Online patient data

Continuing our focus this month on patients, we spoke with Dr James Sawyer of Prism Ideas Ltd, on the challenges associated with using online patient data.

 How can we harness the power of social networks and make use of online patient information and discussions? And how reliable is this information? We spoke with James Sawyer of Prism Ideas Ltd about these topics.

Read his interview to hear more on the associated challenges, and what use this online information can be to us.

Interview summary

RA: Dr Sawyer thank you for taking part in the interview, what challenges are there with using online patient data and how do we overcome them?

JS: There are three areas where there are clearly challenges for pharma when looking at online data. One is in terms of the quality of the data, by which I mean the validity of it in terms of whether what’s been said is accurate, the amount of reports that are there. There’s a lot of noise around all the information that’s available on the internet. So just filtering out all those different aspects is one big challenge.

The next one is in terms of what I might call the provenance of the data. If you’re looking at when patients are talking about drugs, it’s important to make sure that you’re not getting the same thing said twice or picking up people simply repeating what others have said.

Then there are the obligations upon companies to report safety data, about finding out things that they might otherwise not have known about. Once a company is aware of an issue , they need to be passing that information on to the regulatory authorities.

If I was to deal with each of those problems in turn, for data quality you need to be sure that you’ve got good coverage across the internet, so that particular groups or individual forums don’t bias the picture you see. You also have to have a mechanism to understand whether an individual is actually taking a drug or just talking about it in general.

When you’re looking at data provenance it’s very important to be able to identify repletion by one individual on multiple sites and indeed to pick out campaigns from normal discussion. There are a number of ways that high quality natural language processing can enable you to do that.

Finally in terms of the reporting obligations to regulatory authorities, much depends upon the sorts of study you undertake and whether the sort of information you develop. For example observational studies are useful for their overall results not the specific data from particular individuals that they draw upon.


“It’s like trying to find your way through a bowl of spaghetti by following an individual strand…”


RA: Will monitoring the internet in general essentially produce a massive additional burden on companies and regulators when it comes to reporting and evaluating adverse events?

JS: I think this is one of the big worries, one that sometimes opens up a can of worms. If you start following a thread of information, how far does it go? There potentially a massive burden of activity and evaluation if one simply starts reading all the information then trying to follow the threads everywhere you go.

Without good analysis techniques, it’s like trying to find your way through a bowl of spaghetti by following an individual strand. What has to happen is to find an approach that enables us to use the internet as a resource, not increase the burden of work but to inform our decisions. At present it can be hypothesis generating, giving us an indication about where we should undertake traditional activities.

I believe that a consensus will evolve amongst the regulators, pharma companies, and big data analysts that allows this approach to be utilised without creating that additional burden.

RA: How does online patient data related to adverse events compare with FDA labels?

JS: There is clearly a different pattern in terms of the information that patients talk about in lay language compared to the information that’s talked about in medical interactions and the terminology that reaches our product labels.

Nevertheless, if we look at the most commonly reported adverse events in clinical trials we see that there’s really generally a very good match with patients’ experience reported on the internet. This is both in terms of what’s talked about most commonly by patients, and also the frequency at which patients talk about those particular problems.

What we also see is that patients tend to talk more about things which they can be physically or mentally aware of, so aches and pains, stomach disturbances etc, than they might do for a particular laboratory diagnosis that you tend only to get out of clinical trial data.

RA: Why do you think patients are quicker to discuss adverse events online than report through the appropriate channels?

JS: Well the first thing is it’s easier. If one wants to report a suspected side effect through the usual healthcare channels there are forms to complete, there’s a process and one needs to be quite highly motivated and aware to do that.

What’s happening online is people are talking about their life experiences they’re able to do it in the comfort of their home on their iPad or on their PC. They can tap away and make that information available in seconds rather than what might be half an hour in a formalised process.


“…whether or not it had a big impact upon the patient or not is rather lost in terms of the forms and the terminology we use.”


RA: What does internet monitoring tell us about the adverse events associated with a drug that the FDA reports don’t tell us?

JS: The most important thing is the fact that when we have a report made through the standard process to the regulators it’s rather sterile in that we simply learn that something has happened. Whether or not it had a big impact upon the patient or not is rather lost in terms of the forms and the terminology we use.

When one is looking at what patients talk about it’s usually the things that have impacted them the most with regards to a therapy. The things that have had less impact – that might otherwise be picked up in a comprehensive clinical study – simply don’t get recorded. That’s both good and bad, but what it tells us is that the information we see on the internet is what matters to patients and in some ways perhaps that’s the most important thing.

RA: What do you think the future of capturing online adverse events will look like?

JS: I wish I had a crystal ball. One of the things is that we’re seeing quite a lot of change in terms of the way that information is being provided. The interaction between healthcare professional data collection through medical records and other systems is likely to expand.

We already see interfaces directly between patients and healthcare providers when patients record diaries and log into care sites. As people become more and more comfortable with expressing their personal experience on the internet the volume will grow.

With large numbers of patients’ information being analysed on the internet you have an effect whereby the extremes are weeded out and you can look at the overall picture at what patients are saying to make useful decisions.

RA: James, thank you very much for your time today and your insights.

JS: Very welcome, thank you.




About the interviewee:

In a wide ranging career, Dr Sawyer has published over 30 papers across a large range of subjects. He joined the Pharma industry 20 years ago as a clinical research physician. Thereafter, he alternated between medical and commercial roles at big Pharma companies such as Sanofi, AZ and Roche before, in 2001, establishing Prism Ideas which was granted the Queen’s Award for Enterprise in 2010. He has a long history of innovation ranging from clinical trial feasibility through research into stratified care and approaches to clinical trial design. More recently he has been working with experts in natural language analytics to explore how the internet can be used as a resource for pharmacovigilance.

Why are patients quicker to discuss adverse events online than report through the appropriate channels?