Commercial risk will be a critical catalyst of progress – it’s complicated, but is it possible? We think so.

Blog Post

Artificial intelligence reality check: Your team must prepare for these challenges

By Andrew Rebhan

March 5, 2018

    The artificial intelligence (AI) space is popular, immature, and crowded with options, but there are several challenges organizations need to keep in mind as AI adoption in health care moves forward.

    AI has been a focus of significant research and development for decades—but more recently, the rise in digitized data in health care, coupled with an exponential growth of computing power and a flood of investor dollars, have fueled a bevy of health care AI startups and projects by mature companies (e.g., Google, Apple). These initiatives aim to shape exciting new possibilities that could improve patient care and business operations. CB Insights keeps a working list of AI startups in the health care sector that currently contains more than 100 companies that are applying AI techniques to a variety of health care domains, including patient monitoring, medical imaging and diagnostics, and drug discovery.

    AI offers health care organizations the potential to improve decision-making and data interpretation speed, capacity, and consistency. AI can relieve staff of mundane tasks, continuously monitor complex streams of information, improve clinical decision support, and ensure that nothing falls through the cracks: In essence, AI could eventually be a trusted and capable assistant. AI can also help organizations address data overload, as clinicians can take into consideration only so much data at a time.

    Better data interoperability, the addition of patient wearables, genomic data, and publications detailing clinical best practices offer clinicians all sorts of opportunities to improve patient care. However, the advances are happening at a pace too fast for clinicians to make use if in patient care.

    Despite the benefits that AI can offer, there are a number of barriers to AI's adoption and use within health care.

    Business challenges

    • Cost: The average cost of developing, testing, certifying, and implementing AI is unclear at this stage, and without a way to measure return on investment, it can be difficult to justify large-scale AI initiatives.
    • Workflow: The industry is still figuring out how AI will fit into existing workflows in a way that is intuitive and not disruptive. Productive use of AI assistants may require new concepts in EHR user interface design and more seamless ways of integrating outside agents into native EHR processes.
    • Competing priorities: The health care industry is still suffering from "EHR fatigue" while also struggling to address new regulatory mandates and pressing initiatives regarding population health management, margin management, etc. Finding the time and money to focus on AI will be difficult.
    • Staff resistance: Some staff are skeptical that AI technologies can replace human decision-making, even for narrowly targeted tasks. Others are fearful that they will be replaced, or their hard-won skills devalued.

    Legal and ethical Challenges

    • Regulation: AI technology is advancing faster than health IT regulations can keep up, and finding the right balance of innovation and caution is key. It remains to be seen whether the government or FDA will develop specific policies to address AI in health care, or provide guidance on how existing software and device regulations will apply.
    • Liability: Health care is a naturally risk-averse industry, so the thought of machines making autonomous or opaque evaluations regarding patient care can be unsettling from a legal perspective. How do we deal with algorithms that produce unethical responses, or worse, lead to patient harm? There are also concerns regarding data privacy and whether AI will be vulnerable to security threats or hacking.
    • Human/machine interactions: How will patients interact with AI over the long term? Patients are still likely to prefer human interaction when they visit a doctor's office or hospital, and it will take some time before the majority of patients are willing to trust care decisions to an algorithm. Some vendors have started to address these concerns by developing technology that can read emotions and possibly display empathy in the near future.

    Considerations to get started

    Despite the many unanswered questions and challenges facing AI in health care, the field is showing impressive early results and can open up immense opportunity if done right. Below, we've outlined a number of considerations for both your internal operations, but also for if your organization begins to evaluate AI products and services from vendors or potential partners. Let these considerations act as a starting point to get your team thinking about how to strategically prepare yourself for the next wave of digital innovation.

    Comprehensive Planning Can Prevent Headaches down the Road

    Should radiologists embrace or fear machine learning?

    Learn the benefits and challenges of adopting AI into your imaging strategy.

    Download Now

    Have a Question?


    Ask our experts a question on any topic in health care by visiting our member portal, AskAdvisory.