Skip to main content

social mining for drug adverse events – linguistic and cultural considerations

reporting adverse events

Most drug adverse events are not reported by patients to medical professionals, pharmaceutical companies or official agencies, such as the FDA’s Adverse Event Reporting System (FAERS). Mining social media for patient reporting of adverse events is a potentially important tool to increase this knowledge base efficiently and quickly. A 2018 presentation (REF A) by the chair of the FDA Data Mining Council stated that the FDA “recognizes there is a wealth of consumer information from social media that may help our safety surveillance” and is exploring the value of social media mining for this purpose

In one 2017 study, machine learning and natural language processing techniques were used to identify posts of interest regarding 10 known FDA safety signals. Almost a million harvested Facebook and Twitter posts yielded 98,252 adverse events, with 13 selected for detailed review. In one case, the adverse event was documented on social media before it was seen in FAERS, raising the possibility that social media mining may be able to identify AEs more quickly than official reporting channels. Another 2014 study compared patient reported AEs on Twitter to the public FDA adverse event reporting system, finding high concordance.

But designing the algorithms and tools needed to do this work is not easy. Many patients do not casually use specific medical terms in social media posts and may describe their symptoms in culturally specific terms, which vary widely between primary language used, nationality and even ethnicity. Patient reported drug side-effects in people with limited English proficiency (LEP) can also be extra complex when reporting symptoms such as pain, where a patient may lack the vocabulary to effectively communicate the extent of their pain to clinicians, but they may discuss their symptoms on social media in their primary language. Culture also impacts the way in which people report pain, so a lack of understanding of this combined with the ever-evolving uses of social media can lead to vital information being missed. Similarly, language used to describe mental health side effects, emotional distress and even fatigue and cognition can vary widely between different cultural groups

 

leveraging social media in clinical research

The European Union’s WEB-RADR project seeks to find out more about how emerging technologies such as social media can be used to help track adverse drug reactions. However, researchers have encountered difficulties in designing recognition systems with a high level of precision and recall when applied to real-world data. They concluded that widespread mining of Twitter and Facebook posts may not be fortuitous for mining for adverse events. However, they do suggest that designing systems to mine for information about specific medical products, health conditions and patient groups or networks may yield positive results, all of which require bespoke culturally and linguistically informed processes to ensure data integrity, accuracy and quality.

Social media can also be an important source of information for real-time reporting of adverse events in scenarios such as the Covid-19 pandemic. Early in 2021 as Covid-19 as more people received vaccines, women began to use social media to report changes in their menstrual cycles, correlating the changes to their vaccinations. This eventually piqued the interest of two scientists with expertise in menstrual cycle changes and they set up a study to officially document the changes, which had 40,000 respondents as of August 2021.

As well as watching for reports of drug side effects, marketing departments of healthcare companies can also monitor social media for times where patients could, for example, have trouble accessing or affording a medication and might need extra support. Ensuring patient satisfaction helps protect brand reputation and real-time data from social media allows companies to allocate resources to this and strategize more effectively.

To cast a wider net, considering digital behaviors of separate demographics can help guide a more robust social listening strategy. Studies show that different communities leverage different social platforms to receive and share information. For example, Facebook is shown to be a preferred social media platform for Latinos across age groups, while Instagram is more commonly used among younger Latino generations. Additionally, it is useful to identify trends and influencers relevant to the tracked topic to gather a more focused pool of users across target languages.

CQ fluency is your social listening partner

CQ fluency has extensive experience of working with top healthcare and life science companies to search and analyze large volumes of social media content via innovative mining processes, which also reduce bias and optimize data quality. We offer social listening services in over 170 languages and can monitor multilingual/multicultural patient communications, including utilizing linguistic variants, such as acronyms and slang. This robust data mining ensures rich data collection for real world evidence to compliment and elevate your clinical research, providing key insights that drive credibility and foster trust with your patient base.