictive value (NPV) and number needed to screen (NNS). Subgroup analysis of patients aged ≥65 years was also performed.

RESULTS

Of 2,542,732 patients in DISCOVER, the algorithm identified 604,135 patients suitable for risk assessment. Of these, 3.0% (17,880 patients) had a diagnosis of atrial fibrillation recorded before study end. The area under the curve of the receiver operating characteristic was 0.87, compared with 0.83 in algorithm development. The NNS was nine patients, matching the CPRD cohort. In patients aged ≥30 years, the algorithm correctly identified 99.1% of patients who did not have atrial fibrillation (NPV) and 75.0% of true atrial fibrillation cases (sensitivity). Among patients aged ≥65 years (n = 117,965), the NPV was 96.7% with 91.8% sensitivity.

CONCLUSIONS

This atrial fibrillation risk prediction algorithm, based on machine learning methods, identified patients at highest risk of atrial fibrillation. It performed comparably in a large, real-world population-based cohort and the developmental registry cohort. If implemented in primary care, the algorithm could be an effective tool for narrowing the population who would benefit from atrial fibrillation screening in the United Kingdom.

Frontiers of Upstream Stroke Prevention and Reduced Stroke Inequity Through Predicting, Preventing, and Managing Hypertension and Atrial Fibrillation: A Call to Action From the Value in Healthcare Initiative's Predict & Prevent Learning Collaborative

Stroke is one of the leading causes of morbidity and mortality in the United States. While age-adjusted stroke mortality was falling, it has leveled off in recent years due in part to advances in medical technology, health care options, and population health interventions. In addition to adverse trends in stroke-related morbidity and mortality across the broader population, there are sociodemographic inequities in stroke risk. These challenges can be addressed by focusing on predicting and preventing modifiable upstream risk factors associated with stroke, but there is a need to develop a practical framework that health care organizations can use to accomplish this task across diverse settings. Accordingly, this article describes the efforts and vision of the multi-stakeholder Predict & Prevent Learning Collaborative of the Value in Healthcare Initiative, a collaboration of the American Heart Association and the Robert J. Margolis, MD, Center for Health Policy at Duke University. This article presents a framework of a potential upstream stroke prevention program with evidence-based implementation strategies for predicting, preventing, and managing stroke risk factors. It is meant to complement existing primary stroke prevention guidelines by identifying frontier strategies that can address gaps in knowledge or implementation. After considering a variety of upstream medical or behavioral risk factors, the group identified 2 risk factors with substantial direct links to stroke for focusing the framework: hypertension and atrial fibrillation. This article also highlights barriers to implementing program components into clinical practice and presents implementation strategies to overcome those barriers. A particular focus was identifying those strategies that could be implemented across many settings, especially lower-resource practices and community-based enterprises representing broad social, economic, and geographic diversity. The practical framework is designed to provide clinicians and health systems with effective upstream stroke prevention strategies that encourage scalability while allowing customization for their local context.

Recurrent Stroke With Rivaroxaban Compared With Aspirin According to Predictors of Atrial Fibrillation: Secondary Analysis of the NAVIGATE ESUS Randomized Clinical Trial

Importance

The NAVIGATE ESUS randomized clinical trial found that 15 mg of rivaroxaban per day does not reduce stroke compared with aspirin in patients with embolic stroke of undetermined source (ESUS); however, it substantially reduces stroke risk in patients with atrial fibrillation (AF).

Objective

To analyze whether rivaroxaban is associated with a reduction of recurrent stroke among patients with ESUS who have an increased risk of AF.

Design, Setting, and Participants

Participants were stratified by predictors of AF, including left atrial diameter, frequency of premature atrial contractions, and HAVOC score, a validated scheme using clinical features. Treatment interactions with these predictors were assessed. Participants were enrolled between December 2014 and September 2017, and analysis began March 2018.

Intervention

Rivaroxaban treatment vs aspirin.

Main Outcomes and Measures

Risk of ischemic stroke.

