Can AI revolutionize population health management?

Share:

Share on twitter
Share on linkedin
Share on facebook
Share on email

Can AI revolutionize population health management?

In recent years, artificial intelligence’s potential to support healthcare advancement has seemed sky-high. Countless AI-powered tools have appeared on the market and offered to reinvent the way that providers and patients alike approach healthcare.

The digital EHR scribe Suki attentively listens to doctor-patient conversations and creates accurate clinical notes of consultations; Amazon’s Alexa has begun helping certain users to manage, reorder and take their prescriptions. In 2017, IBM’s AI-powered Watson even promised to help oncologists diagnose and treat cancer patients — although the tool has admittedly fallen short of the success that its proponents hoped it would have found by now.

Each one of these advancements is notable — and helpful — in its way; however, the most influential use of AI may be occurring behind the scenes.

In July, the Geisinger health system announced that it would begin using AI solutions to identify and facilitate preventative care to patients who carry a significant risk of developing a high-burden disease. This strategy, Geisinger leaders hope, will empower the health system to improve both its population health management and financial outcomes by giving physicians a better understanding of their patient base. In theory, these AI tools will be able to use clinical information to direct physicians towards high-risk patients before their chronic or acute conditions develop into costly health crises.

To accomplish its AI goals, Geisinger partnered with Medial EarlySign to implement a suite of machine learning tools that could identify specific chronic conditions. According to a report from Health IT Analytics, Medial’s software uses machine learning to parse routine medical and electronic health data. As one reporter explains, “Patients on a high-risk trajectory are identified through routine lab results (e.g., blood tests) and other early signs of risk. The software then flags these patients for providers to indicate which patients would benefit from further evaluation, preventive care, and possible disease management.”

When the partnership began, Medial EarlySign focused exclusively on identifying patients at risk of developing severe lower GI disorders. However, the software has since added prediabetes, chronic kidney disease, and coronary artery disease to its early-detection capabilities.

To quote Ori Geva, the co-founder and CEO of Medial EarlySign, “This [partnership] is the first step of our ultimate goal: enabling healthcare systems to identify and connect with those high-risk patients and engage with them early enough via interventions that may prevent or delay disease progression.”

Geisinger and Medial EarlySign are innovators, but they are not breaking entirely new ground. Other AI services have been used to predict patient outcomes and prompt early intervention. For example, researchers for one 2018 study used deep learning techniques to develop an EMR-based prediction model; they found that AI was better at creating a 30-day hospital readmission projections for heart failure patients than traditional means. Similarly, another research group found that deep learning could be a useful tool for predicting clinically-significant intracranial aneurysms.

All of these capabilities are significant because they empower providers to begin treating potentially costly conditions earlier and avoid taking on the cost and care burden of unnecessary treatment.

Value-based care centers on patients as healthy as possible; doctors are reimbursed not on the number of services provided, but on the overall health of their patient base. As such, they are financially motivated to be proactive about improving their patients’ health. The model works, too; according to United Healthcare’s 2018 Value-Based Care Report, provider organizations that adopted an outcomes-based approach ranked 87% better on top quality measures and saw 17% fewer hospital admissions than their fee-for-service competitors.

In this context, consider a scenario wherein a patient isn’t actively treating a chronic condition, and their disease progresses into a crisis. Over time, the advancing disease could severely damage the patient’s health and create a significant cost burden for the patient and payers alike. However, if providers can identify high-risk patients and intervene quickly, they may be able to limit — or even prevent — healthcare disasters. This capability allows providers to both improve patient outcomes, limit avoidable hospital visits, and potentially achieve high profits.

Some early identification systems are lower-stakes in their intentions. The AI-powered healthcare tool Cyft, for example, uses data from a variety of sources, including EHRs, health assessments, and patient surveys, to identify patients who could benefit from additional outreach. Depending on the patient’s unique case, Cyft may prompt physicians to request an office visit, make a change in medication, refer the patient to specialty care, or even make a simple call.

These measures — though comparably small against Medial EarlySign’s efforts to limit health crises — could be crucial to improving primary care within a value-based system. It can, in theory, empower doctors to take a more proactive approach to care and encourage patients to be more engaged in routine treatment.

The latter is particularly important, given the relative lack of patient engagement in the healthcare sector today. According to one 2016 Insights report from Catalyst, more than 70% of surveyed physicians believe that less than half of their patients were highly engaged in care, while just 9% reported high rates of engagement overall. On average, providers said that only 34% of their patients could be considered highly engaged in care.

A lack of engagement can lead to uncoordinated care, limited follow-ups or even patient nonadherence. The latter is particularly problematic; research has indicated that noncompliance with treatment plans can pose significant health and economic risk to patients. One study published in Therapeutics and Clinical Risk Management found that in some disease conditions, “more than 40% of patients sustain significant risks by misunderstanding, forgetting, or ignoring healthcare advice.”

With AI tools like Cyft and Medial EarlySign, doctors have an increased opportunity to act proactively, engage patients, and improve patient health outcomes. The work these tools do may be less visible than the gains made by Alexa or Watson, by the impact they stand to create for patients and providers is exponentially greater. AI can— and likely will — revolutionize the way we address population health in the future.

 

Click here to read the original article

Share:

Share on twitter
Share on linkedin
Share on facebook
Share on email

Are you looking to solve high-burden disease challenges?