Faculty Bibliography

IMPORTANCE/UNASSIGNED:Increasing numbers of unaffected individuals could benefit from genetic evaluation for inherited cancer susceptibility. Automated conversational agents (ie, chatbots) are being developed for cancer genetics contexts; however, randomized comparisons with standard of care (SOC) are needed. OBJECTIVE/UNASSIGNED:To examine whether chatbot and SOC approaches are equivalent in completion of pretest cancer genetic services and genetic testing. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This equivalence trial (Broadening the Reach, Impact, and Delivery of Genetic Services [BRIDGE] randomized clinical trial) was conducted between August 15, 2020, and August 31, 2023, at 2 US health care systems (University of Utah Health and NYU Langone Health). Participants were aged 25 to 60 years, had had a primary care visit in the previous 3 years, were eligible for cancer genetic evaluation, were English or Spanish speaking, had no prior cancer diagnosis other than nonmelanoma skin cancer, had no prior cancer genetic counseling or testing, and had an electronic patient portal account. INTERVENTION/UNASSIGNED:Participants were randomized 1:1 at the patient level to the study groups at each site. In the chatbot intervention group, patients were invited in a patient portal outreach message to complete a pretest genetics education chat. In the enhanced SOC control group, patients were invited to complete an SOC pretest appointment with a certified genetic counselor. MAIN OUTCOMES AND MEASURES/UNASSIGNED:Primary outcomes were completion of pretest cancer genetic services (ie, pretest genetics education chat or pretest genetic counseling appointment) and completion of genetic testing. Equivalence hypothesis testing was used to compare the study groups. RESULTS/UNASSIGNED:This study included 3073 patients (1554 in the chatbot group and 1519 in the enhanced SOC control group). Their mean (SD) age at outreach was 43.8 (9.9) years, and most (2233 of 3063 [72.9%]) were women. A total of 204 patients (7.3%) were Black, 317 (11.4%) were Latinx, and 2094 (75.0%) were White. The estimated percentage point difference for completion of pretest cancer genetic services between groups was 2.0 (95% CI, -1.1 to 5.0). The estimated percentage point difference for completion of genetic testing was -1.3 (95% CI, -3.7 to 1.1). Analyses suggested equivalence in the primary outcomes. CONCLUSIONS AND RELEVANCE/UNASSIGNED:The findings of the BRIDGE equivalence trial support the use of chatbot approaches to offer cancer genetic services. Chatbot tools can be a key component of sustainable and scalable population health management strategies to enhance access to cancer genetic services. TRIAL REGISTRATION/UNASSIGNED:ClinicalTrials.gov Identifier: NCT03985852.

IMPORTANCE/UNASSIGNED:Virtual patient-physician communications have increased since 2020 and negatively impacted primary care physician (PCP) well-being. Generative artificial intelligence (GenAI) drafts of patient messages could potentially reduce health care professional (HCP) workload and improve communication quality, but only if the drafts are considered useful. OBJECTIVES/UNASSIGNED:To assess PCPs' perceptions of GenAI drafts and to examine linguistic characteristics associated with equity and perceived empathy. DESIGN, SETTING, AND PARTICIPANTS/UNASSIGNED:This cross-sectional quality improvement study tested the hypothesis that PCPs' ratings of GenAI drafts (created using the electronic health record [EHR] standard prompts) would be equivalent to HCP-generated responses on 3 dimensions. The study was conducted at NYU Langone Health using private patient-HCP communications at 3 internal medicine practices piloting GenAI. EXPOSURES/UNASSIGNED:Randomly assigned patient messages coupled with either an HCP message or the draft GenAI response. MAIN OUTCOMES AND MEASURES/UNASSIGNED:PCPs rated responses' information content quality (eg, relevance), using a Likert scale, communication quality (eg, verbosity), using a Likert scale, and whether they would use the draft or start anew (usable vs unusable). Branching logic further probed for empathy, personalization, and professionalism of responses. Computational linguistics methods assessed content differences in HCP vs GenAI responses, focusing on equity and empathy. RESULTS/UNASSIGNED:A total of 16 PCPs (8 [50.0%] female) reviewed 344 messages (175 GenAI drafted; 169 HCP drafted). Both GenAI and HCP responses were rated favorably. GenAI responses were rated higher for communication style than HCP responses (mean [SD], 3.70 [1.15] vs 3.38 [1.20]; P = .01, U = 12 568.5) but were similar to HCPs on information content (mean [SD], 3.53 [1.26] vs 3.41 [1.27]; P = .37; U = 13 981.0) and usable draft proportion (mean [SD], 0.69 [0.48] vs 0.65 [0.47], P = .49, t = -0.6842). Usable GenAI responses were considered more empathetic than usable HCP responses (32 of 86 [37.2%] vs 13 of 79 [16.5%]; difference, 125.5%), possibly attributable to more subjective (mean [SD], 0.54 [0.16] vs 0.31 [0.23]; P < .001; difference, 74.2%) and positive (mean [SD] polarity, 0.21 [0.14] vs 0.13 [0.25]; P = .02; difference, 61.5%) language; they were also numerically longer (mean [SD] word count, 90.5 [32.0] vs 65.4 [62.6]; difference, 38.4%), but the difference was not statistically significant (P = .07) and more linguistically complex (mean [SD] score, 125.2 [47.8] vs 95.4 [58.8]; P = .002; difference, 31.2%). CONCLUSIONS/UNASSIGNED:In this cross-sectional study of PCP perceptions of an EHR-integrated GenAI chatbot, GenAI was found to communicate information better and with more empathy than HCPs, highlighting its potential to enhance patient-HCP communication. However, GenAI drafts were less readable than HCPs', a significant concern for patients with low health or English literacy.

