Using an artificial intelligence software improves emergency medicine physician intracranial haemorrhage detection to radiologist levels

BACKGROUND

Tools to increase the turnaround speed and accuracy of imaging reports could positively influence ED logistics. The Caire ICH is an artificial intelligence (AI) software developed for ED physicians to recognise intracranial haemorrhages (ICHs) on non-contrast enhanced cranial CT scans to manage the clinical care of these patients in a timelier fashion.

METHODS

A dataset of 532 non-contrast cranial CT scans was reviewed by five board-certified emergency physicians (EPs) with an average of 14.8 years of practice experience. The scans were labelled in random order for the presence or absence of an ICH. If an ICH was detected, the reader further labelled all subtypes present (ie, epidural, subdural, subarachnoid, intraparenchymal and/or intraventricular haemorrhage). After a washout period, the five EPs reviewed again the scans individually with the assistance of Caire ICH. The mean accuracy of the EP readings with AI assistance was compared with the mean accuracy of three general radiologists reading the films individually. The final diagnosis (ie, ground truth) was adjudicated by a consensus of the radiologists after their individual readings.

RESULTS

Mean EP reader accuracy significantly increased by 6.20% (95% CI for the difference 5.10%-7.29%; p=0.0092) when using Caire ICH to detect an ICH. Mean accuracy of the EP cohort in detecting an ICH using Caire ICH was found to be more accurate than the radiologist cohort prior to discussion; this difference, however, was not statistically significant.

CONCLUSION

The Caire ICH software significantly improved the accuracy and sensitivity of detecting an ICH by the EP to a level comparable to general radiologists. Further prospective research with larger numbers will be needed to understand the impact of Caire ICH on ED logistics and patient outcomes.

Bilateral medial medullary stroke: "The heart sign"

Bilateral medial medullary stroke is a very rare condition. It is usually associated with severe motor deficits, dysarthria, sensory symptoms, nystagmus, and respiratory disturbances. The most common etiology is atherosclerotic disease of the vertebral and anterior spinal artery or its branches. We present the case and brain imaging of a 48-year-old man with a bilateral medial medullary stroke and its classic "Heart Sign" in the magnetic resonance imaging. This case highlights the anatomy of the rostral medulla, clinical presentation, etiology, and characteristic radiologic findings of this uncommon type of stroke.

A deep learning method to identify and localize large-vessel occlusions from cerebral digital subtraction angiography

BACKGROUND AND PURPOSE

An essential step during endovascular thrombectomy is identifying the occluded arterial vessel on a cerebral digital subtraction angiogram (DSA). We developed an algorithm that can detect and localize the position of occlusions in cerebral DSA.

METHODS

We retrospectively collected cerebral DSAs from a single institution between 2018 and 2020 from 188 patients, 86 of whom suffered occlusions of the M1 and proximal M2 segments. We trained an ensemble of deep-learning models on fewer than 60 large-vessel occlusion (LVO)-positive patients. We evaluated the model on an independent test set and evaluated the truth of its predicted localizations using Intersection over Union and expert review.

RESULTS

On an independent test set of 166 cerebral DSA frames with an LVO prevalence of 0.19, the model achieved a specificity of 0.95 (95% confidence interval [CI]: 0.90, 0.99), a precision of 0.7450 (95% CI: 0.64, 0.88), and a sensitivity of 0.76 (95% CI: 0.66, 0.91). The model correctly localized the LVO in at least one frame in 13 of the 14 LVO-positive patients in the test set. The model achieved a precision of 0.67 (95% CI: 0.52, 0.79), recall of 0.69 (95% CI: 0.46, 0.81), and a mean average precision of 0.75 (95% CI: 0.56, 0.91).

CONCLUSION

This work demonstrates that a deep learning strategy using a limited dataset can generate effective representations used to identify LVOs. Generating an expanded and more complete dataset of LVOs with obstructed LVOs is likely the best way to improve the model's ability to localize LVOs.

