Broderick, Joseph P.

Introduction: Intravenous alteplase (tissue plasminogen activator) has been shown to be cost-effective due to savings in long-term disability. In October of 2005, an increased DRG payment to hospitals for alteplase-treated stroke patients was introduced. We sought to describe the trends in the cost of alteplase over time, in comparison to trends in hospital reimbursement in the United States. Methods: Utilizing publicly available information on the Centers for Medicare and Medicaid Services (CMS) website (www.cms.gov), we obtained CMS quarterly payment amounts between 1/1/2005-10/1/2014 for alteplase, listed as “alteplase recombinant, per mg.” CMS payment amounts are the manufacturer’s average sales price (ASP) + 6% until 4/2014, when it was lowered to + 4.3%. Estimates for DRG base payments were calculated within MEDPAR for FY 2006 (DRG 559) and 2013 (MS DRGs 61,62, and 63) as: (DRG relative weights) x (standardized operating and capital amount). The Consumer Price Index (CPI) was also queried for all prescription drugs, urban areas, during the same study period as reference. Results: The CMS payment amount for alteplase/mg was $30.50 in 1/2005, and $64.30 in 10/2014. Trends in the CMS payment amounts for alteplase increased by 111% between 2005-2014, shown in the Figure. The CPI for all prescription drugs increased by 30.2% in the same time frame. The base payment for alteplase-treated stroke admissions was $11,173 in 2006, and $12,064 in 2013, an 8% increase. Discussion: We found a striking increase in the cost of alteplase over the last decade, with a 100mg vial now with a CMS payment of ∼$6400, a more than 100% increase over ten years. During the same timeframe, the DRG base payment to hospitals increased by only 8%, and alteplase cost increased from 27% of the payment in 2006 to 53% in 2013. Researchers and stroke physicians should be aware of these changes in drug costs and their impact on cost-effectiveness analyses.

Background Meeting Recruitment goals in a research study can be challenging. Recruitment for pedigree-based research, in which data and DNA are collected from genetically important family members, is especially difficult. The Familial Intracranial Aneurysm (FIA) study is a multicenter international study whose goal is to identify genetic and other risk factors for the formation and rupture of intracranial aneurysms (IAs). The chosen sites for the study are major referral centers for the diagnosis and treatment of IAs. In this analysis, we sought to determine what factors caused a recruited family to be incomplete, and what impact the methods used in identifying families had on the successful completion of family enrollment. Methods Three recruitment strategies were utilized to identify and enroll families: prospective monitoring of every new case of IA, subarachnoid hemorrhage, or intracranial hemorrhage; retrospective review of IA cases; and advertisement on the Internet, in scientific journals, and at professional meetings. An unexpected additional source of recruitment was exposure to the study through the news media. To help with public accessibility to the study, a web site and toll free number were established. Families were eligible for the study if they fit one of 4 patterns: [1] ≥2 living siblings affected with IA; [2] ≥2 affected siblings, 1 living and 1 genotype reconstructed from living spouse and children; [3] ≥3 affected members, 2 living and 1 reconstructed; [4] ≥ 3 affected members, 1 living, ≥1 reconstructed, and ≥1 living connecting relative. Exclusion criteria included fusiform-shaped IA; IA part of an AVM; family history of polycystic kidney disease; Ehlers-Danlos, Marfan, or moyamoya syndrome; fibromuscular dysplasia; or failure to obtain informed consent. Results Between 8/2002 and 10/2007, 3771 IA patients were screened for family history, and 681 potential families were identified, of which 441 successfully completed enrollment. There were 240 families which did not complete enrollment, due to refusal to participate (97), inability to locate needed relatives (87), exclusion criteria (50), no reason documented (5), and duplication (1 recruited at two sites). Completion of study enrollment was not significantly different by recruitment category (p=.17): 191 of 311 prospectively-recruited (61%), 66 of 89 retrospective (74%), 95 of 145 advertisement (66%), and 89 of 136 methods not documented (65%) families successfully completed enrollment. Conclusion In our study, the major reason for family exclusion was the inability to enroll needed family members, through refusal or failure to locate. All three methods used to identify potential families were successful in completing eligible families. The combination of recruitment strategies can be effectively implemented in future studies.

