Artificial intelligence-powered technologies are driving meaningful change across the entire patient journey.
From apocalyptic prognostications to impassioned positions — and everything in between — it seems like everyone and their mother has developed an opinion on the role artificial intelligence (AI) will play in shaping society in the coming decades. The disparity between each of these notwithstanding, what’s clear is that “narrow AI” is already starting to have an impact on everything from software development to education to insurance.
Despite multiple dalliances with AI stretching all the way back to the 1970s, my industry, healthcare, has yet to embrace AI with the same vigor as many others. Fortunately, this is finally starting to change.
Consulting firm Frost & Sullivan reports that the healthcare AI market is set to experience a compound annual growth rate of 40 percent through 2021, largely because AI has the potential to improve healthcare outcomes by 30 to 40 percent while simultaneously cutting the costs of treatment in half.
“AI systems are poised to transform how we think about disease diagnosis and treatment,” says Frost & Sullivan Transformational Health Industry Analyst Harpreet Singh Buttar. “Augmenting the expertise of trained clinicians, AI systems will provide an added layer of decision support capable of helping mitigate oversights or errors in care administration.”
The value of AI in the healthcare space is not limited to clinical settings, however. By facilitating medical diagnostics, improving pharmaceutical marketing, and reducing medication nonadherence, AI-powered technologies are driving much-needed change at nearly every stage of the patient journey.
Facilitating Medical Diagnostics
Diagnostic errors play a role in around 10 percent of patient deaths and between 6 and 17 percent of all hospital complications. As exceptionally skilled as most healthcare providers (HCPs) are, in many ways, the human mind remains fallible.
As Andrew Beck, Director of Bioinformatics at Beth Israel Deaconess Medical Center Cancer Research Institute points out, “Identifying the presence or absence of metastatic cancer in a patient’s lymph nodes is a routine and critically important task for pathologists, [but] peering into [a] microscope to sift through millions of normal cells to identify just a few malignant cells can prove extremely laborious using conventional methods.”
That’s why Beck and his team built an automated diagnostic tool using a deep learning algorithm trained to differentiate between cancerous and noncancerous cells. In an evaluation conducted in 2016, the automated tool achieved a diagnostic success rate of 92 percent — just 4 percentage points lower than human pathologists. What’s more, when Beck’s team combined human pathologists’ analyses with the analyses of the automated tool, the diagnostic success rate rose to a remarkable 99.5 percent.
Ultimately, Beck believes that his experiment barely scratches the surface of what a hybrid — that is, human and algorithmic — approach has to offer to medical diagnostics. “Our results…show that what the computer is doing is genuinely intelligent and that the combination of human and computer interpretations will result in more precise and more clinically valuable diagnoses to guide treatment decisions.”
Improving Healthcare Marketing
Once a patient has been diagnosed, the next step is to find a therapy that will cure — or at least mitigate the effects of — their condition. HCPs obviously have an outsized influence over which therapeutic regimen a patient adopts, but the importance of “Ask your doctor about [Drug X]” direct-to-patient messaging shouldn’t be underestimated.
Unfortunately, the pharmaceutical sector often finds itself talking past its core constituencies. In fact, one study found that as many as 45 percent of patients believe that pharmaceutical companies don’t understand their real needs. Not unlike the challenges of traditional medical diagnostics, this disconnect is first and foremost a problem of scale.
In the digital age, gauging patient behavior — the first step toward delivering relevant, tailored messaging — involves aggregating information drawn from a wide variety of datasets, including medical data like hospital records, lab results, and HCP notes and general data like media preferences, internet usage, and demographic information. Healthcare marketers must then draw out salient narratives and insights from their aggregated data — “connecting the dots,” so to speak — not just once, but on a rolling basis over the course of a campaign.
The reality is that executing such a data-driven approach at scale requires a superhuman amount of computing capacity. Not even the best, most experienced team of marketers is capable of organizing and analyzing millions of data points, which is where machine learning-based predictive analytics tools become invaluable.
By leveraging a properly-trained predictive analytics algorithm, a marketing team can gain unparalleled insight into their target patients, facilitating messaging based not on broad-strokes segmentation, but on analyses of the intricate — and often imperceptible to the human eye/mind — ways that a patient’s past behavior, personal characteristics, and current position in the patient journey interact.
Reducing Medication Nonadherence
Between 1988 and 1994, roughly 38 percent of adults living in the United States were taking at least one prescription drug. Over the subsequent two decades, that figure grew to 49 percent, driven in large part by a 100 percent increase in the number of adults taking three or more prescription drugs.
All told, according to research presented to the American Hospital Association in October 2016, “Total net spending on prescription drugs…has accelerated over the past year to $309.5 billion annually, making prescription drugs the fastest growing segment of the U.S. healthcare economy.”
Troublingly, a significant fraction of this $309.5 billion is going to waste. As many as half of the 3.2 billion prescriptions written in the U.S. each year aren’t taken as directed — if they’re even taken at all. This nonadherence leads to over $250 billion dollars of unnecessary costs, or roughly 13 percent of the country’s total annual healthcare expenditures.
But just as pharmaceutical marketers can use algorithmic tools to refine their patient targeting, HCPs can use algorithmic tools to reduce this systemic waste by getting a better sense of which of their patients are most prone to medication nonadherence. Everything from demographics and payer type to out of pocket costs and the prescribing HCP’s area of speciality bear upon the likelihood of a patient deviating from their prescribed regimen, and an AI-based approach is a robust way to take all of these factors into account.
Armed with the predictive outputs of such systems, HCPs are able to pinpoint which patients need additional support in order to remain on course with their treatment. Granted, better, more targeted communication isn’t a comprehensive solution for medication nonadherence, but research published in Medical Care suggests that poor HCP-patient communication results in a 19 percent higher risk of nonadherence. In other words, it’s a start.
Artificial Intelligence, Genuine Results
Artificial intelligence clearly has the potential to drive meaningful change at numerous points along the care continuum. But as Andrew Beck’s experiment made clear, the most effective deployments of AI in the healthcare space are those that augment human capabilities, not replace them.
Cutting-edge technology is inherently intimidating — we humans are evolutionarily predisposed to being wary of the unknown — but healthcare professionals have nothing to fear from the imminent healthcare AI revolution. As always, our proximity to matters of life and death compels us to proceed with more caution than the average AI enthusiast, but as long as we’re careful to only buy into real results — not hype — we’re well-positioned to inspire a much-needed paradigm shift in the healthcare industry at large.