Reveal Hidden Patterns in Healthcare Data: Graph Analytics and the Opioid Crisis
1 like200 views
This document discusses using graph analytics and public healthcare datasets to analyze the opioid crisis. It proposes connecting datasets on prescription drug payments to physicians, prescription drug claims, and opioid distribution through a graph database to identify relationships and patterns. This could help track the source and causes of the opioid epidemic by revealing connections between pharmaceutical company payments to doctors and the opioid prescriptions written. The analysis method involves modeling the data as a labeled property graph to combine Medicaid, CMS, and DEA data sources to analyze the opioid supply chain and determine if payments influenced prescription rates.
Reveal Hidden Patterns in Healthcare Data: Graph Analytics and the Opioid Crisis
- 1. Neo4j Connections Graphs in Life Sciences and Healthcare Reveal Hidden Patterns in Healthcare Data: Graph Analytics and the Opioid Crisis Cynthia Femano – Sr. Systems Engineer, Neo4j
- 3. 3 ~70,000 US deaths/year from drug overdoses ~70% of those deaths from Opioid drugs
- 4. Fentanyl – A highly potent synthetic opioid intended to treat severe pain associated with terminal cancer
- 5. September 29, 2008 Cephalon, Inc. – ACTIQ® (Fentanyl Citrate) “Off Label” Marketing
- 7. ü How can we track down the source & causes of the opioid epidemic? ü Who is doing the over-prescribing and why? ü Are some parts of the country more susceptible than others? This brings to light several questions
- 8. ü Use public domain datasets that are openly available ü Connect these datasets in a graph using common attributes (NPI, NDC, Geolocation) ü Use a combination of Cypher queries and a graph visualization tool (Neo4j Bloom) to derive insights from the data Our analysis method
- 10. Combine data from Medicaid Part D and CMS Open Payments to determine if there is a connection between Payments made by pharma companies to physicians and the Part D Prescriptions these same physicians wrote for opioid class drugs. Use DEA ARCOS data to examine the Opioid Supply Chain.
- 11. CMS Open Payments Data CMS Open Payments Data from the Years 2013-2015. (27M records) This includes information about payments made by drug manufacturers and specialty pharmacies to doctors and teaching hospitals. The payments are for things like meals, entertainment, consulting or speaking fees, etc. Each payment is associated with a particular drug or medical device that they are trying to market. https://openpaymentsdata.cms.gov/
- 12. Medicare Part D Prescriber Data Medicaid Part D Prescriber Data from the Years 2013-2015 (72.3M aggregate records) This includes aggregated data for drugs and devices, showing number of doses prescribed by a certain provider over the course of a year. https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/Medicare-Provider-Charge-Data/Part-D-Prescriber.html
- 13. DEA ARCOS – Pill Distribution of Controlled Substances ARCOS FOIA Data (178.6M transactions) Released by the Washington Post, This data tracks the distribution of oxycodone and hydrocodone pills from manufacturer to pharmacies and physicians in the United States from 2006 through 2012. https://www.washingtonpost.com/graphics/2019/investigations/dea-pain-pill-database/#download-resources
- 14. • The NPPES NPI Provider Registry The NPI Registry Public Search is a free directory of all active National Provider Identifier (NPI) records. • The FDA NDC National Drug Code Directory Drug products are identified and reported using a unique, three-segment number, called the National Drug Code (NDC), which serves as a universal product identifier for drugs. 14 Other Data
- 15. CMS Open Payments / Medicare Part D / DEA ARCOS Schema
- 16. Highest Value Opioid Drugs 2013-2015
- 17. Top Opioid Payees 2013-2015
- 21. Using Neo4j Bloom to explore the graph
- 22. Graph Feature Categories and Algorithms Pathfinding and Search Finds the optimal paths or evaluates route availability and quality Centrality / Importance Determines the importance of distinct nodes in the network Community Detection Detects group clustering or partition options Heuristic Link Prediction Estimates the likelihood of nodes forming a relationship Evaluates how alike nodes are Similarity 22
- 24. INSYS Thereputics: Prioritizing profits over patient’s lives
- 26. Alec Burlakoff: A 60 Minutes investigation into the causes of the American opioid epidemic includes an examination of the tactics of one former top sales executive.
- 27. ü While voluminous, detailed data exists to track payments, prescriptions and the distribution of opioids. ü There is a network of influence between doctors, teaching hospitals and pharma companies. Graph algorithms can reveal this network by finding patterns in the data. ü Graph analytics can help to identify the opioid supply chain and find geographic areas most at risk. In summary…
- 29. Thank you! Cynthia Femano – Sr. Systems Engineer, Neo4j cynthia.femano@neo4j.com