ATP1A1-linked diseases require a malfunctioning protein product from one allele

Heterozygous germline variants in ATP1A1, the gene encoding the α1 subunit of the Na+/K+-ATPase (NKA), have been linked to diseases including primary hyperaldosteronism and the peripheral neuropathy Charcot-Marie-Tooth disease (CMT). ATP1A1 variants that cause CMT induce loss-of-function of NKA. This heterodimeric (αβ) enzyme hydrolyzes ATP to establish transmembrane electrochemical gradients of Na+ and K+ that are essential for electrical signaling and cell survival. Of the 4 catalytic subunit isoforms, α1 is ubiquitously expressed and is the predominant paralog in peripheral axons. Human population sequencing datasets indicate strong negative selection against both missense and protein-null ATP1A1 variants. To test whether haploinsufficiency generated by heterozygous protein-null alleles are sufficient to cause disease, we tested the neuromuscular characteristics of heterozygous Atp1a1+/- knockout mice and their wildtype littermates, while also evaluating if exercise increased CMT penetrance. We found that Atp1a1+/- mice were phenotypically normal up to 18 months of age. Consistent with the observations in mice, we report clinical phenotyping of a healthy adult human who lacks any clinical features of known ATP1A1-related diseases despite carrying a plasma-membrane protein-null early truncation variant, p.Y148*. Taken together, these results suggest that a malfunctioning gene product is required for disease induction by ATP1A1 variants and that if any pathology is associated with protein-null variants, they may display low penetrance or high age of onset.

External validation of an artificial intelligence device for intracranial hemorrhage detection

BACKGROUND

Artificial intelligence applications have gained traction in the field of cerebrovascular disease by assisting in the triage, classification, and prognostication of both ischemic and hemorrhagic stroke. The Caire ICH system aims to be the first device to move into the realm of assisted diagnosis for intracranial hemorrhage (ICH) and its subtypes.

METHODS

A single-center retrospective dataset of 402 head non-contrast CT scans (NCCT) with an intracranial hemorrhage were retrospectively collected from January 2012 to July 2020; an additional 108 NCCT scans with no intracranial hemorrhage findings were also included. The presence of an ICH and its subtype were determined from the ICD-10 code associated with the scan and validated by an expert panel. We used the Caire ICH vR1 to analyze these scans, and we evaluated its performance in terms of accuracy, sensitivity, and specificity.

RESULTS

We found the Caire ICH system to have an accuracy of 98.05% (95% Confidence interval (CI): 96.44%-99.06%) a sensitivity of 97.52% (95% CI: 95.50%-98.81%) and a specificity of 100% (95% CI: 96.67%-100.00%) in the detection of ICH. Experts reviewed the ten incorrectly classified scans.

CONCLUSION

The Caire ICH vR1 algorithm was highly accurate, sensitive, and specific in detecting the presence or absence of an ICH and its subtypes in NCCTs. This work suggests that the Caire ICH device has potential to minimize clinical errors in ICH diagnosis that could improve patient outcomes and current workflows as both a point-of-care tool for diagnostics and as a safety-net for radiologists.

ATP1A1 -linked diseases require a malfunctioning protein product from one allele

Heterozygous germline variants in ATP1A1 , the gene encoding the α1 subunit of the Na + /K + -ATPase (NKA), have been linked to diseases including primary hyperaldosteronism and the peripheral neuropathy Charcot-Marie-Tooth disease (CMT). ATP1A1 variants that cause CMT induce loss-of-function of NKA. This heterodimeric (αβ) enzyme hydrolyzes ATP to establish transmembrane electrochemical gradients of Na + and K + that are essential for electrical signaling and cell survival. Of the 4 catalytic subunit isoforms, α1 is ubiquitously expressed and is the predominant paralog in peripheral axons. Human population sequencing datasets indicate strong negative selection against both missense and protein-null ATP1A1 variants. To test whether haploinsufficiency generated by heterozygous protein-null alleles are sufficient to cause disease, we tested the neuromuscular characteristics of heterozygous Atp1a1 +/- knockout mice and their wildtype littermates, while also evaluating if exercise increased CMT penetrance. We found that Atp1a1 +/- mice were phenotypically normal up to 18 months of age. Consistent with the observations in mice, we report clinical phenotyping of a healthy adult human who lacks any clinical features of known ATP1A1 -related diseases despite carrying a protein-null early truncation variant, p.Y148*. Taken together, these results suggest that a malfunctioning gene product is required for disease induction by ATP1A1 variants and that if any pathology is associated with protein-null variants, they may display low penetrance or high age of onset.

Abstract TP172: Predicting Poor Response To Anti-coagulation Therapy In Cerebral Venous Thrombosis Using A Simple Clinical-radiological Score

Few studies target to identify predictors of poor response to anticoagulation therapy in Cerebral Venous Thrombosis (CVT). We aimed to determine the predictors of poor clinical response to therapeutic anticoagulation in acute CVT.