Results

Among 7112 patients with a mean (SD) age of 67 (9.8) years, the mean (SD) HAVOC score was 2.6 (1.8), the mean (SD) left atrial diameter was 3.8 (1.4) cm (n = 4022), and the median (interquartile range) daily frequency of premature atrial contractions was 48 (13-222). Detection of AF during follow-up increased for each tertile of HAVOC score: 2.3% (score, 0-2), 3.0% (score, 3), and 5.8% (score, >3); however, neither tertiles of the HAVOC score nor premature atrial contractions frequency impacted the association of rivaroxaban with recurrent ischemic stroke (P for interaction = .67 and .96, respectively). Atrial fibrillation annual incidence increased for each tertile of left atrial diameter (2.0%, 3.6%, and 5.2%) and for each tertile of premature atrial contractions frequency (1.3%, 2.9%, and 7.0%). Among the predefined subgroup of patients with a left atrial diameter of more than 4.6 cm (9% of overall population), the risk of ischemic stroke was lower among the rivaroxaban group (1.7% per year) compared with the aspirin group (6.5% per year) (hazard ratio, 0.26; 95% CI, 0.07-0.94; P for interaction = .02).

Conclusions and Relevance

The HAVOC score, left atrial diameter, and premature atrial contraction frequency predicted subsequent clinical AF. Rivaroxaban was associated with a reduced risk of recurrent stroke among patients with ESUS and moderate or severe left atrial enlargement; however, this needs to be independently confirmed before influencing clinical practice.

Diagnosis and Treatment for embolic stroke of undetermined source: Consensus statement from the Taiwan stroke society and Taiwan society of cardiology

Cryptogenic stroke comprises about one-quarter of ischemic strokes with high recurrence rate; however, studies specifically investigating the features and treatment of this stroke subtype are rare. The concept of 'embolic stroke of undetermined source' (ESUS) may facilitate the development of a standardized approach to diagnose cryptogenic stroke and improve clinical trials. Since recent large randomized control trials failed to demonstrate a reduction in stroke recurrence with anticoagulants, anti-platelet agents remain the first-line treatment for ESUS patients. Nevertheless, patients with high risk of stroke recurrence (e.g., those with repeated embolic infarcts despite aspirin treatment) require a more extensive survey of stroke etiology, including cardiac imaging and prolonged cardiac rhythm monitoring. Anticoagulant treatments may still benefit some subgroups of high-risk ESUS patients, such as those with multiple infarcts at different arterial territories without aortic atheroma, the elderly, or patients with high CHA₂D₂-VASc or HOVAC scores, atrial cardiopathy or patent foramen ovale. Several important ESUS clinical trials are ongoing, and the results are anticipated. With rapid progress in our understanding of ESUS pathophysiology, new subcategorizations of ESUS and assignment of optimal treatments for each ESUS subgroup are expected in the near future.

Left atrial diameter thresholds and new incident atrial fibrillation in embolic stroke of undetermined source

BACKGROUND AND PURPOSE

We analyzed consecutive patients with embolic stroke of undetermined source (ESUS) from three prospective stroke registries to compare the prognostic performance of different LAD thresholds for the prediction of new incident AF.

METHODS

We calculated the sensitivity, specificity, positive prognostic value (PPV), negative prognostic value (NPV) and Youden's J-statistic of different LAD thresholds to predict new incident AF. We performed multivariate stepwise regression with forward selection of covariates to assess the association between the LAD threshold with the highest Youden's J-statistic and AF detection.

RESULTS

Among 675 patients followed for 2437 patient-years, the mean LAD was 38.5 ± 6.8 mm. New incident AF was diagnosed in 115 (17.0%) patients. The LAD threshold of 40mm yielded the highest Youden's J-statistic of 0.35 with sensitivity 0.69, specificity 0.66, PPV 0.27 and NPV 0.92. The likelihood of new incident AF was nearly twice in patients with LAD > 40 mm compared to LAD ≤ 40 mm (HR:1.92, 95%CI:1.24-2.97, p = 0.004). The 10-year cumulative probability of new incident AF was higher in patients with LAD>40 mm compared to LAD ≤ 40 mm (53.5% and 22.4% respectively, log-rank-test: 28.2, p < 0.001). The annualized rate of stroke recurrence of 4.0% in the overall population did not differ significantly in patient above vs. below this LAD threshold (HR:0.96, 95%CI:0.62-1.48, p = 0.85).

