Background: Artificial intelligence (AI) is rapidly expanding in medicine despite a lack of consensus on its application and evaluation. Objective: We sought to identify current frameworks guiding the application and evaluation of AI for predictive analytics in medicine and to describe the content of these frameworks. We also assessed what stages along the AI translational spectrum (ie, AI development, reporting, evaluation, implementation, and surveillance) the content of each framework has been discussed. Methods: We performed a literature review of frameworks regarding the oversight of AI in medicine. The search included key topics such as “artificial intelligence,” “machine learning,” “guidance as topic,” and “translational science,” and spanned the time period 2014-2022. Documents were included if they provided generalizable guidance regarding the use or evaluation of AI in medicine. Included frameworks are summarized descriptively and were subjected to content analysis. A novel evaluation matrix was developed and applied to appraise the frameworks’ coverage of content areas across translational stages. Results: Fourteen frameworks are featured in the review, including six frameworks that provide descriptive guidance and eight that provide reporting checklists for medical applications of AI. Content analysis revealed five considerations related to the oversight of AI in medicine across frameworks: transparency, reproducibility, ethics, effectiveness, and engagement. All frameworks include discussions regarding transparency, reproducibility, ethics, and effectiveness, while only half of the frameworks discuss engagement. The evaluation matrix revealed that frameworks were most likely to report AI considerations for the translational stage of development and were least likely to report considerations for the translational stage of surveillance. Conclusions: Existing frameworks for the application and evaluation of AI in medicine notably offer less input on the role of engagement in oversight and regarding the translational stage of surveillance. Identifying and optimizing strategies for engagement are essential to ensure that AI can meaningfully benefit patients and other end users.

Discrete-choice experiments (DCEs) are a frequently used method to explore the preferences of patients and other decision-makers in health. Pretesting is an essential stage in the design of a high-quality choice experiment and involves engaging with representatives of the target population to improve the readability, presentation, and structure of the preference instrument. The goal of pretesting in DCEs is to improve the validity, reliability, and relevance of the survey, while decreasing sources of bias, burden, and error associated with preference elicitation, data collection, and interpretation of the data. Despite its value to inform DCE design, pretesting lacks documented good practices or clearly reported applied examples. The purpose of this paper is: (1) to define pretesting and describe the pretesting process specifically in the context of a DCE, (2) to present a practical guide and pretesting interview discussion template for researchers looking to conduct a rigorous pretest of a DCE, and (3) to provide an illustrative example of how these resources were operationalized to inform the design of a complex DCE aimed at eliciting tradeoffs between personal privacy and societal benefit in the context of a police method known as investigative genetic genealogy (IGG).

Objective: Acute myeloid leukemia (AML) is experiencing a therapeutic renaissance due to the heightened biomedical understanding of AML and patient-focused drug development (PFDD). Many AML patients now live long-term with the side effects of treatment. This study documents the prevalence and severity of AML treatment-related side effects. Methods: A national cross-sectional survey designed with the Leukemia & Lymphoma Society assessed patients’ experiences with short-term (nausea/vomiting, diarrhea, hair loss, mouth sores, infection, rash) and long-term (organ dysfunction, chemobrain, fatigue, neuropathy) treatment side effects. Patient and caregiver participants rated side effect severity (none-severe). Results: Survey participants (N = 1182) were mostly female (65%), AML patients (76%), and had undergone chemotherapy (94%). Eighty-seven percent of participants reported severe short-term effects, and 33% reported severe long-term effects of treatment. Only 11% of respondents did not have any severe effects. Hair loss and fatigue were the most common severe short- and long-term side effects (78%, 33%). There was a moderate correlation between having short- and long-term adverse effects (r = 0.41, P < 0.001). Caregivers were more likely than patients to report severe organ dysfunction, fatigue, & neuropathy (P-values <0.05). Conclusions: Survivors experience a high burden of side effects from AML treatments highlighting the need for the development of less toxic therapies. Differences in patients’ and caregivers’ experiences illustrate the importance of sampling from diverse sources to understand the full burden of AML treatment, and the need for less toxic drugs. This study informs patients, patient-advocacy groups, clinicians, and regulators about AML treatment burdens and provides the community with information to inform PFDD.

Background Many human immunodeficiency virus (HIV) prevention technologies (pre-exposure prophylaxis, microbicides, vaccines) are available or in development. Preference elicitation methods provide insight into client preferences that may be used to optimize products and services. Given increased utilization of such methods in HIV prevention, this article identifies and reviews these methods and synthesizes their application to HIV prevention technologies. Methods In May 2020, we systematically searched peer-reviewed literature in PubMed, CINAHL, and Web of Science for studies employing quantitative preference elicitation methods to measure preferences for HIV prevention technologies among populations of any age, sex, or location. Quality assessment used an existing checklist (PREFS) and a novel adaptation of the Newcastle–Ottawa Scale (PROSPERO #CRD42018087027). Results We screened 5022 titles and abstracts, reviewed 318 full texts, and included 84 studies. Common methods employed were discrete-choice experiment (33%), conjoint analysis (25%), and willingness-to-participate/try/accept (21%). Studies were conducted in 25 countries and had a mean of 768 participants (range = 26–7176), two-thirds of them male. Common HIV prevention technologies included pre-exposure prophylaxis (23%), voluntary testing and counseling (19%), HIV self-testing (17%), vaccines (15%), and topical microbicides (9%). Most attributes focused on product design (side effects, frequency), service design (provider type, location), acceptability or willingness to accept/pay; results are summarized in these categories, by prevention type. Mean quality-adapted Newcastle–Ottawa Scale score was 4.5/8 (standard deviation = 2.1) and mean PREFS scores was 3.47/5 (standard deviation = 0.81). Conclusions This review synthesizes extant literature on quantitative measurement of preferences for HIV prevention technologies. This can enable practitioners to improve prevention products and interventions, and ultimately reduce HIV incidence.