BOOK AND SOFTWARE REVIEWS
Yom-Tov, Elad. Crowdsourced health. How what you do on the Internet will improve medicine. Cambridge, MA: MIT Press, 2016. xxviii, [8], 144 p. ISBN 978-0-262-03450-0. $24.95/£18.95.
Unless your physician is very much old school, you are likely, these days, to find that he or she welcomes the fact that you search the health-related sites on the Internet before setting up an appointment. No general practitioner can be expert in every area of medicine and the information that the patient brings to the appointment can help in the diagnosis of the problem. The result of the availability of health information on the Internet has led to millions of searches every day by people concerned about their health. Too often, of course, they are inadequately prepared for the scientific research sites or for those intended for medical practitioners, but, if they restrict their searches to those sites intended for patients, as well as those run by patients as self-help sites, the results can be extremely useful.
Yom-Tov's subject is not exactly this: he is interested in searching for health information, but from the perspective of mining the data generated by those searches for the further benefit of medical research. To take a simple example: persons at risk of stroke are prescribed blood-thinning drugs, of which the most common and the longest used is warfarin, and the effects and control of warfarin are well researched. However, a new set of drugs, known as 'new oral anti-coagulants', is now available, but there is relatively little information available on how they interact with other drugs a patient may be taking. This is where data-mining the patient sites can be of help, since people will report their experiences of these drugs, including the side effects experienced and their interaction with other drugs.
Of course, before a drug is released for use it undergoes massive testing with, often, thousands of patients. Even big trials, however, cannot encompass all the possibilities of side effects or drug interactions that may exist once the drug is on the market. Data mining health information sites offers pharmaceutical companies a very cost-effective way of gathering information. And, of course, use of the sites, helps the patient directly by revealing that the experienced side effect is well known, or at least shared by others, or that there is no known adverse interaction with another drug.
To illustrate the significance of Internet data, Yom-Tov mentions the site PatientsLikeMe.com, which today 'serves more than 200,000 patients suffering from 1,500 diseases'. Particularly, the site was set up to serve patients with ALS (Amyotrophic Lateral Sclerosis): one study showed that, by comparing patients who used a lithium carbonate supplement with those who did not, the results of a small clinical test of the supplement were shown to be false.
This points to a different way of doing medical research, or, rather, a supplementary way, which is the author's main message. There is, of course, the problem of patient privacy to be dealt with, but users of sites often use pseudonyms and search data can be anonymised (although the experience of AOL in 2006 presents a warning). Yom-Tov comments:
I anticipate that if we can achieve an appropriate balance between privacy and ethics, on the one hand, and usefulness, on the other, more and more Internet data will become available to the research community. This availability will make it easier to collect the data and will help health experts to gain insights from them.
An appendix tells those researchers how to get access to the data by using, for example, the Yahoo Answers API, which allows one to query the site. In sum, this is a very interesting little book, with useful notes and a good index, it will be of interest to many Internet researchers, as well as to the medical research community.
