How AI Fits in the Healthcare Puzzle – Four Things to Know

U.S. healthcare operators, regardless of size or specialty, face increasing pressure to reduce costs while maximizing revenue and cash flow in an already complicated system. Outdated controls, disparate IT systems and complex processes for patient interaction, coding, billing and reimbursement make it difficult to identify and implement change. The increased scrutiny on patient outcomes, safety and satisfaction, along with increasing labor, medical supply and drug costs only make the task of running a successful healthcare business more challenging.

At the intersection of information technology, healthcare operations and business performance improvement, Artificial Intelligence has emerged as a production-level, highly-customizable solution that has driven increased profitability at the largest healthcare providers and payors. Advances in AI tools and technology have made these robust solutions increasingly accessible to early-lifecycle healthcare startups, PE and VC-held growth operators and other middle market healthcare organizations.

What does AI mean for healthcare organizations across the US? How can organizations take practical steps to assess, implement and leverage AI solutions? Here are four things every healthcare organization should know as they consider introducing AI into their IT environment.

(1) What is artificial intelligence?

The term Artificial Intelligence broadly refers to a wide array of capabilities along a spectrum of complexity. This spectrum goes from the familiar such as expert systems, decision support and predictive analytics tools to advanced automation solutions such as machine learning, natural language processing and robot process automation (RPA). The truly autonomous, sentient AI arguably does not exist – as far as we know.  The spectrum of innovation and application is ever-growing and constantly debated. Most of us are already familiar with some version of “AI.” We are comfortable receiving AI guidance on our shopping decisions, health, entertainment or even romantic partners. To accomplish this feat, systems rely on relatively straightforward predictive analytics: recognizing patterns and making assumptions based on large data sets, consumption activities and behaviors. Getting the “system” to detect patterns and understand the relevant signal within the noise requires a great deal of guidance, rules-based logic, a large data set to train the system and manual exception processing. All of this activity – all of these rules and selected inputs – are developed, built and maintained by humans. So, what is AI in healthcare? AI in healthcare refers to a tool or set of tools that, with some level of autonomy, helps resolve business challenges, uncover business insights and automate repeated tasks to free up people to apply their skills to other more productive tasks. AI in healthcare enables humans to spend more time on strategic, value-added activities that computers still can’t handle like maintaining a strong relationship with the other participants in your ACO or taking your client’s AP manager to lunch.

(2) What AI is not (yet?)

Understanding how to incorporate AI solutions in a healthcare environment requires developing a common understanding of AI capabilities and limitations and, perhaps most importantly, understanding what challenges can and cannot be addressed with AI solutions.Artificial Intelligence is not magic. It’s not one size fits all. It won’t solve fundamental business issues and it won’t solve fundamental people or process challenges. It may self-improve and self-correct over time, but AI is built by humans and limited by humans.One of the most recognizable issues with AI is the learning bias problem, in which human bias is transmitted to the system via the selection of data sets and the logic structure of the system’s algorithms. Over time, these omissions and influences can amplify, creating the possibility of serious negative outcomes.As much as AI is enabled and built by humans, it is not always capable of providing a response sufficient for human understanding. AI presents what we call the “black box problem” in which the system is unable to provide a human-context explanation of why a decision was made – it is unable to “show the math.” This deficiency is less of an issue when the struggle is to understand why your AI tool is transposing call center data into your system. It’s a much bigger issue when it is determining what drugs should be administered in a NICU.

(3) Common challenges that AI can address

There are many areas where AI can provide real automation and efficiencies that significantly and positively affect healthcare organizations. Healthcare professionals spend an inordinate amount of time recording, managing and interpreting data from a variety of systems such as off-the-shelf systems, electronic health record systems or customer relationship management tools. The time spent on translating data is time away from patient care and other activities necessary to run a successful business. Considering all that’s required in every medical setting, AI can help solve challenges such as:

• Human capital management:Employees have a critical role in influencing patient outcomes and patient satisfaction. For instance, a nurse practitioner’s time is better spent with patients or training other employees than managing and tracking labor costs or staffing needs. To free up that time, using AI properly can uncover patterns within patient populations, schedule services, and discover seasonality effects to reveal optimal scheduling and staffing structures to meet patient needs.
• Patient matching: To validate information such as patient history or eligibility, most healthcare organizations have more than one system or source of data (EHR, PHR, etc.) in place. Most healthcare organizations also struggle toggling between those different health IT systems. Validating this data – ensuring the data belongs to the same patient and is error-free – is a time-consuming but critical effort. Providing AI the rules and logic for automating patient matching and safeguarding data integrity does more than just prevent a rejected claim – it can help doctors, nurses and technicians avoid critical mistakes that can harm patients.
• Revenue cycle management: For most healthcare provider organizations, regardless of size, revenue cycle management is a complex, disjointed process made up of different systems and multiple sources of data, each with unique variables. A successful healthcare organization needs to consistently capture accurate coding and charges, submit timely billings and claims, manage collections and post payment. AI – in this case RPA – can help streamline and execute repeatable processes, aggregate information while checking for errors, and present metric information in a way that allows human participants to manage the business efficiently.
• Other back-office functions: The same challenges with collecting, manipulating and interpreting data across multiple systems happen in the back office. Continuously juggling between the CRM, ERP, and GL systems in between departments, from accounting and finance to contracting and managed care, leads to inefficient rework and ultimately lost revenue. Using AI to coordinate back office data flow can help alleviate this challenge by enabling faster process execution and uncovering data insights such as lease agreement overlap and staffing inefficiencies that are impacting the business.

(4) How to move forward

It is critical for healthcare operators to fully understand the technology options available, what each solution accomplishes, what hurdles may arise in sensitive clinical and non-clinical environments, and the initial and ongoing costs for the solution. AI solutions are increasingly affordable to most healthcare operators at all stages of the business lifecycle, but it is not a one-size-fits-all solution.

Determine what needs to be accomplished and if AI the best tool to support your initiative. Define key goals, motivations and the organization’s appetite for change. If your goals are too complex, think about how to break the challenge into more achievable, manageable pieces. A series of wins does more to convince skeptics in your organization than one big initiative that requires herculean effort and presents multiple potential points of failure.

Next, highlight champions within the organization that can own a project and cheerlead for success. These are people who have identified an opportunity as well as possible solutions. Look for practical, discrete and compelling challenges and opportunities that have business value and reasonable expectations of success.

Finally, consult with internal and external experts and partners that can provide a clear understanding of what is achievable and can provide options to help you think through and rationalize next steps.

Suffice it to say, AI is not a perfect solution to every challenge nor is it as an existential threat to human existence. Instead, AI is an emerging set of tools and capabilities that can be applied to a variety of healthcare challenges to free personnel to focus on making continuous strategic improvements that improve the healthcare payor and provider environment.

This article first appeared in HealthCare Business News and is reposted with permission.