The Evidence Speaks Series is a recurring feature highlighting the latest in Advancing Health research. This series features summaries of select publications and is designed to keep media and the research community up to date with the Centre’s current research results in the health outcomes field.
To ensure this research is quick and easy to share, you are welcome to save the social cards and use as you see fit.
Could large language models be used to predict breast cancer relapses based on pathology reports?
Natural language processing for local, regional, and distant breast cancer relapse identification in pathology reports. Lee J, Jettinghoff W, Arbour G, Zepeda A, Isaac K, Ng R, Nichol A. CMAJ. 2025.
Large language models (LLMs) are a way to recognise and predict patterns in very large sets of text. LLMs have recently become popular on the public stage in forms such as generative AI, which allows them to write compelling text through realistic predictions of natural human language. Outside of use within general public, LLMs have significant potential applicants in medicine and health research. To test whether LLMs could used to find trends in cancer reports and give early predictions of potential relapses, an investigative team including Advancing Health scientist Dr. Kathryn Isaac collected 1,888 pathology reports from 993 patients diagnosed with breast cancer in British Columbia between January 1, 2005, and December 31, 2014.
The initial pathology reports were manually tagged with information about each patient’s cancer, including how far the cancer had travelled from its initial source, as well as any relapses and their severity. The researchers then assessed the model’s ability to identify these features in reports. They also asked the model to predict if the patient would relapse based on the initial pathology report, and compared the prediction to the real-world data. After fine tuning the model, they found it was extremely accurate in predicting the chance of relapse, with low levels of both false positives and false negatives.
Although still in early stages, this technology could have several beneficial real-world applications, including significantly speeding up the processing and analysis of large report databases. It can also be a tool to help flag cancer patients with higher chances of relapses for more careful monitoring, allowing those cases to be caught early. Future research will attempt to expand this technology to other types of cancer, as well as integrate other sources of data, such as administrative records.
Survey shows that a lack of information and autonomy makes it harder for patients to make decisions about treating their chronic pain
Risk factors of decisional conflict among people living with chronic pain identified through a pan-Canadian survey. Naye F, Tousignant-Laflamme Y, Sasseville M, Cachinho C, Gérard T, Toupin-April K, Dubois O, Paquette J, LeBlanc A, Gaboury I, Poitras M, Li L, Hoens A, et al. Sci Rep. 2025.
Nearly 8 million people across Canada live with chronic pain. Making decisions about how to manage and treat that pain can often be challenging. Decisional conflict can cause delays in patients selecting a treatment, increase chances of regret in their treatment choice, and increases intentions of suing physicians. In this study, Advancing Health scientist Dr. Linda Li, research associate Alison Hoens, and team conducted a survey of 1,649 people with chronic pain across Canada to uncover if they experienced decisional conflict, and what parts of the process caused that difficulty.
Respondents reported experiencing pain that ranged from 3 months to 59 years. Seventeen risk factors for conflict in decision-making were identified. The most serious of these were difficulty understanding health care information, a lack of prior knowledge about options for their pain management, and situations where they felt they were unable to select their preferred treatment option. This research shows the importance of improving health literacy for Canadians and making sure health care providers give sufficient and effective explanations and information to their patients. It also contributes to a growing body of evidence that shows that doctors should place increased emphasis on explaining options in accessible terms, and treating patients as equal collaborators who are given autonomy in choosing a treatment plan that best suits them.
Top TikTok videos on #homelessness are generally positive but often lack nuance.
A summative content analysis: how is #Homelessness portrayed on TikTok? Turuba R, Boniface V, Adair S, Zenone M, and Barbic S. J Soc Distress Homelessness. 2025.
By fostering communication and connection across all walks of life, social media has incredible power to share information and reduce prejudice. Equally, however, social media can also be used to spread and perpetuate misinformation. To explore these themes, a team of scientists, including Advancing Health’s Dr. Skye Barbic, wanted to see how the topic of homelessness was covered on the short-form video platform TikTok. To find out, they ran a content analysis of the top 200 videos tagged #homelessness in July 2022.
The team was able to break the videos down into four main categories by theme. The most common by far (74.6 per cent) was of content creators giving unhoused people money, food, shelter, or other forms of support. Only one of these content creators was unhoused themselves, and they were also the only one to ask people being filmed if they wanted their faces blurred. Ten per cent of the videos featured the content creator helping people become permanently housed, which usually involved getting to know them over the long-term.
The second largest category (28 per cent) of videos were about content creators connecting with people who were unhoused through conversation. Twenty-three per cent of videos were about users sharing their personal experience with homelessness. Only two videos involved the creators asking or thanking their viewers for money.
Most of these videos presented unhoused people in a positive light, emphasizing kindness and gratitude. In this way, these videos could help shift the public perception around homelessness. However, the videos rarely discussed the underlying causes of homelessness. It also over-emphasized street homelessness over forms of being unhoused, such as couch surfing or trading sex for shelter. As “hidden homelessness” is more often experienced by women, gender diverse people, and young adults, videos like these may perpetuate stereotypes and misconceptions. There are also ethical concerns surrounding the filming of homeless people for views and ad revenue, especially without consent. Further efforts are needed to improve public awareness of the causes of homelessness and effective real-world solutions.
