The following is a guest article by Lars Maaløe, Co-Founder and CTO at Corti

The integration of Artificial Intelligence (AI) across healthcare—in clinical workflows, care coordination, business processes, logistics, revenue cycles, first responder and emergency services, and others—is a development that holds immense promise. As a technologist, academic, and entrepreneur deeply immersed in this field, I’ve witnessed firsthand the transformative power of AI in healthcare. As a result, I am deeply optimistic about AI’s benefits as a support tool for doctors, nurses, and other clinicians.

However, the credibility and success of AI in healthcare have reached an inflection point. As an industry, we need to quickly move from “we could do” to “we have done” in the adoption of AI and other intelligent automation tools for this technology to realize its potential.

To bridge this gap, the implementation of AI solutions in healthcare should honor and remain deeply rooted in the industry’s history of evidence-based research as much as possible to achieve its greatest positive impact. As optimistic as the healthcare industry is about AI, its ultimate successful adoption hinges on a crucial factor: how AI is supported by evidence.

Evidence is Foundational to AI’s Long-Term Success

The reliance on evidence-based data and reporting in healthcare is a time-honored tradition critical for making informed, safe, and effective patient care decisions. As AI gains momentum and grows in potential, this reliance becomes even more pivotal. AI systems process vast amounts of data, and their outputs can significantly influence clinical workflows. Therefore, ensuring that these AI systems are grounded in reliable, scientifically validated data is imperative for their acceptance and effectiveness in clinical settings.

A fitting example is a recent study published in Nature titled “A Retrospective Study on Machine Learning-Assisted Stroke Recognition for Medical Helpline Calls,” highlighting AI’s utility in improving stroke diagnosis during medical helpline calls. Stroke, as a major health concern, demands accurate and timely recognition for effective treatment. This study is pivotal for showcasing how AI enhances decision-making in critical situations. It finds that AI-supported workflows can outperform traditional stroke decision-making in meaningful and measurable ways to greatly improve time to relevant care.

Notably, as described in the study discussion, “Our results showed that a machine learning framework can substantially improve stroke recognition in medical helpline calls compared to solely relying on human call-takers. This improvement was observed across all performance metrics and for basic patient demographics (age and sex).”

There is Urgency in Establishing Trust in AI

AI must align with the established practices and standards of healthcare knowledge to gain credibility in this sector. Like any new medical intervention or tool, AI applications in healthcare should undergo stringent development, testing, and validation processes. The study involving the analysis of medical helpline call data for stroke recognition serves as an exemplary model of how AI is being developed and validated using rigorous scientific methods.

For example, ambient clinical voice (ACV) has emerged as a top priority for most health systems and a focus area for Healthcare IT Today. However, for a physician to trust an AI to assist them with ambiently documenting and assigning a set of healthcare codes, e.g., ICD-10, they need to understand the rigorous validation supporting the technology. In my recent video interview on this topic, I explain the research that went into our large language models and predictive modeling tools to improve the clinician-patient conversation and, ultimately, the long tail of healthcare coding, billing, and reimbursement processes.

Strong, evidence-based foundations for AI applications and use cases in healthcare are vital for the current applications of AI and its future advancements. Ensuring AI applications are safe, effective, and dependable through evidence-based research is a responsibility that must be upheld.

It’s Not Just About the Tech, AI Also Benefits from Interdisciplinary Collaboration

The success of AI for stroke recognition also highlights the importance of interdisciplinary collaboration. In this example, neurologists, data scientists, machine learning experts, and emergency medical professionals came together. They combined their expertise to develop a solution that enhances prehospital care quality.

This study is a milestone in a much larger journey and represents the fusion of medical knowledge with cutting-edge technology, driven by a shared commitment to advancing healthcare. Such collaborative efforts are essential for successfully developing and implementing AI solutions in the industry.

Scientific Methods and Rigorous Processes are Stalwart Guides During Transformational Times

As the healthcare industry navigates its way through this AI (r)evolution, the emphasis should be on rigorous validation, ethical considerations, and a focus on improving patient outcomes. These principles will guide the successful integration of AI into healthcare workflows, ultimately enhancing the quality of care provided to patients.

As thought leaders and innovators in this field, it is our duty to ensure the development of AI tools also adheres to the highest standards of scientific rigor and ethical practice. In doing so, we can pave the way for a future where AI and medical expertise work in harmony to deliver innovative, safe, effective, and beneficial patient care.

About Lars Maaløe

Lars Maaløe is the Co-Founder and CTO at Corti and Adjunct Associate Professor of Machine Learning at Technical University of Denmark (DTU).