Dr. Christine Gall, Head of Healthcare Marketing, T-Mobile for Business

There has been a lot of news about what AI is, with OpenAI's ChatGPT having more than 100 million users sign on within two months of launch. Simultaneously, there has been substantial news coverage about potential problems with AI, particularly with Microsoft Bing's AI chatbot's threatening and odd behavior², as well as concerns whether AI will eliminate jobs.

The concerns have even caused several well-known computer and AI experts, including Elon Musk, CEO of SpaceX, Tesla and Twitter, and Steve Wozniak, co-founder of Apple, to pen a letter calling for AI developers to stop development until "we are confident that their effects will be positive, and their risks will be manageable."³

The questions arise: Is AI technology safe for use in healthcare? And in terms of how to advance health equity, can AI bridge the divide?

As of November 2022, the FDA approved more than 520 AI tools to support healthcare delivery.

With resolved healthcare disparities and innovation, McKinsey estimates the U.S. could gain a 10% GDP increase, or $3 trillion, in 20407.

The potential for AI to improve workflows, efficiencies, and accuracy of diagnosis and treatment is easy to imagine. Harvard's School of Public Health estimates AI can save the medical industry up to $150 billion by 2025.

• Create more accurate social and health data to support policies that drive improved health equity. An ongoing National Institutes of Health project, All of Us, is attempting to "bridge the gap for historically under-recognized populations by collecting observational and longitudinal health data on over one million Americans."⁸
   
• Help identify the problems. Health systems collect enormous amounts of data about their patient populations. AI can be used to help these organizations analyze the data to understand where providers should focus for the best return on investment (ROI) for positive patient outcomes.⁷ Writing for HIT Consultant, Dr. John Sargent, founding partner of BroadReach Group, provided an example:
  • A case manager checks into the system and reads an email about a patient, John Doe. An AI-powered algorithm flagged John Doe with two issues that might make managing his diabetes more difficult. First, his current provider is not a native Spanish speaker. And second, John Doe does not currently have a vehicle. These two health inequities could affect John Doe's ability to both acquire the right information and make it to his clinic appointments.  
• Identify "next-best actions." Dr. Sargent noted that identifying the problem was important but needed to be accompanied by successful actions to neutralize the identified barriers. "By using AI to provide predictive and prescriptive recommendations in a culturally sensitive way, we can bridge the equity gap," he wrote.
  • In the case of John Doe, the first recommendation would be to find him a physician that speaks Spanish. The second would be to implement telehealth services if he cannot resolve his transportation problems. As Dr. Sargent wrote, "If Amazon can predict which book on the history of World War II I should read next based on my buying history, certainly we can use similar technology to predict what issues will arise for our patients and what we need to do to intervene."
• Better allocate resources. This relates to ROI. Using AI in healthcare offers a more strategic look at assets, both skills and resources, to meet equity challenges. Within the context of John Doe, here are some questions that may arise from a resource analysis:
  • Is there a need for more Spanish-speaking doctors?
  • Or should they be deployed to other clinics or locations?
  • Should we invest in more telehealth applications, or partnerships with transportation providers to mitigate lack of transportation as a barrier to healthcare access?
• Clinician decision support. AI has the potential to be used by physicians to assist in diagnosing conditions and customizing the optimal plan of care for their patients. It could also potentially decrease clinician administrative burden by providing access to data from multiple sources, such as medical images, EHR data and even consumer devices like smartphones and activity trackers that collect health-related data.