Smart Tools, Uneven Impact: What New Columbia Nursing Studies Reveal About AI and Digital Health in Health Care
Assessing AI in Supporting Family Caregivers: What Current Research Shows
More than one in five people in America are family caregivers, and for many, caring for their loved ones is a heavy burden. Numerous interventions have been developed to help ease this burden, but it remains unclear whether the potential of artificial intelligence (AI) in supporting caregivers has been explored.
To delve further into this topic, postdoctoral research fellow Soojeong Han, PhD, and her colleagues reviewed existing literature on how researchers are evaluating the effectiveness of large language models (LLMs), such as ChatGPT, in helping caregivers with their questions and concerns. The paper, “Exploring Evaluation Measures of Large Language Models for Family Caregiver Use: A Scoping Review,” was published in Digital Health on February 17, 2026.
Through a search of major health and technology databases for studies from 2018 to 2024, the research team identified 10 publications focused on this topic. Most focused on the following core components such as accuracy, reliability, readability, and comprehensiveness, primarily in studies of ChatGPT and similar models.
The studies found AI models were somewhat accurate in their responses, with mixed results in readability and comprehensiveness. This suggests that LLMs as an intervention may not reliably support caregivers, especially in urgent or high-risk situations, and challenges regarding appropriate literacy levels still remain.
Han and her colleagues propose improving readability and accounting for various levels of literacy during the development and evaluation of LLM-based tools. They also recommend generating greater consensus on the frameworks used to evaluate LLMs. These considerations will help inform future research and the practical application of LLMs for family caregiver use, the authors conclude.
How to Bridge the Digital Divide in Heart Health to Reach Black and Latino Communities
More than 75% of adults worldwide now use smartphones, and digital health tools, such as wearable devices and mobile apps, are increasingly used to monitor conditions like atrial fibrillation (AF), an irregular rapid heart rhythm. However, adoption of these technologies has not been consistent, particularly among Black and Latino communities who face higher risks of cardiovascular disease. To examine the barriers and facilitators to using digital health tools for heart health monitoring, assistant professor Meghan Reading Turchioe, PhD, and her colleagues interviewed 25 stakeholders, including patients, caregivers, clinicians, and community health workers, and published findings in a Journal of Cardiovascular Nursing paper, “Community-Driven, Bioethics-Informed Approaches to Digital Inclusion,” on February 12, 2026.
The researchers found that barriers to adoption included low confidence in using technology, financial constraints, language barriers, privacy concerns, and deep-rooted mistrust of the health care system. At the same time, key facilitators included support from community health workers, family involvement, and strong motivation to maintain health. Participants emphasized people are more likely to adopt digital health tools that are accessible, easy to use, and supported by trusted community networks.
The study highlights that even though digital health technologies have the potential to improve detection and management of conditions like AF, a one-size-fits-all approach is insufficient. The authors argue for a community-centered, ethically informed approach that addresses structural barriers, builds trust, and ensures equitable access to technology in underserved communities.
Right Tool, Wrong Time? Study Finds Digital Decision Aids for Heart Conditions Show Promise but Timing and Context Matter
Shared decision-making, in which patients and clinicians work together to make decisions, is critical in health care, especially for complex conditions like AF, where patients must choose between multiple treatment options. To support this process, Turchioe and her colleagues created and evaluated a digital decision aid designed to help patients understand AF rhythm control options, such as medication or catheter ablation. The study involved 75 older adults and examined both how well the tool worked and how easily it could be implemented in clinical settings. The unedited paper ahead of final publication, “Evaluating a digital decision aid for atrial fibrillation rhythm control in a hybrid implementation-effectiveness trial,” was made available online in npj Digital Medicine on March 6, 2026.
The researchers found that while the decision aid was widely used and well received by patients, it did not significantly improve overall decision-making confidence or reduce uncertainty across the entire group. Its effectiveness depended on factors such as patients’ health literacy, digital skills, and when the tool was introduced during the care process.
Screenshot from the AF Rhythm Control Decision Aid, describing changes in a patient’s quality of life after one year of ablation compared to medications when treating AF. Patients can swipe through the tool to understand the efficacy of each treatment option and choose their priorities for treatment. Their preferences are then shared with their provider.
The findings suggest that digital tools alone are not enough to improve decision-making. To ensure these tools are effective, information needs to be delivered to the right patient at the right time. The authors emphasize the need to better integrate these tools into clinical workflows and tailor them to patient needs to maximize their usefulness in real-world care.