JMIR Publications

ABSTRACT

Large Language Models (LLM) continue to exhibit noteworthy capabilities across a spectrum of areas including emerging proficiencies across the healthcare continuum. Successful LLM implementation and adoption is dependent on factors such as digital readiness, modern infrastructure, a trained workforce, as well as privacy, ethical and the regulatory landscape that can vary significantly across the healthcare ecosystem. This perspective discusses the economics of LLM implementation pathways to facilitate their equitable distribution across healthcare organizations. Three broad onboarding pathways (TSP: Training from Scratch Pathway, FTP: Fine-Tuned Pathway and OBP: Out of the Box Pathway) along with risks, benefits, and cost-comparisons across four major cloud service providers (Amazon, Microsoft, Google, Oracle) are presented. TSP provides the most customization but is resource-intensive, FTP balances customization and efficiency, while OBP offers rapid deployment with minimal customization. Although LLMs have the potential to transform healthcare outcomes, their equitable adoption and democratization through these pathways are critical for their long-term success. Understanding the economics and trade-offs of the LLM onboarding pathways can guide healthcare organizations in strategically adopting LLMs to improve patient outcomes.