Improving Patient Understanding of Their Prostate Cancer Diagnosis Using AI
Status
Conditions
Treatments
Details and patient eligibility
About
The purpose of this study is to determine if ChatGPT (an artificial intelligence driven chatbot) can help patients better understand their own pathology results and prostate cancer diagnosis.
Full description
Prostate cancer is the second most common cancer among men and poses a significant challenge due to its varied clinical course, ranging from indolent tumors that only requires surveillance, to aggressive life-threatening metastatic disease.1 Nearly 1 million prostate biopsies are performed in the United States each year, and approximately 25% of these biopsies result in a prostate cancer diagnosis. One out every four patients receiving a prostate biopsy will inevitably have a positive pathology report, and are typically waiting anywhere from 2 days to multiple weeks prior to follow-up with their healthcare provider. It is without question that patients experience substantial anxiety while awaiting pathology results or making treatment decisions, as the uncertainty surrounding their diagnosis can be overwhelming.
Risk stratification in localized prostate cancer requires the patient's pre-biopsy PSA (prostate-specific antigen) value and prostate biopsy results. Once appropriately risk-stratified, depending on their risk level, patients may require additional staging scans, followed by a discussion with their urologist or oncologist regarding treatment options. Treatment options for localized prostate cancer may include active surveillance, radical prostatectomy, radiotherapy, or focal therapy.
Chat GPT, developed by Open AI, is a large language model based on the GPT (Generative Pre-trained Transformer) architecture. It leverages deep learning techniques, specifically transformer networks, to generate human-like text responses. Trained on vast amounts of diverse data, it uses pattern recognition and contextual understanding to engage in meaningful conversations. Initially aimed at general natural language processing tasks, Chat GPT's capabilities have been adapted for various applications, including healthcare, where it can assist in patient communication, decision support, and educational initiatives. A recent study has shown promising results with several large language model chatbots such as Chat GPT, which was able to simplify pathology reports for the average layperson with 97% accuracy, and an extremely low rate of hallucination (0.26%).2 Technology such as this could empower patients to not only better understand their disease, but disease but also increase participation in shared decision-making. In addition, to test how well AI-powered chatbots perform, when asked some of the most queried questions about the five most common cancers, including prostate cancer, a recent study in JAMA oncology reported that the quality of information generated by four different AI chatbots were good (median DISCERN score of 5, indicating high quality of information), and no misinformation was identified.3
In the current era, where information is so readily available and accessible, patients are often able to 'Google' their own diagnosis before seeing the clinician for follow-up. However, there is no quality-control for the information that is presented online. Often, when patients input keywords such as 'prostate cancer' into search engines, they are led down a rabbit-hole of information, that is neither accurate, nor easy to understand. Loeb et al, performed a comprehensive overview of available prostate cancer videos on screening and treatment on the You tube platform, and deemed the quality of information as only moderate, with 77% of videos containing potentially misinformative or biased content, and there was a significant negative correlation between scientific quality and viewer engagement (viewer reviews via thumbs up, and number of views).4 An additional study which examined the readability of prostate cancer information online found that the large majority of information available on the Internet about prostate cancer are difficult to understand, particularly for older adults - who may already have compromised capacity to understand health information, who happens to also be the population at greatest risk for prostate cancer.5 Another study showed that when prompted, the free version of Chat GPT was able to paraphrase health information to a grade six reading level, which is associated with increased readability for the general public.6 Given the challenges of locating reliable, and patient-friendly information online regarding a new prostate cancer diagnosis, investigator identified a potential gap that may be filled by AI-powered chatbots.
Although high-quality patient-oriented websites (such as the Cleveland Clinic or Mayo Clinic websites) do exist, along with other sources of vetted patient education materials (such as the Cancer Support Community, and Facing Our Risk of Cancer Empowered - FORCE), these are often not what patients turn to at first, especially given the influence of social media platforms and news outlets.
Our proposed Prostate Cancer chatbot is not meant to replace these vetted resources, but to enhance patient comprehension in addition to what is available, by having information presented in a different format - interactive and customized. In fact, in a recent white paper published by the ACCC (Association of Community Cancer Centers), one listed future goal of patient-centered reporting of pathology results, is to include interactive web pages with customized educational content - which is exactly what our chatbot hopes to achieve.7 The main goal of this project is to create a tool which can enhance understanding of pathology reports, which is filled with medical terminology and grading systems.
Enrollment
Sex
Ages
Volunteers
Inclusion criteria
- Participants must be >18 years or age
- Participants must have received a prostate biopsy at Cleveland Clinic main campus within the last 30 days, and have not received follow-up on their pathology results
- Participants must be willing to attend an in-person follow-up to review their pathology
- Participants must be English speaking
- Participants must provide informed consent
Exclusion criteria
• Participants who do not meet the inclusion criteria, as stated above
Trial design
Primary purpose
Allocation
Interventional model
Masking
50 participants in 1 patient group
Trial contacts and locations
Loading...
Central trial contact
Zeyad Schwen, MD
Data sourced from clinicaltrials.gov
Clinical trials
Research sites
Resources
Legal