Studies from Columbia Nursing
AI can transform how nurses document patient care plans, potentially leading to better health outcomes, according to a new study led by Postdoctoral Research Scientist Fabiana Dos Santos, PhD. However, Santos emphasized careful validation of AI models is essential. The research team developed and validated a method for structuring ChatGPT prompts to generate high-quality nursing care plans. ChatGPT-generated plans prioritized critical aspects of care for a hypothetical lung cancer patient, such as oxygenation, infection prevention, fall risk, and emotional support, while also providing thorough explanations for each suggested intervention, making them valuable tools for nurses.
A study led by Assistant Professor Maryam Zolnoori, PhD, examined the capabilities of ADscreen, a computerized speech processing algorithm, to assist clinicians in detecting and tracking the progression of Alzheimer’s disease and related dementias (ADRDs). The study looked at how the technology analyzed acoustic and linguistic features in voice recordings of ADRD patients, identifying early cognitive impairment signs. Zolnoori and her colleagues found ADscreen to be sensitive, cost-effective, and easily integrated into clinical workflows for early diagnosis, which could improve treatment outcomes. For the next phase of the study focused on fine-tuning the tool, the research team has been collaborating with VNS Health, one of the nation’s largest home- and community-based health care nonprofits, and where both Zolnoori and her colleague, Associate Professor Maxim Topaz, PhD, of Columbia Nursing, conduct clinical research.
Patient Perspectives on AI for Mental Health Care: A Cross-Sectional Survey Study
Natalie Benda, PhD, assistant professor, Meghan Turchioe, PhD, assistant professor, and colleagues investigated patient perspectives on using AI in mental health care in this study. The survey of 500 U.S. adults found that 49.3% of participants viewed AI as beneficial for mental health care, with African Americans and individuals with lower health literacy more likely to hold this opinion, while women were less likely. The study also revealed that 81.6% of participants believed health care professionals should be responsible for misdiagnoses in mental health conditions. Concerns centered on AI tools leading to incorrect diagnoses, inappropriate treatments, reduced provider interaction, and potential breaches of confidentiality.
In a follow up study, Benda, Turchioe, and colleagues compared patient views on somatic and mental health AI applications using two national cross-sectional surveys conducted in 2019 and 2022. Participants in the mental health survey expressed greater skepticism towards AI than those in the somatic health survey, although participants in both surveys had high discomfort with AI delivering diagnoses.
ChatGPT could help novice students overcome programming barriers, thus saving time and reducing frustration and cognitive load, a study led by Turchioe, finds. The study examined the feasibility of using generative AI tools, such as ChatGPT, to help teach nursing PhD students coding and programming skills. Students interacted with ChatGPT for help with programming tasks related to data science and visualization. ChatGPT helped novice students complete tasks like building regression models and generating data visualizations faster and with more confidence.
Another study led by Turchioe explored the integration of generative AI tools in health care, focusing on how PhD nursing students can be equipped to assess and use these technologies effectively. The study aimed to develop and evaluate a brief intervention to increase students’ knowledge of AI applications in health care. Researchers introduced generative AI to nursing PhD students through a combination of lectures and laboratory activities. Students completed tasks related to clinical decision support, patient decision support, and scientific communication using ChatGPT. Results showed students identified both strengths and limitations of ChatGPT and reported increased self-rated competency.