CONCLUSIONS

The LAD threshold of 40 mm has the best prognostic performance among other LAD values to predict new incident AF after ESUS. The diagnostic yield of prolonged cardiac rhythm monitoring in patients with LAD ≤ 40 mm seems low; therefore, such patients may have lower priority for prolonged cardiac monitoring.

Neuro-arrhythmology: a challenging field of action and research: a review from the Task Force of Neuro-arrhythmology of Italian Association of Arrhythmias and Cardiac Pacing

: There is a growing interest in the study of the mechanisms of heart and brain interactions with the aim to improve the management of high-impact cardiac rhythm disorders, first of all atrial fibrillation. However, there are several topics to which the scientific interests of cardiologists and neurologists converge constituting the basis for enhancing the development of neuro-arrhythmology. This multidisciplinary field should cover a wide spectrum of diseases, even beyond the classical framework corresponding to stroke and atrial fibrillation and include the complex issues of seizures as well as loss of consciousness and syncope. The implications of a more focused interaction between neurologists and cardiologists in the field of neuro-arrhythmology should include in perspective the institution of research networks specifically devoted to investigate 'from bench to bedside' the complex pathophysiological links of the abovementioned diseases, with involvement of scientists in the field of biochemistry, genetics, molecular medicine, physiology, pathology and bioengineering. An investment in the field could have important implications in the perspectives of a more personalized approach to patients and diseases, in the context of 'precision'medicine. Large datasets and electronic medical records, with the approach typical of 'big data' could enhance the possibility of new findings with potentially important clinical implications. Finally, the interaction between neurologists and cardiologists involved in arrythmia management should have some organizational implications, with new models of healthcare delivery based on multidisciplinary assistance, similarly to that applied in the case of syncope units.

External Performance of the HAVOC Score for the Prediction of New Incident Atrial Fibrillation

Background and Purpose—

The HAVOC score (hypertension, age, valvular heart disease, peripheral vascular disease, obesity, congestive heart failure, coronary artery disease) was proposed for the prediction of atrial fibrillation (AF) after cryptogenic stroke. It showed good model discrimination (area under the curve, 0.77). Only 2.5% of patients with a low-risk HAVOC score (ie, 0–4) were diagnosed with new incident AF. We aimed to assess its performance in an external cohort of patients with embolic stroke of undetermined source.

Methods—

In the AF-embolic stroke of undetermined source dataset, we assessed the discriminatory power, calibration, specificity, negative predictive value, and accuracy of the HAVOC score to predict new incident AF. Patients with a HAVOC score of 0 to 4 were considered as low-risk, as proposed in its original publication.

Results—

In 658 embolic stroke of undetermined source patients (median age, 67 years; 44% women), the median HAVOC score was 2 (interquartile range, 3). There were 540 (82%) patients with a HAVOC score of 0 to 4 and 118 (18%) with a score of ≥5. New incident AF was diagnosed in 95 (14.4%) patients (28.8% among patients with HAVOC score ≥5 and 11.3% among patients with HAVOC score 0–4 [age- and sex-adjusted odds ratio, 2.29 (95% CI, 1.37–3.82)]). The specificity of low-risk HAVOC score to identify patients without new incident AF was 88.7%. The negative predictive value of low-risk HAVOC score was 85.1%. The accuracy was 78.0%, and the area under the curve was 68.7% (95% CI, 62.1%–73.3%).

Conclusions—

The previously reported low rate of AF among embolic stroke of undetermined source patients with low-risk HAVOC score was not confirmed in our cohort. Further assessment of the HAVOC score is warranted before it is routinely implemented in clinical practice.

Usefulness of the ACTEL Score to Predict Atrial Fibrillation in Patients with Cryptogenic Stroke

INTRODUCTION

To assess the probability of undetected atrial fibrillation (AF) in patients with ischemic stroke, we previously compared patients who were first diagnosed with AF with patients with large or small artery disease and obtained the MrWALLETS 8-item scoring system. In the present study, we utilized cryptogenic strokes (CS) as the control group, as AF is normally sought among CS patients.

