Natural language processing and big data: The prescription for saving big bucks in healthcare?

The health industry could really do with a shot in the arm when it comes to cost-cutting, and big data could be the answer.
Written by David Shamah, Contributor

With health care costs spiraling across much of the world, figuring out ways to cut costs is a matter of financial life and death for insurers, healthcare providers, and governments, who (via the taxpayer) end up ultimately footing the bill for health costs.

Fortunately, there's a lot of data to mine in the health business, and an Israeli firm called CliniWorks is aiming to put it to work. This month, CliniWorks started a new project with drug giant Pfizer to help healthcare organizations identify areas where they could cut spending.

While CliniWorks' technology could be used in several industries, according to CEO Nitzan Sneh there's so much to do in healthcare, the company couldn't focus on another business right now even if it wanted to. "Our system analyzes free and unstructured text using natural language processing, and traditionally the medical business has been built on such data," said Sneh. "Pfizer had the idea of gathering up this data and analyzing it in order to see how costs could be cut and service improved, and it took them a year to figure out how big a job it was — and to call us in."

For years, doctors would take notes on patients' conditions and progress, recording them in files that in many cases grew to be rather thick — essentially useless for analysis, because handwriting can be difficult to read and enter into a database for analysis.

One would think that with the shift to electronic record-keeping, with medical staff entering information on tablets and computers directly into the database, that the situation would be improved. Pfizer thought so too — but the company forgot one important detail, said Sneh.

"It turns out that even with all the reams of data flowing into the database — and doctors are recording a lot more information now than they did in the pencil and pen era — most of it is unstructured, and therefore useless in a database."

If the electronic forms provided doctors include a variety of fields regarding patients and their conditions, it turns out that medical staff record much of the information incorrectly — putting data in the wrong fields, or leaving the bulk of their analysis for the unstructured 'comments' field. As a result, a database designed to analyze information in patient records according to database categories is going to miss a great deal of information.

Enter CliniWorks, which makes a system that peruses all data fields, and assigns data to where it belongs using natural language processing. "It's a lot deeper than just identifying keywords. We scan free text at very high speeds, and using specialized algorithms we figure out where the information a doctor wrote will be most useful in analysis. Once the data is properly classified, it can be analyzed to make suggestions on more effective treatment, which drugs are better for specific situations, where there is waste that can be cut, and so on. The information is gathered into a HIPAA-compliant searchable repository, enabling rapid data query, analysis, and reporting."

Pfizer is best-known as a drug company, but it also has a healthcare management division, and that's the division behind the joint project with CliniWorks.

"Believe it or not, this is the first time anyone in the US is trying to tie patient outcomes with expenditures during care, scientifically analyzing whether the money we lavish on health care is being spent effectively," Sneh said. The usual system is for healthcare providers to negotiate with insurance companies over costs of care, but the cost-effectiveness of care — that is, whether patients were actually getting better, surviving longer and so on — had never really been seriously studied.

Sneh has a long history in the health tech business. In 1991, he launched the world's first digital video-capture technology for minimally invasive surgery as well as a non-invasive technology for early detection of cervical cancer, and CliniWorks has been instrumental in helping drug companies analyze their development of new drugs, helping to shave millions off the cost of development.

"We have somewhat of a reputation in this area – Pfizer solicited us to work with them on this," said Sneh. "We've been talking to hundreds of doctors and hospitals, and our pilot program has proven to be very successful so far. There's no question that our methods could save companies in any industry money — health and life sciences is a $3tn market worldwide and growing at 10 percent a year. If we can harness big data just in that business and use it to save companies money, I would say we did a good day's work."

Read more on big data


Editorial standards