OHSUG 2014: How to Detect Safety Reports in Social Media for Processing in Oracle Argus?
Download as PPTX, PDF1 like720 views
This document discusses how social media can be monitored to detect safety reports and discusses integrating social media data into pharmacovigilance processes. Regulators are beginning to recommend that companies screen social media for potential adverse event reports. New technologies use natural language processing and machine learning on large amounts of social media data to identify mentions of adverse events, though manual review is still needed. Practical considerations for companies include experimenting with these approaches and establishing standard operating procedures for handling potential adverse event reports found in social media.
OHSUG 2014: How to Detect Safety Reports in Social Media for Processing in Oracle Argus?
- 1. How to Detect Safety Reports in Social Media for Processing in Oracle Argus? Brad Gallien November Research Group
- 2. Summary A quick look into the presentation Current Regulatory Framework Approaches to Monitoring Social Media Integrating Social Media into Your Business Process What’s Next
- 3. 3 Problem Statement • Let’s start at the beginning… – Identifiable reporter – Identifiable patient – Medical product – An adverse event or fatal outcome suspected to be cause by the product • Is this an AE? From a tweet: “My mom just got taken to hospital after passing out #tylenolsux” • It sure feels like one… – Reporter: child of patient, with a twitter account – Patient: mom of twitter user – Product: Tylenol – Event: Initial hospitalization following syncope
- 4. 4 Why Does It Matter? • The infamous WOMMA study – Nielson Buzzmetrics randomly sampled 500 “healthcare” messages – Only 1 message had all four criteria • Woo hoo! We are safe! • So what do we have to worry about??? – Current daily volume of tweets: 500 million – Current daily volume of Facebook posts: at least that many, more counting comments – Proliferation of sponsored and unsponsored health blogs (e.g. patientslikeme.com, adverseevents.com, etc.)
- 5. 5 What Are the Regulators Saying? • EMA: Guideline on good Pharmacovigilance Practices, Module IV – VI.B.1.1.4. Information on suspected adverse reactions from the internet or digital media • Marketing authorization holders should regularly screen internet or digital media under their management or responsibility, for potential reports of suspected adverse reactions • If a marketing authorization holder becomes aware of a report of suspected adverse reaction described in any non-company sponsored digital medium, the report should be assessed to determine whether it qualifies for reporting. • Unsolicited cases of suspected adverse reactions from the internet or digital media should be handled as spontaneous reports. The same reporting time frames as for spontaneous reports should be applied
- 6. 6 More Regulatory Opinions • CIOMS – Section IId (p.55) of Current challenges in pharmacovigilance: pragmatic approaches, (report of CIOMS Working Group V) states8: • A procedure should be in place to ensure daily screening by a designated person(s) of the website(s) in order to identify potential safety case reports • The working group does not believe it necessary for regulators or companies routinely to ‘surf ’ the internet beyond their own sites for individual spontaneous reports.
