CLINICAL AI

The intersection of artificial intelligence and geriatric medicine has emerged as a critical frontier in healthcare transformation, driven by demographic imperatives and regulatory mandates that are fundamentally altering how hospitals approach care for older adults. Dr. April Ehrlich, an assistant professor at the University of Arizona College of Medicine and chief of clinical research development in the Division of General Internal Medicine, Geriatrics & Palliative Medicine, exemplifies a new generation of clinician-researchers bridging the gap between technological innovation and frontline clinical practice. Her recognition as a distinguished scholar at the inaugural AI, Technology, and Aging Summit hosted by the Johns Hopkins Artificial Intelligence and Technology Collaboratory for Aging Research underscores the growing national focus on leveraging computational approaches to address the complex needs of hospitalized elderly patients.
The regulatory landscape driving this transformation centers on the CMS Age-Friendly Hospital Measure, finalized in July 2024 as part of the FY2025 Inpatient Prospective Payment System rule. This mandatory measure requires all hospitals participating in the Hospital Inpatient Quality Reporting Program to implement protocols across five domains: eliciting patient healthcare goals, responsible medication management, frailty screening and intervention including cognition and mobility assessment, social vulnerability assessment, and designation of age-friendly leadership. Hospitals failing to meet these reporting requirements face significant financial penalties through reduced annual Medicare payment updates, creating powerful incentives for systematic adoption of evidence-based geriatric care principles. The measure explicitly aligns with the 4Ms Framework—What Matters, Medication, Mentation, and Mobility—developed through the Age-Friendly Health Systems initiative led by The John A. Hartford Foundation and the Institute for Healthcare Improvement.
Dr. Ehrlich's research provides empirical support for these policy developments, with two of her papers cited in the CMS Age-Friendly Hospital Specifications Guide. Her work on multispecialty geriatric surgery pathways at Johns Hopkins demonstrated that comprehensive geriatric assessment integrated into surgical care reduced inpatient costs by approximately $2,000 per patient while simultaneously improving clinical outcomes. More specifically, her propensity-matched analysis of frail surgical patients revealed a 20 percent reduction in total hospitalization costs and significant decreases in major complications and loss of independence. These findings challenge the persistent misconception that comprehensive geriatric assessment represents an additional cost burden rather than a value-generating intervention, providing health systems with concrete evidence to justify resource allocation for specialized geriatric services.
The application of artificial intelligence to geriatric care extends beyond surgical pathways to address one of the most challenging complications in hospitalized older adults: delirium. Dr. Ehrlich's current research focuses on delirium onset in emergency departments, where ED boarding—the practice of holding admitted patients in the ED while awaiting inpatient beds—has been associated with increased delirium risk. Recent studies have demonstrated that each hour of ED boarding increases the odds of developing delirium or severe agitation during inpatient admission, with this risk magnified substantially in patients with pre-existing dementia. AI-powered prediction models are emerging as tools to identify high-risk patients and trigger specialized assessment protocols. A landmark study from Mount Sinai demonstrated that an AI model analyzing electronic health record data and clinical notes through natural language processing increased delirium detection rates from 4.4 percent to 17.2 percent monthly, representing a 400 percent improvement in case identification. The model achieved diagnostic accuracy of 93.07 percent while reducing prescriptions of potentially inappropriate medications in older adults, illustrating how AI can simultaneously improve detection and modify clinical behavior to prevent iatrogenic complications.
The technical architecture underlying these AI applications in geriatric care encompasses multiple computational approaches, each addressing different clinical challenges. Machine learning algorithms excel at integrating heterogeneous data sources—from electronic health records and wearable device outputs to imaging studies and laboratory results—to generate risk predictions that would be computationally impossible for clinicians to synthesize in real-time. Natural language processing enables extraction of clinically relevant information from unstructured clinical notes, capturing subtle observations about mental status changes that individual providers might document without recognizing their significance in the broader context of delirium risk. Computer vision applications are being deployed to analyze gait patterns and detect early signs of frailty through objective measurement of mobility parameters including stride length, stride time variability, and double support duration. These objective assessments address a fundamental limitation of traditional frailty screening tools, which often rely on clinician judgment and face-to-face evaluation that may be time-consuming and subject to interobserver variability.
The Johns Hopkins Artificial Intelligence and Technology Collaboratory for Aging Research, which recognized Dr. Ehrlich's contributions, represents a $40 million National Institute on Aging investment in developing and validating AI applications specifically designed for older adults. This collaboratory model, replicated at the University of Pennsylvania and University of Massachusetts Amherst, provides pilot funding for promising AI-driven technologies addressing healthy aging and Alzheimer's disease and related dementias. The infrastructure includes access to study sites, diverse datasets, expertise in AI implementation design, and connections to venture capital networks, creating an ecosystem that accelerates translation of research findings into commercially viable products. Importantly, the program emphasizes stakeholder engagement throughout the development process, recognizing that AI tools designed without meaningful input from older adults, family caregivers, and frontline clinicians risk creating solutions that are technically sophisticated but clinically impractical or culturally inappropriate.
