ASSESSMENT OF MIGRAINE PROGRESSION AND PREDICTORS OF PROGRESSION USING A CLAIMS- AND ELECTRONIC HEALTH RECORD-INTEGRATED DATABASE

Author(s)

Li L1, Mueller L1, Hitchcock C1, Zyoud R1, Bell J2, Chang SC3
1GNS Healthcare, Cambridge, MA, USA, 2Teva Pharmaceuticals, Frazer, PA, USA, 3GNS Healthcare, Cambridge, USA

OBJECTIVES: To determine the rate and identify predictors of episodic migraine (EM) to chronic migraine (CM) progression using machine learning and structured data from a large, geographically diverse claims and electronic health record (EHR) database.

METHODS: A retrospective cohort study design was applied to an integrated claims/EHR data set indexed on first EM diagnosis date. Descriptive analyses were conducted assessing progression rates. Using a hypothesis-free, Bayesian machine learning analytic platform (GNS Healthcare REFS™: Reverse Engineering and Forward Simulation), an ensemble of progression-free survival models was built to predict progression, computationally exploring thousands of variables and their interactions. Accuracy of prediction models was evaluated using Harrell’s C-statistic.

RESULTS: The derivation cohort comprised 39,680 adults with an EM diagnosis between 2007 and 2017. In newly diagnosed EM patients, the progression rate after 1-year follow-up was 2.5%. By the end of follow-up, 4.5% of patients had progressed to CM with a median time to event of 377 days. Patients using preventive medications prior to first EM diagnosis had the largest median hazard ratio (HR) for progression; including those with prescriptions for beta-blockers (HR=1.78), anticonvulsants (HR=1.72), and antidepressants (HR=1.52). Other identified risk factors included migraine-related outpatient visits (2+ vs 0, HR=1.52), neurology visits (2+ vs 0, HR=1.49), physical therapy exercise (HR=1.36) and spondylosis (HR=1.32). Male gender (HR=0.69) and older age (75+ vs 18-34, HR=0.24) were protective against progression. All median P-values for listed predictors were <0.001. The mean C-statistic was 0.68 in held-out testing data.

CONCLUSIONS: Using 10-year real-world data, we determined progression rates for extended follow-up periods, replicated established progression risk factors, and identified novel predictors of progression. Predictors can help clinicians identify at-risk patients and implement timely interventions, thereby reducing migraine progression, improving clinical outcomes, and lowering economic burden.

Conference/Value in Health Info

2019-05, ISPOR 2019, New Orleans, LA, USA

Value in Health, Volume 22, Issue S1 (2019 May)

Code

PND70

Topic

Methodological & Statistical Research

Topic Subcategory

Artificial Intelligence, Machine Learning, Predictive Analytics

Disease

Neurological Disorders

Explore Related HEOR by Topic


Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×