Key takeaways from Harvard Medical School's AI in Health Care: From Strategies to Implementation program

For over two decades as an orthopedic surgeon, I’ve specialized in hand surgery at Proliance Orthopedic Surgeons in Washington. Concurrently, for the last four years I’ve worked at Cohere Health , and this past summer I received Harvard Medical School’s certification of completion.

Harvard Medical School Executive Education’s AI in Health Care: From Strategies to Implementation program is designed to equip health care leaders, like me, with the knowledge and strategies to design, pitch, and implement AI-driven solutions that bring about transformative change to their organizations.

Through case studies, real-world examples, and a capstone project, I engaged with Harvard Medical School faculty and guest speakers about how best to:

• Evaluate existing AI systems in health care to understand their strengths and weaknesses
• Identify new opportunities for AI in health care to address unmet needs
• Assess the ethical implications and potential biases of AI technologies in health care settings

As the Vice President of Clinical Strategy for Musculoskeletal Health at Cohere Health , I spearhead various programs aimed at improving patient care through the use of AI and machine learning (ML) technologies. These programs include: identifying high-quality providers; implementing comprehensive episodic authorizations (to avoid needing multiple prior authorizations for a care episode); improving our rules to allow more real-time auto-approvals of prior authorization requests; and enhancing Cohere’s care-path approach, which enables patients with access to the safest and best care available. During this course, I gained a better understanding of:

• How and when to use evaluation measures to ensure effective use of AI models in health care settings
• What is the level of certainty required to take a clinical action based on an AI recommendation/prediction
• How to assess the ethical implications and potential biases of AI technology in health care settings

The program’s curriculum provided exceptional examples of how to make informed and strategic decisions about AI’s implementation in the complex cultural, economic, and regulatory context of today’s health care landscape. The potential of AI to change and revolutionize health care is vast; however, there are high stakes and challenges that must be considered when implementing this technology.

During the Transparency, Reproducibility, and Generalizability in AI session, Dr. Karandeep Singh discussed best practices for evaluating AI predictive models, specifically those used to predict whether a patient has breast cancer. He spoke about the vital considerations and measures that must be used to decide whether an AI model is performing within the right parameters and standards to predict the intended outcome. When determining the best approach to using these models in a clinical setting, it is crucial to establish the appropriate probability or risk threshold to ensure the successful deployment of the AI model.

Listening to Dr. Singh’s session reinforced my philosophy that the responsible use of AI technology will never replace the work of physicians or the art of medicine, nor should it be used to deny care. Ultimately, technology accelerates the prior authorization process, paving the way to an appropriate “yes,” and benefiting patients, providers, and health plans alike.

Physicians, like myself, are key collaborators in building and training AI and ML models. At Cohere, we’ve built four core principles for utilizing responsible AI within the prior authorization process.

We are leveraging AI to transform prior authorization. It’s time for prior authorization to stop focusing on transactions and start creating a more patient-centric approach by tailoring care plans to individuals’ needs and circumstances.

Cohere’s intelligent prior authorization platform processes over 9 million requests annually, positively impacting more than 15 million health plan members and 492,000 health care providers nationwide. Currently, Cohere’s solutions are utilized by several major health plan clients nationwide, including Humana , Geisinger , and Medical Mutual . To date, Cohere’s intelligent prior authorization platform has:

• Enabled 70% faster access to appropriate care (roughly equivalent to five days) for patients
• Improved providers ability to schedule patient care by 81%?
• Automatically approved 89% of prior authorization requests*
• Decreased time providers spent on prior authorization submissions by 61%
• Reduces medically unnecessary surgeries (arthroscopy) by 35%

*calculation based on a health plan's implementation of Cohere Complete?