At the New Jersey Technology Council’s (NJTC) 2012 N.J. Health Information Technology Summit in July, Christine Kretz, Solutions Executive for Healthcare at IBM, gave the N.J. healthcare industry a look into the future, focusing on IBM’s Watson, the natural language processor that won Jeopardy last year.

Kretz told an audience of approximately 170 healthcare IT professionals at the N.J. Hospital Association Conference Center in Princeton, that IBM decided that healthcare will be the first industry it will focus on to see if Watson can be used practically.

Electronic Health Records (EHR) are great, she said, “but the majority of the important things you might want to know are buried in the nurses and doctors notes, in the unstructured texts.” Doctors are overloaded with data and have little spare time to get into the depths of the information just jotted down during a conversation.

However, Watson isn’t just a data mining tool for unstructured data. It needs to be trained with the right answers, so that it has background to use to make decisions and determine probabilities, Kretz said.

For its first test of Watson, IBM began a small project with Wellpoint, a health benefits company, which is using the system to improve best practices for patient treatment.  Normally, when a patient is about to undergo treatment, the company uses an experienced nurse to read patient information and compare it to preset rules for best practices to authorize patient treatments, Kretz said.

In this project, Wellpoint took 20,000 sets of patient data along with the right answers supplied by experienced nurses, and trained Watson, she said. Watson now is being tested to see if the answers the computer comes up with compare favorably against those developed by trained individuals. A nurse is still needed to evaluate the results, but Watson can go through the data and make comparisons faster. The goal is to make quicker approvals for treatments and provide better patient care. If all goes well, in December, Wellpoint will use Watson into real life situations, she said.

IBM is also cooperating with New York-based Memorial Sloan Kettering Cancer Center (MSK)to use Watson to help make the right choice in drug or radiation protocols. In oncology, if you give a wrong drug or radiation protocol, you usually don’t have a second chance to fix it, Kretz explained, so doctors have been looking for a system that would bring consistent good results to patient outcomes. MSK has very smart doctors who are training Watson “so we always get it right the first time, specialized to the patient.” Watson will not only have specialized patient information, but industry wide information to work with.

This has implications for connected medicine, she noted. What if, after MSK has programmed Watson with its knowledge, it could continually update it with new information and share Watson’s knowledge with doctors all over the world? “The implications are staggering,” she said. Groups of oncology experts worldwide could collaborate on a tool that everyone can share to make outcomes consistently better.

IBM also envisions Watson being used in a more generalized portal approach­, a sort of WebMD that knows your medical background. So when a web visitor asks questions about a particular problem, Watson gives alternatives based on that person’s medical background. Or Watson could be used to back up an “ask a nurse” call center.

The difficulties in making this future come true are vast. Watson is very expensive; it works on a really big platform. Further, it takes a lot of time and labor (which translates to money) to train Watson and build up a corpus of accurate information from which it can make determinations. A lot of the content Watson has to learn isn’t free. IBM is negotiating one-by-one with content companies to allow their information to be used and shared, Kretz said.

In the meantime, some pieces of Watson are being used in smaller more discrete projects, which Kertz said are faster to bring to market and monetize. One hospital is using the system for predictive analytics, figuring out which of the patients hospitalized for congestive heart failure will be back in the hospital in 30 days. The hospital looked at EHRs and went through nurses’ and doctors’ notes to find out what caused patients to be readmitted.

While some factors were obvious, some were not. The hospital found 18 predictors of readmission, six that could be affected by changing something. A person’s living condition was one of those. Patients who lived alone or couldn’t drive would take their medication until it ran out and then not renew because they couldn’t get to the pharmacy. The hospital was able to do outreach and make sure that patients got prescriptions sent to their homes.