Our Online Dialogue colleague and analyst Arend Zwaneveld is for Webanalists.com started a series of articles. He looks at several real-world cases of Big Data examples and what we can learn from them. Today his second article: Big Data example: tailored medical care with omniscient Big Data physician assistant

Watson -the now world-famous Jeopardy-playing supercomputer-understands spoken word, jokes and can learn. So well, in fact, that after its illustrious victory, the computer was retrained as a physician assistant. The former game show winner now provides doctors with medical [second] opinions using Big Data.

But doctor ... what are you doing now?

Even in the role of physician assistant, Watson -called “Watson-as-a-service for Hospitals” in full by IBM- has now had its first successes: when taking the history of a pregnant patient, the retrained supercomputer prescribed a drug of which the accompanying physician-of-meat-and-blood was alarmed to say “that drug may not be prescribed to pregnant women at all”.

Had the supercomputer made a mistake? No!

As it turned out, when the flesh-and-blood doctor got his degree, the prevailing belief was still that the prescription drug would not be suitable for women who were pregnant ... but this truth has since been proved outdated. Physician assistant Watson was up to date on the most current treatment methods and developments in medical science, the doctor-of-flesh-and-blood was not!

Behind every good doctor is ... good Big Data!

Big data enables patient-specific treatment

Whether a standard treatment method (protocol) is effective or not, a single physician in a practice or hospital cannot say on the basis of one, two or ten individual patients. Only on the basis of larger numbers of patients treated can a physician -after a while- Make a statement about the effectiveness of a treatment in general.

Physician assistant Watson, on the other hand, can find patterns in the results of global (new/experimental) treatment methods direct discover and determine the optimal treatment (drug, dose, length of course) for the individual patient.

For example, men and women respond differently to drugs. However, approval of new drugs is still based on biomedical research on male subjects.

“Even with medications that are known to have significant differences between how men and how women respond to them, according to a 2005 study, there is usually no sex-specific dosage on the container. This will undoubtedly play a role in why women are one and a half times more likely to have an adverse reaction to medications than men.”

Big data accelerates knowledge sharing

The way physician assistant Watson arrived at his custom treatment advice comes is thus a lot faster and more effective than the alternative: the lengthy -not standardized- process of establishing best practices in treating diseases (protocols).

• Testing new medical treatment method.
• Medical meta-studies (studies on studies).
• Medical conferences at which these results are shared and discussed.
• Establish new best practice in protocol.
• Communicate new protocols to all physicians (e.g. incorporate into medical education programs).

Then, through conferences and professional journals, these latest insights -hopefully- reach your own physician.

Predictive analytics based on historical input: big data!

In this example, large amounts of historical data with which Watson has used, based on real-time information is immediately given useful advice. The system does not limit itself to answering concrete questions such as “would there be a connection between BMI and the effectiveness of drug X?”, the system itself discovers (subtle) connections that you can only discover when you have large amounts of data at your disposal: predictive analytics. Sometimes a lot of data is better than a good model!

Sometimes a lot of data is better than a good model

Specifically, the following data are combined with each other:

• Medical patient records
• Anamnesis: what complaints did patients come to the doctor with?
• Prescription: what treatment did the doctor prescribe for them? (drug,dose, cure or lifestyle advice)
• What side effects were reported?
• How effective was the prescribed treatment on health status of each patient?
• Background information from patients
• Age
• Gender
• Lifestyle
• Living conditions
• Education level
• Real-time input:
  • Patient's health complaints (anamnesis).
  • Doctor's diagnosis
  • Background information from patient

Results:

• Treating patients based on the latest medical knowledge (treatment method/drug, dose, duration)
• More effective medical treatment by tailoring to individual patient (gender, age, BMI, etc...)
• Reduction in medical misses
• Saving medical costs through more effective treatments (that's what health insurer WellPoint was all about in the first place!)

Conclusions

1. Big Data is going to help doctors treat patients in a tailored way.
2. Ordinary data is often static, difficult to access and dusty. Big data is proactive and relevant.
3. Big Data will also bring knowledge sharing and
   -development improve.*
4. One thing even Big Data will not be able to solve: deciphering the doctor's handwriting! 🙂

Big Data as a Game changer of the labor market

Not only the profession of medicine, but also more and more other professions -including that of web analyst!- will change content under the influence of Big Data.

Read more about this on Friday in the article “Big Data as a Game changer of the labor market“.

Originally posted on July 10, 2012 at webanalists.com

Resources

[1] Frans Bentlage, “Smarter Analytics Leader Benelux” at IBM
http://new.livestream.com/eventproducent/onlinetuesday

[2] IBM's ‘Watson-as-a-Service’ Ready to Crunch Big Data, March 12, 2012

http://www.wired.com/cloudline/2012/03/ibm-watson-cloud/

[3] WellPoint, Cigna and large hospital chains expected to actively engage in medical home partnerships

http://www.ft.com/cms/s/2/424170a0-2037-11e0-a6fb-00144feab49a.html#axzz1xxHwWy7G

[4] Researchers ignore women - Patient care suffers from gender inequality

http://www.wetenschap24.nl/nieuws/artikelen/2010/juni/Onderzoekers-negeren-vrouwen.html

[5] Improving medical protocols by formal methods.

http://www.cs.vu.nl/~frankh/abstracts/AIM06.html