This study introduces a Generative Artificial Intelligence (GenAI) assistant designed to address key challenges in Remote Patient Monitoring (RPM) for hypertension. After a comprehensive needs assessment from clinicians and patients, we identified pivotal issues in RPM data management and patient engagement. The GenAI RPM assistant integrates a patient-facing chatbot, clinician-facing smart summaries, and automated draft portal messages to enhance communication and streamline data review. Validated through six rounds of testing and evaluations by ten participants, the initial prototype was positively received, highlighting the importance of personalized interactions. Our findings demonstrate GenAI's potential to improve RPM by optimizing data management and enhancing patient-provider communication.

BACKGROUND:During the COVID-19 pandemic, acute respiratory infection (ARI) antibiotic prescribing in ambulatory care markedly decreased. It is unclear if antibiotic prescription rates will remain lowered. METHODS:We used trend analyses of antibiotics prescribed during and after the first wave of COVID-19 to determine whether ARI antibiotic prescribing rates in ambulatory care have remained suppressed compared to pre-COVID-19 levels. Retrospective data was used from patients with ARI or UTI diagnosis code(s) for their encounter from 298 primary care and 66 urgent care practices within four academic health systems in New York, Wisconsin, and Utah between January 2017 and June 2022. The primary measures included antibiotic prescriptions per 100 non-COVID ARI encounters, encounter volume, prescribing trends, and change from expected trend. RESULTS:At baseline, during and after the first wave, the overall ARI antibiotic prescribing rates were 54.7, 38.5, and 54.7 prescriptions per 100 encounters, respectively. ARI antibiotic prescription rates saw a statistically significant decline after COVID-19 onset (step change -15.2, 95% CI: -19.6 to -4.8). During the first wave, encounter volume decreased 29.4% and, after the first wave, remained decreased by 188%. After the first wave, ARI antibiotic prescription rates were no longer significantly suppressed from baseline (step change 0.01, 95% CI: -6.3 to 6.2). There was no significant difference between UTI antibiotic prescription rates at baseline versus the end of the observation period. CONCLUSIONS:The decline in ARI antibiotic prescribing observed after the onset of COVID-19 was temporary, not mirrored in UTI antibiotic prescribing, and does not represent a long-term change in clinician prescribing behaviors. During a period of heightened awareness of a viral cause of ARI, a substantial and clinically meaningful decrease in clinician antibiotic prescribing was observed. Future efforts in antibiotic stewardship may benefit from continued study of factors leading to this reduction and rebound in prescribing rates.

INTRODUCTION/UNASSIGNED:Shared decision-making (SDM) is a method of care by which patients and clinicians work together to co-create a plan of care. Electronic health record (EHR) integration of SDM tools may increase adoption of SDM. We conducted a "lightweight" integration of a freely available electronic SDM tool, CV Prevention Choice, within the EHRs of three healthcare systems. Here, we report how the healthcare systems collaborated to achieve integration. METHODS/UNASSIGNED:This work was conducted as part of a stepped wedge randomized pragmatic trial. CV Prevention Choice was developed using guidelines for HTML5-based web applications. Healthcare systems integrated the tool in their EHR using documentation the study team developed and refined with lessons learned after each system integrated the electronic SDM tool into their EHR. CV Prevention Choice integration populates the tool with individual patient data locally without sending protected health information between the EHR and the web. Data abstraction and secure transfer systems were developed to manage data collection to assess tool implementation and effectiveness outcomes. RESULTS/UNASSIGNED:Time to integrate CV Prevention Choice in the EHR was 12.1 weeks for the first system, 10.4 weeks for the second, and 9.7 weeks for the third. One system required two 1-hour meetings with study team members and two healthcare systems required a single 1-hour meeting. Healthcare system information technology teams collaborated by sharing information and offering improvements to documentation. Challenges included tracking CV Prevention Choice use for reporting and capture of combination medications. Data abstraction required refinements to address differences in how each healthcare system captured data elements. CONCLUSION/UNASSIGNED:Targeted documentation on tool features and resource mapping supported collaboration of IT teams across healthcare systems, enabling them to integrate a web-based SDM tool with little additional research team effort or oversight. Their collaboration helped overcome difficulties integrating the web application and address challenges to data harmonization for trial outcome analyses.