METHODS

We retrospectively examined 191 ischemic stroke patients (72.5 ± 12.6 years), 68 with first diagnosed AF and 123 with CS, who had undergone 2 brain CT scans, echocardiography, carotid/vertebral ultrasound, continuous electrocardiogram monitoring and anamnestic/laboratory search for cardiovascular risk factors.

RESULTS

In logistic regression, 5 variables were independently associated with AF, forming the "ACTEL" score: Age ≥75 years (OR 2.42, 95% CI 1.18-4.96, p = 0.02; +1 point); hyperCholesterolemia (OR 0.38, 95% CI 0.18-0.78, p = 0.009; -1 point); Tricuspid regurgitation ≥ mild-to-moderate (OR 4.99, 95% CI 1.63-15.27, p = 0.005; +1 point); left ventricular End-diastolic volume <65 mL (OR 7.43, 95% CI 2.44-22.6, p = 0.0004; +1 point); Left atrium ≥4 cm (OR 4.57, 95% CI 1.97-10.62, p = 0.0004; +1 point). The algebraic sum of these points may range from -1 to +4. For AF identification, the area under the receiver operating characteristic curve was 0.80 (95% CI 0.73-0.87). With a cutoff of ≥2, positive predictive value was 80.8%, specificity 92.7% and sensitivity 55.9%.

CONCLUSIONS

The ACTEL score, a simplified and improved version of the MrWALLETS score, allows the identification of patients with first diagnosed AF, in the context of CSs, with a high positive predictive value.

A boosting inspired personalized threshold method for sepsis screening

Sepsis is one of the biggest risks to patient safety, with a natural mortality rate between 25% and 50%. It is difficult to diagnose, and no validated standard for diagnosis currently exists. A commonly used scoring criteria is the quick sequential organ failure assessment (qSOFA). It demonstrates very low specificity in ICU populations, however. We develop a method to personalize thresholds in qSOFA that incorporates easily to measure patient baseline characteristics. We compare the personalized threshold method to qSOFA, five previously published methods that obtain an optimal constant threshold for a single biomarker, and to the machine learning algorithms based on logistic regression and AdaBoosting using patient data in the MIMIC-III database. The personalized threshold method achieves higher accuracy than qSOFA and the five published methods and has comparable performance to machine learning methods. Personalized thresholds, however, are much easier to adopt in real-life monitoring than machine learning methods as they are computed once for a patient and used in the same way as qSOFA, whereas the machine learning methods are hard to implement and interpret.

Development of a novel score to predict newly diagnosed atrial fibrillation after ischemic stroke: The CHASE-LESS score

BACKGROUND AND AIMS

Prompt detection of atrial fibrillation (AF) is essential for optimal secondary stroke prevention, but routine long-term cardiac monitoring of all ischemic stroke patients is neither practical nor affordable. We aimed to develop and validate a risk score to identify patients at risk for newly diagnosed AF (NDAF) after ischemic stroke.

METHODS

Information on adult patients hospitalized for ischemic stroke without known AF was retrieved from a nationwide database. Primary outcome was NDAF within one year following index stroke. A stepwise Cox model was used to screen for predictors. Beta coefficients for the independent predictors were converted to integer points, which were summed to create a risk score.

RESULTS

We identified 4 positive predictors and 3 negative predictors. The CHASE-LESS score (Coronary, Heart failure, Age, stroke SEverity, - LipidEmia, Sugar, prior Stroke) comprises coronary artery disease (1 point), congestive heart failure (1 point), age (1 point for every 10 years), stroke severity (National Institutes of Health Stroke Scale; 1 point for 6-13 and 4 points for ≥14), hyperlipidemia (-1 point), diabetes (-1 point), and prior history of stroke or transient ischemic attack (-1 point). Overall, 6.0% (1029/17,076) of patients developed NDAF. The incidence rate ranged from 8/1000 person-years (CHASE-LESS ≤3) to 240/1000 person-years (CHASE-LESS ≥10). The model achieved a c-index of 0.730 in the development cohort and 0.732 in the validation cohort.

CONCLUSIONS

The CHASE-LESS score could aid clinicians to identify patients at risk of developing NDAF and help prioritize patients for advanced cardiac monitoring.

