Despite the ongoing buzz around artificial intelligence (AI), healthcare leaders are finding it difficult to proceed confidently with new solutions.
First, AI is not a single methodology or approach, but rather an entire category. Second, AI cannot make a positive impact by itself without changing underlying organizational processes. And lastly, people must learn to work with AI tools. Taken together, healthcare organizations must therefore deduce the best combination of people, process, and technology to tackle specific problems.
Given this complexity, it is important for organizations to apply AI in settings with fewer unknowns and a clear definition of value, such as revenue cycle management (RCM). This allows hospitals to gain a clearer understanding of AI and how to effectively measure its benefits, which in turn will help hospitals use this technology in more complex and difficult-to-measure domains like clinical care. Below are three guidelines healthcare leaders should apply when building their strategy.
Define the expected value and how it will be measured
For any new investment—technology or otherwise—it’s important to start with a clear understanding of the value you hope to achieve – and ensure you have a way to measure and attribute that value accurately. New initiatives are often undertaken to deliver value from efficiency gains, cost savings, revenue capture, and more, but it’s important to define up front which value is the primary goal and how that will be measured and communicated to organization leadership.
For example, clinician burnout is a common pain point and thus target for AI solutions. However, many organizations find it challenging to measure results – if you make improvements to documentation efficiency, but then add more patient value, have you really reduced burnout? If clinicians report less burnout, how exactly does that translate to attrition and turnover? Another common value is cost savings via automation (specifically, reducing FTEs). But, as every leader knows, “work expands to fill the time allotted” and oftentimes employees will be re-assigned to other tasks. Is the value from those tasks greater than the cost of automation?
A clear example of value is net new revenue. One example is to use precisely targeted AI for prebill review, scanning charts for additional documentation and coding opportunities after final coding but before billing. The attribution is well-defined as AI does its work after existing processes and each dollar can be matched to added documentation and diagnosis codes. Hospitals using AI in this way can measure and attribute financial results clearly and have realized millions of dollars in found revenue. (Savings on average have been reported at $2 million per 10,000 discharges.)
Evaluate the true cost
The true cost of implementing an AI solution is not just what is paid to the vendor. Other costs — such as staff time, training, system replacement and IT costs — must also be taken into account. For example, any change to physician workflow that requires education, monitoring, and training for hundreds or thousands of physicians may well cost millions of dollars. Leaders need to account for these dollars and not vendor fees.
The other often-misunderstood cost is IT integration. People often think the costs increase with the amount of data, but it actually scales with the depth of integration. Sending gigabytes of data periodically as one-way file transfers is typically easier than real-time bi-directional integration of even a few data fields. For clinicians at the bedside using AI tools, timely patient data and deeper integration is needed. But AI solutions focused on revenue cycle processes often work outside existing workflows and rely on straightforward data transfers rather than deep integration.
Minimize financial risk
Finally, consider contracting options to minimize financial risk. The most common contracting model over the last decade was Software as a Service (SaaS)—which typically requires multi-year lock-ins and upfront costs. If the organization discovers six months later that they are not achieving the expected value, they are still required to pay. However, modern AI solutions often offer different payment models that help minimize financial risk. For example, some solutions have a use-based model where the organization pays only for actual usage. Another is a contingency model, where the vendor is paid a percentage of cost savings achieved or new revenue captured. Organizations have much to gain from AI solutions when they share risk and reward with their vendor.
Conclusion
AI solutions have enormous potential to transform healthcare. Those leaders who proceed with full understanding of the benefits and potential pitfalls are better equipped to implement AI solutions strategically for both near-term and long-term gains. With RCM processes as the focus, organizations can benefit from AI’s ability to analyze vast amounts of data quickly and effectively while mitigating common risks. Narrow applications help organizations take action to reduce costs, increase revenue and reinvest those gains into optimizing patient care.