Gilad Kuperman

PURPOSE: Ambient artificial intelligence (AI) scribes, widely implemented to reduce clinician burden and dissatisfaction, are not optimized for oncology. We evaluated this technology at-scale at a high-volume cancer center. METHODS: This prospective mixed-methods study used surveys, electronic health record (EHR), and utilization data to assess adoption, burden, and satisfaction with an ambient AI scribe that was implemented from June to September 2025. Descriptive statistics categorized clinicians as adopters (3+ weeks of use), attempted adopters (1-2 weeks), and never-users (0 weeks). Generalized estimating equations assessed the adoption's effect on the likelihood of closing notes on visit days and pajama time. Pre- and postimplementation surveys obtained clinicians' insights into satisfaction and documentation burden. RESULTS: In all, 190 (19.8%) of 958 clinicians used the tool, most often with new, video, and in-person visits. Adopters (n = 124) and attempted adopters (n = 66) were more commonly physicians from medicine and surgery. Pre- to post-rollout, the likelihood of closing notes on visit days remained unchanged among never-users (70% v 70%) and decreased among attempted adopters (73% v 70%) and adopters (56% v 54%). Pajama time remained unchanged among attempted adopters (50.6 v 50.1 minutes per week) and decreased among never-users (69.5 v 62.2) and adopters (86.9 v 81.4). Clinicians completing both pre- and postrollout surveys (n = 45) showed a decrease in high/excessive note-writing time (78% v 60%; P = .043). Postrollout surveys (n = 77) showed largely positive experiences, with 73% reporting they would recommend the tool to colleagues. CONCLUSION: This study of an ambient AI scribe at-scale showed positive satisfaction but mixed burden outcomes among oncology clinicians. Notably, despite enterprise-wide implementation and training, adoption was low and varied by clinician and visit characteristics. Future research should explore adoption barriers and tool optimization to support oncology documentation workflows.

Purpose: Artificial intelligence (AI) is increasingly incorporated into health care, yet patient perspectives on AI in oncology remain underexplored. Cancer presents an important clinical context, characterized by high stakes, uncertainty, and longitudinal clinician-patient relationships. Understanding patient perspectives is essential to inform human-centered AI implementation in oncology. Patients and Methods: In this scenario-based attitudinal assessment at a cancer center in New York City, survey questions captured patients' fixed-choice responses (very/somewhat positive or negative, unsure) to five scenarios involving AI in diagnostic/screening imaging, treatment planning, triage of patient portal messages about symptoms, survivorship risk assessment, and clinical trial matching. Group discussion then delved deeply into participants' views. Sessions were audio-recorded, transcribed, and analyzed to explore context-dependent attitudes and other factors shaping acceptability. Saturation was achieved with three sessions. Results: Attitudes of participants (N = 20: 14 female, 16 White, 18 non-Hispanic, average age 60 years) varied substantially by scenario. While AI-assisted imaging was seen as providing quality assurance, AI triage of patient-reported symptoms elicited concerns about missed urgency, delayed care, and loss of human contact. Treatment planning and survivorship risk assessment scenarios highlighted the importance of human interpretation and the patient-physician relationship. AI-assisted clinical trial matching was viewed positively although some participants raised concerns about accuracy and bias. Across scenarios, participants articulated conditions for acceptable AI use, most notably clinician oversight, transparency, and equity. Conclusion: Patients' views of AI in oncology are nuanced and context-dependent. While many see promise in AI across the care continuum, acceptance is contingent on preserving clinician involvement and attending to concerns of depersonalization, errors, and inequity. These novel insights can guide additional research and inform effective, trustworthy, human-centered AI implementation in oncology.

PURPOSE: Access to clinical notes enhances patient engagement and trust, and the 21st Century Cures Act enabled immediate electronic patient access in April 2021. Yet, technological advances may perpetuate disparities, which remain understudied. Understanding whether inequities in note access exist in oncology would highlight challenges around making this foundational health information available to all patients receiving ongoing, complex medical care. MATERIALS AND METHODS: This study at a high-volume specialty cancer center explored disparities around clinical notes posted to patients' portal accounts from September 1, 2021, to August 31, 2022, and accessed by March 19, 2024. Logistic and Poisson regression were used to identify patient characteristics associated with note access and note opening rates. RESULTS: The study included 124,554 patients and 815,104 clinical notes, 43.7% of which (356,290) were accessed. Although modest differences in access rates emerged around sex, age, and marital status, larger disparities appeared for ethnicity, race, and language: Black patients (odds ratio [OR], 0.63 [95% CI, 0.60 to 0.66]; P < .001; incidence rate ratio [IRR], 0.74 [95% CI, 0.73 to 0.76]; P < .001), Hispanic patients (OR, 0.85 [95% CI, 0.80 to 0.90]; P < .001; IRR, 0.90 [95% CI, 0.89 to 0.92]; P < .001), and non–English-preferred language speakers (OR, 0.78 [95% CI, 0.72 to 0.84]; P < .001; IRR, 0.82 [95% CI, 0.80 to 0.84]; P < .001) were, respectively, 37%, 15%, and 22% less likely to open at least one note, and opened 26%, 10%, and 18% percent fewer notes, compared with white, non-Hispanic, and English-preferred patients, respectively. CONCLUSION: This analysis highlighted disparities, by race, ethnicity, and language, in cancer patients' accessing clinical notes. Tailored interventions are crucial to ensure that diverse groups benefit from digital health care resources.

