Author Of 2 Presentations
P0856 - Collecting real world MRI in MS: preliminary results from FlywheelMS (ID 1226)
Abstract
Background
Real-world evidence can be used to better characterize the course of multiple sclerosis (MS), care provision and outcomes in clinical practice. Magnetic resonance imaging (MRI) that occurs in the context of usual care is an important source of information that can inform clinical decision-making. Guidelines exist to enhance the clinical impact of routine MRI in MS, but it is unclear whether MRIs acquired as part of routine care in the United States adhere to these guidelines.
Objectives
To describe the clinical routine brain MRIs from patients with MS across different US imaging sites.
Methods
FlywheelMS is a novel patient-centered study that aims to extract and digitize health information not readily available in existing electronic health records of patients with MS. Up to 5,000 consenting adults with a confirmed MS diagnosis will be enrolled. Brain MRI data were retrieved, and summary statistics were computed to describe the sessions, imaging sites, scanner field strengths and slice thickness of T1-weighted and FLAIR (fluid-attenuated inversion recovery) images. Longitudinal acquisition consistency (i.e. MRIs acquired from the same center with the same scanner) was also assessed.
Results
Out of 2,389 patients enrolled, 1555 brain MRI data were retrieved from the first 492 patients (female, 81%; mean age at consent, 49±11 years). The mean number of MRI sessions per patient was 3.2±2.4, and data were captured between 1999 and 2018. Sessions were acquired at 598 different imaging sites, using mainly 1.5T scanners (61.3%), followed by 3T (32.7%) and lower field-strength magnets (3.4%; not available, 2.6%). The mean slice thickness of T1-weighted (3.1±1.7 mm) and FLAIR images (3.1±1.3 mm) was similar. Of the 352 patients (72%) that had more than one MRI session, 85 (24.1%) had consistent acquisition (i.e. same site/scanner), 153 (43.5%) had one site or scanner change, and 114 (32.4%) had more than one site and/or scanner change.
Conclusions
The novel, patient-centered approach of FlywheelMS can successfully extract imaging data from medical records of patients with MS across US imaging sites. These data will help us in describing the clinical routine MRI, determining the compliance to guidelines and understanding which measure (e.g. lesion volume and/or atrophy) could be potentially extracted from MRI data.
P0875 - FlywheelMS: The prevalence of multiple sclerosis subtypes in digitized health records (ID 1882)
Abstract
Background
Data generated from electronic health records (EHRs) offer insight into real-world care of people with multiple sclerosis (MS). Data extracted most readily from EHRs include templated or administrative health information (e.g., MS International Classification of Diseases codes). However, clinical data like disease subtype and characteristics are unlikely to be captured systematically. FlywheelMS is a novel patient-centered study with the aim of digitizing health records of patients with MS and extracting information not readily available in existing EHRs.
Objectives
To evaluate patient characteristics and the prevalence of MS subtypes (i.e., relapsing-remitting MS [RRMS], secondary progressive MS [SPMS], primary progressive MS [PPMS], progressive relapsing MS [PRMS]) in the FlywheelMS cohort and to compare them with existing real-world data sources.
Methods
Adults with MS are recruited across the US via advocacy groups, social media and healthcare professionals. Supervised machine learning with human curation is used to retrieve, digitize and abstract medical records, which are collected as far back as are available and prospectively up to 5 years after enrollment. The most recent non-negated MS subtype from neurology visit records was used as a proxy for the prevalent subtype. Summary statistics were calculated and compared with other MS cohorts.
Results
As of March 1, 2020, 2,389 patients with MS with 24,362 neurology visits across 3,093 neurologists have enrolled in FlywheelMS. Data on MS subtype were available for 973 patients (40.7%); this proportion will increase as abstractions continue. RRMS accounted for 78.9% of patients, followed by SPMS (12%), PPMS (7.3%) and PRMS (1.7%). These findings were comparable to the MSBase Registry (RRMS=76.9%, SPMS=13.0%, PPMS=8.0%, PRMS=2.2%; Kister et al., J Neurol Sci 2012) and NARCOMS Registry (RRMS=65.6%, SPMS=25.1%, PPMS=9.3%; Salter et al., Mult Scler 2018). Mean [SD] age at mention of MS subtype and percent female distribution were as follows: RRMS (46.4 [10.9] years, 80.4%), SPMS (56.4 [9.6] years, 81.2%), PPMS (53.9 [10.7] years, 62.0%), PRMS (mean 53.9 [5.5] years, 82.4%).
Conclusions
The prevalence of MS subtypes in the digitized health records of patients in FlywheelMS was comparable to other real-world data sources. Digitizing and machine-learning guided abstraction of patient healthcare records in MS yields important data about clinical features not readily available in other EHR data sets.
Presenter Of 1 Presentation
P0875 - FlywheelMS: The prevalence of multiple sclerosis subtypes in digitized health records (ID 1882)
Abstract
Background
Data generated from electronic health records (EHRs) offer insight into real-world care of people with multiple sclerosis (MS). Data extracted most readily from EHRs include templated or administrative health information (e.g., MS International Classification of Diseases codes). However, clinical data like disease subtype and characteristics are unlikely to be captured systematically. FlywheelMS is a novel patient-centered study with the aim of digitizing health records of patients with MS and extracting information not readily available in existing EHRs.
Objectives
To evaluate patient characteristics and the prevalence of MS subtypes (i.e., relapsing-remitting MS [RRMS], secondary progressive MS [SPMS], primary progressive MS [PPMS], progressive relapsing MS [PRMS]) in the FlywheelMS cohort and to compare them with existing real-world data sources.
Methods
Adults with MS are recruited across the US via advocacy groups, social media and healthcare professionals. Supervised machine learning with human curation is used to retrieve, digitize and abstract medical records, which are collected as far back as are available and prospectively up to 5 years after enrollment. The most recent non-negated MS subtype from neurology visit records was used as a proxy for the prevalent subtype. Summary statistics were calculated and compared with other MS cohorts.
Results
As of March 1, 2020, 2,389 patients with MS with 24,362 neurology visits across 3,093 neurologists have enrolled in FlywheelMS. Data on MS subtype were available for 973 patients (40.7%); this proportion will increase as abstractions continue. RRMS accounted for 78.9% of patients, followed by SPMS (12%), PPMS (7.3%) and PRMS (1.7%). These findings were comparable to the MSBase Registry (RRMS=76.9%, SPMS=13.0%, PPMS=8.0%, PRMS=2.2%; Kister et al., J Neurol Sci 2012) and NARCOMS Registry (RRMS=65.6%, SPMS=25.1%, PPMS=9.3%; Salter et al., Mult Scler 2018). Mean [SD] age at mention of MS subtype and percent female distribution were as follows: RRMS (46.4 [10.9] years, 80.4%), SPMS (56.4 [9.6] years, 81.2%), PPMS (53.9 [10.7] years, 62.0%), PRMS (mean 53.9 [5.5] years, 82.4%).
Conclusions
The prevalence of MS subtypes in the digitized health records of patients in FlywheelMS was comparable to other real-world data sources. Digitizing and machine-learning guided abstraction of patient healthcare records in MS yields important data about clinical features not readily available in other EHR data sets.