ADAndrea Downing

Andrea Downing

*This is an algorithmically generated author record based on publicly available information.

Andrea Downing has a total of 69 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Advancing Patient‐Centred AI With Adaptive Machine Learning

    by Dimitrios Kalogeropoulos, Paul Barach, Andrea Downing, Stefaan Verhulst, Maryam Lustberg

    Healthcare services and data ecosystems remain fragmented, inequitable and misaligned with patient, clinician and public health needs. Three pathways to patient‐centred Artificial Intelligence (AI) have emerged to address this: data collaboration, Whole System Approaches and digital inclusion. Each has advanced the field. However, none delivers the continuous, participatory and outcome‐grounded learning that equitable AI demands. The populations whose data drives AI learning must also be the populations who benefit from it. This principle has not been structurally embedded in any existing pathway. This narrative review defines a framework for operationalising Adaptive Machine Learning (AML) as a fourth, integrative pathway to patient‐centred AI. AML is defined as continuous learning from real‐world individual‐level data to generate personalised, population‐level insights returning direct benefit to the communities who produce them. A narrative review of PubMed and Google Scholar (2018–2025) was conducted across five thematic domains: AI and machine learning in healthcare; real‐world data and evidence; patient‐centred care and digital health; data governance and AI regulation; and adaptive clinical trial design. Search terms covered healthcare AI applications and adaptive, continuously learning systems, emphasising prediction‐oriented rather than static diagnostic approaches. Eligibility was determined through iterative thematic analysis. Eight systemic barriers to patient‐centred AI are identified and the three existing pathways critically examined for limitations in addressing bias, validation and equitable implementation. AML is proposed as the mechanism that closes this gap, connecting systems medicine, personalised medicine and precision medicine through continuous, outcome‐based learning, identifying three original conceptual contributions as enablers: Integrative Data Governance (IDG), Adaptive Clinical Trial (ACT) designs and Evidence Sandbox Facilities (ESFs). Progress requires globally standardised Healthcare Data Recycling frameworks, adaptive evaluation methodologies and ESFs enabling multi‐stakeholder validation. Future research should address federated AML validation, collective consent mechanisms and alignment of regulatory frameworks with international AI governance standards.

  • Open Access

    Safety Risk Prioritization for Patient-Facing Agentic Voice AI in Clinical Care: A Two-Round Modified Delphi Study (Preprint)

    by Arooshi Kumar, Anil Saldanha, Erin Palm, Melanie Fernando, Chethan Sarabu, Vishnu Ravi, Maritza Suarez, Peter McCaffrey, Chiraag Kulkarni, Bryon Frost, et al.

    BACKGROUND Patient-facing agentic voice artificial intelligence (AI) is moving from pilot to production across US health systems, with autonomous agents managing chronic disease check-ins, post-acute follow-up, and quality outreach. Unlike diagnostic models or ambient documentation tools, these agents initiate patient contact, interpret verbal responses in real time, and act with limited synchronous clinician oversight. Existing AI governance frameworks were not designed for autonomous systems of this kind. We previously developed a qualitative risk framework as a multi-stakeholder voice AI safety taskforce. OBJECTIVE This study aimed to establish multidisciplinary expert consensus on the relative priority of safety risks for patient-facing agentic voice AI, stratified by clinical use case, to support risk-proportionate oversight by health systems, payers, and vendors. METHODS A 14-member expert panel spanning 9 constituencies, including clinical practice, health system leadership, clinical informatics, industry, payer operations, patient safety research, and patient advocacy, rated that framework through a two-round modified Delphi conducted across 2025 and 2026. The instrument comprised 21 risk categories across 4 themes (6 agent-level, 5 data-level, 5 patient-level, 5 clinician-level), each rated against 5 clinical use cases, yielding 105 risk-by-use-case cells. Panelists scored likelihood and impact on 1-to-5 scales anchored in ISO 31000 descriptors and case-specific vignettes. Dual consensus was defined a priori as at least 70% of raters within 1 point of the median on both dimensions, per the RAND/UCLA Appropriateness Method. Round 2 non-responder ratings were carried forward, with prespecified sensitivity analyses. RESULTS All 14 panelists completed Round 1 (2940 ratings); 11 of 14 completed Round 2. Dual consensus was reached on 95 of 105 cells (90.5%) after Round 1 and all 105 (100%) after Round 2. Of the 10 Round 1 non-consensus cells, 8 reflected impact disagreement alone, 1 likelihood alone, and 1 both. By theme, clinician-level risks ranked highest (median likelihood by impact 10.5, IQR 9.0-12.0), followed by agent-level (9.5, IQR 9.0–12.0), patient-level (9.0, IQR 8.0–12.0), and data-level (9.0, IQR 6.0–10.5). Hypertension monitoring and post-stroke follow-up ranked highest among use cases; HEDIS gap closure ranked lowest. Identity mis-verification accounted for 3 of the 10 Round 1 non-consensus cells. Theme rank order was preserved in sensitivity analyses excluding framework developers and excluding non-responders. CONCLUSIONS A multidisciplinary panel reached complete consensus on a use-case-stratified prioritization of safety risks for patient-facing agentic voice AI. Clinician-level risks ranked highest, directing attention toward the clinician interface rather than agent-internal failure modes, and risk profiles varied substantially across use cases, indicating uniform governance is unlikely to be efficient. These priorities may inform risk-proportionate oversight, pre-deployment testing, and vendor contracting, although prospective validation against observed incidents is required before they can be used as predictive estimates of harm.

