Displaying One Session

Scientific Communications
Session Type
Scientific Communications
Room
Hall K
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM

COST-EFFECTIVENESS OF A QUALITY IMPROVEMENT PROJECT REDUCING DOOR-TO-NEEDLE TIMES IN STROKE THROMBOLYSIS

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
02:30 PM - 02:40 PM

Abstract

Background And Aims

Rapid revascularization in acute ischemic stroke is crucial to reduce the burden of stroke including societal costs. A quality improvement project including protocol revision and simulation-based training was followed by a considerable reduction in median door-to-needle time (27 to 13 min) and improved patient outcomes in stroke thrombolysis at our centre. Here, we present a formal cost-effectiveness analysis of the quality improvement project.

Methods

Costs for initiating and maintaining quality improvement were assessed using recognized frameworks for cost reporting in quality improvement and simulation-based training. Effectiveness was calculated from previously published outcome measures. Cost-effectiveness was presented as annual costs per minute door-to-needle time reduction and costs per averted death over an estimated 5-year period. Costs were calculated including and excluding donated (unpaid) time and are presented in Euros (€).

Results

Including costs of donated time, total costs were approximately 64 200 € during the first year, and 20 200 € the 5th year (Table 1). The estimated effects were a mean reduction of 17.7 min per patient and 6,36 averted deaths annually. Estimated spending per death averted in the 5th year was 1000 €. The estimated cost-effectiveness over time is presented in Figure 1A and B.

table1.png

figure1a.png

figure1b.png

Conclusions

Costs of quality improvement declined quickly while the effects could be maintained, resulting in higher cost-effectiveness over time. Estimated spending per death averted was below 2000 € during year 3-5. Economic consequences of quality improvement projects and simulation-based training interventions are rarely reported. The presented cost-effectiveness data might help guide decisionmakers planning to implement similar interventions.

Trial Registration Number

Not applicable

Hide

DOOR TO DECISION IN UNDER 30! A NATIONAL QUALITY IMPROVEMENT COLLABORATIVE TO IMPROVE THE QUALITY OF CARE FOR PATIENTS PRESENTING WITH ACUTE ISCHAEMIC STROKE

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
02:40 PM - 02:50 PM

Abstract

Background And Aims

Intravenous thrombolysis and thrombectomy are standard of care in appropriately selected patients presenting with acute ischaemic stroke (AIS). Time is brain, therefore reducing time to treatment is critical to improve patient outcomes.

The National Thrombectomy Service (NTS), in conjunction with the RCSI & the RCPI, developed a national collaborative Quality Improvement programme ‘Door to Decision in under 30!’

The aim of this collaborative was to utilise QI methodology and shared learning to improve the efficiency of acute stroke care in the initial period when a patient presents to hospital with an AIS. The target was to reduce to <30 minutes the time from arrival at the hospital (door) to the time of a decision regarding thrombectomy.

Methods

The collaborative used a modified breakthrough series approach, based upon the IHI Breakthrough Series (2003).

Multi-disciplinary teams from 12 acute hospitals across Ireland attended learning sessions and completed a QI project over 10 months. qi methods image.png

Supported by a QI Facilitator, each team mapped processes, identified areas for improvement, generated change ideas and tested interventions using PDSA cycles.

qi interventions image.png

Results

All hospitals have realised improvements in median door to decision times. Three hospitals have attained the target door to decision time of <30mins (median), with a further five hospitals currently achieving <40mins median door to decision time.

Conclusions

By engaging and supporting teams in a collaborative QI programme across multiple sites, novel ideas can be readily shared, tested and implemented, resulting in improved quality of care, patient safety and outcomes from stroke.

Hide

TECHNOLOGY FOR IN SILICO STROKE TRIALS

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
02:50 PM - 03:00 PM

Abstract

Group Name

The INSIST investigators

Background And Aims

In silico trials are patient-specific simulations on a cohort of virtual patients to improve the development and evaluation of medical devices, drugs and treatment. These in silico trials can refine or reduce cost and partially replace current clinical trials or animal experimentation. In silico trials are composed of multiple computational models on different scales. Integrating these models is not straightforward; several challenges have to be overcome for a successful in silico trial. We present the design and implementation of an in silico trial for treatment of acute ischemic stroke.