PURPOSE: Incorporating patient-reported outcomes (PROs) into health care processes can improve engagement with patients; however, adopting PROs at scale is challenging. The aim of this study was to describe the design, development, and adoption at scale of a comprehensive PRO program for standard of care and research at a cancer center. METHODS: Requirements for a PRO program were obtained from multiple stakeholders. Components of the program included a governance process to assure a consistent and satisfactory experience for patients completing PRO questionnaires, tools to create and manage questionnaires and related content, methods to send questionnaires to relevant patients at the appropriate time, interactive tools for patients to complete the questionnaires as part of their portal experience, and integration of PRO data into workflows for clinicians. We used descriptive statistics to assess the use of the program from 2016 to 2023. RESULTS: From program launch (on February 1, 2016) until December 31, 2023, 189 unique questionnaires were developed (101 for standard-of-care, 70 for research, and 18 for quality improvement). Of the 432,497 unique patients who were assigned at least one questionnaire, 314,685 (73%) completed at least one. Of 5,948,464 questionnaires sent, 3,098,574 (52%) were completed. The median completion time was 2 minutes. CONCLUSION: Large-scale adoption of PROs at a cancer center is feasible. Key considerations for success include governance processes, attention to patient experience and clinician workflow, and the ability to manage complex inclusion criteria and timing of delivery of questionnaires. These principles should be disseminated so the full potential of PROs in health care can be realized.

PURPOSE: Clinical trials are integral for patients with cancer but remain inaccessible to many because of barriers including geographic and transportation challenges. This study aimed to evaluate cancer patients' preferences for telemedicine versus in-person visits for clinical trial discussions and informed consent (IC). METHODS: An electronic survey was administered to first-time telemedicine users at Memorial Sloan Kettering Cancer Center from 2021 to 2023. The survey assessed patients' preferences for telemedicine versus in-person visits for the IC process and their comfort with discussing clinical trials virtually. The primary outcome was the proportion of patients who indicated that they preferred a telemedicine visit for the IC process. Patient comfort and preference for discussing clinical trials through telemedicine versus an in-person visit was also explored. Structured responses provided quantitative data over the 2021-2023 observation period and demographic variations. To gain a more detailed understanding, unstructured free-text responses describing clinical trial discussions were also analyzed through language modeling. RESULTS: Overall, 57% of patients (540/955) preferred telemedicine, 26% (249/955) had no preference, and 17% (166/955) preferred in-person visits for the IC process. The preference for telemedicine remained consistent across the 2021-2023 observation period. Most patients reported no difference between a telemedicine versus in-person visit for clinical trial discussion, including asking questions, sharing concerns, declining participation, and asking for more time to make a decision. Language modeling analysis revealed areas for improvement. CONCLUSION: A majority of patients at a comprehensive cancer center who participated in clinical trial discussions through telemedicine reported a preference for telemedicine to complete the IC process. Telemedicine thus represents a valuable tool for reducing barriers to clinical trial participation, particularly in reducing travel and time barriers.

