Overcoming Roadblocks to AI Implementation in Healthcare

Updated on April 29, 2022
Artificial intelligence, Healthcare, Robots in Healthcare, Healthcare Technology

By John Schneider, CTO, Apixio

For several years now, we’ve been talking about the potential for artificial intelligence (AI) solutions to transform healthcare by making back office and point of care workflows more efficient, and by augmenting decision making, leading to better health and cost outcomes. Professionals in the space largely agree according to a recent KPMG report, where over 80% of healthcare and life sciences leaders stated they wanted to see their organizations be more aggressive in adopting AI technology. 

There are several benefits driving this enthusiasm. There is good evidence that AI can:

  • Provide medical professionals access to more accurate patient information. Healthcare AI can be instrumental in building a patient’s longitudinal medical history and more importantly help keep it free from clutter and contradiction.
  • Surface insights that allow providers to address issues proactively and reduce errors and misdiagnoses. For example, AI could automatically alert a provider to previous negative drug reactions or potential complications due to certain health conditions, even if the patient forgets to mention them, or AI can backstop radiologists who might be asked to evaluate even larger caseloads.
  • Help speed decisions on whether to authorize care by helping identify cases where a request for care is generally aligned with best practices of the organization.
  • Help reduce spending on unnecessary or even harmful services. By grouping patients into cohorts and associating care decisions with outcomes, we can find care paths that best work for patients potentially eliminating procedures that add no value and highlighting potential options a provider might not know about. 
  • Create a better patient experience by liberating our physicians from their EMRs while they are interacting with us. Right now, AI can read the medical chart ahead of time and provide summaries for the physician that highlight what is already known about the patient, saving the physician valuable time spent reading the chart. At some point soon, the AI will actually listen in on the conversation and automatically transcribe it and create the clinical note for the physician to review. This will save time and importantly aid in standardizing the information capture, making it richer and more useful in downstream AI processes
  • Help personalize care especially in the management of chronic diseases. Predictive modeling can be used to suggest the most personalized course of management based on specific patient data compared to other patients with the same or very similar medical history. It could also identify the candidates with risk of complications and suggest to the care team diagnostics or preventative measures to mitigate those risks. 

So, what’s stopping us from accessing these, and many more promises of AI in healthcare?

  1. The primary data available are claims, which fuel the primary data workflow in the healthcare industry and have for decades now. While there are a lot of them, they don’t say very much about the patient and they don’t always tell the truth. For example, a condition might be stated to justify a procedure, but there’s no requirement that the condition be true. Another problem is that claims report only what is claimed for payments, not what is known, so a lot of information is not captured.
  2. Data from EMRs is a welcome addition, but it’s a mess because there are no standards. No two EMRs capture the same data, and two instances from the same vendor likely have differences in the data that will break things down stream. To compound matters, physicians all use different language and note taking techniques, so the variability of the text in clinical notes makes them an enormous challenge to process. 
  3. Regulations that are in place to protect patients, actually hurt our ability to share data, and this doesn’t have to be the case. HIPAA, often misunderstood to be about privacy, is actually the Health Insurance Portability and Accountability Act. One of the five titles in the act, Title 2, has two sections about privacy and the other about security. The problem we have today is that there is an enforcement section, which can be very punitive, that drives healthcare organizations holding the data to act very conservatively and default to not share. HIPAA was written in 1996; that’s eons ago in the information age and very little was understood about how the data could be used and protected. We are now constrained by 27-year-old thinking, and it’s time to dive back in and update the regulations.
  4. Healthcare organizations and software vendors are often not good actors in this equation. They see ownership of the data or the walled garden they sell as a competitive advantage and are often a key roadblock to data being used to benefit the larger healthcare community.

It’s not all bad news though, there are already many successful uses of AI in healthcare, such as smart image reading that can help radiologists detect breast cancer, or vascular neurologists identify blood clots in stroke patients. The management of many chronic diseases benefit from “nudge” technologies that focus on behavioral change to improve health and wellness.

Use of full patient longitudinal records also shows a lot of promise. AI techniques are uniquely suited to succeed with this data because information needed to assess the outcomes of a care path are buried in unstructured clinical notes. Natural Language Processing (NLP) now has the power to uncover these hard to access outcomes, and AI can handle sorting through the likely millions of potential care paths and outcomes where a typical provider only ever sees a few thousand.

These longitudinal records will be augmented by all of the remote wearable and home health devices that live on the edge. Importantly, the large scaled healthcare data platforms like Apixio’s APICare are helping payers and providers make sense of their data and operationalize it to solve their real world problems. Advantages of these platforms are that they create a unified patient record from any number of sources that can be leveraged and solve the security and privacy issues that have been baked in from the start.

There is the potential for AI to revolutionize the healthcare system. The technology is available. In order to realize the potential, there needs to be a shift in how we process and use healthcare records. In many respects, the medical system has to get out of its own way. It’s time to adjust the regulatory infrastructure to unlock valuable medical insights. 

14556571 1295515490473217 259386398988773604 o

The Editorial Team at Healthcare Business Today is made up of skilled healthcare writers and experts, led by our managing editor, Daniel Casciato, who has over 25 years of experience in healthcare writing. Since 1998, we have produced compelling and informative content for numerous publications, establishing ourselves as a trusted resource for health and wellness information. We offer readers access to fresh health, medicine, science, and technology developments and the latest in patient news, emphasizing how these developments affect our lives.

1 thought on “Overcoming Roadblocks to AI Implementation in Healthcare”

  1. This Email has nothing to do with the above article and everything to do with a clear wondering
    thought. I am wondering if by chance you have any AI information about people who have a
    very poor memory and comprehension problems in the books. As a share holder of GATEH
    if you have any products for sale and care to share them with me with prices please let me know. Just wondering. Thank You for the answers before hand if you by chance give me a reply. Sincerely, James P. Dill

Comments are closed.