Artificial Intelligence Considerations for Healthcare Leaders
Written by Joachim Roski and Scott Barr
Written by Joachim Roski and Scott Barr
Note: The below article includes excerpts from the September 2018 Health Affairs article “Implementing and Scaling Artificial Intelligence Solutions: Considerations for Policy Makers and Decision Makers.” The full article can be accessed through the Health Affairs Journal, and was developed in partnership with Eric Just, MS, from Health Catalyst; and Bruce L. Gillingham, MD, CPE, FAOA RADM, MC, USN, Rear Admiral in the US Navy.
The New England Journal of Medicine, Harvard Business Review, and other publications, noted that Artificial Intelligence (AI) and machine learning can achieve breakthroughs in improving patient safety, health, and reducing waste. AI infers patterns, relationships, and rules directly from large volumes of data in ways that can exceed human cognitive capabilities. Today, opportunities for greater deployment of AI in health are made possible by increased electronic data availability from mobile devices, sensors, cameras, and electronic health records; faster data processing capabilities; and newly developed computing techniques. AI solutions are still only used sparingly in health/healthcare organizations. However, those organizations able to increase their AI capabilities will likely be better-positioned for the transition to value-based care and achieving desired results with greater efficiency.
While there are complex technical and methodological details that underlie specific AI applications, there are three basic types of AI solutions including: simple task automation (Type 1), pattern recognition (Type 2), and contextual reasoning (Type 3). All AI solution types can offer significant improvements across multiple use cases in health including clinical support, population/public health, patient/consumer experience, and administrative/business processes. Health organizations will likely benefit from critical, data-driven improvements across all use-cases to achieve desired results effectively and efficiently.
Decision makers in health organizations need to decide if and how to adopt AI technologies for their organizations. They need to take a wide array of considerations into account including AI solution types, requirements of different user groups or use cases, varying technical/analytic complexities, the ability to scale and sustain solutions across their enterprise, and return on investment.
As federal healthcare officials, private sector health executives, and other health leaders look to integrate AI into their organizations, we offer three main considerations:
Consideration 1. Entry Costs Can Be Low. First, entry costs to health organizations for implementing individual AI solutions are often not high and can yield significant ROI quickly. Type 1 AI solutions (simple task automation) are commercially widely available, are sometimes embedded and available in modular fashion as part of larger software solutions (e.g., EHRs), or can be quickly developed and implemented de-novo. These solutions strive to reduce the burden associated with repetitive, time-consuming tasks often performed by clinicians, analysts, administrators. It has been estimated that $18 billion dollars could be saved across the health care industry by applying AI solutions to eliminate certain administrative processes (e.g., automating writing chart notes, prescribing medications, and ordering tests).
Consideration 2. Success With Implementation May Not Translate Easily Across Different AI Solutions. Success in implementing simple task automation solutions may not translate easily into success with more complex (Type 2 and 3) AI solutions. Many Type 1 AI solutions we have developed and implemented target efficiency gains by eliminating administrative or other support processes that operate in the “background (e.g., the Finance/Audit Department). The stakes significantly increase when developing and implementing AI solutions that support clinical care or population health processes that could negatively impact the care experience, outcomes, or patient safety.
To reduce time-to-development/deployment and resource-intensity, some organizations may benefit from crowd-sourcing solution development. For example, in collaboration with the Broad Institute at MIT and Harvard, Booz Allen hosted the Data Science Bowl competition inviting a global community of data scientists to develop AI Type 2 solutions in the field of image processing to accelerate medical research (prior years focused on cancer and cardiovascular disease diagnostic support). Within 90 days, more than 18,000 data scientists submitted 68,000 algorithms focused on analyzing images of cells and identifying nuclei to help speed development of new drugs. The most promising algorithms will be made publicly available to researchers, possibly saving thousands of hours of effort per year and advancing critical biomedical research.
Consideration 3. Sustaining Enterprise-Wide Impact Requires The Long-Term View. Finally, scaling AI solutions and sustaining broad organizational impact across applications and use case requires significant, sustained leadership focus and investment. Most health organizations we have worked with including large and small public and private sector organizations, are at early stages of piloting and implementing Type 1 and Type 2 AI solutions. Moreover, as these organizations gain experience in piloting and implementing such solutions, they are discovering that scaling AI solution development and deployment across their enterprise proves to be an even greater lift than getting started. Some organizations have adopted a long-term view on needed enterprise-wide changes to accommodate AI solution development, deployment, and long-term sustainment. They focus on a number of critical changes including (1) enterprise-view on solutions; (2) modernizing systems; and (3) focusing and integrating expertise.
Given the likely slow pace of policy development, decision makers in health organizations will need to apply sound judgment in interpreting existing regulations and consider consumer acceptance when deploying AI solutions. As organizations move from the implementation of isolated AI solutions to deployment at scale across their enterprise, organizational culture often impedes more rapid progress. Hence, it will likely be critical that organizations carefully assess and evolve their organizational maturity across multiple dimensions in becoming a learning or data-driven organization.