Introduction: Low rates of enrollment into stroke clinical trials are an ongoing challenge for the stroke research community. We sought to retrospectively describe the association between medical eligibility and recruitment rates for two acute clinical trials. Methods: Within a population of 1.3 million, we screened local hospital admissions in 2010 using ICD-9 discharge codes 430-436, and performed chart review. Within the region are 15 adult acute care hospitals, all of which are covered by the same stroke team for clinical care and trial enrollment. Medical eligibility, the number of eligible patients/total number of patients, was calculated for the IMS-III trial and the CLEAR-ER trials using their inclusion/exclusion criteria. The number of patients enrolled in the trial per month of open enrollment was determined for each hospital. Statistical analysis included paired signed rank test and regression. Results: In 2010, 2035 ischemic strokes (IS) were hospitalized in the region. Regional medical eligibility was 1.7% for IMS III and 4.5% for CLEARER (p=0.001), overall regional hospital enrollment/month was 0.81 for IMS III and 1.84 for CLEAR-ER (p=0.001). Eligibility rates per hospital ranged from 0-9.5%, and recruitment rates per hospital/mo ranged from 0-0.38. Hospital-level eligibility and recruitment rates were significantly correlated (see figure for both trials combined, adjusted r2 = 0.57, p<0.001.) Discussion: Within our population, medical eligibility explained 57% of the variability in hospital-level enrollment rates for two acute IS trials. The use of population-based epidemiology in trial planning may identify optimal trial enrollment criteria, which must be weighed against proper patient selection. Many other factors impact enrollment rates, such as competing trials, which were not considered here. Further evaluation of our population-based eligibility rates in sites outside our region and with additional clinical trials is needed.

Background: Fundamental advances in stroke care will require pooling imaging phenotype data from multiple centers, to complement the current aggregation of genomic, environmental, and clinical information. Sharing clinically acquired MRI data from multiple hospitals is challenging due to inherent heterogeneity of clinical data, where the same MRI series may be labeled differently depending on vendor and hospital. Furthermore, the de-identification process may remove data describing the MRI series, requiring human review. However, manually annotating the MRI series is not only laborious and slow but prone to human error. In this work, we present a recurrent convolutional neural network (RCNN) for automated classification of the MRI series. Methods: We randomly selected 1000 subjects from the MRI-GENetics Interface Exploration study and partitioned them into 800 training, 100 validation and 100 testing subjects. We categorized the MRI series into 24 groups (see Table). The RCNN used a modified AlexNet to extract features from 2D slices. AlexNet was pretrained on ImageNet photographs. Since clinical MRI are 3D and 4D, a gated recurrent unit neural network was used to aggregate information from multiple 2D slices to make the final prediction. Results: We achieved a classification accuracy (correct/total cases) of 99.8%, 98.5% and 97.5% on the training, validation and testing set, respectively. The averaged F1-score (percent overlap between predicted cases and actual cases) over all categories were 99.8% 98.2% and 94.4% on the training, validation and testing set. Conclusion: We showed that automated annotation of MRI series by repurposing deep-learning techniques used for photographic image recognition tasks is feasible. Such methods can be used to facilitate high throughput curation of MRI data acquired across multiple centers and enable scientifically productive collaboration by researchers and, ultimately enhancing big data stroke research.

Introduction: White matter hyperintensity volume (WMHv) is an important and highly heritable cerebrovascular phenotype; however, manual or semi-automated approaches to clinically acquired MRI analysis hinder large-scale studies in acute ischemic stroke (AIS). In this work, we develop a high-throughput, fully automated WMHv analysis pipeline for clinical fluid-attenuated inversion recovery (FLAIR) images to facilitate rapid genetic discovery in AIS. Methods: Automated WMHv extraction from multiple subjects relies on significant pre-processing of medical scans, including co-registration of the images. To reduce the effects of anisotropic voxel sizes, each FLAIR image is upsampled using bi-cubic interpolation. Brain extraction is performed using RObust Brain EXtraction (ROBEX). Images are then registered to an in-house FLAIR template using Advanced Normalization Tools (ANTs). The spatial covariation of WMH is learned through principal component analysis (PCA) of manual outlines from 100 subjects. Areas of leukoaraiosis are identified and separated from other lesions using the PCA modes. Volumes are then computed using non-interpolated slices for each subject. Standard deviation (SD) in WMHv (9 subjects; 6 raters each) is calculated as a measure of variability. Good agreement between automated and manual outlines is assessed in 358 subjects (automated WMHv within 3SD of manual WMHv). Results: As part of the MRI - Gen etics I nterface E xploration (MRI-GENIE) study, WMHv were calculated on a set of 2703 FLAIR images of patients from 12 independent AIS cohorts (sites). Results are shown in Figure 1. Comparing manual and automated WMHv shows that 88% of the automated WMHv fall within 3 SD from the manual WMHv, suggesting good agreement. Conclusion: WMHv segmentation using a fully-automated pipeline for analysis of clinical MRIs is both feasible and accurate. Ongoing analysis of the extracted WMHv is expected to advance current knowledge of risks and outcomes in AIS.