Methods: We performed a retrospective analysis of patients therapeutically anticoagulated for acute CVT. We defined poor clinical outcomes as death, need for mechanical thrombectomy during the hospitalization, or a modified Rankin Scale (mRS) >3 at clinical follow-up. Multivariate analyses identified factors associated with poor outcomes in anticoagulated patients for acute CVT, and we used the identified factors to create the PRACT-CVT (Poor Response to Anticoagulation Therapy in CVT) score.

Results: We included 109 patients anticoagulated with acute CVT. Twenty-one (19%) patients had poor clinical outcomes. Age >65 or <10 years (OR: 3.16, 95% CI: 1.06-9.44), a GCS ≤12 upon presentation (OR: 19.2, 95% CI: 4.05-91.4), focal motor deficits at admission (OR: 5.03, 95% CI: 1.64-15.44), clinical deterioration following admission (OR: 28.18, CI: 4.81-164.86), seizures following admission (OR: 5.59, 95% CI: 1.27-24.51), evidence of brain bleeding/ischemia on admission (OR: 4.67, 95% CI: 1.42-15.34), involvement of the superior sagittal sinus (OR: 3.88, CI: 1.33-11.32), or involvement of both transverse sinuses (OR: 3.87, 95% CI: 1.01-14.90) predicted poor clinical outcome despite therapeutic anticoagulation. A PRACT-CVT score (0-22 points) of ≥7 points provided a sensitivity of 71% and a specificity of 95% for predicting poor clinical outcomes with anticoagulation alone.

Conclusion: Patients with acute CVT, presenting with a GCS ≤12, with focal motor deficits, showing clinical deterioration after admission, having seizures during hospitalization, with brain bleeding/ischemia on initial imaging, involvement of the superior sagital sinus, or involvement of both transverse sinuses had poor response to anticoagulation.

Deep Learning System Boosts Radiologist Detection of Intracranial Hemorrhage

BACKGROUND

Intracranial hemorrhage (ICH) requires emergent medical treatment for positive outcomes. While previous artificial intelligence (AI) solutions achieved rapid diagnostics, none were shown to improve the performance of radiologists in detecting ICHs. Here, we show that the Caire ICH artificial intelligence system enhances a radiologist's ICH diagnosis performance.

METHODS

A dataset of non-contrast-enhanced axial cranial computed tomography (CT) scans (n=532) were labeled for the presence or absence of an ICH. If an ICH was detected, its ICH subtype was identified. After a washout period, the three radiologists reviewed the same dataset with the assistance of the Caire ICH system. Performance was measured with respect to reader agreement, accuracy, sensitivity, and specificity when compared to the ground truth, defined as reader consensus.

RESULTS

Caire ICH improved the inter-reader agreement on average by 5.76% in a dataset with an ICH prevalence of 74.3%. Further, radiologists using Caire ICH detected an average of 18 more ICHs and significantly increased their accuracy by 6.15%, their sensitivity by 4.6%, and their specificity by 10.62%. The Caire ICH system also improved the radiologist's ability to accurately identify the ICH subtypes present.

CONCLUSION

The Caire ICH device significantly improves the performance of a cohort of radiologists. Such a device has the potential to be a tool that can improve patient outcomes and reduce misdiagnosis of ICH.

Predicting poor response to anti-coagulation therapy in cerebral venous thrombosis using a simple clinical-radiological score

BACKGROUND

Multiple studies have attempted to determine predictors of poor clinical outcomes in cerebral venous thrombosis (CVT). Fewer studies target to identify predictors of poor response to anticoagulation therapy in CVT.

OBJECTIVE

We aimed to determine the predictors of poor clinical response to therapeutic anticoagulation in patients with acute CVT.

METHODS

We performed a retrospective analysis of patients therapeutically anticoagulated for acute CVT. We defined poor clinical outcomes as death, need for mechanical thrombectomy during the hospitalization, or a modified Rankin Scale (mRS) > 3 at clinical follow-up. Bivariate and multivariate analyses identified factors associated with poor outcomes in anticoagulated patients for acute CVT, and we used the identified factors to create the PRACT-CVT (Poor Response to Anticoagulation Therapy in CVT) score.

