For decades, the computer world has embraced the phrase “garbage in, garbage out” (GIGO) to illustrate the concept that the quality of information delivered by a computer is determined by the quality of information that’s received. In the early days of computing, you could create plenty of garbage by accidentally key-punching a stray hole in your programming card.
Today’s computers are exponentially more sophisticated than those of yesteryear, but the basic tenet of GIGO still holds true: the information produced by a computer is only as good as the information entered by human beings – even if the computer is the IBM Watson supercomputer.
Three years ago, IBM introduced its Watson for Oncology platform, which was designed to recommend the best treatments for cancer. According to IBM, more than 50 hospitals worldwide have used Watson to support care for 14,000 individuals. IBM claims there is “no question” that Watson is having a positive impact on oncology care worldwide.
Despite Watson’s contributions to better cancer care, a recent article by STAT writers Casey Ross and Ike Swetlitz concludes that Watson still has not met the high expectations originally set by IBM. In fact, the authors claim that, “IBM, in its rush to bolster flagging revenue, unleashed a product without fully assessing the challenges of deploying it in hospitals globally.”
Perhaps Watson has not yet revolutionized medicine, but the potential for Watson is still exciting to consider. Watson’s biggest challenge right now is fixing the GIGO phenomenon—and no, I am not suggesting that Watson is producing “garbage.” Instead I am pointing out that Watson’s output is only as good as the input.
Watson is a powerful platform but needs some tweaks in order to make a stronger impact on healthcare. Consider some of the current issues:
Watson’s knowledge. Watson offers treatment recommendations that are based on the insights of the human beings who feed Watson their medical opinions. Specifically, the guidance is based on the input of a couple dozen physicians from the renown Memorial Sloan Kettering Cancer Center. The physicians’ recommendations are biased in favor of their own training and experience, which may vary from the training and experience of doctors in other institutions across the world. Furthermore, Watson delivers recommendations that mirror existing physician recommendations, which may not be closely linked to evidence-based outcomes. The guidance Watson provides may be solid, but it doesn’t preclude the possibility that alternate therapies may also be effective.
Updating Watson. Treatment guidelines for cancer can change rapidly with the introduction of new drug therapies or innovative technologies. For every update, a human must analyze the details, determine what is relevant, and then train Watson – which can be a time-consuming process and requires a certain level of technical expertise. In general, making changes in Watson is cumbersome, which is also an issue for organizations trying to customize the system to reflect their unique treatment practices, differences in available drugs, or financial factors.
Watson’s insights. Physicians need access to Watson’s treatment recommendations within existing workflows and at the point care – and should not have to access multiple systems to analyze treatment options. Instead, clinicians need the ability to review Watson’s insights from within the patient’s record when the patient is in front of them.
Furthermore, users need easier options for inputting their own insights. Like a human, Watson doesn’t know what it doesn’t know. While its recommendations are based on the opinions of some of oncology’s top experts, other great opinions from other experts may be overlooked because users can’t easily add new information. If Watson’s insights are not constantly updated, physicians risk missing out on ground-breaking therapies that could be life-saving for certain patients.
Watson enhancements will advance cancer care
To truly revolutionize cancer care, IBM needs to incorporate additional tools to facilitate the addition of new knowledge into Watson, as well as make its recommendations more readily available to physicians at the point of care. Data that is input into Watson must be in a structured format, so ideally users need the ability to capture high-quality coded clinical data at the point of care. This eliminates the need for users to manually review free-text sections of a patient’s chart to identify relevant information, and then convert those details into a structured format for input into Watson. By giving clinicians the ability to create coded information as a by-product of the documentation process, users can more easily give Watson precise data for a given condition.
IBM can further enhance the usability of Watson by making its treatment recommendations available to physicians from within their existing workflows. This requires a translation of Watson’s data into a format that is actionable for physicians at the point of care.
Watson for Oncology holds great promise for supplementing the wisdom of clinicians and aiding in the fight to eradicate cancer. Patients will be the ultimate beneficiaries of IBM’s efforts to further enhance the power and usability of Watson.
By Dr. Jay Anders – Chief Medical Officer at Medicomp Systems
The views, opinions and positions expressed within these guest posts are those of the author alone and do not represent those of Becker's Hospital Review/Becker's Healthcare. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them.