Can we use an online health community forum to find out about what side effects of steroids are most important to patients?

A summary of findings by Dr Arani Vivekanantham, a Rheumatology Academic Clinical Fellow based at the Centre for Epidemiology Versus Arthritis at The University of Manchester. 

The following project was conducted under the excellent supervision of Professor Will Dixon last year. It has been presented at the Annual European Congress of Rheumatology in Madrid last year and has been recently submitted for consideration for publication.

Manchester University logo

Patients are increasingly sharing their health-related experiences online on forums and social media websites, such as Twitter, Facebook and NRAS’ own forum on HealthUnlocked. These posts and responses include information about the onset and progression of symptoms, the impact of living with a health condition, and the experiences of treatment, including both the benefits and side effects. Whilst Twitter has been widely used by researchers, the 280-character message length restriction is limiting; in health forums, posts can be much longer. This provides an opportunity for much richer description of an individual person’s experience, and is a vast source of information when collated across tens of thousands of people living with disease. We wanted to see whether a computer system that have been developed previously to automatically spot mentions of medications and side effects within short posts in Twitter could be used to identify side effects of steroids from NRAS posts in HU.  

HU provided us with over 35,000 de-identified posts from NRAS from December 2015- December 2016 after we informed the HU NRAS community of the research plans, reminding them about how they could opt out of data sharing. All personally identifiable information like names were removed from the posts. The computer system found just over 300 of these posts to contain information about steroid side effects. However, when we manually checked the work of the computer system, we found that it did not work perfectly. It missed some of the posts reporting on side effects that we were interested in. It also mistakenly identified some posts as examples of side effects, when they were not.

When we individually looked at a hundred posts we found important themes of discussions about steroid side effects. For example, not only were we able to understand which side effects were most commonly discussed by patients (such as fractures and infections) but also understand the nature of these side effects (e.g. “I gained four stone in four months due to 30mg of steroids”), the impact of these side effects on patients’ quality of life (“with steroid induced diabetes, I lost a stone in three days, it was grim”) and how these changed over time (“huge mood swings settles after a while”).

We were also able to appreciate how patients benefitted from taking steroids (“my pain subsided with steroids”), to note that they experienced both benefits and harms ("wonderful to not feel like I had RA in the first month of having steroids, but now I have more acne then when I was a teenager") and how they balanced these benefits and side-effects ("my taste buds are making everything taste strange, either salty, metallic, or plain awful … but I cope with it, as hardly any pain with steroids”). We also found some signs of frustration about lack of information about certain side effects (“I had two eye ops for cataracts, no one told me steroids caused cataracts”).

When we manually reviewed a selected number of posts we found important themes that may not be obtained using traditional methods of drug safety reporting.  However, further work needs to be done to improve the way our computer system identifies steroid side effects from the longer HU posts. With improved computerised systems, this vast and growing source of data could be used to help healthcare professionals better understand side effects of drugs most important to patients and the impact of this on patients’ quality of life. This new information could be used to help doctors counsel patients about side effects that are most concerning to them (as well as the more serious side effects) when starting patients on steroids and also facilitate patients to share their experiences with others in similar situations.