Healthcare AI's Elusive ROI
Originally published in Forbes.
Healthcare AI holds great promise, but promises do not pay the bills. In Bain and McKinsey surveys, American healthcare leaders reported expecting—even requiring—a positive return on their AI investments. However, demonstrating ROI is challenging and will remain a significant barrier to adoption for the foreseeable future.
AI's Potential Benefits And Costs
Healthcare organizations could harness AI for various benefits, including improving quality of care, enhancing patient and staff experiences, accelerating research, and extracting more insights from data.
They may also use AI to directly increase revenue by boosting volume, speeding throughput, increasing risk adjustment and service level coding, and improving revenue cycle management. Meanwhile, AI could cut costs by reducing staffing needs, decreasing staff turnover, and improving supply chain efficiency.
PROMOTED
Of course, AI also adds costs. Evaluating different AI products takes time and effort, and organizations must consider the opportunity costs (AI could distract from other activities) and reputational risks (due to possible adverse events).
Additionally, implementing AI is complex, requires significant resources, and is fraught with potential pitfalls. Productivity often drops temporarily ("switchover costs"). Once implemented, ongoing software, monitoring, and data infrastructure expenditures exist.
Why Analyzing AI's ROI Is So Difficult
Translating abstract concepts like quality, efficiency, and productivity into numbers is challenging and requires asking many hard questions.
For one, from whose perspective is ROI assessed? The interests of healthcare's various stakeholders do not always align. For example, a nurse manager may find AI that triages patients appealing if it reduces staffing needs, but patients may bristle at having to do more work on their own.
Second, who pays for AI, and who does it affect? Often, those making AI purchasing decisions differ from those it impacts most. For instance, a C-suite executive may push an AI tool that increases risk coding, but physicians may resist it if they must change how they document care.
Third, what is the time horizon? History tells us that organizations initially use technology to make their existing processes more efficient, limiting the total benefits. It takes many years to unlock new, better ways of producing goods and services. We see this today as health systems race to adopt AI that enables clinicians to write the same (typically shoddy) notes faster rather than rethinking clinical documentation entirely.
Fourth, what is the baseline performance? Most organizations know little about the time and effort clinicians and staff spend on various tasks (e.g., writing a discharge summary) and even less about the quality of their output (e.g., are the summaries accurate and readable?).
Finally, which metrics best evaluate AI's impact? Healthcare data is siloed and incomplete, and it is hard to measure—and assign value to—constructs like quality and clinician wellness.
Why A Positive ROI May Be Elusive
As Goldman Sach's Head of Global Equity Research, Jim Covello, explained, "The substantial cost to develop and run AI technology means that AI applications must solve extremely complex and important problems for enterprises to earn an appropriate return on investment." But it may be too much to expect today's AI to solve complex and important healthcare problems.
For one, generative AI tools are typically too unreliable and error-prone to apply to high-value tasks. So, most organizations use them to offload "drudgery," such as writing clinical notes and completing prior authorizations. Yet, AI could paradoxically make these tasks harder. For example, physicians at UC San Diego Health who used ChatGPT to respond to patient messages paradoxically spent 22% more time on this task than those who did not use AI.
Even AI solutions that save time may not boost productivity. Consistent with Parkinson's Law–that "work expands to fill the time allotted for its completion"–three in four British clinicians reported they would not spend time that AI frees up caring for patients. American physicians now adopting tools like AI scribes are likely no different.
Likewise, AI does not work in a vacuum. Organizations must relieve various downstream constraints to realize AI's benefits. For example, automating patient scheduling will not improve access if doctors already have full schedules. Similarly, algorithms that identify inpatients ready for discharge are useless if there is nowhere to send patients post-hospitalization.
Payment models pose additional challenges. Most healthcare payments are fee-for-service, with payers very rarely reimbursing AI software. Thus, to break even financially, organizations adopting AI must increase service volume, often substantially, given their low single-digit operating margins. Yet most AI tools – for example, those that summarize clinical records, identify patients at high risk for deterioration, or help detect precancerous colon polyps – do not impact volume.
Across industries, technology transformation programs typically realize less than one-third of their expected value. AI in healthcare will be no different.
Looking Ahead
None of this is to say that AI in healthcare is worthless. AI can help make care more accessible, effective, and sustainable. Still, ROI pressures will intensify as AI's hype (and attendant FOMO) wears off.
Consequently, AI will most rapidly penetrate back-office financial areas, such as revenue cycle management. Startups in this space have already seen some of the highest maturity rates, valuations, and exits.
Likewise, many clinician-facing AI products will extend into activities directly impacting finances. For instance, documentation and summarization tools will start recommending risk codes and charges.
For other AI products, organizations may require clinical teams to see more patients or reduce staffing. The point is that while AI holds great promise and could transform care over the long term, there will be no free lunch.