Contributions
Abstract: P350
Type: Poster
Abstract Category: Clinical aspects of MS - 5 Epidemiology
Background: Electronic medical record (EMR) systems hold various data types, included easily extractable structured data and unstructured data requiring more sophisticated extraction procedures. Although research was never a primary goal of EMR system architecture, clinicians and researchers increasingly have the opportunity to leverage clinical data for research-based analyses. However, no matter how large the datasets are, health-related research requires high-quality information and validation. While previous EMR studies successfully identified patients and extracted data, none have assessed the quality of these EMR real-life data compared to well-curated research databases.
Goals: To
(1) use the University of California, San Francisco (UCSF) EMR system in order to identify patients with specific neurological disease, multiple sclerosis (MS), and extract their clinical data,
(2) compare EMR-extracted data with gold-standard data from research to demonstrate that information extracted from EMR is of high-quality and can be used for health-related research, and
(3) compare the EMR MS population characteristics to expected MS natural history.
Methods: A classification algorithm to identify MS patients in the UCSF Health System EMR was implemented under an IRB-approved protocol and clinical data algorithmically extracted and de-identified. Classification specificity and extraction performance were assessed by manual clinician record review. EMR data was compared to research cohort data in a subset of patients.
Results: We identified 4,142 MS patients with 95.9% sensitivity. Clinical data were extracted from free-text with high positive predictive values. We showed good concordance between EMR and research values for Expanded Disability Status Scale (EDSS) and Timed-25-foot walk (ICC=0.87 and 0.79 respectively) and for MS subtype (Kappa=0.65). We replicated several expected epidemiological features of MS: higher EDSS for patients with progressive forms compared to relapsing-remitting patients (p< 2.2x10-16) and for male compared to female patients (p=1.1x10-18); and an increase of EDSS with age at examination (p=1.3x10-229) and with disease duration (p=1.7x10-86). We also replicated an expected decrease in lymphocyte counts and increase in liver enzyme testing following fingolimod initiation.
Conclusions: Large real-world cohorts algorithmically extracted from EMR data are expanding opportunities for clinical research in MS
Disclosure:
Vincent Damotte received a postdoctoral fellowship from ARSEP foundation
Antoine Lizée has no disclosures.
Matthew Tremblay has no disclosures
Alisha Agrawal has no disclosures
Pouya Khankhanian has no disclosures
Adam Santaniello has no disclosures
Refujia Gomez has no disclosures
Robin Lincoln has no disclosures
Wendy Tang has no disclosures
Tiffany Chen has no disclosures
Nelson Lee has no disclosures
Pablo Villoslada reports personal fees from Hiedleberg Engineering, Roche, Novartis, Health Engineering, other from Bionure, Mint-Labs, Spire Bioventures, grants from Genzyme, outside this work.
Jill Hollenbach has no disclosures
Carolyn D. Bevan has no disclosures
Jennifer Graves reports grants from Race to Erase MS, Genentech, Biogen, outside this work
Riley Bove received research support from NIH and NMSS
Douglas Goodin has no disclosures
Ari J Green reports grants from National MS Society, Novartis, grants and other from Inception 5 Biosciences, other from Mediimune, personal fees and other from Perkins Coie LLP/Mylan outside the submitted work.
Sergio E. Baranzini has no disclosures
Bruce Cree reports personal fees from Abbvie, Biogen, EMD Serono, MedImmune, Novartis, Sanofi Genzyme, Shire, and Teva, outside the submitted work
Roland Henry reports grants from Hoffman La Roche, other from Novartis, ABBVIE, Genzyme, MEDDAY, outside this work
Stephen L. Hauser serves on the Scientific Advisory Boards for Annexon, Symbiotix, Bionure, Neurona, and Molecular Stethoscope. Dr. Hauser also has received travel reimbursement and writing assistance from F. Hoffman-La Roche Ltd. for CD20-related meetings and presentations.