- 7. 7 And More… • FDA – Has published guidance on advertising – Has yet to weigh in on adverse events – But has suggested a “safe harbor” for sponsored sites that include an approved link to FDA or company AE reporting site • WEBAE project (Public/Private) – The WEBAE project (Web Adverse Events) aims to build on these trends and form a specialist public private consortium that undertakes research into the appropriate policy and technology solutions that enable the leverage of such web based media mining and crowd-sourcing technologies in pharmacovigilance to strengthen the protection of public health. • The ABPI (Association of the British Pharmaceutical Industry) has a good overview article – http://www.abpi.org.uk/our-work/library/guidelines/Documents/ABPI Guidance on PV and Digital Media.pdf
- 8. 8 The Perfect “Big Data” Storm • What is “big data” and why does everyone keep talking about it? – The broad area of concern is deriving knowledge out of unstructured data (like tweets, blog posts, publications, etc.) – It is “big” because it typically encompasses the entire set of “stuff” accessible on the knowledge – It is “data” IF you can make sense out of it • How can this technology help? – The suite of big data technology can search a social media “feed” like Twitter, Facebook, and health blogs and use technology to identify potential adverse events based on computer intelligence
- 9. 9 How Does It Work? • Big Data uses a few core technologies in order to deduce that a string of text might pertain to an adverse event: – NLP (Natural Language Processing): In order for a computer to understand text, it needs to know how we humans talk. Computer scientists have spent decades developing this basic framework (driven from the field of voice activated computing). And just as they figured it out, we invented a new language: tweeting – Semantic search: This component of the technology is used when asking questions of text. A library of “triples” consisting of a subject , a predicate, and an object (aspirin relieves pain, viagra is known to cause vision coloration). The triples support the NLP processing. – Ontologies: In this context, we forget the meta, and focus on the physical. Ontologies for the basis for describing things that we are looking for in the data. The idea here is to build a universe of synonyms to help find an object of interest. Examples include MedDRA, SNOMED, WHO-DRL, etc. but the technology also allows you to build custom ontologies based on human input and computer learning
- 10. 10 Who Is Doing it? • There are several companies now that are applying this technology to adverse event detection – IMS: Nexxus AE Tracker. Part of their library of Nexxus tools, AE Tracker identifies potentially reportable adverse events – Epidemico: MedWatcher for Enterprise. A spin off from MIT, Harvard Medical School and Boston Children’s Hospital offers a subscription service to companies • Is the technology enough? – No. Both companies include “manual curation” of the result sets. This refines the results as well as feeds their custom ontologies. • What are they finding? – The example on the Epidemico site shows approximately 300,000 “mentions” that the algorithms tagged with 4500 potential AEs (this is for xanax over a 3 month period…)
- 11. 11 Yikes! What Do I Need to Do? • Clearly the volume of potential adverse events is staggering, but the actionable adverse events are probably MUCH lower (remember the WOMMA study…) • Many companies are approaching these in an experimental fashion – Purely exploratory – Post-marketing surveillance studies • The regulatory agencies have not mandated this, but consider literature sources and how they have evolved • In my opinion, it is inevitable that there will be some movement towards standard monitoring of these information feeds
- 12. 12 Some Practical Considerations • Give it a try! There are companies out there that can assist you as your explore this emerging area • Consider your operational response – Be sure that you have SOPs in place for sponsored sites (patient registries, etc.) – Consider products that would benefit from added surveillance (products under a REM for example) – Stay engaged • These are potentially reportable adverse events – You need a staging area for them outside your global PV system • Affiliate module, ARISg IRT, PRIMO, etc. – Develop a set of criteria for following up on these
- 13. 13 References and Links • Innovative Medicines Initiative (WEBAE): http://www.imi.europa.eu/sites/default/files/uploads/documents/9th_Call/Cal ll_9_Text.pdf • ABPI Document: http://www.abpi.org.uk/our-work/ library/guidelines/Documents/ABPI Guidance on PV and Digital Media.pdf • Bart Colbert article: http://www.telerx.com/blog/collecting-adverse-events- with-social-media/ • Bloomberg Law: http://www.bna.com/pharma-challenges-adverse-event-reporting- and-social-media/ • Eye For Pharma: http://social.eyeforpharma.com/patients/patients-social-media- and-adverse-event-reporting • John Mack article: http://www.news.pharma-mkting.com/pmn93- article04.pdf
- 14. 14 Continuing the Conversation… • How are companies here handling this? • Are any companies doing this? • How concerned are you about the potential wave of work?
- 15. 15 Speaker Bio Brad Gallien is a Vice President of Product Development at November Research Group, a professional services company focused on the implementation and support of pharmacovigilance systems. He is responsible for product development at November Research Group. Brad has been focused on the pharmacovigilance business for over 15 years, leading product development and implementation of global pharmacovigilance systems. This focus has provided him with a broad understanding of industry best practices and trends. Brad joined November Research Group in October 2005, after three years at Oracle Corporation as Director of Life Sciences Strategy and the Product Manager for Oracle AERS. Prior to Oracle, he was Vice President of NetForce, following nine years in clinical research and development at Syntex. Brad has a BA in Biology from UC Berkeley and MS in Zoology from University of Hawaii. Contact Info Email: brad.gallien@novemberresearch.com Phone: +1 415-279-9107