RESULTS

We included 109 patients anticoagulated with acute CVT. The mean patient age was 37 years old (SD 19); nine patients were > 65 years, ten patients were < 10 years, and 64 (59%) were female. Twenty-one (19%) patients had poor clinical outcomes. Age > 65 or < 10 years (OR: 3.16, 95% CI: 1.06-9.44), a GCS ≤ 12 upon presentation (OR: 19.2, 95% CI: 4.05-91.4), focal motor deficits at admission (OR: 5.03, 95% CI: 1.64-15.44), clinical deterioration following admission (OR: 28.18, CI: 4.81-164.86), seizures following admission (OR: 5.59, 95% CI: 1.27-24.51), evidence of brain bleeding/ischemia on admission (OR: 4.67, 95% CI: 1.42-15.34), involvement of the superior sagittal sinus (OR: 3.88, CI: 1.33-11.32), or involvement of both transverse sinuses (OR: 3.87, 95% CI: 1.01-14.90) predicted poor clinical outcome despite therapeutic anticoagulation. A PRACT-CVT score (0-22 points) of ≥ 7 points provided a sensitivity of 71% and a specificity of 95% for predicting poor clinical outcomes with anticoagulation alone.

CONCLUSION

Patients with acute CVT aged > 65 or < 10 years old, presenting with a GCS ≤ 12, with focal motor deficits, showing clinical deterioration after admission, having clinical seizures during hospitalization, with brain bleeding/ischemia on initial neuroimaging, involvement of the superior sagittal sinus, or involvement of both transverse sinuses had poor response to anticoagulation. Clinicians may employ the PRACT-CVT score to predict poor response to anticoagulation for acute CVT.

Demographic and institutional predictors of stroke hospitalization mortality among adults in the United States

Introduction

Stroke remains a primary source of functional disability and inpatient mortality in the United States (US). Recent evidence reveals declining mortality associated with stroke hospitalizations in the US. However, data updating trends in inpatient mortality is lacking. This study aims to provide a renewed inpatient stroke mortality rate in a national sample and identify common predictors of inpatient stroke mortality.

Methods

In this cross-sectional study, we analyzed data from a nationwide database between 2010 and 2017. We included patient encounters for both ischemic (ICD9 433–434, ICD10 I630–I639) and hemorrhagic stroke (ICD9 430–432, ICD10 I600–I629). We performed an annual comparison of in-hospital stroke mortality rates, and a cross-sectional analytic approach of multiple variables identified common predictors of inpatient stroke mortality.

Results

Between 2010 and 2017, we identified 518,185 total stroke admissions (86.6% ischemic stroke and 13.4% hemorrhagic strokes). Stroke admissions steadily increased during the studied period, whereas we observed a steady decline in in-hospital mortality during the same time. The inpatient stroke mortality rate gradually declined from 4.8% in 2010 (95% CI 4.6–5.1) to 2.1% in 2017 (95% CI 2.0–2.1). Predictors of higher odds of dying from ischemic stroke were female (OR 1.059, 95% CI 1.015–1.105, p = 0.008), older age (OR 1.028, 95% CI 1.026–1.029, p < 0.001), and sicker patients (OR 1.091, 95% CI 1.089–1.093, p < 0.001). Predictors of higher odds of dying from hemorrhagic stroke were Hispanic ethnicity (OR 1.459, 95% CI 1.084–1.926, p < 0.001), older age (OR 1.021, 95% CI 1.019–1.023, p < 0.001), and sicker patients (OR 1.042, 95% CI 1.039–1.045, p < 0.001). All census regions and hospital types demonstrated improvements in in-hospital mortality.

Conclusion

This study identified a continuous declining rate in in-hospital mortality due to stroke in the United States, and it also identified demographic and hospital predictors of inpatient stroke mortality.

•

Stroke remains a leading cause of morbidity and mortality in the United States

•

Stroke hospitalization mortality trends are important to guide efforts in acute stroke care

•

Vascular risk factors are still prevalent in the population admitted due to stroke and continue to be associated with higher odds of death

•

There are important regional disparities in stroke hospitalization deaths in the United States

•

Hospital characteristics influence odds of death from a stroke independent of stroke etiology

Antihypertensive use for stroke in United States emergency departments

OBJECTIVE

Timely emergency department (ED) control of hypertension in the acute phase of stroke is associated with improved outcomes. It is unclear how emergency physicians use antihypertensive medications to treat severe hypertension associated with stroke. We sought to determine national patterns of antihypertensive use associated with ED visits for stroke in the United States.

