The RxView Blog

Real world data, analytics, and insights for healthcare decision makers

Advera Health Analytics, Inc.

Social Media Monitoring for Pharmacovigilance and Drug Safety

February 2, 2018

It is now well established that the rapid expansions of the internet and computing power have opened up the use of social media and Internet forums for pharmacovigilance. These sources contain untapped, albeit noisy, safety and benefit information. As noted in Social Media Listening for Routine Post-Marketing Safety Surveillance, published in the journal Drug Safety, methods exist to reduce noise and make the data suitable for post-marketing safety surveillance. However, the use of these data to date have been constrained by limitations in how to best implement novel methods without disrupting traditional signal detection management and evaluation work flow.

Social Media in Pharmacovigilance Workflow

Advera Health and Booz Allen Hamilton, a global management and technology consulting firm, recently announced a new partnership that will provide Advera’s clients with access to Booz Allen’s health vigilance data that examines publicly-available social media for adverse drug side effects within Advera’s Evidex drug safety analytics platform. This partnership is key to enabling the industry to implement social media monitoring into pharmacovigilance workflows. 

social tec

 

Advera Health’s Evidex data aggregation platform is a scalable, flexible, and agile way for pharmacovigilance professionals in the healthcare industry to access and analyze drug safety data. Powered by Advera’s proprietary RxFilter data optimization technology, Evidex supports a broad array of safety signal detection, signal management, and risk mitigation use cases. Evidex’s integration of Booz Allen’s data allows for active monitoring of public social media posts for product side effects alongside clinical trial results, spontaneous reporting, and other real-world data. The data are fully mapped to all existing Evidex data to facilitate early signal detection and hypothesis generation.

Technology Enabled Process

Booz Allen’s social media data and analytic capability observes and classifies online discussions of drug side effects. Public data collected from social media platforms are processed with an extensive library of natural language processing and machine learning tools that are tailored to the health domain. These automated aggregation processes are combined with manual curation to detect potential safety issues. In 2014, Booz Allen acquired Epidemico, Inc. which was working with regulatory agencies such as FDA and EMA, as well as pharmaceutical safety and epidemiology teams for early signal detection and hypothesis generation regarding both labeled and unlabeled safety events.

 

bah placeholder

 

Reduce Noise Through Manual Curation

bah processWhile technological advances have powered the ability to process large amounts of social data, the results can be “noisy”. Therefore, the Epidemico process uses curation, a manual review of a subset of data by human annotators, immediately following automated processing. The curation step is critical in order to provide the highest quality data and ensure accurate representation at the aggregate level. This curation step also serves to reduce false positives and false negatives, improving the automated Bayesian classifier through positive and negative feedback loops, and ensuring the ability to recognize emerging syntax and slang. Human curators are trained to never make medical interpretation beyond the stated social media text and to adhere to MedDRA coding standards and MedDRA’s ICH-endorsed guidelines. 

Social Media Bias

bah processThere are three dimensions of bias that come from different drug safety datasets: Patient reported outcomes, seriousness, and completeness. These dimensions differ across social media, clinical trials, spontaneous reports, and EHR records. Further details on how certain data skew, can be found here.

Reportability Requirements

One of the biggest hurdles to using social media data has been the risk of having the data contribute to already tough-to-manage case processing workloads. The Booz Allen Epidemico process automatically detects and removes personally identifiable information to protect the patients’ identity, therefore eliminating the reportability requirement. Evidex further provides for the ability to display only aggregate views without the verbatim text displays for additional safeguards.

Pharmacovigilance 2.0

Booz Allen Epidemico social media data and analysis, integrated in Evidex alongside proprietary clinical trials safety outcomes data, optimized post-approval spontaneous reporting data from the FDA Adverse Event Reporting System (FAERS), and other real-world data sources provides pharmacovigilance professionals with the software, data, and analytics to track emerging safety issues through multiple data sets, validate signals seen in spontaneous reporting, and engage across these various data sets in a dynamic and proactive manner.

Follow this link if you are interested in learning more.

 

Access to the Drug Safety publications on social media monitoring mentioned in this post below as well as further academic work here:

Social Media Listening for Routine Post-Marketing Safety Surveillance (Powell GE, Seifert HA, Reblin T, Burstein PJ, Blowers J, Menius JA, Painter JL, Thomas M, Pierce CE, Rodriguez HW, Brownstein JS, Freifeld CC, Bell HG, Dasgupta N.) Drug Safety. 2016 Jan 21. 

 

Digital Drug Safety Surveillance: Monitoring Pharmaceutical Products in Twitter (Freifeld CC, Brownstein JS, Menone CM, Bao W, Filice R, Kass-Hout T, Dasgupta N.) Drug Safety. 2014 May

 

Evaluation of Facebook and Twitter Monitoring to Detect Safety Signals for Medical Products: An Analysis of Recent FDA Safety Alerts (Pierce CE, Bouri K, Pamer C, Proestel S, Rodriguez HW, Le HV, Freifeld CC, Brownstein JS, Walderhaug M, Edwards IR, Dasgupta N.) Drug Safety. 2017 Jan 2.

 

Topics: Pharmacovigiance 2.0

By Jim Davis
Find me on:
New Call-to-action

Subscribe to Email Updates