Navigating change: Embracing process automation in healthcare
A deep dive into the opportunities, challenges and emerging trends in bringing advanced computer capabilities to healthcare.
Each year, principals in the healthcare industry grow a little more confident that process automation will bring meaningful benefits in such areas as revenue cycle management, supply chain operations and even clinical procedures.
But when Galen Healthcare Solutions hosted a focus group of healthcare executives, all CHIME members, last summer, the participants acknowledged that limited budgets, unwieldy, undistributed data, and the preference for the familiar, reinforced resistance and prevented significant advances in the acceptance and adoption of this potentially revolutionary technology.
How much has changed in the past year? Has resistance continued? For what reasons? Have advances occurred? In what areas has automation become more prevalent and how successful was it?
One would think that support for automation in health care organizations would be easy to obtain. After all, KLAS reported earlier this year that more than three quarters of healthcare providers who had implemented some form of process automation experienced improved financial performance and greater staff efficiency.
But willingness to embrace artificial intelligence in its many forms and applications remains, for most, a bridge too far. A new study from Bain & Company finds that only 6 percent of hospitals have adopted a strategy for the use of AI.
Galen conducted another focus group this summer, comprised of healthcare IT professionals, CIOs and other executives from various healthcare organizations. Their discussion revealed what might be described as positive hesitancy – that is, an increasing acceptance of the value of process automation, a desire to apply it in its many forms to ameliorate the numerous problems that beset caregivers in their efforts to enhance patient safety while controlling costs.
At the same time, this support is tempered by various concerns: Will the return on investment be large enough? Will the data spewed out by this new technology be reliable and transparent enough to facilitate decision-making? Will doctors accept and stand behind medical judgments made by a machine?
Positive hesitancy
“We talk a lot more about it than I think we practice. We have practical examples broadly. It's a tough thing getting the humans to catch up with the technology. I would say that that's the barrier.” – Chief Information Officer
“There are lots of areas in which there's very clear value. But what's the tradeoff of getting to a place where there's at least some form of standardization where you can automate that across your enterprise, especially if it's a large integrated health system?” – Emergency Physician
Artificial intelligence is crowding the environmental crisis and other urgent matters for space and time in the media. As the public becomes more familiar with its potential, anxiety is as common as enthusiasm. This is particularly apparent in healthcare.
A new paper in Nature: Digital Medicine on the acceptance of artificial intelligence among healthcare professionals in hospitals delineates reactions to several different automated applications. For example, caregivers in acute hospital settings agreed that clinical decision support systems reduced the rate of medical errors through warnings and recommendations.
Others disagreed, claiming clinical decision support systems (CDSS) were liable to induce errors, especially in the ER.
When automated technology is perceived as intuitive and easily understood, it was well regarded. If it was seen as complex and the source of additional tasks, it was opposed. Which reaction is more common?
An article in the Journal of Medical Systems showed that 38 percent of users of machine learning algorithms were unable to integrate the system into their clinical routines. On the other hand, 82 percent of anesthesiologists in a different study declared that new technology made available to them was easy to learn. In short, “When participants believed the AI-based system was aligned with their tasks, had consistent reporting of values and required minimal time and effort, they welcomed it.”
Acceptance of automation in healthcare is not just a matter of whether individual caregivers or their teams find a particular innovation convenient. Another study indicates that physicians are concerned that CDSS systems will reduce the amount of time they can spend with their patients. Pediatric nurses agree, fearing a reduction in human touch and connection.
And looming over all of this is the never-ending effort to keep pace with regulatory changes and legal matters. In an analysis of a diagnostic CDSS, caregivers warned that it was unclear who would be accountable in case there was a system error. In another part of the hospital, only 5 percent of radiologists said they would be willing to assume legal responsibility for imaging result interpretations provided by AI.
A study for The Brookings Institute compared AI adoption in healthcare to other industries. “Even for the relatively skilled job postings in hospitals, which includes doctors, nurses, medical technicians, research lab workers and managers, only approximately 1 in 1,250 job postings required AI skills. This is lower than other skilled industries such as professional, scientific or technical services, finance and insurance and educational services.”