METHODS

We analyzed data from the National Hospital Ambulatory Medical Care Survey (NHAMCS) 2008-2017. We included ED visits associated with ischemic stroke (ICD9 433-434, ICD10 I630-I639) or hemorrhagic stroke (ICD9 430-432, ICD10 I600-I629). We estimated the number and proportions of stroke ED visits with triage blood pressure meeting treatment thresholds (triage systolic blood pressure [SBP] ≥180 mm Hg). We identified the frequency of antihypertensive use, as well as the most commonly used agents.

RESULTS

Between 2008-2017, of a total 135,012,819 ED visits, 619,791 were associated with stroke (78.3% ischemic strokes and 21.7% hemorrhage strokes). Of all stroke visits, 21.8% received antihypertensive medications. Of the identified visits, 9.0% (95% confidence interval [CI] = 6.0%, 13.1%) ischemic stroke visits and 58.2% (95% CI = 49.0%, 66.9%) hemorrhagic stroke visits met criteria for BP reduction. A total of 47.6% (95% CI = 29.1%, 66.7%) of eligible ischemic stroke visits and 41.5% (95% CI = 30.5%, 53.3%) of eligible hemorrhagic strokes visits received antihypertensives. The most common agents used in ischemic stroke were beta-blockers, calcium-channel blockers, and ACE inhibitors. The most common agents used in hemorrhagic stroke included calcium-channel blockers, beta-blockers, and vasodilators.

CONCLUSION

In this national sample, less than half of strokes presenting to the ED with hypertension received antihypertensive therapy.

Neurological manifestations of COVID-19 and other coronavirus infections: A systematic review

BACKGROUND

Increasing research reports neurological manifestations of COVID-19 patients. SARS-CoV-2 shares homology with other human coronaviruses that have also had nervous system involvement.

OBJECTIVE

To review the neurological aspects of SARS-cov2 and other coronavirus, including transmission pathways, mechanisms of invasion into the nervous system, and mechanisms of neurological disease.

METHODS

We conducted a systematic review of articles in PubMed, SCOPUS and EMBASE data bases. Reviewed evidence is presented in sections of this manuscript which includes pathogenesis, neuro-invasion, encephalitis, Guillain-Barré, ADEM, multiple sclerosis, polyneuropathy, and cerebrovascular disease.

RESULTS

A total 67 studies were included in the final analysis of experimental studies, case reports, series of cases, cohort studies, and systematic reviews related to neurological manifestations of SARS- CoV-2 and other human coronavirus infections. The SARS-CoV-2 receptor is expressed in the nervous system. Common reported symptoms included hyposmia, headaches, weakness, altered consciousness. Encephalitis, demyelination, neuropathy, and stroke have been associated with COVID-19. Infection through the cribriform plate and olfactory bulb and dissemination through trans-synaptic transfer are some of the mechanisms proposed. Invasion of the medullary cardiorespiratory center by SARS-CoV-2 may contribute to the refractory respiratory failure observed in critically-ill COVID-19 patients.

CONCLUSION

An increasing number of reports of COVID-19 patients with neurological disorders add to emergent experimental models with neuro-invasion as a reasonable concern that SARS-CoV-2 is a new neuropathogen. How it may cause acute and chronic neurologic disorders needs to be clarified in future research.

Outcome of Patients with Venous Thromboembolism and Factor V Leiden or Prothrombin 20210 Carrier Mutations During the Course of Anticoagulation

BACKGROUND

Individuals with factor V Leiden or prothrombin G20210A mutations are at a higher risk to develop venous thromboembolism. However, the influence of these polymorphisms on patient outcome during anticoagulant therapy has not been consistently explored.

METHODS

We used the Registro Informatizado de Enfermedad TromboEmbólica database to compare rates of venous thromboembolism recurrence and bleeding events occurring during the anticoagulation course in factor V Leiden carriers, prothrombin mutation carriers, and noncarriers.

