Is AI in healthcare a massive money pit? Maybe, maybe not.

It’s too soon to characterize the economic impact of AI across Western healthcare with anything more than rough guesstimates. This is so for two reasons. 

One, direct research into the matter is thin bordering on skeletal. And two, the issues confronting those who set out to assess AI’s effect on healthcare economics are knotty. How knotty? Enough to preclude the degree of clarity needed to aid cost-conscious decisionmakers even with broad strokes. 

A review of the relevant English-language literature produced these conclusions. The research was conducted in Finland and is slated for publication in the January 2026 edition of the International Journal of Medical Informatics.

“In line with past literature covering the economic evaluation of AI technologies in healthcare, the findings of this review identified significant gaps in reporting essential details related to economic evaluations,” write Hanna von Gerich, MNSc, of the University of Turku and colleagues. That shouldn’t be surprising, though, since “half of the included articles did not recognize the study as an economic evaluation in the title, nor was economic evaluation the main aim of the studies.”

In their discussion section, the team offers observations that may guide the needs, aims and designs of future studies into the economic impact of AI on healthcare. Here are three. 

1. As the significance of AI in healthcare grows, so does the need for consistent economic evaluations.

This need “urges the development or complementation of existing technology development frameworks with unified and established economic evaluation and reporting guidelines,” von Gerich and co-authors remark. “Likewise, previous reviews have reported insufficient and superficial reporting of technological details of the evaluated systems.” 

‘More than half (n = 10, 55.6 %) of the studies included in this review reported the methods used as simply ML or deep learning, and the comparison of the studies to the CHEERS-AI reporting standards indicated a lack of key details related to specific characteristics or impacts of AI.’

2. AI-based systems are not isolated or standalone tools. They involve complex interdependencies between technologies, healthcare providers and healthcare users.

The performance of learning systems progressively changes over time, the researchers point out. The resulting inconstancy affects “not only the outcomes provided by [such systems] but also the behavior of the healthcare professionals using them.” Meanwhile, “complementing the economic evaluations with specific features and characteristics of the used systems is vital to assess and truly understand their real impacts.” 

‘AI is commonly framed as a cost and resource-saving tool, overlooking the support and maintenance it requires to operate successfully, including regular evaluations of the system’s performance or the associated costs for integrating it into novel or existing healthcare systems. Most studies in this review did not cover the purchase or other cost components related to the technologies.’

3. Issues in reporting AI-specific details in economic evaluations might partly stem from a lack of comprehensive understanding of the potential impacts of these systems. 

“This could reflect a lack of engagement between the system developers, who understand the technical details and requirements of the AI systems,” von Gerich and colleagues surmise, “and the system users and evaluators, who understand the healthcare environments and specific features related to the favorable outcomes of the systems.”

‘Future research could further investigate how to support better preparation for and account for more in-depth system evaluations, starting from the early planning phases of the system development life cycle.’

Read the rest.

• Other newsworthy research: 
   
  • Icahn School of Medicine: New AI tool addresses accuracy and fairness in data to improve health algorithm
     
  • Technical University of Munich: How patients view medical AI