Implementation of AI-enhanced geriatric care faces substantial challenges that extend beyond technical performance metrics to encompass workforce development, organizational culture, and ethical considerations. Healthcare provider engagement emerges as a critical determinant of success, with studies demonstrating that clinicians serve as both facilitators and barriers to AI adoption depending on their confidence, training, and involvement in system design. Educational interventions paired with workflow integration have shown promise, with one study reporting that a one-day workshop for clinicians improved delirium screening rates and increased referrals for geriatrics evaluation and home care services. However, reminder systems without accompanying provider education demonstrated limited impact, suggesting that effective AI implementation requires more than technical deployment of alerts and requires fundamental changes in clinical practice patterns supported by ongoing training and feedback mechanisms.
The issue of algorithmic bias presents particularly concerning challenges in geriatric applications, where training datasets may systematically underrepresent the diversity of older adult populations. Studies have documented that AI systems trained predominantly on data from younger, healthier, or more socioeconomically advantaged populations can produce biased outputs when applied to elderly patients with multiple comorbidities, cognitive impairment, or limited English proficiency. Technical bias can occur when detection algorithms perform less reliably for certain demographic groups—for example, when visual assessment tools for detecting falls or mobility impairment are optimized for lighter skin tones and fail to perform equivalently for patients with darker skin. Label bias arises when the parameters used to train algorithms incorporate proxy variables that correlate with protected characteristics, potentially reproducing historical patterns of discrimination in clinical decision-making. Addressing these challenges requires intentional efforts to ensure training datasets adequately represent population diversity, continuous monitoring of algorithm performance across demographic subgroups, and transparency in disclosing known limitations and potential biases to end users.
Dr. Ehrlich's appointment to research committees of both the American Delirium Society and the American Geriatrics Society positions her to influence the development of practice standards and research priorities that will shape how AI tools are validated and integrated into clinical guidelines. Her concurrent involvement with Banner Health's Delirium Power Plan Working Group and Identification & Management of Delirium Working Group demonstrates the importance of simultaneous engagement in national standard-setting and local implementation efforts. This dual role is essential for ensuring that research advances translate into operational changes that improve patient outcomes, while real-world implementation challenges inform the research agenda to address practical barriers encountered in clinical settings. Banner Health's recent $100,000 grant from the Institute for Healthcare Improvement to implement the Neurological Orientation and Verbal Response Assessment for Delirium tool exemplifies how health systems are investing in AI-enabled diagnostic technologies specifically designed for geriatric populations.
The Age-Friendly Health Systems movement has achieved remarkable scale, with 5,477 hospitals, medical practices, and long-term care organizations recognized as Age-Friendly Health Systems Participants or Committed to Care Excellence as of October 2025. Participation in Age-Friendly action communities, led by the American Hospital Association and Institute for Healthcare Improvement, provides structured support for implementing the 4Ms framework through seven-month virtual learning collaboratives that enable teams to test interventions and share lessons learned. Health systems report multiple benefits from participation including improved patient and provider satisfaction scores, reduced length of stay for geriatric patients, and better alignment of organizational resources with the needs of older adults. Importantly, the framework emphasizes redeploying existing resources rather than requiring substantial new investments, making adoption feasible even for resource-constrained institutions.
The convergence of AI technology, evidence-based geriatric assessment, and federal quality measurement represents a paradigm shift from episodic, disease-focused care toward comprehensive, person-centered management of older adults' complex health needs. Dr. Ehrlich's work on geriatric surgery pathways demonstrated that this approach reduces not only clinical complications but also loss of independence, defined as discharge to post-acute care facilities rather than home. For frail surgical patients, implementation of the geriatric surgery pathway was independently associated with a 70 percent reduction in risk of loss of independence and a 69 percent reduction in major complications. These outcomes extend beyond the immediate hospitalization to affect long-term trajectories of functional decline, with implications for patients' quality of life, caregiver burden, and healthcare system costs associated with institutional long-term care.
Looking forward, the integration of wearable technologies and remote monitoring systems promises to extend AI-enhanced geriatric care beyond hospital walls into home and community settings. Studies have demonstrated that consumer-grade wearable devices can predict frailty status among home care clients by analyzing patterns in daily step count, sleep quality, and heart rate variability. These continuous monitoring capabilities enable detection of gradual functional decline before acute decompensation necessitates emergency department visits or hospitalization. The Nordic countries have pioneered integration of AI and Internet of Things technologies for remote health monitoring, with Sweden's WASP program and Finland's AI-based video monitoring solutions demonstrating feasibility of real-time health data analysis to support aging in place. However, adoption of these technologies among older adults faces barriers including limited digital literacy, concerns about privacy and data security, and inadequate internet infrastructure particularly in rural areas. Addressing these barriers requires attention to user-centered design, provision of training and ongoing technical support, and engagement of healthcare providers as trusted advisors who can explain the benefits and address concerns about digital health tools.
The path forward for AI in geriatric medicine requires sustained collaboration among clinicians, technology developers, health system administrators, policymakers, and importantly, older adults themselves and their family caregivers. Dr. Ehrlich's participation in the Tideswell Emerging Leaders in Aging Scholars Program reflects recognition that leadership development is essential for building the workforce capacity to drive this transformation. The program's emphasis on authenticity, leading others, and scaling for impact provides frameworks that emerging geriatric leaders can apply to navigate the organizational and cultural changes required to implement AI-enhanced care delivery models. As hospitals respond to the CMS Age-Friendly Hospital Measure mandate, the clinical research and implementation science conducted by leaders like Dr. Ehrlich will provide the evidence base and practical guidance necessary to ensure that AI technologies genuinely improve outcomes for older adults rather than simply adding complexity to already-strained healthcare systems.