In November 2020, the Government Accountability Office (GAO) and the National Academy of Medicine (NAM) jointly released a report on the current use of artificial intelligence (AI) in healthcare, the implications for wider adoption, and the challenges and opportunities for enhanced patient care provided by this technology.
Titled Artificial Intelligence in Health Care: Benefits and Challenges of Technologies to Augment Patient Care, the report was intended to “explore AI in augmenting patient care both inside and outside traditional clinical settings, assess its implications, and identify key policy options available for optimizing its use.” In the first part of the report, the GAO authors addressed 3 key topics: the potential benefits of current and near-future AI tools that can augment patient care, challenges to the development and adoption of these tools, and policies that may maximize benefits and reduce or eliminate challenges to the use of AI tools in patient care.
The report focused on 5 areas in which AI tools have shown the most promise in benefitting providers and patients: monitoring patients, recommending treatments, guiding surgical care, predicting health trajectories, and supporting population health management.
As an example of using AI tools for monitoring patients, the report discusses applications being developed to support patient fall prevention by analyzing movements in a patient’s room and alerting the healthcare team when the risk of fall increases. The use of AI for treatment recommendations could include mechanical ventilator clinical decision support tools that could better predict when to successfully wean patients from ventilators. Tools for guiding surgical care could include AI applications that provide clinical decision support for surgical planning and postoperative care that can predict the risk of complications such as infections or in-hospital mortality. AI tools that help predict a patient’s prognosis and health trajectory could offer support in areas such as sepsis prediction by anticipating sepsis and septic shock before they occur. Population health management projects and initiatives could be augmented with AI support to identify broad health risks and treatment opportunities for patient groups or communities.
The report also identified administrative categories in which AI tools have shown the potential “to reduce provider burden and increase the efficiency of patient care,” including automating time-consuming tasks, optimizing operational processes, and recording digital clinical notes. Repetitive and routine tasks, such as entering patient data, could be taken over by AI applications. An example of an administrative AI tool that is already in fairly widespread use is speech recognition software used for digitally recording clinical notes. Applications are being developed that could go beyond word-for-word transcription and allow users to add structured data to electronic health records. Scheduling and other operational processes could be improved through the use of AI by optimizing efficient resource use by practices and providers, which could decrease delays and other errors and improve the patient experience.
The report outlined several challenges that could affect the use of AI for patient care and delay wider adoption, including difficulty accessing high-quality data used in the development of AI tools, data bias that can reduce effectiveness and accuracy, difficulties in scaling technologies and integrating them into new settings, limited transparency that can make it difficult to understand an AI tool’s decision-making process, dispersion of potentially sensitive patient data into the hands of more people and organizations, and uncertainty about potential liability issues that could arise with the use of AI tools in patient care.
The GAO also developed 6 policy options that policymakers and stakeholders could adopt to address the challenges previously outlined and enhance the use of AI in patient care. The report authors noted that the first 5 options “would aim to do so by encouraging collaboration, improving data sharing, encouraging the development of best practices, fostering interdisciplinary education, and clarifying oversight mechanisms.” The sixth option is to maintain the status quo and allow current efforts to develop and implement AI applications to proceed without intervention. For each policy option, the report offers detailed analysis of the potential benefits and outcomes that could ensue if each were adopted, as well as potential negative considerations to consider.
Part 2 of the report “presents the NAM publication Advancing Artificial Intelligence in Health Settings Outside the Hospital and Clinic discussing the use and challenges associated with AI technology in the delivery of health care services outside of settings where health care providers are employed, including the home.” The analysis and recommendations outlined in this section will be covered in an upcoming article on this site.
Source
US Government Accountability Office. Artificial intelligence in health care: benefits and challenges of technologies to augment patient care. Published November 2020. www.gao.gov/assets/720/711471.pdf. Accessed March 24, 2021.