  • Open Access

    Safety of Patient-Facing Agentic AI: a Consensus Framework for Risk Assessment and Mitigation (Preprint)

    by Sudha Jayaraman, Arooshi Kumar, Erin Palm, Melanie Fernando, Chethan Sarabu, Michael Choma, Kevin Wang, Vishnu Ravi, Maritza Suarez, Peter McCaffrey, et al.

    UNSTRUCTURED Deploying agentic AI systems without adequate plans for human supervision raises serious concerns about patient safety, privacy, and equity. To address this gap, a group of experts across industry, academic and clinical informatics interested in AI and safety convened a voice AI taskforce to discuss and develop consensus on the impact of agentic AI in healthcare. Through this collaboration, we developed a consensus framework to determine potential risks and plan mitigation efforts based on potential clinical use cases to aid health care delivery organizations assess, implement and evaluate AI agents to meet their needs. Based on five diverse use case examples, we identified common themes of risk at the level of the agent, data, patient and clinician as well as the mitigation strategies needed to address them. Agent-level risks include robust transcription validation, knowledge-grounded responses, mandatory conversation checklists, demographic bias testing, and red-teamed escalation triggers. At the data level, secure identity verification, high‑quality data, interoperable standards and rigorous governance form the foundation of safety. Patient‑level risks include equitable access, patient suitability and clear escalation paths. Finally, clinician-level risks include alert prioritization, defined liability frameworks, workflow-integrated outputs, and preserved clinical override authority. Robust symptom recognition and a thoughtful precision–recall balance are also essential aspects to consider. These guardrails, supported by multidisciplinary oversight and continuous evaluation, can enable AI agents to contribute to patient care without compromising safety, privacy or equity. This framework aims to address the uncertainties in risks to patient safety that should be considered by healthcare delivery organizations to safely apply these technologies to address healthcare needs.


  • Building The Health Chatbot Users’ Guide

    by Duaa O. Khair, Aditya U. Kale, Ruth Agbakoba, Ethan Goh, Bilal A. Mateen, Ariel Y. Ong, Sara-Jane McAteer, Susan C. Shelmerdine, Maxine Mackintosh, Xiaoxuan Liu, et al.


  • Open Access

    Patient Participation in AI for Health Curriculum

    by Kirsten Ostherr, Waverly Huang, Ana Park, Tom Punnen, Bhavik Tadigotla, Valencia Robinson, Andrea Downing

    The adoption of artificial intelligence (AI) in health care has outpaced education of the clinical workforce on responsible use of AI in patient care. Although many policy statements advocate safe, ethical, and trustworthy AI, guidance on the use of health AI has rarely included patient perspectives. This gap leaves out a valuable source of information and guidance about what responsible AI means to patients. In this viewpoint coauthored by patients, students, and faculty, we discuss a novel approach to integrating patient perspectives in undergraduate premedical education in the United States that aims to foster an inclusive and patient-centered future of AI in health care.