RESULTS

Between March 2001 and December 2015, 10,139 patients underwent thrombophilia testing. Of these, 1384 were factor V Leiden carriers, 1115 were prothrombin mutation carriers, and 7640 were noncarriers. During the anticoagulation course, 160 patients developed recurrent deep vein thrombosis and 94 patients developed pulmonary embolism (16 died); 154 patients had major bleeding (10 died), and 291 patients had nonmajor bleeding. On multivariable analysis, factor V Leiden carriers had a similar rate of venous thromboembolism recurrence (adjusted hazard ratio [HR], 1.16; 95% confidence interval [CI], 0.82-1.64), half the rate of major bleeding (adjusted HR, 0.50; 95% CI, 0.25-0.99) and a nonsignificantly lower rate of nonmajor bleeding (adjusted HR, 0.66; 95% CI, 0.43-1.01) than noncarriers. Prothrombin mutation carriers and noncarriers had a comparable rate of venous thromboembolism recurrence (adjusted HR, 1.00; 95% CI, 0.68-1.48), major bleeding (adjusted HR, 0.75; 95% CI, 0.42-1.34), and nonmajor bleeding events (adjusted HR, 1.10; 95% CI, 0.77-1.57).

CONCLUSIONS

During the anticoagulation course, factor V Leiden carriers had a similar risk for venous thromboembolism recurrence and half the risk for major bleeding compared with noncarriers. This finding may contribute to decision-making regarding anticoagulation duration in selected factor V Leiden carriers with venous thromboembolism.

The Clinical Course of Venous Thromboembolism May Differ According to Cancer Site

BACKGROUND

We hypothesized that the clinical course of venous thromboembolism in patients with active cancer may differ according to the specificities of primary tumor site.

AIM AND METHODS

We used data from RIETE (international registry of patients with venous thromboembolism) to compare the clinical venous thromboembolism-related outcomes during the course of anticoagulation in patients with one of the 4 more frequent cancers (breast, prostate, colorectal, or lung cancer).

RESULTS

As of September 2014, 3947 cancer patients were recruited, of whom 938 had breast, 629 prostate, 1189 colorectal, and 1191 lung cancer. Overall, 55% had metastatic disease (42%, 36%, 53%, and 72%, respectively). During the course of anticoagulant therapy (mean duration, 139 days), the rate of thromboembolic recurrences was similar to the rate of major bleeding in patients with breast (5.6 [95% confidence interval (CI), 3.8-8.1] vs 4.1 [95% CI, 2.7-5.9] events per 100 patient-years) or colorectal cancer (10 [95% CI, 7.6-13] vs 12 [95% CI, 9.4-15] per 100 patient-years). In contrast, in patients with prostate cancer, the rate of venous thromboembolic recurrences was half the rate of major bleeding (6.9 [95% CI, 4.4-10] vs 13 [95% CI, 9.2-17] events per 100 patient-years), whereas in those with lung cancer, the rate of thromboembolic recurrences was twofold higher than the rate of major bleeding (27 [95% CI, 22-23] vs 11 [95% CI, 8.6-15] per 100 patient-years).

CONCLUSIONS

Significant differences in the clinical profile of venous thromboembolic-related outcomes were observed according to the site of cancer. These findings suggest the development of cancer-specific anticoagulant strategies as an area for further research.

A prognostic score to identify low-risk outpatients with acute deep vein thrombosis in the lower limbs

BACKGROUND

No prior studies have identified which patients with deep vein thrombosis in the lower limbs are at a low risk for adverse events within the first week of therapy.

METHODS

We used data from the Registro Informatizado de la Enfermedad TromboEmbólica (RIETE) to identify patients at low risk for the composite outcome of pulmonary embolism, major bleeding, or death within the first week. We built a prognostic score and compared it with the decision to treat patients at home.

RESULTS

As of December 2013, 15,280 outpatients with deep vein thrombosis had been enrolled. Overall, 5164 patients (34%) were treated at home. Of these, 12 (0.23%) had pulmonary embolism, 8 (0.15%) bled, and 4 (0.08%) died. On multivariable analysis, chronic heart failure, recent immobility, recent bleeding, cancer, renal insufficiency, and abnormal platelet count independently predicted the risk for the composite outcome. Among 11,430 patients (75%) considered to be at low risk, 15 (0.13%) suffered pulmonary embolism, 22 (0.19%) bled, and 8 (0.07%) died. The C-statistic was 0.61 (95% confidence interval [CI], 0.57-0.65) for the decision to treat patients at home and 0.76 (95% CI, 0.72-0.79) for the score (P = .003). Net reclassification improvement was 41% (P < .001). Integrated discrimination improvement was 0.034 for the score and 0.015 for the clinical decision (P < .001).

CONCLUSIONS

Using 6 easily available variables, we identified outpatients with deep vein thrombosis at low risk for adverse events within the first week. These data may help to safely treat more patients at home. This score, however, should be validated.