Guest Column | March 6, 2018

3 Myths About AI And Healthcare

By Anthony Macciola, CIO, ABBYY

GFSI Myths

The use of artificial intelligence (AI) in healthcare is finally becoming a reality. According to CB Insights, from 2012 to 2017 more than $2 billion was invested specifically in healthcare AI with companies that leverage machine learning algorithms. Additionally, technology giants Apple, Microsoft and Google are investigating entering the healthcare market, and Amazon’s secret health tech team 1492 is exploring a platform for electronic medical record data, telemedicine and health apps.

You’re not quite going to find doctor Amazon Alexa diagnosing patients in homes – yet, but we are excited by the many advancements in AI to help detect diseases earlier and analyze vast data streams to identify risks and outcomes. AI in this capacity can truly change the healthcare delivery and diagnosis model in the future. Today, we are dealing with AI to help onboard new patients, make the patient management more efficient, and improve our data outcomes for patients. You can also find immediate challenges addressed with analyzing unstructured data such as in claims processing.

However, to better understand the potential of AI in healthcare, certain myths need to be dispelled.

Myth: AI Is A Standalone Software

Artificial intelligence has become a catch-all phrase to mean a lot of capabilities. It’s reminiscence of the “Big Data” trend the industry experienced a few years ago. When discussing specific use case scenarios, healthcare IT executives describe a system that watches and learns what knowledge workers are doing as part of their daily routines – especially routines they are considering automating with robotic process automation (RPA). They want a system that can recommend various courses of action based on learned behavior. Finally, they want the knowledge worker to be able to direct the system to automate the learned behavior after getting comfortable with past recommendations. There is no one size fits all software that does all this. To me, this is machine learning technology.

In fact, it is RPA combined with machine learning technology that is enables AI in healthcare. RPA is the application of technology that allows administrators and clinicians to configure a software “robot” to capture and interpret existing applications for processing a transaction, manipulating data, triggering responses and communication with other internal digital systems. In claims processing, many healthcare organizations are using RPA to pull data from legacy electronic medical record systems and push to new digital platforms and telemedicine applications. It’s a great way to leverage and automate the sharing of patient information across several internal silos and outside to authorized healthcare partners.

Myth: RPA Will Fix The Data

Healthcare organizations receive data from various channels in a variety of formats. New patient forms, handwritten notes, faxes, medical claim forms, X-rays and prescription notes all exemplify the challenges of managing unstructured data and making the data useful in adjacent systems. There is no magic “AI wand” that will correct the ingestion of data – this illustrates the importance of modern capture software that can correct and perfect data as it comes into the organization, enables document classification and automated data extraction. With capture combined with RPA, healthcare organizations can climb five levels to achieving AI:

  1. The orchestration of basic repetitive tasks. For example, instead of an administrator manually inputting data from the patient information form, the data is automatically populated into the system.
  2. The extraction of metadata from forms. Especially ensuring that information provided by patients in their new patient onboarding process is captured, validated, and actionabale in the electronic medical records system. In claims processing, forms such as HCFA/CMS-1500 and UB-04/ CMS-1450 are accurately processed.
  3. The inclusion of unstructured content – such as correspondence, explanation of benefits, previously inputted worker’s compensation, X-ray and prescription information included within appropriate forms for processing.
  4. The learning of human interaction with RPA robots and the eventual recommendation and automation of action. This can include automatic treatment plans and referrals to specialists based on a diagnosis, triggering a warning for prescription allergies and exceptions in claims processing.
  5. Better understanding of the impact automation has with a patient and within an organization. At this level healthcare organizations can customize services based on patient-generated data, offer social community support through partnerships to coordinate care, and obtain process intelligence to know real-time status of patient tracking, claims processing and operational efficiency.

Myth: AI Will Replace Nurses

Healthcare providers needn’t worry that AI will replace their job. While software robots will continue to do largely routine or repetitive jobs, humans will continue to do tasks that require creativity, problem solving and flexibility. Characteristics like empathy, creativity, judgment, and critical thinking are lacking in robots, so in the areas where human compassion is essential, AI-powered robots or solutions won’t be a replacement.

The overarching truth is AI-enabling solutions will continue to automate basic tasks performed by healthcare workers and allow providers to focus on patient-centered care.

About The Author

Anthony Macciola is the Chief Innovation Officer at ABBYY, a global provider of intelligent capture solutions that improve business outcomes. He holds more than 45 patents for technologies in mobility, text analytics, image processing, and process automation.