Risk factors for recurrence of atrial fibrillation

OBJECTIVE

Atrial fibrillation (AF) is a progressive disease, associated with increased risk of mortality, stroke, heart failure, and worsens quality of life. There is a high incidence of AF recurrence despite the treatment. The aim of the study was to assess the time to recurrence of AF after sinus rhythm restoration with electrical or pharmacological cardioversion and to identify the risk factors.

METHODS

This study included 101 patients with AF (56% females) at a mean age of 68.02±7 years, after sinus rhythm restoration in a clinical observation of 1-year placebo-controlled treatment with spironolactone (1: 1). The patients were analyzed on the basis of AF recurrence, hospitalization, demographic parameters, comorbidities, embolic risk, and value of biomarker galectin-3 (Gal-3).

RESULTS

The average number of AF recurrences was1.62 per patient per year. The median time of occurrence of at least one new episode was 48 days, 95% confidence interval (CI) 14.24-81.76. Female patients experienced significantly more recurrences than male-53.3% vs. 28.6% hazard ration (HR) =1.76, 95% CI 1.02-3.03, p=0.036. The recurrences were more common with increased age, although not significantly. Patients with arterial hypertension had a threefold risk of recurrences than those without hypertension (p=0.025), independently of the treatment. CHA2DS2-VASc score was significantly associated with AF recurrent episodes. Patients with gout had a twofold increased risk, without statistical significance (p=0.15). There was no difference in the AF episodes according to treatment with spironolactone. The levels of Gal-3 did not affect the number of AF recurrences (p=0.9).

CONCLUSION

AF is associated with frequent recurrences after restoration of sinus rhythm in the majority of the patients. Most of them occurred within the first 3 months. Female sex, arterial hypertension, and CHA2DS2-VASc score were significant predictors of AF recurrence. Spironolactone did not reduce AF recurrences.

Emergency management of stroke in the era of mechanical thrombectomy

Emergency management of stroke has been directed at the delivery of recombinant tissue plasminogen activator (tPA) in a timely fashion. Because of the many limitations attached to the delivery of tPA and the perceived benefits accrued to tPA, its use has been limited. Mechanical thrombectomy, a far superior therapy for the largest and most disabling strokes, large vessel occlusions (LVOs), has changed the way acute strokes are managed. Aside from the rush to deliver tPA, there is now a need to identify LVO and refer those patients with LVO to physicians and facilities capable of delivering urgent thrombectomy. Other parts of emergency department management of stroke are directed at identifying and mitigating risk factors for future strokes and at preventing further damage from occurring. We review here the most recent literature supporting these advances in stroke care and present a framework for understanding the role that emergency physicians play in acute stroke care.

Potential Utility of Neurosonology in Paroxysmal Atrial Fibrillation Detection in Patients with Cryptogenic Stroke

Background: Occult paroxysmal atrial fibrillation (PAF) is a common and potential treatable cause of cryptogenic stroke (CS). We sought to prospectively identify independent predictors of atrial fibrillation (AF) detection in patients with CS and sinus rhythm on baseline electrocardiogram (ECG), without prior AF history. We had hypothesized that cardiac arrhythmia detection during neurosonology examinations (Carotid Duplex (CDU) and Transcranial Doppler (TCD)) may be associated with higher likelihood of AF detection. Methods: Consecutive CS patients were prospectively evaluated over a six-year period. Demographics, clinical and imaging characteristics of cerebral ischemia were documented. The presence of arrhythmia during spectral waveform analysis of CDU/TCD was recorded. Left atrial enlargement was documented during echocardiography using standard definitions. The outcome event of interest included PAF detection on outpatient 24-h Holter ECG recordings. Statistical analyses were performed using univariate and multivariate logistic regression models. Results: A total of 373 patients with CS were evaluated (mean age 60 ± 11 years, 67% men, median NIHSS-score 4 points). The rate of PAF detection of any duration on Holter ECG recordings was 11% (95% CI 8%–14%). The following three variables were independently associated with the likelihood of AF detection on 24-h Holter-ECG recordings in both multivariate analyses adjusting for potential confounders: age (OR per 10-year increase: 1.68; 95% CI: 1.19–2.37; p = 0.003), moderate or severe left atrial enlargement (OR: 4.81; 95% CI: 1.77–13.03; p = 0.002) and arrhythmia detection during neurosonology evaluations (OR: 3.09; 95% CI: 1.47–6.48; p = 0.003). Conclusion: Our findings underline the potential utility of neurosonology in improving the detection rate of PAF in patients with CS.

