P58 Patient discussions of glucocorticoid related side effects within an online health community forum

Abstract

Background: Social media websites are an important, largely untapped source of data about patients' experience of disease and treatment, including the occurrence, nature and impact of drug side effects (SEs). However, there are few studies using such data. HealthUnlocked (HU), Europe's largest social media network for health, hosts the National Rheumatoid Arthritis Society (NRAS). Using the example of glucocorticoid (GC) therapy, we explored the potential of HU posts in providing information about drug SEs. Objectives: Evaluate the accuracy of a computerised system for automated suspected adverse drug reaction (sADR) detection from HU posts compared to human annotation. Explore themes of discussion about GC-related ADRs. Methods: HU provided a dataset of de-identified posts from NRAS. Posts mentioning GCs were processed by Natural Language Processing software, which identified the drug and health issues, mapped them to Medical Dictionary for Regulatory Activities dictionary and categorised as a sADR or not. A sample (n=50) of sADR posts were randomly selected and reviewed to determine whether they were true ADRs. Additionally, a sample (n=50) of the posts that included GC and labelled as having a health issue but not an ADR, were assessed for true ADRs. Posts identified as true ADRs from manual analysis were reviewed to identify themes. Results: Of 35,904 posts, 2,409 posts mentioned GCs, of which 324 posts were identified as containing a sADR. After manual review of these posts, only 36% contained a true ADR. Of the 50 sampled posts that included a mention of GCs and a health issue but not a sADR, 28% contained true ADRs. Thematic analysis of the 32 posts containing true GC ADRs found the most frequently mentioned ADRs were fractures (n=6), infect ion (n=5) and headaches (n=3). Posts included rich descriptions about nature of SEs (“my weight tripled in size”) and how these changed with time (“huge mood swings settle after a while”). Users described how ADRs impacted their quality of life (“with steroid induced diabetes, I lost a stone, it was grim”), and their value judgements about importance of SEs (“my taste buds are making everything taste strange.but I cope with it, as hardly any pain”) Posts described patients frustrations (“I had ops for cataracts, no one mentioned steroids caused cataracts”). Patients commented on benefits of treatment (“pain subsided”) and the balance between benefits and harms (“feel wonderful after steroids, but now I have more acne than a teenager”). Conclusion: Current machine learning models need improvements to identify sADRs in health forum data. Nonetheless, manual review shows there are important themes that may not be obtained using traditional methods. With improved models, this rich data source may be useful to identify ADRs most important to patients and impact on quality of life.