Center Researchers Present 5 Abstracts at the Diagnostic Excellence 2024 Meeting
October 2024
Helen Haskell speaking in a panel discussion on Measuring Diagnostic Excellence at DEX24.
The Diagnostic Excellence 2024 Meeting (DEX2024) brought together researchers, diagnostic safety experts, and patients from across the globe. Co-sponsored by the University of Minnesota and the UCSF Coordinating Center for Diagnostic Excellence (CoDEx), this conference was a hub for sharing the latest breakthroughs and innovations in diagnostic safety. Topics ranged from the role of artificial intelligence (AI) in diagnosis to advancing clinical reasoning. Researchers also had the opportunity to present their latest work through posters and oral presentations.
We are proud to share that the Patient Partnered Diagnostic Center of Excellence (the Center) had a strong presence at the meeting. Our researchers presented five of the 27 poster abstracts showcased, and Dr. Andrew Zimolzak delivered an oral abstract presentation. Additionally, Helen Haskell, who leads Workstream 4 at the Center, participated in a thought-provoking panel discussion on Measuring Diagnostic Excellence. Below, we’ve summarized the posters presented by our team, offering a glimpse into our work from 2024.
Poster Summaries
Using Natural Language Processing (NLP) to Uncover Patient Insights from Free-text Survey Responses
By: Haoyan Zheng, Kewei Ni, PhD, MS, Kristen E. Miller, DrPH, Myrtede Alfred, PhD, Kelly M.
Smith, PhD, MSc
This study used topic modeling, a type of artificial intelligence (AI), to find themes in patient responses to a survey. The survey asked patients at a community hospital in Toronto, Ontario, what “patient engagement,” “diagnostic safety,” and “patient engagement in diagnostic safety” meant to them. Patients described patient engagement as shared decision-making, where they work with care teams to plan treatments. They saw diagnostic safety as understanding test results, getting accurate diagnoses, and clear communication with their care teams. Patients viewed engagement in diagnostic safety as care teams listening to them and respecting their opinions. While the computer program isn’t perfect, the study showed how it can help researchers understand patient feedback to improve safety.
Haoyan Zheng presenting on the use of topic modeling to identify themes in patient surveys.
Talking Tech: Patient Perspectives and Communication of AI in Healthcare
By: Kristen Miller, DrPH, Laura Schubel, MPH, Josh Biro, PhD, Garrett Foresman, Alberta Tran,
PhD, RN, Sadaf Kazi, PhD, Kate MacRae, Adam Visconti, MD, MPH
In this study, we explored patient perspectives on using AI in healthcare, focusing on how AI can be used to make and communicate a diagnosis. Patients participated in a workshop where we discussed different AI tools such as digital scribes, radiology readers, and virtual telehealth assistants. While participants appreciated AI's potential to improve healthcare efficiency and support providers, they emphasized the importance of transparency, patient data privacy, and maintaining human oversight in decision-making. Patients preferred AI as a tool to assist clinicians rather than replace them and expressed a strong need for clear communication about AI's role in their care. These findings underscore the importance of involving patients in designing AI tools to ensure they address concerns, build trust, and enhance patient-provider relationships
Diagnosing Disparities: Exploring Gaps in Diagnostic Safety and Equity
By: Traber Giardina, PhD, Abirami Srivarathan, PhD, Sheryl Jefferson, PhD, Kelly Smith, PhD,
MSc, Kristen Miller, DrPH, Helen Haskell, MA
In this study, we reviewed research on diagnostic errors in mental health disorders to understand disparities impacting historically underserved patients in the U.S. We analyzed 38 studies published between 2015 and 2024, focusing on how timely, accurate, and well-communicated diagnoses were. The studies looking at delays in diagnosis mostly focused on neurodevelopmental disorders. The studies looking at accuracy of diagnosis were mostly about mood, anxiety, or personality disorders. Many disparities in diagnosis were found, such as delayed autism diagnosis among
Black children and increased odds of misdiagnosis of borderline personality disorder among
sexual minorities. Our research findings show that diagnosis disparities exist, but many current studies fail to explain where the diagnostic process breaks down. Future research should explore these breakdowns to improve mental health diagnosis for underserved communities.
Disparities in Diagnostic Errors in Mental Health Disorders: A Systematic Review
By: Abirami Srivarathan, PhD, Sheryl Jefferson, PhD, Layla Heimlich, Kelly Smith, PhD, MSc,
Kristen Miller, DrPH, Helen Haskell, MA, Traber Giardina, PhD
We reviewed current research on diagnostic disparities in mental health disorders to figure out where in the diagnostic process things go wrong. Twenty research studies published from January 2015 to August 2024 were reviewed by our team. All of the studies were focused on disparities experienced by patients in the United States. Overall, most disparities we identified were related to delays in diagnosis. Most commonly, diagnosis errors were due to barriers in accessing care and provider bias among historically marginalized and underserved patients. Culturally inequivalent assessment approaches to testing and diagnosing also led to wrong or delayed diagnosis among Black and Hispanic patients.
Dr. Abirami Srivarathan presenting the systematic review of diagnostic disparities in mental health disorders.
By: Bat-Zion Hose, PhD, Amanda Rounds, PhD, Ishaan Nandwani, Deanna-Nicole Busog,
Kristen Miller, DrPH
In this study, we explored how people with spinal cord injuries or diseases, multiple sclerosis, or spina bifida use ChatGPT to manage urinary symptoms. We interviewed 30 participants who used ChatGPT to understand symptoms like changes in urination patterns, urine characteristics, and severity of symptoms to decide when to seek medical care. Most participants found ChatGPT useful for quick and accessible advice, especially for non-urgent symptoms, and some noted that it aligned with their doctor’s medical advice. However, some participants were concerned about the lack of personalized responses and sources or information. The study highlights ChatGPT's potential as a helpful tool for managing urinary symptoms and the need for more personalized and tailored AI tools in healthcare.
Dr. Bat-Zion Hose presenting on the use of ChatGPT in urinary symptom management.
At the Patient Partnered Diagnostic Center of Excellence, we are dedicated to improving diagnostic safety by listening to patients, leveraging innovative technologies, and addressing disparities in care. The work presented at DEX2024 showcases our commitment to these goals. We look forward to continuing partnering with patients in this important research and making diagnosis safer and more equitable for all.
Mary A. Hill, Dr. Kristen Miller, Helen Haskell, Dr. Traber Giardina, and Dr. Abirami Srivarathan at the DEX24 Meeting.
Haoyan Zheng and Helen Haskell at the DEX24 Meeting.