BACKGROUND:Digital diabetes prevention programs (dDPPs) are effective "digital prescriptions" but have high attrition rates and program noncompletion. To address this, we developed a personalized automatic messaging system (PAMS) that leverages SMS text messaging and data integration into clinical workflows to increase dDPP engagement via enhanced patient-provider communication. Preliminary data showed positive results. However, further investigation is needed to determine how to optimize the tailoring of support technology such as PAMS based on a user's preferences to boost their dDPP engagement. OBJECTIVE:This study evaluates leveraging machine learning (ML) to develop digital engagement phenotypes of dDPP users and assess ML's accuracy in predicting engagement with dDPP activities. This research will be used in a PAMS optimization process to improve PAMS personalization by incorporating engagement prediction and digital phenotyping. This study aims (1) to prove the feasibility of using dDPP user-collected data to build an ML model that predicts engagement and contributes to identifying digital engagement phenotypes, (2) to describe methods for developing ML models with dDPP data sets and present preliminary results, and (3) to present preliminary data on user profiling based on ML model outputs. METHODS:Using the gradient-boosted forest model, we predicted engagement in 4 dDPP individual activities (physical activity, lessons, social activity, and weigh-ins) and general activity (engagement in any activity) based on previous short- and long-term activity in the app. The area under the receiver operating characteristic curve, the area under the precision-recall curve, and the Brier score metrics determined the performance of the model. Shapley values reflected the feature importance of the models and determined what variables informed user profiling through latent profile analysis. RESULTS:We developed 2 models using weekly and daily DPP data sets (328,821 and 704,242 records, respectively), which yielded predictive accuracies above 90%. Although both models were highly accurate, the daily model better fitted our research plan because it predicted daily changes in individual activities, which was crucial for creating the "digital phenotypes." To better understand the variables contributing to the model predictor, we calculated the Shapley values for both models to identify the features with the highest contribution to model fit; engagement with any activity in the dDPP in the last 7 days had the most predictive power. We profiled users with latent profile analysis after 2 weeks of engagement (Bayesian information criterion=-3222.46) with the dDPP and identified 6 profiles of users, including those with high engagement, minimal engagement, and attrition. CONCLUSIONS:Preliminary results demonstrate that applying ML methods with predicting power is an acceptable mechanism to tailor and optimize messaging interventions to support patient engagement and adherence to digital prescriptions. The results enable future optimization of our existing messaging platform and expansion of this methodology to other clinical domains. TRIAL REGISTRATION/BACKGROUND:ClinicalTrials.gov NCT04773834; https://www.clinicaltrials.gov/ct2/show/NCT04773834. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)/UNASSIGNED:RR2-10.2196/26750.

Digital health implementations and investments continue to expand. As the reliance on digital health increases, it is imperative to implement technologies with inclusive and accessible approaches. A conceptual model can be used to guide equity-focused digital health implementations to improve suitability and uptake in diverse populations. The objective of this study is expand an implementation model with recommendations on the equitable implementation of new digital health technologies. The Digital Health Equity-Focused Implementation Research (DH-EquIR) conceptual model was developed based on a rigorous review of digital health implementation and health equity literature. The Equity-Focused Implementation Research for Health Programs (EquIR) model was used as a starting point and merged with digital equity and digital health implementation models. Existing theoretical frameworks and models were appraised as well as individual equity-sensitive implementation studies. Patient and program-related concepts related to digital equity, digital health implementation, and assessment of social/digital determinants of health were included. Sixty-two articles were analyzed to inform the adaption of the EquIR model for digital health. These articles included digital health equity models and frameworks, digital health implementation models and frameworks, research articles, guidelines, and concept analyses. Concepts were organized into EquIR conceptual groupings, including population health status, planning the program, designing the program, implementing the program, and equity-focused implementation outcomes. The adapted DH-EquIR conceptual model diagram was created as well as detailed tables displaying related equity concepts, evidence gaps in source articles, and analysis of existing equity-related models and tools. The DH-EquIR model serves to guide digital health developers and implementation specialists to promote the inclusion of health-equity planning in every phase of implementation. In addition, it can assist researchers and product developers to avoid repeating the mistakes that have led to inequities in the implementation of digital health across populations.