Methods

The in silico trial is implemented as an event-based simulation for stroke and treatment, coupled to a continuous tissue model. The simulation includes models of arterial blood flow, tissue perfusion, oxygen transport, thrombectomy, thrombolysis, and tissue state. A statistical population model generates cohorts of virtual patients. Sophisticated 3-D models are replaced by surrogate and statistical models.

Results

The in silico trial is designed to be compartmental to aid development and reproducibility; the computational models are installed in their own Docker containers. The containers contain all the requirements of the models thereby increasing reproducibility, flexibility and user friendliness.

Conclusions

The design and implementation for an in silico trial for treatment of acute ischemic stroke is presented. Ways to overcome some of the challenges and difficulties in setting up such in silico (stroke) trials are discussed.

Trial Registration Number

Not applicable

Hide

CAN STROKE UNIT CARE BE IMPLEMENTED IN UNDER-RESOURCED SETTINGS IN RWANDA? RESULTS FROM THE ORGANIZED STROKE CARE ACROSS INCOME LEVELS (OSCAIL) STUDY

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:00 PM - 03:10 PM

Abstract

Group Name

The Organized Stroke Care Across Income Levels (OSCAIL) Study Collaborators

Background And Aims

Stroke unit care has become established as the central component of a modern stroke service to improve patient outcomes, but it requires resources which may not be available in under-resourced settings. We explored whether key elements of stroke unit care could be implemented in Rwanda.

Methods

Local champions led stroke patient recruitment and data collection at two urban hospitals, for a before and after implementation trial. Data were collected on usual stroke care practices for common key performance indicators (KPIs), then feedback and training were provided to address areas requiring improvement. Interviews were conducted with hospital directors.

Results

We recruited 106 participants for each of the two study phases. A shortage of infrastructure and personnel with specific training were common reasons for not performing KPIs. The mean KPIs implementation rate was 45.9% and 55.3% before and after the intervention respectively. After case mix adjustment for stroke severity, stroke type and delay to hospital, we found consistent trends associating the implementation intervention with both an increase in recording of KPIs, and better patient outcomes. However, the results were statistically significant only for the use of standardized assessment tools (OR: 2.98, 95% CI 1.36-6.51), swallowing assessment (OR: 5.73, 2.08-15.74), mobilization (OR: 2.30, 1.16-4.56) and multidisciplinary team meetings (OR: 9.04, 2.74-29.86). In-hospital (OR: 2.97, 1.25-7.05) and 3-month (OR: 2.30, 1.10-4.78) survival also significantly improved. One hospital is establishing the first geographic stroke unit in Rwanda.

Conclusions

Stroke KPIs implementation and patient outcomes in Rwanda can be improved with simple intervention but further studies are needed.

Trial Registration Number

Not applicable

Hide

IMPACT OF MEDICAL APPLICATION BASED POST STROKE CARE STRATEGY (MAPSS) ON QUALITY OF LIFE AND CAREGIVER BURDEN

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:10 PM - 03:20 PM
Presenter

Abstract

Background And Aims

Since trained caregivers are scarce in low middle income countries, post stroke complications, patients quality of life and caregiver burden are major challenges. App based care strategies may provide low cost solution for these challenges.

We aimed to investigate impact of MAPSS on delivering post stroke care, patients quality of life and caregiver burden.

Methods

Patients within 3 months of stroke with mRS>3, caregivers with internet enabled smartphone willing to use the MAPSS as healthcare tool were prospectively randomised in open labelled blinded endpoint (PROBE). Control group was given routine care with booklet and MAPSS group were given addition stroke home care app and trained to use the app. Blinded three months followup was done.

