Home   >   Blog   >   Preparing for an Artificial Intelligence-Driven Future

Preparing for an Artificial Intelligence-Driven Future

by Eric Sullivan, Senior Vice President, Innovation & Data Strategies, Inovalon, on July 25, 2019
Eric Sullivan

This article originally appeared in Inside Digital Health on July 17, 2019.

Artificial intelligence has found application in almost every industry today, but the stakes are arguably highest in healthcare because mistakes can affect patient care. The tsunami of health data being produced on a daily basis, along with chronic inefficiencies across the system, make the industry ripe for artificial intelligence-driven innovation. However, the most significant hurdles in realizing artificial intelligence’s full potential in healthcare is identifying the right applications and earning the trust of users.

A substantial portion of healthcare data generated today is unstructured, making it challenging to use to effectively train algorithms. Moreover, healthcare definitions and policies are constantly in flux. A machine learning model trained on data about diabetic patients, for example, would need to be adjusted every time the definition of diabetes changes, which could be as often as every year.

Despite these challenges, artificial intelligence has proven its ability to help inform clinical decisions based on many variables and huge data sets more efficiently than humans. As more data and data sources come to the fore, artificial intelligence will be an important tool to help make sense of it all.

Successful Artificial Intelligence Applications

Many artificial intelligence applications in healthcare are already showing promising results and may offer insights for how to best leverage artificial intelligence in the future.

  • Reviewing medical records: Using artificial intelligence, health plans and providers can identify and address issues like gaps in care, such as an overdue appointment, more efficiently.
  • Predicting adverse events and outcomes: Applying artificial intelligence to a set of claims and/or clinical data can help predict adverse patient events that are often avoidable, such as a fall in a hospital.
  • Informing staffing decisions: Artificial intelligence supports acute and post-acute facility operations by not only predicting the number of admitted patients, but also their length of stay.
  • Increasing pharmacy and specialty pharmacy effectiveness and efficiency: Artificial intelligence technologies can predict which patients are at risk of medication non-adherence and identify patients who are good candidates for clinical trials.
  • Tiering population risk: Artificial intelligence enables more efficient risk stratification within a population and can surface the most effective way to engage at-risk patients/members.

The Role of the Cloud in Healthcare Artificial Intelligence

The application of artificial intelligence in healthcare is quickly picking up steam, and cloud technology has been central to its growth. Cloud adoption was initially slow in healthcare but gradually gained wide acceptance because it increased access to data while decreasing build, processing and storage costs.

The benefits of larger centralized storage and faster processing times are required for large-scale artificial intelligence training. Additionally, once an artificial intelligence tool is developed, cloud technology makes it readily available to as many users as needed and as fast as possible.

Looking Ahead to Artificial Intelligence’s Future

As artificial intelligence continues to gain traction in healthcare, it’s critical that providers and others who use it to make decisions can validate and substantiate the recommendations. An artificial intelligence model might be highly advanced, but it must earn the trust of its end user to have real value and gain broader adoption. Having medical professionals join the training process can help ensure confidence in the model.

Additional considerations for future applications of artificial intelligence in healthcare include:

  • Artificial intelligence solutions should be applicable to a high volume of people to benefit as many patients as possible and move the needle on outcomes.
  • Too many data inputs increase the likelihood of introducing inaccuracies, biases and too many dependencies. In some instances, smaller data sets are more reliable.
  • Artificial intelligence models must be flexible enough to continually adapt to the rapidly changing healthcare environment. Continuous and diverse data streams are key to ensure models keep up with evolving trends and policies.

Over the past few years, the Centers for Medicare & Medicaid Services and other industry stakeholders have been encouraging the adoption of artificial intelligence by publishing guidelines, making deidentified data more readily available to the academic community and introducing innovation challenges. There is no doubt the healthcare ecosystem is flush with opportunities for artificial intelligence innovation, but all stakeholders, from patients to providers and payers, must be confident in its application to embrace artificial intelligence and enable it to achieve its full potential.

© 2019 Inovalon. All rights reserved.