Jeffrey M. Gelfand reports personal fees for consulting from Genentech and from medical legal consulting, and research support to UCSF from Quest Diagnostics, Genentech and MedDay outside this work
Pierre-Antoine Gourraud is the founder of Methodomics.com and has served as consultant for Sanofi-Genzyme and Biogen France SAS
Abstract: P350
Type: Poster
Abstract Category: Clinical aspects of MS - 5 Epidemiology
Background: Electronic medical record (EMR) systems hold various data types, included easily extractable structured data and unstructured data requiring more sophisticated extraction procedures. Although research was never a primary goal of EMR system architecture, clinicians and researchers increasingly have the opportunity to leverage clinical data for research-based analyses. However, no matter how large the datasets are, health-related research requires high-quality information and validation. While previous EMR studies successfully identified patients and extracted data, none have assessed the quality of these EMR real-life data compared to well-curated research databases.
Goals: To
(1) use the University of California, San Francisco (UCSF) EMR system in order to identify patients with specific neurological disease, multiple sclerosis (MS), and extract their clinical data,
(2) compare EMR-extracted data with gold-standard data from research to demonstrate that information extracted from EMR is of high-quality and can be used for health-related research, and
(3) compare the EMR MS population characteristics to expected MS natural history.
Methods: A classification algorithm to identify MS patients in the UCSF Health System EMR was implemented under an IRB-approved protocol and clinical data algorithmically extracted and de-identified. Classification specificity and extraction performance were assessed by manual clinician record review. EMR data was compared to research cohort data in a subset of patients.
Results: We identified 4,142 MS patients with 95.9% sensitivity. Clinical data were extracted from free-text with high positive predictive values. We showed good concordance between EMR and research values for Expanded Disability Status Scale (EDSS) and Timed-25-foot walk (ICC=0.87 and 0.79 respectively) and for MS subtype (Kappa=0.65). We replicated several expected epidemiological features of MS: higher EDSS for patients with progressive forms compared to relapsing-remitting patients (p< 2.2x10-16) and for male compared to female patients (p=1.1x10-18); and an increase of EDSS with age at examination (p=1.3x10-229) and with disease duration (p=1.7x10-86). We also replicated an expected decrease in lymphocyte counts and increase in liver enzyme testing following fingolimod initiation.
Conclusions: Large real-world cohorts algorithmically extracted from EMR data are expanding opportunities for clinical research in MS
Disclosure:
Vincent Damotte received a postdoctoral fellowship from ARSEP foundation
Antoine Lizée has no disclosures.
Matthew Tremblay has no disclosures
Alisha Agrawal has no disclosures
Pouya Khankhanian has no disclosures
Adam Santaniello has no disclosures
Refujia Gomez has no disclosures
Robin Lincoln has no disclosures
Wendy Tang has no disclosures
Tiffany Chen has no disclosures
Nelson Lee has no disclosures
Pablo Villoslada reports personal fees from Hiedleberg Engineering, Roche, Novartis, Health Engineering, other from Bionure, Mint-Labs, Spire Bioventures, grants from Genzyme, outside this work.
Jill Hollenbach has no disclosures
Carolyn D. Bevan has no disclosures
Jennifer Graves reports grants from Race to Erase MS, Genentech, Biogen, outside this work
Riley Bove received research support from NIH and NMSS
Douglas Goodin has no disclosures
Ari J Green reports grants from National MS Society, Novartis, grants and other from Inception 5 Biosciences, other from Mediimune, personal fees and other from Perkins Coie LLP/Mylan outside the submitted work.
Sergio E. Baranzini has no disclosures
Bruce Cree reports personal fees from Abbvie, Biogen, EMD Serono, MedImmune, Novartis, Sanofi Genzyme, Shire, and Teva, outside the submitted work
Roland Henry reports grants from Hoffman La Roche, other from Novartis, ABBVIE, Genzyme, MEDDAY, outside this work
Stephen L. Hauser serves on the Scientific Advisory Boards for Annexon, Symbiotix, Bionure, Neurona, and Molecular Stethoscope. Dr. Hauser also has received travel reimbursement and writing assistance from F. Hoffman-La Roche Ltd. for CD20-related meetings and presentations.
Jeffrey M. Gelfand reports personal fees for consulting from Genentech and from medical legal consulting, and research support to UCSF from Quest Diagnostics, Genentech and MedDay outside this work
Pierre-Antoine Gourraud is the founder of Methodomics.com and has served as consultant for Sanofi-Genzyme and Biogen France SAS