Among the most significant causes of this rather surprising situation is the concern that artificial intelligence and process automation produce solutions with flaws that are often discovered too late because they employ algorithms that are “black boxes.” Lack of transparency quite naturally shatters trust in the technology and alarms doctors who are most at risk to be held accountable for decisions involving AI that cause harm.
Moreover, in healthcare, we are still struggling to capture accurate and relevant medical data. It remains challenging to pool data from across hospitals or caregiving facilities. Without reliable data sets, it’s harder to build useful AIs and more difficult to secure trust for the adoption of newer tools and applications.
Proofs of concept
Process automation has been compared with other disruptive technologies that showed great promise. More than a few of these failed. But many others spawned successful enterprises and industries. In the current healthcare climate, few organizations have extra dollars to invest in technology – that means, of course, that the investment had better generate returns.
Happily, examples abound, supporting faith in RPA. In Singapore, the country’s flagship hospital has successfully implemented process automation in 36 cases since 2019. A report in Healthcare IT News states that the technology saved 52,500 hours and added more than $1.8 million in productivity. Within the organization’s physiotherapy department, the technology saved its caregivers 1,300 hours that would otherwise have been spent on manual calculations and documentation.
One panelist characterized revenue cycle management as the low-hanging fruit for the application of process automation. Another declared that “the investment made on the revenue cycle side paid for itself and now allows us to focus on clinical aspects of automation.” Why?
Revenue cycle management depends on the completion of repetitive tasks and processes. These can be performed with robotic process automation, bringing a huge financial benefit. The Council for Affordable Quality Healthcare estimates that automation for administrative tasks could generate $17.6 billion in annual savings. Internally, healthcare organizations can improve data accuracy and integration as well as foster greater employee satisfaction by utilizing staff for strategic high-value initiatives rather than routine tasks and manual processes.
In addition, RPA can streamline appointment scheduling. Missed appointments in healthcare have been valued at $150 billion annually. Should another pandemic occur and again overwhelm healthcare staff and systems, RPA bots can schedule appointments based on diagnoses, provider availability, location, health coverage and other considerations, providing front-end support to collect and process data.
The claims management process is another area where process automation can provide a return on investment in a reasonable period. There are tens of thousands of diagnostic and procedure codes in the claims management process. If these codes contain inaccuracies – as in the designation of coded claims, any one error will lead to denials, each of which triggers a review, which produces recoding, and refiling. One errant key stroke leads to time-consuming re-entries and delays in revenue collection. Healthcare organizations can rely on RPA bots to collect revenue and keep their organizations in operation.
Many panelists wanted to discuss the potential for automation in clinical workflow. Here, advances have already occurred and are making a difference in diagnosing patients more quickly. For example, medical images can be fed into AI-based systems to assist with the identification of diseases, thus adding efficiency to the operations of pathology departments.
Elsewhere, finding donors and matches for transplants has often necessitated as many as 500 phone calls, emails and texts – often from non-HIPAA-compliant devices – to make one procedure possible. We are now seeing the use of platforms that can create detailed audit trails of all conversations and case activity allowing transplant teams to mitigate risk, improve outcomes, and drive process and quality improvement.
While we are far from realizing the full potential of project automation, the hesitancy we've seen appears to be a function of uncertainty about what it can do, how costly it will be and what consequences it will generate. The focus group discussion reflected the perspectives that were mentioned in the studies cited in this article:
- • Challenges to adopting process automation include physician ambivalence, need for clear evidence of a positive return on investment and lingering lack of trust in artificial intelligence.
- • Potential exists for revenue cycle management to function as a proof of concept for robotic process automation.
- • Interest in and desire to apply automation to clinical workflow and supply chain deficiencies should be pursued.
In short, this specific focus group and thought leaders in the healthcare information technology community in general, share a cautious optimism about the potential for process automation to improve efficiency, reduce costs and enhance patient care. But challenges remain in terms of adoption, trust, and standardization.
Justin Campbell is vice president of Galen Healthcare Solutions.