Skip to main content

3 ways to use intelligent automation in healthcare IT architecture

Intelligent automation helps healthcare organizations reach better architectural outcomes while adapting to new business challenges.
Image
Black and white structure medical

Photo by Tara Winstead from Pexels

IBM defines intelligent automation (IA) as "the use of automation technologies—artificial intelligence (AI), business process management (BPM), and robotic process automation (RPA)—to streamline and scale decision making across organizations." IDC uses the term intelligent automation, while analyst firm Gartner uses hyperautomation to describe the same thing: a convergence of tools and technologies for AI and automation to reach better architectural outcomes while adapting to new business challenges.

Why intelligent automation is a step forward

Healthcare organizations are under constant pressure to meet compliance, budgetary, information-access, and other requirements. If you layer on the typical business operating requirements to keep costs down and maximize profits, you have the formula for a very complex IT environment.

To meet these demands, healthcare organizations adopt various off-the-shelf tools to automate routine processes and analysis tools to examine data retrospectively, looking for insights they can use to drive positive patient outcomes and drive down the cost of care delivery.

This typical "customized off the shelf" (COTS) approach has contributed to the current state of tool fragmentation and data in silos. Healthcare organizations rely primarily on their electronic medical record (EMR) vendor to drive the bulk of their application strategy. If the EMR vendor has requirements for data access or infrastructure hosting, the entire infrastructure tends to be driven by those requirements.

A significant challenge of these silos is interoperability in data access and processing. Data is typically in separate repositories and in different formats. So much time and money gets spent managing and securing the data that drives the organizational structure into siloes. This further leads to each silo trying to derive value from their data their own way, causing redundant solutions and increasing the cost of ongoing maintenance and innovation.

Organizations may have various tools for process automation, orchestration, data analysis, and machine learning (ML), and a separate set of tools and processes for native application development. Healthcare organizations are increasingly looking for a standard, interoperable platform of tools with a consistent app development and operations experience.

What is intelligent automation?

Understanding the role these automation technologies play in IA is the first step for enterprise architects working to create a consistent experience for the development, operations, and other professionals working with the organization's IT systems. To that end, I'll break down some of the ways healthcare organizations use artificial intelligence to improve patient care, business, and other areas.

Artificial intelligence and machine learning

AI is a set of methods to enable machines to approximate human intelligence in specific ways. ML is a rapidly evolving branch of AI dedicated to emulating human intelligence by deriving insights from the surrounding environment, typically represented by datasets collected based on changing environmental conditions. In healthcare, you can apply AI and ML to both clinical and revenue cycle workloads.

[ Learn more about transforming your organization with automation by downloading the free eBook: The automated enterprise. ]

An example of using AI in healthcare that is easy to understand is image processing. Any provider that has to review x-rays for specific disease conditions typically has a backlog at the radiology department. The majority of films that need review are for common conditions like bone fractures or pneumonia. An AI algorithm can be "trained" to examine thousands of images for patterns corresponding to various conditions and give a score (% likelihood) that a given image contains that pattern.

Think of this as triage for images. The image, along with the score(s), can then be passed to a radiologist or clinician for further analysis if necessary. This approach not only breaks the logjam of the initial review for many images, but it drives consistency and thus more positive patient outcomes. A well-implemented AI model like this also captures feedback from the reviewers to become smarter and more accurately score future images.

Business process management

Business process management (BPM) is the discipline of improving a business process from end to end by analyzing it, modeling how it works, and making continuous improvements based on feedback. BPM is commonly also used to indicate the automation of the processes, but that is more accurately referred to as business process automation (BPA). BPA use cases in healthcare are numerous and, like AI and ML, can apply to both clinical and business operations. The challenge is implementing "workflows" (automated business processes) that reflect changing business conditions and orchestrating multiple workflows across a complex infrastructure.