Subclinical and Device-Detected Atrial Fibrillation: Pondering the Knowledge Gap: A Scientific Statement From the American Heart Association

The widespread use of cardiac implantable electronic devices and wearable monitors has led to the detection of subclinical atrial fibrillation in a substantial proportion of patients. There is evidence that these asymptomatic arrhythmias are associated with increased risk of stroke. Thus, detection of subclinical atrial fibrillation may offer an opportunity to reduce stroke risk by initiating anticoagulation. However, it is unknown whether long-term anticoagulation is warranted and in what populations. This scientific statement explores the existing data on the prevalence, clinical significance, and management of subclinical atrial fibrillation and identifies current gaps in knowledge and areas of controversy and consensus.

Duration of Implantable Cardiac Monitoring and Detection of Atrial Fibrillation in Ischemic Stroke Patients: A Systematic Review and Meta-Analysis

BACKGROUND AND PURPOSE

Current guidelines do not provide firm directions on atrial fibrillation (AF) screening after ischemic stroke (IS). We sought to investigate the association of implantable cardiac monitoring (ICM) duration with the yield of AF detection in IS patients.

METHODS

We included studies reporting AF detection rates by ICM in IS patients with negative initial AF screening. We excluded studies reporting prolonged cardiac monitoring with devices other than ICM, not providing AF detection rates or monitoring duration, and reporting overlapping data for the same population. The random-effects model was used for all pooled estimates and meta-regression analyses.

RESULTS

We included 28 studies (4,531 patients, mean age 65 years). In meta-regression analyses, the proportion of AF detection by ICM was independently associated with monitoring duration (coefficient=0.015; 95% confidence interval [CI], 0.005 to 0.024) and mean patient age (coefficient=0.009; 95% CI, 0.003 to 0.015). No associations were detected with other patient characteristics, including IS subtype (cryptogenic vs. embolic stroke of undetermined source) or time from IS onset to CM implantation. In subgroup analyses, significant differences (P<0.001) in the AF detection rates were found for ICM duration (<6 months: 5% [95% CI, 3% to 6%]; ≥6 and ≤12 months: 21% [95% CI, 16% to 25%]; >12 and ≤24 months: 26% [95% CI, 22% to 31%]; >24 months: 34% [95% CI, 29% to 39%]).

CONCLUSION

s Extended duration of ICM monitoring and increased patient age are factors that substantially increase AF detection in IS patients with initial negative AF screening.

Prognostic models for predicting incident or recurrent atrial fibrillation: protocol for a systematic review

BACKGROUND

Atrial fibrillation (AF) is the arrhythmia most commonly diagnosed in clinical practice. It is associated with significant morbidity and mortality. Prevalence of AF and complications of AF, estimated by hospitalisations, have increased dramatically in the last decade. Being able to predict AF would allow tailoring of management strategies and a focus on primary or secondary prevention. Models predicting recurrent AF would have particular clinical use for the selection of rhythm control therapy. There are existing prognostic models which combine several predictors or risk factors to generate an individualised estimate of risk of AF. The aim of this systematic review is to summarise and compare model performance measures and predictive accuracy across different models and populations at risk of developing incident or recurrent AF.

METHODS

Methods tailored to systematic reviews of prognostic models will be used for study identification, risk of bias assessment and synthesis. Studies will be eligible for inclusion where they report an internally or externally validated model. The quality of studies reporting a prognostic model will be assessed using the Prediction Study Risk Of Bias Assessment Tool (PROBAST). Studies will be narratively described and included variables and predictive accuracy compared across different models and populations. Meta-analysis of model performance measures for models validated in similar populations will be considered where possible.