Results

166 patients (83 each group; mean age: 54.98±15.76 years; 26.5% females; 53% ischemic strokes) were enrolled. Bedridden patients (mRS 4-5) had median NIHSS 13 (Both groups matched). At 3 months, controls had higher composite post stroke complication (56.4% vs 33.8%; P=0.004); and bedsores (15.4% vs 5%; P=0.031). Median HRQOL score showed better QOL in MAPSS [68(IQR: 50-80) vs 50(IQR: 38-61); P=0.001]. Caregivers in control group had severe stress (27.6% vs 5.7%; P=0.001); Median Caregiver Stress Index(CSI) showed reduced stress in MAPSS [9(IQR: 5-13) vs 15(IQR: 13-20); P=0.001].

Conclusions

MAPPS was associated with lesser post stroke complications, better QOL in severe strokes and reduced caregiver stress.

Hide

UTILIZING MACHINE LEARNING TECHNIQUES TO EVALUATE QUALITY OF CARE FROM THE REGISTRY OF STROKE CARE QUALITY (RES-Q) DATA

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:20 PM - 03:30 PM

Abstract

Background And Aims

Quality of stroke care varies between countries, but it is difficult to compare inter-country for overall performance. Machine learning clustering algorithms provide optimal and unbiased analyses for multi-objective optimization problems. Using this algorithm, the aim was to group countries with similar quality of care based on selected performance metrics.

Methods

Using the Python programming language, we applied the hierarchical agglomerative clustering algorithm from the scikit-learn library to our data, encompassing 38 countries with at least 30 enrolled patients in January - April 2019. Using the Elbow method, we optimized the analysis to 7 clusters/groups of countries based on 3 key performance metrics with the least amount of entropy: door-to-needle time, total recanalization rate, and rate of patients obtaining CT or MRI under 1 hour after admission into hospital.

Results

The hierarchical agglomerative clustering algorithm distributed countries into 7 clusters, based on data from 20,000 patients. 6 out of the 7 clusters contained countries from different geographic regions. The first cluster consisted of primarily Eastern European countries, with additional Asian countries. Western European countries were clustered with outlier African countries.

resq_clustering.png

Conclusions

The clustering algorithm was able to group countries with similar quality of care, utilizing an unbiased computational method. Countries within each cluster were often from the same geographic region, such as Eastern Europe or Southeastern Europe. However, these clusters also included countries from outside of the geographic grouping, which was an expected result. The validity of this approach requires further investigation.

Trial Registration Number

Not applicable

Hide

AI POWERED DIAGNOSIS OF ACUTE ISCHEMIC STROKE - LVO DETECTION FROM NATURAL SPEECH AND FACIAL ANALYSIS: A CROSS-CENTER STUDY

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:30 PM - 03:40 PM

Abstract

Background And Aims

CVA Flow (CVAid LTD. Tel-Aviv, Israel) is a novel multi-modal remote, AI decision-support tool for stroke diagnosis using mobile devices. In previous work, we showed that CVAid’s AI facial analysis objectively determine facial palsy severity, stroke probablity and predicted LVO. In this work, we tested a novel component of speech analysis for LVO prediction.

Methods

Stroke patients and healthy controls from 2 different stroke centers were included. Patient’s face and speech were recorded while performing the NIHSS test, and audio signal was extracted, followed by automatic analysis of speech articulation at the sub-phonetic level, leveraged by machine learning analysis. The auditory data was then classified for likelihood of LVO based on comprehensive AI analysis.

Results

88 Stroke patients (50 with confirmed LVO) and 43 controls were included: mean age 69.8 years old, mean time from symptom onset 25.3 hours, median NIHSS of 5.00. Optimal sensitivity of LVO detection was of 89% and specificity 73%. Analysis of facial features done separately resulted with average sensitivity of 91% and specificity of 86% for LVO detection.