A common scenario for using BPM in healthcare involves prior authorizations for particular interventional services needed by patients. The typical approach is manual, labor-intensive, and prone to human error.

What usually happens is that the doctor and patient discuss the treatment needed. The doctor sends a prior authorization request (often it's still paper fax, believe it or not). The payer then reviews it according to their policies and the patient's specific insurance coverage, then approves or denies the request. The adjudication of the request is often sent back to the provider manually as a fax.

A more efficient process is to encode the policies into a BPM workflow, accept prior authorization requests electronically, and then send the results the same way. (This has been possible for many years, but some companies still do it manually.) The BPM workflow takes in the request data and compares it to predefined rules for completeness and accuracy, then looks up the patient's policy and decides if the service requested is covered. It then formats an electronic message and sends it back to the provider, where it appears in the EMR system.

This sounds great, but what if the incoming request is not in a machine-readable format (like fax)? This is an example where you can use AI/ML and BPM together. You can run every incoming fax into an electronic queue that feeds an ML algorithm that does optical character recognition (OCR). It scans the fax for usable data, creates an electronic record, and then feeds it into the BPM workflow. Once adjudicated, it can generate a fax electronically and send it to the provider's fax machine. This example is very common and works in many organizations.

Robotic process automation

Robotic process automation (RPA) rose to automate workflows that involve legacy systems that lack APIs. It has evolved into myriad tools to build and manage "robots" that emulate some aspect of human behavior, particularly related to digital systems and software. Robots can do very human-like tasks, such as reading what's on a screen, typing text into an input buffer (like a keyboard), or doing a "screen scrape" (copying text from a screen). You can use all these "inputs" to carry out a series of predefined actions more quickly and efficiently than a human can.

[ You might also be interested in this on-demand webinar: Choose your cloud-native path—an executive checklist. ]

I'll give a quick example using RPA with a "legacy" system—not built to take input or deliver output to anything other than a human. Such systems are still in common use in payers, providers, and life-sciences organizations. Many of them run on specialized hardware or older mainframe systems, and some have been certified by a government standards body, so they are difficult and expensive to update. An RPA process includes rules and logic similar to a BPM workflow, but it also grabs text from a screen or printer. It can then use that text to trigger a set of logical steps similar to a workflow in BPM. It can also call a BPM workflow directly, so long as there is an API available.

Here's an example of patient scheduling. It may seem like a simple operation, but not when you factor in a doctor's availability and the need to look up the patient's medical history records to make sure they see the right provider. If anything changes with the doctor's availability, the patient must be notified. Another complicating factor is when the patient's or doctor's information is in a system that users can't easily access.

RPA can automate this process. It can access the necessary data, no matter where it is, then execute workflows that notify the patient, doctor, or other clinicians and update all the appropriate systems. If something else changes in the real world, it can do it all again.

Conclusion

Data, analytics, and artificial intelligence promise new insights to transform businesses, and healthcare is no exception. Additionally, modern architectures such as edge computing enable analytics and AI workload processing to take place closer to where the data resides and new data gets generated.

Any of these tools can give answers that are predictive or prescriptive. They are all data-driven and can be knowledge-driven, meaning they can incorporate conclusions (for example, scores) from prior iterations or other tools. Any good toolset should also allow you to simulate models before moving them into production.

Intelligent automation is an increasingly influential concept. As the sophistication of the various types of tools evolves and their capabilities overlap, it becomes increasingly important to adopt a coherent strategy that orchestrates the use of all of these capabilities in a common platform. This enables you to choose the right tool for the job.

What to read next

Topics:   Automation   Industries   Edge computing  
Author’s photo

Marc Mangus

Marc brings 30 years of experience in Health IT, Product Management, Software Development and Systems Architecture. Prior to Red Hat, Marc served as a Product Manager at Anthem where he led one of the largest clinical application development projects for the company. More about me

Related Content

OUR BEST CONTENT, DELIVERED TO YOUR INBOX