DISCUSSION

To the best of our knowledge, this will be the first systematic review to collate evidence from all studies reporting on validated prognostic models, or on the impact of such models, in any population at risk of incident or recurrent AF. The review may identify models which are suitable for impact assessment in clinical practice. Should gaps in the evidence be identified, research recommendations relating to model development, validation or impact assessment will be made. Findings will be considered in the context of any models already used in clinical practice, and the extent to which these have been validated.

SYSTEMATIC REVIEW REGISTRATION

PROSPERO ( CRD42018111649 ).

Automated Survival Prediction in Metastatic Cancer Patients Using High-Dimensional Electronic Medical Record Data

BACKGROUND

Oncologists use patients' life expectancy to guide decisions and may benefit from a tool that accurately predicts prognosis. Existing prognostic models generally use only a few predictor variables. We used an electronic medical record dataset to train a prognostic model for patients with metastatic cancer.

METHODS

The model was trained and tested using 12 588 patients treated for metastatic cancer in the Stanford Health Care system from 2008 to 2017. Data sources included provider note text, labs, vital signs, procedures, medication orders, and diagnosis codes. Patients were divided randomly into a training set used to fit the model coefficients and a test set used to evaluate model performance (80%/20% split). A regularized Cox model with 4126 predictor variables was used. A landmarking approach was used due to the multiple observations per patient, with t0 set to the time of metastatic cancer diagnosis. Performance was also evaluated using 399 palliative radiation courses in test set patients.

RESULTS

The C-index for overall survival was 0.786 in the test set (averaged across landmark times). For palliative radiation courses, the C-index was 0.745 (95% confidence interval [CI] = 0.715 to 0.775) compared with 0.635 (95% CI = 0.601 to 0.669) for a published model using performance status, primary tumor site, and treated site (two-sided P < .001). Our model's predictions were well-calibrated.

CONCLUSIONS

The model showed high predictive performance, which will need to be validated using external data. Because it is fully automated, the model can be used to examine providers' practice patterns and could be deployed in a decision support tool to help improve quality of care.

Embolic strokes of undetermined source: theoretical construct or useful clinical tool?

In 2014, the definition of embolic strokes of undetermined source (ESUS) emerged as a new clinical construct to characterize cryptogenic stroke (CS) patients with complete vascular workup to determine nonlacunar, nonatherosclerotic strokes of presumable embolic origin. NAVIGATE ESUS, the first phase III randomized-controlled, clinical trial (RCT) comparing rivaroxaban (15 mg daily) with aspirin (100 mg daily), was prematurely terminated for lack of efficacy after enrollment of 7213 patients. Except for the lack of efficacy in the primary outcome, rivaroxaban was associated with increased risk of major bleeding and hemorrhagic stroke compared with aspirin. RE-SPECT ESUS was the second phase III RCT that compared the efficacy and safety of dabigatran (110 or 150 mg, twice daily) to aspirin (100 mg daily). The results of this trial have been recently presented and showed similar efficacy and safety outcomes between dabigatran and aspirin. Indirect analyses of these trials suggest similar efficacy on the risk of ischemic stroke (IS) prevention, but higher intracranial hemorrhage risk in ESUS patients receiving rivaroxaban compared to those receiving dabigatran (indirect HR = 6.63, 95% CI: 1.38-31.76). ESUS constitute a heterogeneous group of patients with embolic cerebral infarction. Occult AF represents the underlying mechanism of cerebral ischemia in the minority of ESUS patients. Other embolic mechanisms (paradoxical embolism via patent foramen ovale, aortic plaque, nonstenosing unstable carotid plaque, etc.) may represent alternative mechanisms of cerebral embolism in ESUS, and may mandate different management than oral anticoagulation. The potential clinical utility of ESUS may be challenged since the concept failed to identify patients who would benefit from anticoagulation therapy. Compared with the former diagnosis of CS, ESUS patients required thorough investigations; more comprehensive diagnostic work-up than is requested in current ESUS diagnostic criteria may assist clinicians in uncovering the source of brain embolism in CS patients and individualize treatment approaches.

