AI in Healthcare
Slow to embrace artificial intelligence (AI), healthcare organizations are accelerating the pace of adoption by establishing the organizational and technical infrastructure required and delivering use cases across the entire enterprise.
Early application of AI in healthcare was often a result of grant-based research to promote innovation in clinical care. The results of many of these projects were not commercialized and thus remain in the annals of peer-reviewed journals. As healthcare providers see the benefits of applying AI to deliver insights that help avoid adverse events such as readmissions, their comfort level has grown.
This white paper is based on the results of the AI in Healthcare Survey conducted by IDC and sponsored by InterSystems. The respondents of the AI in Healthcare Survey identified a broad set of AI use cases indicating that AI adoption is extending across many areas of the hospital.
While healthcare lags other industries in the deployment of AI, healthcare organizations are building the organizational infrastructure to manage the funding, staffing, data management, and patient privacy policies and procedures for AI.
Unlike other IT investment decisions that are funded increasingly by line of business, AI funding is primarily determined by an annual IT budget, and in some cases, funding decisions are made solely by IT. This may be an early-stage situation as organizations get their “feet wet,” but it is not sustainable. IDC believes that it is a best practice to use interdisciplinary teams to drive AI initiatives from inception to production. The teams must include business and/or clinical representatives; respondents to the AI in Healthcare Survey support this belief.
As with so many innovations, good, clean, well-organized data is key to success, and poor quality data often represents a challenge. The survey respondents identified volume and quality of data for training as a barrier and recognized that sufficient data volume and confidence in the data are critical success factors. The challenges of data and the lack of skilled data scientists may well be the drivers for providers’ preference to use third-party vendors to develop AI capabilities. Few organizations are developing AI in house, and a marked decline in the use of commercial off the shelf (COTS) as a primary development approach is expected from 2020 to 2023. IDC believes that it is also key to have a data management technology that supports both transactions and analytics as the acceptable latency from insight to action is shortening.
The breadth of use cases identified by respondents is an indication of how widespread the use of AI is becoming in hospitals. While one-third of organizations in the AI in Healthcare Survey identified grants as their primary funding source, the use of AI now goes beyond academic research. Today AI is being used to improve data quality using inferencing, to improve back-office productivity, and to read images to assist in diagnosis and predict adverse events. The breadth of use cases gives reason for optimism that healthcare organizations will continue their march toward maturity in the use of AI to transform their businesses.
About the Study
This white paper is based on the results of the AI in Healthcare Survey conducted by IDC and sponsored by InterSystems. The survey evaluates the relative maturity of artificial intelligence adoption for healthcare providers in Germany, the United States, and the United Kingdom, including an understanding of the operational and organizational structure of AI initiatives and top priority AI use cases. Data from additional relevant IDC surveys is also included and noted by source.
The AI in Healthcare Survey was fielded in May 2020. A total of 210 hospitals were surveyed across three countries. The sample size for each country is as follows: 105 respondents in the United States, 54 respondents in Germany, and 51 respondents in the United Kingdom. Readers should be aware of the implications of small sample sizes. Data from the AI in Healthcare Survey should be considered directional. The Methodology section provides a profile of respondents.