Online Program

PS23-GeoMapr: An Analytic Dashboard for Prescription Drug Utilization with Geographically Referenced Data Enrichment and Machine Learning (301127)

View Presentation

Grace Chai, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
*Meilan Chen, University of Massachusetts Amherst, Department of Mathematics and Statistics; US Food and Drug Admi 
Yong Ma, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Translational Scienc 
Shekhar Mehta, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
Rose Radin, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
Travis Ready, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
Jaejoon Song, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Translational Scienc 
Saranrat Wittayanukorn, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
Corinne Woods, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Surveillance and Epi 
Yueqin Zhao, US Food and Drug Administration, Center for Drug Evaluation Research, Office of Translational Scienc 

Keywords: geoMapr, machine learning, spatial model, temporal model, prescription drug dispensing, naloxone, buprenorphine, infectious disease

In post-market drug safety surveillance, pharmacy dispensing data provide valuable insights to FDA for oversight of drug utilization. We have developed a web-based interactive tool, called geoMapr, to analyze proprietary, nationally projected data for prescription drug dispensing. Our tool, deployed as an FDA-internal webpage, can be used to perform descriptive to complex analysis of prescription drug utilization data through data enrichment with other geographically referenced, publicly available, demographic, socioeconomic, or healthcare service data. As a proof-of-concept, we have investigated naloxone and buprenorphine prescriptions dispensed from U.S. retail, mail-order/specialty pharmacies from January 2014 through December 2018. Our software performed a five-step analysis pipeline to: i) visualize dispensing trends over time and across geographical locations, ii) enrich the dispensing data with geographically referenced public data sources, iii) implement decision tree-based ensemble algorithms to explore geographical factors associated with prescription dispensing, and to perform iv) clustering of temporal patterns or v) spatial modeling to identify important factors associated with temporal or spatial variations. Results of this exploratory analysis can inform further investigations, as drug dispensing data do not measure the product’s ultimate use. The geoMapr is continuously updated to address important needs in regulatory decision-making. A planned enhancement is to explore the feasibility of signal detection of infectious disease outbreaks related to prescription opioid use by ingesting blood-borne infections data from the National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention (NCHHSTP) database (AtlasPlus).