You can imagine that if our data consistently failed to reflect the existing scientific literature, it would raise questions about the general applicability of our findings to the real world.

Conversely, if we ARE able to demonstrate multiple replications of prior findings – particularly otherwise unexpected associations – then any future genuinely novel findings from our data may be quite useful in generating new hypothesis that could be independently tested.

Crowdsourced Health Confirms Infertility-Asthma Finding

It’s not clear to me what the new idea or hypothesis was in this particular example. And maybe this was more of a “test case” in terms of seeing that your data indeed is confirming long-held research hypotheses about this link. If that’s the case, it’s a good datapoint, demonstrating similar findings as more traditional research holds.

I can’t wait for the future efforts from sites like yours and PLM. This is an exciting time when the old meets the new, seeing the interaction (and synergies) between the two.

I wonder if you couldn’t publish something on your site about your data findings, and yet still publish it in a journal as well in order to reach more traditional researchers and professionals? I imagine the new Journal of Participatory Medicine would be very interested in such studies, for instance.

Yes, our analysis is not as robust as a randomized clinical trial – ours was an observational study, and RCTs usually require some kind of intervention. We are exploring how to give patients the tools to self-randomize and conduct more robust, crowdsourced RCTs.

Certainly, additional work needs to be done to determine how representative CureTogether members are of the general population. This can be done by comparing CureTogether data with previously published results from other scientific studies – something we are also working on.

We are partnering with researchers who have extensive data expertise at several universities, and will soon be opening up the entire, de-identified CureTogether dataset for them (and other applicants) to analyze. Their findings would generally be published in peer-reviewed journals.

As for how we disclose our own findings, we agree it is important for research to go through the peer review process. At the same time, we want to balance it with your excellent point of quickly disseminating data and opening up even preliminary results to spark new research questions.

The major purpose of CureTogether is to generate hypotheses and new ideas, so our observations at this point in time are not meant to stand on their own as conclusive evidence.

Much appreciated, please keep the comments coming!

I emphasize “current state”, since the field of scholarly publishing is undergoing a transformation that is also driven by changes in information technology. (See my piece: http://www.healthcontentadvisors.com/2009/06/22/health-content-is-rapidly-losing-its-value/)

Having access to a large pool of data on post-marketing outcomes could certainly help our understanding of a drug’s longer-term efficacy, don’t you think? But, as you point out, work has to be done to make sure the data are of high quality for research purposes.

As for whether this opens a new door for epidemiological research is still yet to be determined. I think it has great potential, but it still needs to be published in peer-reviewed journals for it to be accepted on the same level of prior scientific data.

So while I think stuff like this from CureTogether is also great:

http://bit.ly/2BJQ6C

A conference presentation + press release is still not the same thing as getting it through a peer-review process and published. I think it’s great to see the results disseminated more quickly than the traditional scientific process, but I still believe there’s value in going through that process.

Unanswered by the CureTogether press release are questions like, “Are CureTogether users like the general population? If not, how do they differ? Are they “more sick” or “less sick” than traditional patients, e.g., do they report greater numbers of severity of symptoms than a traditional patient sample? Etc. etc.”

It’s all a *very good* start, however, and I hope will produce more interesting and quicker data than traditional research.

In my view, infodemiology data will be of most value when used in conjuction with other research data, for example, to enhance and help test results from controlled clinical studies.

The explosion of data that will be available from the increased digitization of medical records, healthcare transactions, and individual patient behavior (and genomics) is still in the early stages. There will be bumps in the road as we figure out how best to collect, organize, and incorporate new sources of data into medical research. Still, it’s exciting to consider how all these new data sources can be applied to transform medical research–and ultimately transform healthcare delivery.

1. Most survey research conducted on patient communities show that the majority of people post very rarely. My pointing out the 63% number was more that with the PLM website, I simply was expecting virtually all patients to be posting messages back and forth to one another because of what it is — a social networking website for health. To find a significant minority still not using the social aspects of the site in what I think is the “expected” manner is interesting to me.

2. PLM was specifically set up, from listening to its founders talk about PLM, to encourage population-based data that could be analyzed and significant trends discovered. So indeed, not only are these not the last studies to come from PLM, they are really only the first of what will likely be dozens more in the future.

3. People who are putting their data into the PLM engine know they are exposing themselves to such research, which may bias the sampling of PLM patients versus a truly randomized sample. This may have implications for future research done on PLM users (especially if one doesn’t look and quantify the differences between ‘active’ PLM users and ‘inactive’ users, but use both sets of data in one set).

PLM has a lot of very smart and talented people working there, so I suspect they’ll figure most of these issues out. I look forward to reading future studies from them.

The finding that “only 63% of the registered users posted a comment from scratch to the site” pinged my memory of the CHESS studies. My recollection is that, overall, a small group posted to the bulletin boards, but many benefited from those messages (and from the library of resources made available). So the 37% not posting messages on PatientsLikeMe may just be lurkers (and it’s amazing that the lurker group is in the minority since it’s usually the reverse).

Another study to check out as a model is Meier et al’s ACOR study which “systematically sampled 9% of all archived messages for each group” yielding a nice, big sample of 2,775 messages.

I keep hearing about how we need more randomized, controlled trials of internet interventions to prove (or disprove) their efficacy (and I’m collecting a list of these studies). Meantime, I think we can expect to see more studies coming out of orgs like PatientsLikeMe, ACOR, and others.