Creating a more deliberate approach to AI in healthcare

1. The costs of healthcare will continue to rise,
2. Rising costs and regulations such as MACRA will force payers and providers to align reimbursement and financial incentives with improvements in quality, and
3. These pressures will drive the search for more aggressive ways to manage and/or shift financial risk.

Is the industry equipped to meet these certain pressures ahead?

While risk-sharing and value-based care reimbursement gained more traction in the past year, they did not necessarily translate to successes in profit for hospital systems. To thrive in value-based care models, organizations will need to become experts on use of their data, not just administrators of it. Pressures faced by insurers and hospital systems will require more deliberate adoption of artificial intelligence (AI) that equips them with cost-effective resource management, robust analytics, and precise population health management.

AI as an imperative to survive

First, like most innovations, we need to separate the hype from reality. Much has been written about AI in the past year and its capability to spark revolutionary change. While we are at the beginning of what some might call “the golden age of machine learning,” we still are far from a world of virtual treatment environments where machines can independently diagnose a patient with accuracy. However, a recent survey conducted with health IT executives indicates a general belief that AI will be a game-changer across patient engagement, chronic disease management, clinical decision support, and sophisticated analytics for diagnostics, population health management, and financial modeling. But where do you begin?

In working with several large healthcare organizations, we have found that before diving too far into the technology weeds, it is helpful to start by understanding your current pain points and developing a framework for where AI can have the greatest impact in operations.

We are starting to see the strongest possibilities for machine learning and deep learning applications within three areas:

1. Developing stronger capabilities in preparing and managing data,
2. Reducing costs by automating labor-intensive and time-consuming administrative tasks, and
3. Creating analytics that can uncover retrospective trends and gain predictive capabilities for preventative care, reducing utilization and risk and revenue management. This is, arguably, the most complex area.

How can AI do the heavy lifting for you?

At the very onset, data preparation and management is a massive undertaking and a primary challenge for many organizations. Many of healthcare’s legacy technologies are unable to handle the vast amounts of information in new data sets with speed and efficiency. As a result, analysts can spend up to 90% of their time managing the data rather than analyzing it. Data preparation solutions now are using advanced machine learning and deep learning techniques to unify, clean, standardize, and prepare data for modeling and quality reporting. This augments the ability of IT teams, data analysts, data scientists, and business analysts while improving reimbursement and revenue management opportunities.

What operations can you automate? 

Administrative activities are a major cost burden for most organizations. One-third of the industry’s $3 trillion annual spend is wasted on administrative costs, over-treatment, and red tape. But in this “golden age of machine learning,” can we use AI to automate costly and routine administrative tasks, such as billing, coding, prior authorization, and analyzing swaths of documents?

A recent report by Mckinsey indicates that 45% of routine, manual tasks that individuals perform can be automated with AI. We already are seeing this implementation across the healthcare spectrum. Natural language processing is being used by doctors for clinical documentation improvement, converting PDFs of care summaries into text that can be manipulated, or ICD-10 coding. Prior authorization—a time-consuming process that can cost a large health plan up to $90 million annually—can be automated using machine learning, which provides recommendations within seconds as opposed to three to five days when performed manually. And, virtual health apps that use text-to-speech and search technologies are being deployed by insurers and hospitals to improve customer service interactions with members and patients.

How can AI improve payer-provider collaboration?

Greater financial risk also demands greater collaboration, transparency, and data sharing between payers and providers—areas that are often cited as the weakest link in population health management initiatives. To succeed in value-based care arrangements, payers need to help providers overcome data fatigue, provide individual and population-level analytics that have clinical efficacy, and surface the most high-need patients. Predictive models that use advanced machine learning and deep learning techniques make use of the swaths of data sets available to understand, predict, and influence the health of individuals and populations.

Studies have shown that AI frameworks can improve patient outcomes and reduce costs by as much as 50% and reduce unnecessary lab tests for intensive care patients. AI applications also can enable risk stratification to improve population health management. By stratifying the highest risk patients, AI technologies can reduce treatment and care management “guessing” by helping providers be more precise with their care choices, coordinating care in a way that reduces waste, and aligning priorities between payers, providers, and care managers.

As we move into 2017, organizations taking on more risk will start to look at AI for better care management, population health management, and administrative operations. But to see tangible benefits beyond the hype, we need to be more deliberate in how we think about AI in healthcare. The framework above presents high-level considerations of the questions to ask when thinking about AI deployment. Keep in mind that AI is a technology that can be applied to a variety of business challenges; it is therefore crucial to develop a systematic approach to identifying the opportunities where it will have measurable impact in the near term and long term.