Artificial Intelligence: What Does It Mean for Nursing?

Abstract

In one of Dr. Diane Skiba’s last columns, she described the “invisible health care provider” as a way to illustrate health care made possible by transforming big data collected from personal and medical devices into actionable information using artificial intelligence (AI; Skiba, 2018). This column presents basic information about AI and identifies opportunities with AI and associated risks. AI is the science of using computers to mimic human abilities (Nevala, 2017): speech, sensory abilities, movement, and so much more. Machine learning is a subset of AI; the computer uses a large training set of data to look for relationships and later applies the relationships to a new set of data to make predicted outcomes (Nevala, 2017). AI and machine learning were first discussed in research literature in the 1970s; however, because of limitations in computing power, few tangible advances in technology occurred until recently (Kulikowski, 2019). Now, AI is embedded in the human experience without our realizing it. For example, most email and texting applications suggest word choices, mimicking the speech patterns of their human users. Online shopping sites now suggest buying items based on past purchases. These advanced technologies work in the background and have the potential to change human society in the future (Anderson, Rainie, & Luchsinger, 2018). But what does this mean for health care and nursing? In health care, the promise of AI includes the discovery of disease etiologies; the ability to select a treatment option for patients based on predicted outcomes; and calculation of disease risk based on genetic, environmental, and social factors (Galimova, Buzaev, Ramilevich, Yuldybaev, & Shaykhulova, 2019). The reality is more complicated. In order to use the vast amount of data created from personal and medical devices, the complexity of aligning of data must be addressed because data are different in terms of what the Vendome Group (2015) calls the 6Vs: volume (amount of data), variety (different structures of data), veracity (the trustworthiness of the data), velocity (frequency of data coming into the system), visualization (presentation of outcomes in an actionable manner), and value (usefulness in improving health care outcomes). In addition, when new data are quite different from the training set, the algorithm produced from machine learning could become flawed over time unless new training sets are used to update algorithms (Coiera, 2019). Because of the challenges associated with AI, nurses need to understand the data produced from nursing care and become involved in the governance of big data (Meehan, 2017). We can start with questions about the purpose of applying AI to care processes, the outcomes to be predicted, the potential bias in data or algorithms, and the appropriateness of applying AI to forecast real-life situations. It takes nursing knowledge to determine if the data and the predicted outcomes make sense based on clinical experience before accepting erroneous algorithms. There are inherent risks with every new technology. AI will produce the next wave of innovation in health care, but nurses need the educational preparation to be ready to participate as AI is applied to nursing and nursing care (Glasgow, Colbert, Viator, & Cavanagh, 2018).