Conclusions

CVAid’s smartphone-based data acquisition system, powered by AI solution is sensitive to detect LVO, even when using speech signal alone. Integrating among additional modalities and NIHSS tests will result in a highly sensitive decision-support tool for pre-hospital teams. To the best of our knowledge, there are no published reports on automatic analysis of natural speech at ER for LVO prediction.

Hide

DETECTION OF INTRACRANIAL HEMORRHAGE USING PORTABLE, BEDSIDE, LOW-FIELD MAGNETIC RESONANCE IMAGING IN AN INTENSIVE CARE UNIT

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:40 PM - 03:50 PM

Abstract

Background And Aims

Radiographic diagnosis of intracranial hemorrhage (ICH) is a critical determinant of stroke care pathways. Recent advances in low-field MRI have made it possible to obtain clinically useful imaging at the point-of-care (POC). Our aim was to obtain preliminary data regarding the ability of a bedside POC MRI scanner to detect ICH.

Methods

We studied 52 patients (42% female, ages 21-96) with a diagnosis of ICH (n=27) or ischemic stroke (n=25). Five neurologists independently evaluated T2W and FLAIR exams acquired on a 64 mT, portable bedside MRI system. Presence of ICH was determined by majority consensus. Kappa coefficients (κ) were computed as a measure of inter-rater agreement. Ground truth was obtained from the clinical report of the closest conventional imaging study (19.6 ± 11.8 hours).

Results

Intraparenchymal hemorrhage (IPH) volume ranged from 2 cc to 65 cc. Subdural hemorrhage (SDH) thickness ranged from 7 mm to 28 mm. Exams were acquired within 7 days of symptom onset (53.7 ± 24.8 hours). Overall diagnostic accuracy was 88% with agreement among raters (κ = 0.64, p<0.001). ICH sensitivity and specificity were 76% and 100%, respectively. When limited to supratentorial IPH, overall accuracy was 98% and sensitivity was 95%. Sensitivity for SDH was 80%. Intraventricular hemorrhage and midline shift were detected with 100% sensitivity. Results were consistent across sequential exams of the same patient.

figure.png

Conclusions

We present the first use of a portable, low-field MRI system to detect hemorrhagic stroke at the bedside. Further work is needed to evaluate this approach in the hyperacute setting.

Trial Registration Number

Not applicable

Hide

DEEP NEURAL NETWORK-BASED DETECTION OF BRAIN ANEURYSMS USING CT ANGIOGRAPHY

Session Type
Scientific Communications
Date
07.11.2020, Saturday
Session Time
02:30 PM - 04:00 PM
Room
Hall K
Lecture Time
03:50 PM - 04:00 PM

Abstract

Background And Aims

Brain Aneurysms (BAs) may cause a life-threatening intracranial hemorrhage. Machine learning algorithms have been used to detect large vessel occlusion and other vascular brain conditions. We developed an algorithm using deep neural network to detect and assist BAs.

Methods

We developed an algorithm using 3D convolutional neural network modeled as U-net to detect BAs. We used positive and negative CTAs from two institutions from 2015-2017. The data was annotated an experienced neuroradiologist. The algorithm construction initially used 179 CTA datasets containing 230 BAs as a training set. For the algorithm optimization, 528 CTAs containing 674 BAs and 2400 normal scans were used to validate the algorithm. We performed a blind test on the algorithm to assess its accuracy on the detection of BAs using a test set of 300 positive CTAs with BAs and 900 negative scans as controls. We used ROC curves and Pearson correlation tests to assess the algorithm.

Results

We are submitting preliminary results of a blind test of 50 positive CTAs and 150 controls. The aneurysm mean size was 8.1752 mm/median 6.95 mm + 3.70mm in both anterior and posterior locations . The algorithm achieved a sensitivity of 92% and a specificity of 94% (AUC 0.983). At the the conference, we aim to present the complete analysis.

Conclusions

The Viz.ai aneurysm algorithm was able to accurately detect the majority of brain aneurysms from our dataset as well as able to consistently report the negative scans. Further training should improve the accuracy of the algorithm particularly on small aneurysm sizes.

Hide