Contributed Commentary by Dr. John Showalter

May 21, 2021 | Medication can sometimes be a hard pill to swallow — just ask the estimated 50% of patients who don’t take their medication as prescribed. Skipping a pill here and there may seem harmless, but in many cases it can mean the difference between wellness and suffering, and in some cases, even life or death. In the United States, it’s estimated that medication nonadherence accounts for up to 50% of treatment failures, around 125,000 deaths, and up to 25% of hospitalizations each year.

However, patients are not alone in bearing the consequences of nonadherence. As the industry shifts to value-based care models, providers and pharmacy benefit managers (PBMs) can face financial penalties if nonadherence results in readmissions or other adverse outcomes. Meanwhile, health insurers end up paying more as well when their members deteriorate as a result of nonadherence. All told, nonadherence is estimated to cost the healthcare system $100-$289B annually.

To save lives and lower the costs of healthcare, providers, PBMs and payers have a duty to prevent nonadherence wherever they can. The challenge, of course, is that these healthcare organizations are managing the care of thousands of patients, each with their own potential reasons for nonadherence. Knowing whether any one patient is at risk for nonadherence, let alone why, is hard enough, and the challenge only gets harder the more patients are involved.

Surfacing Clues for Nonadherence Risk in the Patient Record

The secret to predicting medication nonadherence is hidden in a sea of data—data that care managers lack the time or the resources to effectively parse through. Fortunately, clinical artificial intelligence (AI) can analyze this data millions of times faster than a human and reveal risk factors for nonadherence that care teams could otherwise overlook, augmenting their decision-making.

A lot of this comes down to the data about a patient’s medical history. In many cases, nonadherence is related to patients’ other medication or clinical conditions. Based on their history, some patients will be more prone to unpleasant side effects or drug interactions that make them less likely to take their medication. Or, they may suffer from mental or behavioral issues that present barriers to adherence. For example, a patient with depression may not have the motivation to take their medication, while a patient with dementia may simply forget.

By using the clinical data in the electronic health records, clinical AI can map specific patients against cohorts of patients with similar histories and their adherence outcomes. From there, clinical AI can predict not only which patients are at risk for nonadherence, but also the clinical factors driving their risk.

What About Non-Clinical Risk Factors?

Of course, clinical factors aren’t the only risk drivers for medication nonadherence. Socioeconomic and behavioral factors can be just as impactful in determining nonadherence risk as clinical factors, if not more so.

For example, if patients can’t afford their medication, then they probably won’t be taking it. Or maybe they can afford it, but they don’t have a pharmacy within walking distance and they don’t have access to a car or public transportation. In other cases, patients (particularly older ones) may live alone or lack the social support they need to help them take their medication.

Health literacy is another major risk driver. About 8 in 10 patients have below average health literacy, which means they may not understand the importance of taking their medication as directed, or the consequences of not doing so. Oftentimes patients think their condition is healed and they no longer need their medication, so they stop taking it and their condition comes roaring back. Patients with low health literacy may also have trouble communicating complications from their medication with their provider.

The problem is these social, economic and behavioral risk factors are usually unknown to care teams. Chances are patients don’t understand how these factors are relevant to their health and don’t discuss them with their providers in the short window they’re together. These factors are rarely captured by the medical record either, making it difficult to track them.

However, clinical AI can pull data on these factors from external sources, such as publicly available databases from the census and other government agencies, as well as commercial data sources. Combined with the clinical data in the medical record, these data can provide a powerful, comprehensive view of patients’ risk for nonadherence.

Taking Action on Nonadherence

By identifying the full scope of potential risk factors for nonadherence—clinical, socioeconomic and behavioral—clinical AI can go even further by providing guidance on the interventions that will address these risk factors. This form of clinical AI is known as prescriptive AI, in contrast to traditional predictive analytics that stop at predicting risk.

To illustrate, I’ll return to some of my previous examples. If the data shows that a patient is unlikely to have access to a pharmacy, a prescriptive AI tool could recommend that the patient be enrolled in mail-order pharmacy. In the case of low health literacy, the recommendation could be as simple as taking more time to educate the patient and make sure they understand how to take their medication.

With even more data, clinical AI can also shine a light on the best channel and time to engage patients to help them take their medication. Studies suggest that a simple text message can double the chances of adherence, raising the overall adherence rate from 50% to 67.8%. Every patient will be different, of course, but clinical AI can help care teams maximize the chances of success for their outreach.

These prescriptive insights can save care teams valuable time, not only by providing a more comprehensive snapshot of the patients at risk for nonadherence, but by identifying the actions they can take to prevent nonadherence. In turn, providers, payers and PBMs can mitigate the downstream consequences of nonadherence, improving outcomes overall and lowering the costs of care. As a result, patients get treated in a way that is more empathetic to their needs and empowering to their health.

John Showalter, MD, MSIS, is Chief Product Officer at Jvion. Dr. Showalter brings visionary thought leadership on the application of advanced information technology to improving outcomes for patients. His unique education in biomedical engineering, physiology, clinical informatics and internal medicine has allowed him to work at the intersection of those fields to positively impact patient care and health system efficiency. His work has been recognized with cross-industry awards including ComputerWorld’s Premiere 100 IT Leaders and health IT awards such as the CHIME Collaboration Award. Dr. Showalter is dedicated to using his passion and knowledge to ensure that Jvion’s machine has the maximum positive impact for patients. He can be reached at john.showalter@jvion.com.