Evaluation of a clinical score for predicting atrial fibrillation in cryptogenic stroke patients with insertable cardiac monitors: results from the CRYSTAL AF study

Background

The HAVOC score was previously developed to predict the risk of atrial fibrillation (AF) after cryptogenic stroke (CS) or transient ischemic attack (TIA). The purpose of this study was to apply the HAVOC score to patients who received insertable cardiac monitors (ICMs) in the CRYSTAL AF study.

Methods

All patients from the CRYSTAL AF study who received an ICM were included. HAVOC score (one point each for peripheral vascular disease and obesity with body mass index >30, two points each for hypertension, age ⩾ 75, valvular heart disease, and coronary artery disease, 4 points for congestive heart failure) was computed for all patients. The primary endpoint was AF detection by 12 months of ICM monitoring.

Results

A total of 214 patients who received ICM were included. AF was detected in 40 patients while the remaining 174 patients were AF negative. The HAVOC score was significantly higher among patients with AF [median 3.0 with interquartile range (IQR) 2-4] than those without AF [median 2.0 (IQR 0-3)], p = 0.01. AF increased significantly across the three HAVOC score groups: 11% in Group A (score 0-1), 18% in Group B (score 2-3), and 32 % in Group C (score ⩾ 4) with p = 0.02.

Conclusions

The HAVOC score was shown in this post hoc analysis of CRYSTAL AF to successfully stratify AF risk post CS or TIA. The 11% AF rate in the lowest HAVOC score group highlights the significance of nontraditional contributors to AF and ischemic stroke.

Atrial fibrillation and stroke: a practical guide

Neurologists and stroke physicians will be familiar with atrial fibrillation as a major cause of ischaemic stroke, and the role of anticoagulation in preventing cardioembolic stroke. However, making decisions about anticoagulation for individual patients remains a difficult area of clinical practice, balancing the serious risk of ischaemic stroke against that of major bleeding, particularly intracranial haemorrhage. Atrial fibrillation management requires interdisciplinary collaboration with colleagues in cardiology and haematology. Recent advances, especially the now-widespread availability of direct oral anticoagulants, have brought opportunities to improve stroke care while posing new challenges. This article gives an overview of the contemporary diagnosis and management of atrial fibrillation, and the associated evidence base. Where there is uncertainty, we describe our own approach to these areas, while highlighting ongoing research that will likely guide future practice.

A scalable discrete-time survival model for neural networks

There is currently great interest in applying neural networks to prediction tasks in medicine. It is important for predictive models to be able to use survival data, where each patient has a known follow-up time and event/censoring indicator. This avoids information loss when training the model and enables generation of predicted survival curves. In this paper, we describe a discrete-time survival model that is designed to be used with neural networks, which we refer to as Nnet-survival. The model is trained with the maximum likelihood method using mini-batch stochastic gradient descent (SGD). The use of SGD enables rapid convergence and application to large datasets that do not fit in memory. The model is flexible, so that the baseline hazard rate and the effect of the input data on hazard probability can vary with follow-up time. It has been implemented in the Keras deep learning framework, and source code for the model and several examples is available online. We demonstrate the performance of the model on both simulated and real data and compare it to existing models Cox-nnet and Deepsurv.

Probabilistic Prognostic Estimates of Survival in Metastatic Cancer Patients (PPES-Met) Utilizing Free-Text Clinical Narratives

We propose a deep learning model - Probabilistic Prognostic Estimates of Survival in Metastatic Cancer Patients (PPES-Met) for estimating short-term life expectancy (>3 months) of the patients by analyzing free-text clinical notes in the electronic medical record, while maintaining the temporal visit sequence. In a single framework, we integrated semantic data mapping and neural embedding technique to produce a text processing method that extracts relevant information from heterogeneous types of clinical notes in an unsupervised manner, and we designed a recurrent neural network to model the temporal dependency of the patient visits. The model was trained on a large dataset (10,293 patients) and validated on a separated dataset (1818 patients). Our method achieved an area under the ROC curve (AUC) of 0.89. To provide explain-ability, we developed an interactive graphical tool that may improve physician understanding of the basis for the model's predictions. The high accuracy and explain-ability of the PPES-Met model may enable our model to be used as a decision support tool to personalize metastatic cancer treatment and provide valuable assistance to the physicians.