DRIVING TRENDS IN DIGITAL HEALTH
The advent of digital health promises to transform healthcare as we know it. In fact, artificial intelligence and machine learning algorithms promise to decipher big data, delivering highly personalized therapeutics, which is heralded by many as a revolution more significant than anything witnessed in the history of medicine.1 While the veracity of this notion has yet to be proven, it’s safe to say that digital health and its budding technological frameworks will continue to advance and permeate the continuum of care at all levels.
It’s an incredibly exciting time to be in the healthcare profession, but as technology delivers data, that data breeds awareness. Many patients are becoming more attuned to their health and better informed of the conditions affecting their well-being. Heightened public access to knowledge places pressure on clinicians to deliver a standard of care that is customized to the personal needs and lifestyles of their patients.
To meet this growing demand, we must develop tools that help clinicians better recognize their patients as unique individuals by bridging the gap between data and insights. Two important trends driving this change across the digital health landscape include the consumerization of healthcare and artificial intelligence.
Consumerization of Healthcare
For more than a century, medical devices were handled only by highly trained healthcare personnel and administered within clinical environments. Over the past twenty years, however, we’ve witnessed an explosion of healthcare devices becoming widely available at retail, which is intended for untrained consumers outside of clinical settings. Consequently, patients are now better equipped to manage a rapidly widening range of conditions without the direct involvement of medical practitioners. While this approach has become commonplace with a glut of analog medical devices, the emerging “digital” layer delivers new ways of capturing, synthesizing and sharing this data to ensure that medical devices are being used properly and consistently. In theory, this promotes compliance and improves patient outcomes by enhancing user experience, uncovering key insights, and providing timely opportunities for clinical intervention as needed.
While this seems ideal, disparate and unconnected data sources pose a major challenge to reaching accurate diagnoses and prescribing bespoke treatments. For example, one device identifies that a patient has become sedentary while a separate device indicates that her blood pressure is trending upward. If fortunate enough to be provided with both inputs, a clinician might blindly suggest that she increases her exercise level and daily step count. Worse yet, he might simply increase the dosage of her blood pressure meds. What the doctor hasn’t discovered is why the patient has become sedentary. While it may be something easily discernable and temporary like a flu virus or an ankle injury, the root cause is often more complex and difficult to recognize. Perhaps she is feeling unusually fatigued, possibly suggesting an undiagnosed condition. Perhaps her diet is insufficient. Perhaps she is depressed and lacks the emotional fortitude to exercise.
In this case, technology has not supplanted the need for interaction with a qualified practitioner. Useful, but limited, data points fail to deliver the critical insights needed to identify and address the underlying cause for her malaise. While, at present, there is no all-in-one device that monitors every aspect of human vitality, we can easily supervise a wide range of consequential behaviors and key vital signs, albeit through multiple systems. Eliciting valuable insights from these systems requires a level of interoperability that is currently lacking in most scenarios. To establish a connected framework from which critical insights can be drawn, we must develop systems capable of compiling and synthesizing data from multiple sources while we continue pursue greater integration and interplay between our digital devices. Given that we’re faced with reams of data coming from disparate sources, could Artificial Intelligence (AI) and machine learning be the solution?
Artificial Intelligence (AI)
Many scientists are eager to posture artificial intelligence as the single most promising advance ever made in medicine. Considering that computers can readily digest and process virtually infinite amounts of data, it’s justifiable to conclude that machines will diagnose afflictions and prescribe optimal therapies more accurately and efficiently than the greatest minds in medicine. While it seems plausible, that vision is far from being realized. When IBM applied AI to transform massive data sets into better and more accurate diagnoses, the company quickly ran into trouble. Ultimately, Ajay Royyuru, IBM’s vice president of healthcare and life sciences research stated, “Diagnosis is not the place to go…That’s something the experts do pretty well. It’s a hard task, and no matter how well you do it with AI, it’s not going to displace the expert practitioner.”2 Following this failed attempt, IBM turned Watson’s computing power toward developing personalized treatments for cancer. If deep learning has the capacity to absorb and analyze wide-ranging data sets supplied by wearable sensors, blood labs, genomic information, and literature on medical research, surely it will identify important variables, subtleties, and patterns that are indiscernible to the human eye. Again, the experiment fell short of expectations, but not because of an inability to scan and process data. AI proved incapable of deciphering the ambiguity of medical records and understanding the ‘point’ of published journal articles.2 Although AI can capture and process statistics, it cannot think abstractly to garner insights, at least in its current state.
This is not intended to suggest that AI has no place in healthcare, nor is it intended to suggest that AI is incapable of achieving its lofty vision sometime in the future. It simply hasn’t reached a point where it can logically curate the myriad of inputs and enormous variabilities in a data structure to develop meaningful insights. In genomics, where data is highly structured and often binary, AI is already saving lives. Medical geneticists now use AI to sequence and interpret genomes at a record-setting pace.3 This enables clinicians to identify the root causes of rare and serious afflictions quickly. In turn, they can swiftly apply the appropriate therapies fast enough to prevent irreparable damage or death.
At present, AI is not an “Ultra Doctor” that replaces human practitioners, and in fact, it never should be. Instead, AI and deep learning algorithms are best used as powerful tools that accelerate critical but mundane tasks that distract clinicians from spending quality one-on-one time with their patients. In this combined approach, AI offers the extraordinary power to humanize medicine by restoring the much-needed connection between patients and practitioners.1
- Topol, Eric., “Deep Medicine,” New York, NY, Basic Books. 2019
- Strickland, Eliza., “How IBM Watson Overpromised and Underdelivered on AI Health Care,” IEEE Spectrum, 02 April 2019.
- Sisson, Paul., “Rady Children’s Institute Sets Guinness World Record,” San Diego Union Tribune. 2018.
RELATED PROJECT WORK
WANT TO LEARN MORE?
Start a conversation with Radius to learn how we can help you take your Healthcare project to the next level.