3 Trends to Consider Before AI Deployment
By 2026, artificial intelligence could create $150 billion in annual savings for the U.S. healthcare economy, according to a recent Accenture analysis. So, it’s no surprise that healthcare organizations are making AI a priority.
“AI” could mean dozens of different things in the healthcare space, just as it does for consumers. For instance, booking a hotel room online, navigating a drive with Google Maps, and adjusting your household’s smart thermostat as you prepare to leave for a business trip, are all instances in which we interact with AI. In healthcare, AI has just as vast a range of uses, forms and goals.
Some of the most popular uses for AI in healthcare exist in the clinical space, such as AI-assisted robotic surgery, clinical decision support, image analysis and medical diagnostics. Although there are powerful real-life examples of AI transforming each of these clinical undertakings, the algorithms that back these technological capabilities do not yet have years of proven reliability or refinement. As a result, executives may feel cautious about decisions to implement such solutions as the ROI can sometimes be hard to measure or recognize quickly.
Similarly, other types of AI are picking up steam in healthcare that have nothing to do with clinical duties. Instead, these solutions take on repetitive, high-volume workflows that don’t involve much variation so administrative staff can focus their time and energy on more people-centric tasks. AI can assist most, if not all, functions of the hospital where high-volume, and often error-prone administrative tasks are required, including the revenue cycle. Organizations that deploy AI for these purposes expand the bandwidth of their staffs by offloading laborious tasks, thereby strategically positioning their organizations for the future.
Trends in AI right now
Hospitals and health system executives are likely encountering the following trends in their path to researching and potential implementation of AI:
Seeking AI education and guidance.
Although most health system CFOs and CIOs recognize the need for AI in their 3- to 5-year strategic plan, most are still educating themselves on AI and determining where it would be most appropriately deployed within their organizations. Health systems possess many core competencies, and AI has not traditionally been one of them. Thirty-five percent of U.S. life sciences CEOs said a lack of internal skills and knowledge is the biggest barrier to implementing AI in their organization, according to a 2017 KPMG survey. CFOs and CIOs have two primary places to draw from for AI case studies and success stories: clinical AI and operational AI. To thoroughly understand how AI can transform their businesses, executives need a basic understanding of AI’s components, including robotic process automation, computer vision and machine intelligence. Many workflows and tasks within a hospital require several of these capabilities at once. We will share an example of RPA, CV and MI working together toward the end of this article.
Assessing clinical and operational AI to support organizational strategy in different ways.
Many organizations are considering clinical AI to improve patient care, but ROI on improvements in the patient care experience can be difficult to prove. On the other hand, “blue-collar AI,” which removes administrative burdens in the revenue cycle to improve cash flow and operational efficiency, represents a more clearly defined opportunity for organizations to quantify and measure value. Revenue cycle functions that are ripe for AI include eligibility checking, claims status updates and procedural prior authorization. These AI applications may not be seen as revolutionary and cutting edge as tumor-detecting algorithms, but they are nonetheless poised to change healthcare by transforming hospitals’ productivity, workforces, revenue cycles and more.
Debating homegrown AI or AI with outside partners.
Since AI and automation are still somewhat new to the healthcare market, executives are weighing the costs of building these solutions internally, hiring consultants to assist in the implementation or leveraging an external partner to build the solutions, implement and manage them from soup to nuts. The best approach depends on the hospital’s maturity, needs and appetite for speed, but organizations will experience the most success by selecting any vendors carefully.
Key steps to identify the right partner or partners for your AI implementation are to:
a.) Define the end goal before you start vendor selection. Is your intent to displace staff? Reallocate to new projects? Drive key financial drivers? The right vendor will understand your intent, and have experience impacting it for others.
b.) Assess your resources. Many health systems do not have the technical resources with the right skill sets to implement AI internally, so making the investment in an implementation partner is worth it. For others, funds are a constraint and thus are willing to add new skill sets to existing IT resources to manage it internally.
c.) Consider ongoing support. Inevitably your payers’ regulations are going to change, one of your tools’ UI’s will update…some resource is required to maintain automations once they are built. Lastly – regulatory requirements are continuing to tighten in healthcare. Select a vendor who knows healthcare and can help you to navigate these complexities through your implementation, or prepare to handle this internally.
An example AI use case
The revenue cycle is marked with error-prone processes and tasks that take up too much of employees’ time. The best workflows for automation are those marked with repetitive, high-volume, rules-based tasks — which are prone to human error. Metrics to assess the ROI of automation include human hours required, error vulnerability, potential to increase revenue or profit, and volume per day.
Reallocation of resources provide a strong potential return for organizations as well. For example, one hospital used automation to manage the insurance eligibility check process. With the time that was subsequently freed up, the hospital was able to switch to more of a concierge model for insurance eligibility verification, which has improved productivity and patient satisfaction.
Here at Olive, Olive (our AI solution by the same name) uses RPA, computer vision and machine learning to tackle prescription prior authorizations. First, Olive is alerted when the provider receives a fax from a pharmacy initiating the prior authorization process. She then extracts the information needed to complete the prior authorization from the hospital’s e-fax software and uses its EHR to fill in any additional information before sending it on to the pharmacy benefits manager. If the PBM has additional questions for the provider to complete the prior authorization, Olive notifies the provider that there is an incomplete prior authorization in the hospital’s queue.
By automating this process, Olive takes on the copy-and-paste steps of completing the prior authorization form, as well as contacting the right PBM or payer and faxing of the forms.
The revenue cycle — a hospital’s financial backbone — contains numerous steps and processes, like prior authorizations, that are a logical use for AI. Through automation, organizations can reallocate staff and other precious resources to quickly recognize and measure ROI.