Risky business: a scoping review for communicating results of predictive models between providers and patients

Abstract

Objective Given widespread excitement around predictive analytics and the proliferation of machine learning algorithms that predict outcomes, a key next step is understanding how this information is—or should be—communicated with patients. Materials and Methods We conducted a scoping review informed by PRISMA-ScR guidelines to identify current knowledge and gaps in this domain. Results Ten studies met inclusion criteria for full text review. The following topics were represented in the studies, some of which involved more than 1 topic: disease prevention (N = 5/10, 50%), treatment decisions (N = 5/10, 50%), medication harms reduction (N = 1/10, 10%), and presentation of cardiovascular risk information (N = 5/10, 50%). A single study included 6- and 12-month clinical outcome metrics. Discussion As predictive models are increasingly published, marketed by industry, and implemented, this paucity of relevant research poses important gaps. Published studies identified the importance of (1) identifying the most effective source of information for patient communications; (2) contextualizing risk information and associated design elements based on users’ needs and problem areas; and (3) understanding potential impacts on risk factor modification and behavior change dependent on risk presentation. Conclusion An opportunity remains for researchers and practitioners to share strategies for effective selection of predictive algorithms for clinical practice, approaches for educating clinicians and patients in effectively using predictive data, and new approaches for framing patient-provider communication in the era of artificial intelligence.

Recentering responsible and explainable artificial intelligence research on patients: implications in perinatal psychiatry

“…To further enhance clinical utility, researchers have also explored visualizations of these explainability measures ( 54 ). However, as with RAI research in general, nearly all of the research into XAI techniques considers clinicians to be the primary end users ( 38 ). Therefore, the unique information needs and bioethical concerns of patients who may be directly interacting with AI in the course of their care have not been fully considered.…”

“…In the near future, perinatal patients may interact with AI during clinical decision-making ( 38 ), in their patient portals ( 39 ), or through AI-powered chatbots delivering psychotherapy ( 40 ). Current discussions of RAI are limited in their consideration of the patient as an active participant with AI.…”

“…There are many envisioned use cases of patient-centered AI in perinatal psychiatry. Patient engagement and shared decision-making continue to rise in importance in healthcare, making it likely that AI-generated risk predictions will be discussed or presented to patients in clinical care as part of the decision-making process ( 38 ). Additionally, new United States policies prohibiting “information blocking” have also led to a deluge of information being returned to patients through patient portals ( 39 ), and this information may soon include AI-generated risk predictions that are integrated into EHRs.…”

Human-Centered Design to Address Biases in Artificial Intelligence

“…Communication scientists play a vital role in ensuring that AI systems are developed, deployed, and communicated in ways that are aligned with the needs, values, and perspectives of the people who will be affected by them [ 38 , 39 ]. They can help to bridge the gap between the technical aspects of AI and the social and human aspects of its development and deployment to develop effective strategies for communicating about AI, its capabilities and limitations, and how it is being used in various domains, and to identify and mitigate potential ethical and societal implications of AI, such as issues related to privacy, bias, and fairness.…”

Patient Perspectives on AI for Mental Health: With Great [Computing] Power, Comes Great Responsibility

“…17,[22][23][24][25][26][27] However, patient perspectives on general and predictive AI have been understudied. [28][29][30] While some predictive AI developers may not intend patients to view the AI's output on their own, it has become more likely that patients may have access to predictive AI output, due to recent advances in patient data ownership and access. The U.S. 21st Century Cures Act, for example, prevents information blocking from patients, requiring health organizations and insurance providers to give patients access to their electronic health information without delay or expense.…”