Much excitement swirls in health care as organizations start to pull in artificial intelligence (AI) and generative AI (GenAI) to improve everything from their billing practices and scheduling to patient engagement and clinical operations. Many already have started introducing AI to their back office strategies and are articulating savings.
But as the industry tries to realize the full potential of AI and GenAI, use cases can easily grow more complex, and organizations may find themselves trying to manage a web of technology blending automation, AI and GenAI that can be expensive and difficult to operate.
For example, say a health organization uses AI or GenAI to recommend and schedule a patient with an appropriate specialist for a given condition. This combines both “simple” steps – such as AI reviewing patient-provided information or EHR records – with more complex moves that may introduce risk, such as algorithms identifying the most likely diagnosis and specialist type. Quantifying the return on investment (ROI) can get complicated as the value may be realized by the patient in terms of convenience and outcomes, and harder to measure for the organization in traditional terms. As health care organizations layer more use cases throughout their operations, managing the risk, cost, data storage and maintenance grows more of a burden and must be properly assessed so the resulting fabric is not too expensive and chaotic.
ROI of Health AI Strategies
Research into the ROI for AI in health care strategies is still lacking, with one recent study in March 2024 citing “the urgent need” for such research and ROI calculators to help health organizations make these decisions. A February 2024 EY survey of 100 US health care executives confirms this feeling in the market around digital health solutions in general. Almost 90% of respondents acknowledge the potential cost reduction through digital health solutions; however, seven in 10 have yet to see any ROI to date, with most pointing the blame at siloed tracking metrics.
In this evidential void, some organizations are taking a “let’s just try it and see” approach. Sixty percent of survey respondents are actively investing in AI applications. Where is the most activity? An EY review of health organizations’ publicly available data around their AI use cases from January 2023 through April 2024 shows the most payer activity in data integration. The most common use cases for providers targeted testing and diagnosis. While health care organizations must embrace the potential of AI to improve quality, outcomes, cost and experience, they also must be strategic about when the rewards outweigh the risks.
During interviews for the new EY report, How to build a foundation in AI to accelerate health transformation, global executives stressed that the use of AI and GenAI in health care requires the right governance strategies. The report also suggested a wide range in the maturity of health organizations in assessing how to effectively govern the risks unique to AI, and in their organizational readiness for AI and GenAI.
Incorporating GenAI in Healthcare
With AI, some health organizations such as Highmark Health have been deploying AI for over a decade, and the anticipated costs are more clear, said Dr. Onyinyechi U. Daniel, Vice President, Data & Analytics Strategy for Highmark Health, who was interviewed by EY for this article. “There are a lot of considerations not yet known because of the maturity of GenAI, so it can be hard to put a number to it in terms of costs.” Organizations must make other decisions that may not be as clear because of the early stage of GenAI, such as whether you rely on an external large language model (LLM) or an internally built LLM.
The incorporation of GenAI ultimately “has to tie to an organization’s value story,” Dr. Daniel said.
While initial development and implementation costs are often the focus of new AI capability development, the total cost of ownership is a critical aspect of AI deployment decisions that should be considered as organizations mature their strategy. There are four factors for health care organizations to consider as they integrate AI into their operations for the long term:
- Operational costs are often overlooked in the early stages of AI development. This includes the costs of data storage, both for your current operations and a future that may include vast amounts of data from new sensors, wearables and other technologies. How will you scale your infrastructure? Where will you store and access the data – in the cloud or on premises? If in the cloud, what is your strategy for controlling cloud consumption? Which data can and should be archived? It will be critical to compare the value the organization is realizing from implementing AI against the cost of having all that data available and being able to access it.
- Maintenance costs also need to be taken into consideration, including software upgrades, system maintenance, ongoing support and troubleshooting. While vended solutions may offer some amount of ongoing support and maintenance, these often come with costs, and internally developed solutions must consider these factors in total costs of ownership.
- Ongoing performance monitoring, model retraining and testing can pose significant costs if not properly planned for. This includes compliance monitoring, particularly when access to large volumes of protected health information is needed. How do you remain compliant and secure, beyond just storage? What sort of infrastructure do you need to deploy and maintain for the data scientists to do their work? How do you demonstrate compliance with regulatory schemes given the need for broad data access?
- Human resource needs are often considered too late in the process. Thinking through who within your organization has the skills necessary to support and take responsibility for new AI health capabilities and strategies is crucial. In many cases, these may be new skills that don’t currently exist within the organization or may require upskilling current staff.
Finally, and perhaps most importantly, AI and GenAI strategies should not inflict a cost to the human relationship between clinician and patient. “Health care is ultimately about human engagement, human touch and empathy. People value genuine connection and don’t want AI to replace human interaction,” Dr. Daniel said. “Let’s not get overexcited about the technology and what we can do; we have to preserve the art and humanness of health care.”
The views reflected in this article are the views of the author and do not necessarily reflect the views of the global EY organization or its member firms.
Authors:
Aloha McBride, EY Global Health Leader
Sezin Palmer, EY Global Health Sector AI Leader