Systems Science for Spondyloarthritis Epidemiologic Research and Clinical Care

2018 SPARTAN Annual Meeting, “Unmet Needs” Breakout Session

Session Leader: 
Maureen Dubreuil, MD, MSc, Boston University School of Medicine, Clinical Epidemiology and Rheumatology, mdubreui(at)bu.edu

Overview:
Large electronic health record and claims databases are important sources of data for epidemiologic studies, but have inherent challenges. Researchers may use emerging methods to identify persons at risk for development of spondyloarthritis, to accurately phenotype patients with SpA, and to assess SpA outcomes. This group will discuss existing challenges and near-future opportunities for research in SpA using large datasets.

Rationale: 
Large electronic health record (EHR) and Claims databases are becoming more common sources for epidemiologic research.  The large number of patients with SpA make these datasets ideal for research into specific SpA phenotypes, rare but severe manifestations (eg- aortitis), and common but small effects related to treatment (eg- reductions in cardiovascular disease incidence) that may be important on a public health level. Many studies that are possible using large datasets may be too lengthy or costly to be performed using other cohort studies or other prospective designs. Additionally large EHR systems may present the opportunity to develop referral rules for patients at higher probability of having SpA, which may improve clinical outcomes. However systems-based research presents specific challenges that warrant consideration by spondyloarthritis (SpA) experts, including case identification and validation, case phenotyping, meshing of increasingly available genotyping data with phenotypes, and ascertaining clinically important outcomes.

Topics for discussion:

• 5 minute review: recent efforts to identify patients with SpA in existing North American EHR and claims datasets (eg- VA, Medicare, MarketScan, Rochester Epi Project)
• Discuss emerging methods to establish SpA cases or patients at increased risk for SpA:
  • Structured data element algorithms (diagnosis, prescription, lab, billing data)
  • Natural language processing (clinical notes, complex radiology reports)
• Can EHR-based referral rules be implemented for patients with moderate probability of SpA? How and in what healthcare systems?
• Define SpA phenotypes of greatest interest for epidemiologic studies or phenome-wide association studies
• How could EHRs be better designed to enable research in to SpA? What data is not captured? What data is unstructured?
• Discuss priorities for systems-based research in the next 5 years

Reference:
https://www.ncbi.nlm.nih.gov/pubmed/27792870, PDF