  • Open Access

    "Why Patients Should Teach Medical Professionals How to Use AI in Healthcare" (Preprint)

    by Kirsten Ostherr, Waverly Huang, Ana Park, Tom Punnen, Bhavik Tadigotla, Valencia Robinson, Andrea Downing

    UNSTRUCTURED Efforts to regulate AI and educate clinicians to implement ethical and trustworthy AI have rarely included patient perspectives on how and when AI should be used. This gap puts patients at risk and leaves trainees unprepared to integrate AI into patient care. We discuss the need for new approaches that integrate patient perspectives into the training of health professionals to foster an inclusive and patient-centered future of AI in healthcare.

  • Open Access

    Banned Tracking Technology Use Among Medical Device, Pharmacy, and Hospital Webforms: A Cross-Sectional Study (Preprint)

    by Andrea Downing, Jill Holdren

    BACKGROUND Tracking technologies are frequently employed to gather and examine data regarding user interactions with websites or mobile applications of regulated health-related entities. These technologies then illegally share user data with third parties that use it to target ads to patients across social media platforms. This collection and sharing constitutes a data leak of massive proportions. Specifically, some pharmacies, medical device companies, and hospitals utilize banned tracking or surveillance technologies on their websites in a way that exposes patients’ prescriptions, medical device information, doctor appointments and contact details to third parties. OBJECTIVE This study aimed to assess the use of prohibited tracking technologies on unauthenticated URL’s within three types of entities: pharmacies, medical companies, and health systems. METHODS We identified the largest-by-revenue medical device companies, pharmacies, and hospitals in the United States, using a scanning tool we developed based on existing open source software to detect the presence or absence of five banned tracking technologies on a sample of them. RESULTS In total, we included 341 URLs associated with three different types of HIPAA covered entities in our scan sample. Medical device company (n= 147) webpages comprised 43.1% of our overall sample. Pharmacy webpages (n=96) comprised 28.2% of our sample, and hospitals/health system URLs (n=98) comprised 28.7% of our sample. 63.9% of the medical device company URLs scanned contained at least one banned surveillance technology. 59.2% of pharmacy URLs scanned had at least one banned surveillance technology installed. 59.8% out of hospital URLs scanned contained at least one banned surveillance tracker. The most common tracker found on medical device company sample was Google Audience (39.5%), followed closely by Facebook Pixel (36.1%) There were a number of device companies, pharmacies, and hospitals scanned that had none of the banned trackers. n=53 URLs or 36.1% did not have any banned trackers identified. 40.8% (44) of pharmacies did not have any trackers. 67.0% of hospital URLs scanned did not have any trackers identified. CONCLUSIONS This study demonstrates the presence of health trackers on many health-related sites despite the laws that prohibit them, and further examines the ways PHI may be shared with social media platforms or third parties via unauthenticated landing pages. Future studies are needed to assess the impact of leaking sensitive data belonging to millions of patients to third party vendors. CLINICALTRIAL n/a

  • Open Access

    Advancing Digital Health Innovation in Oncology: Priorities for High-Value Digital Transformation in Cancer Care

    by Smit Patel, Jennifer C Goldsack, Grace Cordovano, Andrea Downing, Karen K Fields, Cindy Geoghegan, Upinder Grewal, Jorge Nieva, Nikunj Patel, Dana E Rollison, et al.

    Although health care delivery is becoming increasingly digitized, driven by the pursuit of improved access, equity, efficiency, and effectiveness, progress does not appear to be equally distributed across therapeutic areas. Oncology is renowned for leading innovation in research and in care; digital pathology, digital radiology, real-world data, next-generation sequencing, patient-reported outcomes, and precision approaches driven by complex data and biomarkers are hallmarks of the field. However, remote patient monitoring, decentralized approaches to care and research, “hospital at home,” and machine learning techniques have yet to be broadly deployed to improve cancer care. In response, the Digital Medicine Society and Moffitt Cancer Center convened a multistakeholder roundtable discussion to bring together leading experts in cancer care and digital innovation. This viewpoint highlights the findings from these discussions, in which experts agreed that digital innovation is lagging in oncology relative to other therapeutic areas. It reports that this lag is most likely attributed to poor articulation of the challenges in cancer care and research best suited to digital solutions, lack of incentives and support, and missing standardized infrastructure to implement digital innovations. It concludes with suggestions for actions needed to bring the promise of digitization to cancer care to improve lives.


  • Open Access

    Health advertising on Facebook: Privacy and policy considerations

    by Andrea Downing, Eric Perakslis