If you’ve been following the health care debate in the US, it’s become fairly clear that the current trajectory of medical costs will soon be unsustainable for the economy. The latest government figures has the average US health care spend per person at over $8,000, and is projected to top $13,000 by 2018. Whether the latest health care legislation will do much to curb these costs is debatable.
If that $13,000 per capita figure holds up, that means about 20 percent of the nation’s GDP will be spent on medical bills. Other developed nations are currently about twice as efficient as the US, but even there health care cost are outrunning incomes. Fortunately, economic forces that strong have a way of disrupting the status quo.
Probably the lowest hanging fruit for optimizing the health care sector is in information technology. Even though we think of medicine as a high-tech endeavor, it’s mostly based on 30-year-old IT infrastructure overlaid with a manual labor approach to data collection and analysis. Essentially we have a system using 20th century computing technology, but with 21st century wages.
Just going to a doctor’s office and filling out a medical history form (on paper!) for the 100th time should give you some idea of how antiquated the health care industry has become. It’s as if the Internet was never invented.
But it’s not just about your medical records ending up in isolated silos. The amount of data that can be applied to your health is actually growing by leaps and bounds. The results of medical research, genomic studies, and clinical drug trials are accumulating at an exponential rate. Like most sectors nowadays, health care revolves around data.
In general though, your health care provider doesn’t do anything with all this information since the analysis has to done by a time-constrained, high-paid specialist, i.e., your doctor. But that could soon change. The latest advanced analytics technologies are looking to mine these rich medical data repositories and transform the nature of health care forever. Not surprisingly, IT companies are lining up to get a piece of the action.
IBM, in particular, has been pushing its analytics story for all sorts of medical applications. Last week, the compay announced it was expanding its Dallas-based Health Analytics Solution Center with additional people and technology.
Part of this is about sliding the IBM Watson supercomputing technology into a medical setting. With it’s impressive Jeopardy performance under its belt, IBM is now applying HPC-type analytics to understand medical text. Specifically, they want to combine Watson’s smarts with voice recognition technology from Nuance Communications to connect doctors to their patients’ medical data via a handheld device like a tablet or smart phone. From the press release:
By using analytics to determine hidden meaning buried in medical records, pathology reports, images and comparative data, computers can extract relevant patient data and present it to physicians, ultimately leading to improved patient care.
Analytics vendor SAS is also in the game. In May, they unveiled a new Center for Health Analytics and Insights organization that is designed to apply advanced analytics across health care and life sciences. Although the specifics were a little thin, the group will focus on “evidence-based medicine, adaptive clinical research, cost mitigation and many aspects of customer intelligence.”
It’s not all about clinical care though. One of the most expensive undertakings of the health care industry is ensuring drug safety. Both the FDA and pharma have had some spectacular failures in this area, the most recent being Vioxx, a pain-relief drug that was pulled from the market in 2004 after it was discovered that it was causing strokes and heart attacks in some patients.
A recent study by the RAND Corporation suggests data mining can be used to find some of these dangerous drugs before they enter into widespread usage. RAND CTO Siddhartha Dalal and researcher Kanaka Shetty developed an algorithm to search the PubMed database to uncover these bad players. The software employed machine learning algorithms in order to provide the sophistication necessary to differentiate truly dangerous compounds from ones that only looked suspicious (false positives). According to the authors, the algorithm uncovered 54 percent of all detected FDA warnings using just the literature published before warnings were issued.
A more ambitious medical technology is envisioned by the X PRIZE Foundation, a non-profit devoted to encouraging revolutionary technologies. Recently they teamed with Qualcomm to come up with the Tricorder X PRIZE, offering a $10 million award to develop “a mobile solution that can diagnose patients better than or equal to a panel of board certified physicians.” In other words, make the Star Trek tricorder a reality.
The device is intended to bring together wireless sensors, cloud computing, and other technologies to perform the initial diagnosis, and direct them to a “real” doctor if the situation warrants. Presumably the cloud computing component will support the necessary data mining and expert system intelligence, while the tricorder itself would mostly act as the data collection interface and do some medical imaging perhaps. The X PRIZE Foundation will publish the specific design requirements later this year, with the competition expected to launch in 2012.
None of these solutions are being promoted as substitutes for doctors or other medical professionals. Inevitably though, if these technologies become established, these jobs will be very different. With powerful analytics available, doctors won’t have to memorize all the information about the biology, drugs, and medical procedures any more. In truth, they can’t even do that today; there is already far too much data, and it continues to expand.
In an analytics-supported health care system, medical practitioners will need to do less data collection and analysis and more meta-data analysis. Just as today, writers don’t need to know how to spell words (remember, 50 years ago a spell checker was a person, not a piece of software) doctors will not need to memorize which drugs are applicable to which diseases. And that means a lot fewer doctor and less supporting staff. Essentially we’ll be replacing very expensive PhD’s with very cheap computer cycles.
If that seems like a scary prospect, consider the more frightening scenario of a health care system that bypassed this technology and tried to burden medical practitioners with the data deluge. Also consider that without advanced analytics, the majority of the population will be burdened by the long-term costs of sub-standard medical care.
Beyond that, advanced analytics will also be involved in propelling other health care technology forward, including drug discovery, genomics, and the whole field of personalized medicine. Many of these advances will enable medical conditions like heart disease, cancer and diabetes to be prevented, which is a far less expensive proposition than treatment.
It’s reasonable to be optimistic here. Nature abhors a vacuum — in fact, any sort of stark discontinuity. Our problematic health care model will eventually be transformed by technologies that make economic sense. Advanced analytics is poised to be a big part of this.