BACKGROUND:Up to 50% of antibiotic prescriptions for upper respiratory infections (URIs) are inappropriate. Clinical decision support (CDS) systems to mitigate unnecessary antibiotic prescriptions have been implemented into electronic health records, but their use by providers has been limited. OBJECTIVE:As a delegation protocol, we adapted a validated electronic health record-integrated clinical prediction rule (iCPR) CDS-based intervention for registered nurses (RNs), consisting of triage to identify patients with low-acuity URI followed by CDS-guided RN visits. It was implemented in February 2022 as a randomized controlled stepped-wedge trial in 43 primary and urgent care practices within 4 academic health systems in New York, Wisconsin, and Utah. While issues were pragmatically addressed as they arose, a systematic assessment of the barriers to implementation is needed to better understand and address these barriers. METHODS:We performed a retrospective case study, collecting quantitative and qualitative data regarding clinical workflows and triage-template use from expert interviews, study surveys, routine check-ins with practice personnel, and chart reviews over the first year of implementation of the iCPR intervention. Guided by the updated CFIR (Consolidated Framework for Implementation Research), we characterized the initial barriers to implementing a URI iCPR intervention for RNs in ambulatory care. CFIR constructs were coded as missing, neutral, weak, or strong implementation factors. RESULTS:Barriers were identified within all implementation domains. The strongest barriers were found in the outer setting, with those factors trickling down to impact the inner setting. Local conditions driven by COVID-19 served as one of the strongest barriers, impacting attitudes among practice staff and ultimately contributing to a work infrastructure characterized by staff changes, RN shortages and turnover, and competing responsibilities. Policies and laws regarding scope of practice of RNs varied by state and institutional application of those laws, with some allowing more clinical autonomy for RNs. This necessitated different study procedures at each study site to meet practice requirements, increasing innovation complexity. Similarly, institutional policies led to varying levels of compatibility with existing triage, rooming, and documentation workflows. These workflow conflicts were compounded by limited available resources, as well as an implementation climate of optional participation, few participation incentives, and thus low relative priority compared to other clinical duties. CONCLUSIONS:Both between and within health care systems, significant variability existed in workflows for patient intake and triage. Even in a relatively straightforward clinical workflow, workflow and cultural differences appreciably impacted intervention adoption. Takeaways from this study can be applied to other RN delegation protocol implementations of new and innovative CDS tools within existing workflows to support integration and improve uptake. When implementing a system-wide clinical care intervention, considerations must be made for variability in culture and workflows at the state, health system, practice, and individual levels. TRIAL REGISTRATION/BACKGROUND:ClinicalTrials.gov NCT04255303; https://clinicaltrials.gov/ct2/show/NCT04255303.

OBJECTIVES/OBJECTIVE:To evaluate demographic biases in diagnostic accuracy and health advice between generative artificial intelligence (AI) (ChatGPT GPT-4) and traditional symptom checkers like WebMD. MATERIALS AND METHODS/METHODS:Combination symptom and demographic vignettes were developed for 27 most common symptom complaints. Standardized prompts, written from a patient perspective, with varying demographic permutations of age, sex, and race/ethnicity were entered into ChatGPT (GPT-4) between July and August 2023. In total, 3 runs of 540 ChatGPT prompts were compared to the corresponding WebMD Symptom Checker output using a mixed-methods approach. In addition to diagnostic correctness, the associated text generated by ChatGPT was analyzed for readability (using Flesch-Kincaid Grade Level) and qualitative aspects like disclaimers and demographic tailoring. RESULTS:ChatGPT matched WebMD in 91% of diagnoses, with a 24% top diagnosis match rate. Diagnostic accuracy was not significantly different across demographic groups, including age, race/ethnicity, and sex. ChatGPT's urgent care recommendations and demographic tailoring were presented significantly more to 75-year-olds versus 25-year-olds (P < .01) but were not statistically different among race/ethnicity and sex groups. The GPT text was suitable for college students, with no significant demographic variability. DISCUSSION/CONCLUSIONS:The use of non-health-tailored generative AI, like ChatGPT, for simple symptom-checking functions provides comparable diagnostic accuracy to commercially available symptom checkers and does not demonstrate significant demographic bias in this setting. The text accompanying differential diagnoses, however, suggests demographic tailoring that could potentially introduce bias. CONCLUSION/CONCLUSIONS:These results highlight the need for continued rigorous evaluation of AI-driven medical platforms, focusing on demographic biases to ensure equitable care.