Futurism logo

Lessons Learned from IBM Watson’s Failure to Transform Medicine

IBM Watson’s failure provides lessons to guide the future of remote patient monitoring.

By IQV CLOUDPublished 2 years ago 6 min read
Like
IBM Watson’s Failure to Transform Medicine

A decade ago, investors touted the acquisition of Watson Health, an artificial intelligence (AI) powered healthcare analytics platform, by IBM as the strategic chess move that would revolutionize healthcare. In sharp contrast to enthusiastic predictions, synergies failed to actualize and the merger became a liability, an embarrassment, a confusing investment. IBM was forced to sell Watson Health on the auction block at salvage value and it is now owned by a private equity firm, Francisco Partners. Years of datasets and healthcare analytics products since Watson’s inception in 2011 were liquidated for a fraction of their worth. So, what happened?

Watson’s AI system was never intended to diagnose medical conditions independently. Rather, it was meant to provide one tool in the toolbox, to aid care teams who are working against the clock, and to decrease burden. The thought was that Watson’s AI system could parse data points from treatment guidelines, academic research papers, and medical records. Then it would recognize patterns in the data points it had access to. Those patterns could then help providers in their clinical decision making. For example, the AI algorithm might note that patients with diabetes were more likely to be hospitalized with COVID-19, and then flag patients with high glucose levels to not receive immunosuppressive therapies like steroids. In any mathematical formula the inputs determine the outputs. The same holds true with AI algorithms. Predictions from AI algorithms become inaccurate and possibly even dangerous when poor-quality data is inputted to train the algorithms. Watson’s system was relying heavily on electronic medical records (EMRs). Are EMRs reliable, precise, and consistently accurate?

Healthcare’s Shortcomings Present Opportunities

Medical records are crucial for the collection, organization, and communication of health data. They serve as a receptacle for lab values, vital signs, radiographs, medications and clinical notes. However, they have significant short-comings. This is particularly true with electronic medical records, where critical health information is inputted by clicking buttons and choosing from drop-down lists. With an accidental click, the record is inaccurate. Furthermore, patients and the ever-complex human body cannot be reduced to this level of simplicity.

EMRs have been blamed for provider burn-out. For every one hour of face-to-face patient care, providers spend two hours in clinical documentation and in responding to electronic charting needs. This documentation is organized for insurance, administrative, and liability needs. In other words, the notes are not tailored to optimizing clinical care; instead they are optimized for billers and administrators. Medical records ignore some of the most clinically useful data elements. For instance, social determinants—a strong predictor of healthcare outcomes—are missing in EMRs. These include a variety of socioeconomic variables: birthplace, current living situation, culture, education level, income, gender, safety, access to transportation, food accessibility, work-status, race, religion, and more. If social determinants are recognized and addressed preventatively, we can predict and prevent disease. Novel remotely-collected data streams are filling healthcare gaps by recording patients’ social determinants and allowing for providers to measure their impact.

Remote Care as an Extension of the Electronic Health Record

The increased demand for healthcare delivered at home during the COVID-19 pandemic has illustrated the power of digital and AI-powered applications to close healthcare gaps, including those related to socioeconomic disparities. Remote care platforms offer the ability for providers to easily monitor their patients even when they are not in the office. Patients enter information through validated assessments from their smart phones, tablets or computers and these data points communicate critical health information to providers in-between office visits. Wearable devices further enhance remote platforms by allowing for continuous monitoring of subjective elements such as vital signs, number of steps, and sleep quality. When AI algorithms combine data from EMRs, remote care platforms, and wearables, they suddenly become more capable of connecting dots in ways that are accurate. Perhaps if Watson Health and IBM had embraced remote platforms, they would never have been in the current predicament.

Digital health data streams from remote patient monitoring platforms are starting to become integrated in patient’s charts, and seen as an extension of the traditional medical record. More widespread use is critical to solving our nation’s healthcare problems. These digital health records are separate from what occurs at in-person visits and instead provide detail into a patient’s life outside of a clinic. A patient who has high blood pressure only during clinic visits due to medical anxiety has a condition known as white coat hypertension. Such patients do not have high blood pressure when at home. Unless a care team has insight into the home vital signs, the patient is likely to erroneously end up on a lifelong prescription when it is not necessary.

Remote Data Streams Close Healthcare Gaps

When we develop inputs for AI algorithms, we must carefully consider the outcomes that we are looking for. A comprehensive algorithm needs comprehensive inputs. Let us consider the case of chronic pain, which affects 20% of Americans and is a complex, biopsychosocial phenomenon that requires multiple points of intervention. We can easily find the diagnosis of chronic pain in a patient’s chart, but cannot decipher the most significance outcome— the effect of chronic pain on that patient’s life. Often missing in patient’s charts are the subjective assessments that are tied to disease.

Subjective outcomes are key in medicine. To assess the effectiveness of medications, procedures or devices, the highest academic journals and the FDA require that standardized and validated assessments are filled out by patients frequently in the context of clinical trials. These subjective assessments are known as patient-reported outcomes and are currently all the buzz. Patient-reported outcomes are central to research, clinical practice, quality control and benchmarking.

New remote technologies offer the opportunity to combine subjective and objective assessments. These assessments, combined with traditional EMR data, provide the comprehensive data required for AI algorithms to function effectively. Patient Premier, Inc. has effectively bridged this healthcare gap through its product Pain Scored (painscored.com). Pain Scored is a web and mobile platform delivering validated pain, mood, and functional assessments from patients while at home to their providers. These insights are recorded and trended over time to create visual reports representing a patient’s chronic pain condition tracked over time. These reports are accessible on demand to the patient’s care team to provide enhanced clinical decision-making capabilities. They can be integrated into EMRs and combined with objective, continuous data from wearables.

Lessons Learned

The failure of IBM Watson to transform medicine offers an opportunity to perform a post-mortem analysis. We can expand the inputs for AI algorithms to include subjective data from remote platforms and objective data from wearables. Remote platforms can even be tailored to include critical information on social determinants of health. We don’t need a crystal ball to predict the benefits to providers, patients, insurers and healthcare organizations.

artificial intelligence
Like

About the Creator

IQV CLOUD

IQV Cloud ranked top among best cyber security companies in Los Angeles USA. Our Data Security Experts and Analyst provide.

For More Information Please visit Our

Website=https://iqvcloud.com/

Reader insights

Be the first to share your insights about this piece.

How does it work?

Add your insights

Comments

There are no comments for this story

Be the first to respond and start the conversation.

Sign in to comment

    Find us on social media

    Miscellaneous links

    • Explore
    • Contact
    • Privacy Policy
    • Terms of Use
    • Support

    © 2024 Creatd, Inc. All Rights Reserved.