Because every place worth going can be found on a topo map.
This week in medical visualizations…
I can’t make any claim to understand the the full implications of real time, detailed visualization in medicine, but the x-ray specs certainly seem to show promise, at least on the high-end party favor circuit. Though the computing power and finesse was awe inspiring, I was left with an impression, perhaps unfairly, that even the creators of the software were uncertain about how exactly it could/should be used. To paraphrase the closing remarks of this video, the underlying motavation behin many of these developments is the observation that every great advance in science is predicated on a new tool that allows for that advance. These sorts of high-end real time visualizations are certainly a new tool, and combined with modern micro-surgery techniques I am certain will lead, someday, to a robot removing my brain tumor cell by cell leaving the rest untouched. Or maybe not. Really, I’m fine with the cool pictures of lungs and I’ll leave it to the MDs to tell me if this stuff is actually useful. They often seem to know what they’re doing.
Jamie Jeywood, though, definitely speaks my language. His PatientsLikeMe project is a wonder of crowd-sourcing, statistical modeling, research methodology and data visualization. To briefly explain, patientslikeme is a website, currently with 40,000+ contributers, which patients, suffering from ailments from HIV to MS, can submit their medical data over time to create a on-going health profile and chart the progression of their disease, including notes of medicines, treatments, side effects and quality of life. This is nothing that isn’t included in a regular patient file, but the beauty of this site is it allows aggregation of patients with similar backgrounds and medical histories and treatments in order to create a predicted disease progression for each individual, dependent on treatment.
The two main possibilities this data provides, especially organized as accessibly as it is on the site, are for patients to make better treatment decisions given their specific profiles, especially concerning interacting side-effects, and for doctors to evaluate treatments already approved for use to see if they may be better or worse than initially expected. In accomplishing this, patientslikeme effectively violates nearly every major methodological rule: There is no random selection or assignment to conditions. There is almost certainly selection bias in who joins the site and drop-out bias in who leaves. How strong these effects are, which is to say, how generalizable the information on patienslikeme is, is an empirical question that will require a series of comparisons between its findings and more traditionally rigorous methods.
What these effects do not mater for is the first possible use of the site: Patients who end up on patientslikeme will usually look like each other. The selection bias which may harm generalizability should ensure a greater homogeneity in the patient community, making the data more, not less, applicable for current members. The drop out rate is of greater concern, but there are statistical methods for dealing with that as well, so hopefully Jeywood et al. are being smart about that. If I, or someone close to me, am ever unlucky enough to suffer from a serious ailment, I will definitely turn to this site to advise me and my physician as we work out treatment options.
All in all, these videos gave me hope for the future, but in the meantime I’ll keep up the exercise and eating right just in case.