Background: Remote patient monitoring (RPM) aims to improve patient access to care and communication with clinical providers. Overall, understanding the usability of RPM applications and their influence on clinical care workflows is limited from the perspectives of clinician end users at a cancer center in the Northeastern United States. Objective: This study aims to explore the usability and functionality of RPM and elicit the perceptions and experiences of oncology clinicians using RPM for oncology patients after hospital discharge. Methods: The sample included 30 of 98 clinicians (31% response rate) managing at least 5 patients in the RPM program and responding to the mHealth usability between March 2021 and October 2021. Overall, clinicians responded positively to the survey. Item responses with the highest proportion of disagreement were explored further. A nested sample of 5 clinicians who responded to the study survey (30% response rate) participated in interview sessions conducted from November 2021 to February 2022, averaging 60 minutes each. Results: Survey responses highlighted that RPM was easy to use and learn and verified symptom alerts during follow-up phone calls. Areas to improve identified practice changes from reporting RPM alerts through digital portals and its influence on clinicians’ workload burden. Interview sessions revealed 3 main themes: clinician understanding and usability constraints, patient constraints, and suggestions for improving the program. Subthemes for each theme were explored, characterizing technical and functional limitations that could be addressed to enhance efficiency, workflow, and user experience. Conclusions: Clinicians support the value of RPM for improving symptom management and engaging with providers. Improvements to address RPM challenges include functional changes to enhance the program’s utility, such as input from patients about temporal changes in their symptoms and technical resources for home monitoring devices.

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). In this study, we evaluated the performance of a retrieval-augmented language model, powered by GPT-4, to recommend appropriate clinical trial recommendations for a head & neck cancer population at the Memorial Sloan Kettering Cancer Center. We demonstrated that retrieval-augmented LLM could achieve moderate performance, exceeding the historical performance of untrained LLMs to provide oncology treatment recommendations by 4-20 folds. Our study provided insights into the rarely measured performance of retrieval-augmented LLM using real-world patient cases in comparison to physician expert recommendations.

Modern cancer care is costly and logistically burdensome for patients and their families despite an expansion of technology and medical advances that create the opportunity for novel approaches to care. Therefore, there is a growing appreciation for the need to leverage these innovations to make cancer care more patient centered and convenient. The Memorial Sloan Kettering Making Telehealth Delivery of Cancer Care at Home Efficient and Safe Telehealth Research Center is a National Cancer Institute–designated and funded Telehealth Research Center of Excellence poised to generate the evidence necessary to inform the appropriate use of telehealth as a strategy to improve access to cancer services that are convenient for patients. The center will evaluate telehealth as a strategy to personalize cancer care delivery to ensure that it is not only safe and effective but also convenient and efficient. In this article, we outline this new center’s research strategy, as well as highlight challenges that exist in further integrating telehealth into standard oncology practice based on early experiences.

Background: Although family caregivers play a critical role in care delivery, research has shown that they face significant physical, emotional, and informational challenges. One promising avenue to address some of caregivers’ unmet needs is via the design of digital technologies that support caregivers’ complex portfolio of responsibilities. Augmented reality (AR) applications, specifically, offer new affordances to aid caregivers as they perform care tasks in the home. Objective: This study explored how AR might assist family caregivers with the delivery of home-based cancer care. The specific objectives were to shed light on challenges caregivers face where AR might help, investigate opportunities for AR to support caregivers, and understand the risks of AR exacerbating caregiver burdens. Methods: We conducted a qualitative video elicitation study with clinicians and caregivers. We created 3 video elicitations that offer ways in which AR might support caregivers as they perform often high-stakes, unfamiliar, and anxiety-inducing tasks in postsurgical cancer care: wound care, drain care, and rehabilitative exercise. The elicitations show functional AR applications built using Unity Technologies software and Microsoft Hololens2. Using elicitations enabled us to avoid rediscovering known usability issues with current AR technologies, allowing us to focus on high-level, substantive feedback on potential future roles for AR in caregiving. Moreover, it enabled nonintrusive exploration of the inherently sensitive in-home cancer care context. Results: We recruited 22 participants for our study: 15 clinicians (eg, oncologists and nurses) and 7 family caregivers. Our findings shed light on clinicians’ and caregivers’ perceptions of current information and communication challenges caregivers face as they perform important physical care tasks as part of cancer treatment plans. Most significant was the need to provide better and ongoing support for execution of caregiving tasks in situ, when and where the tasks need to be performed. Such support needs to be tailored to the specific needs of the patient, to the stress-impaired capacities of the caregiver, and to the time-constrained communication availability of clinicians. We uncover opportunities for AR technologies to potentially increase caregiver confidence and reduce anxiety by supporting the capture and review of images and videos and by improving communication with clinicians. However, our findings also suggest ways in which, if not deployed carefully, AR technologies might exacerbate caregivers’ already significant burdens. Conclusions: These findings can inform both the design of future AR devices, software, and applications and the design of caregiver support interventions based on already available technology and processes. Our study suggests that AR technologies and the affordances they provide (eg, tailored support, enhanced monitoring and task accuracy, and improved communications) should be considered as a part of an integrated care journey involving multiple stakeholders, changing information needs, and different communication channels that blend in-person and internet-based synchronous and asynchronous care, illness, and recovery.