Novel AI model may predict which patients with cancer are at risk for chronic pain
Click Here to Manage Email Alerts
A multi-institutional team of researchers developed an AI model designed to predict which patients with breast cancer are most likely to experience chronic pain.
The model also may help clinicians addressing underlying conditions that contribute to risk for chronic pain.
“Our goal is to personalize care so the conditions a person has — or previously had — will determine the trajectory of treatment,” Lisiane Pruinelli, PhD, MS, RN, FAMIA, associate professor of family, community and health systems science at University of Florida College of Nursing, told Healio. “For example, we found depression to be one of the key factors for developing chronic pain. Why not address depression early in hopes of reducing the chance patients will develop chronic pain?”
Pruinelli and colleagues used demographic, diagnostic and social survey data from NIH’s All of Us program to conduct a retrospective observational study to assess how the model predicted risk for chronic pain among people with breast cancer.
Results of the study — which included data from 1,131 patients — showed the model had 72.8% accuracy and an area under the receiver operating characteristic curve of 82%.
Researchers also identified anxiety and depression, prior cancer diagnosis and certain infections as factors most closely associated with development of chronic pain.
Healio spoke with Pruinelli about the model, how it has performed so far and the role it may play in clinical practice.
Healio: How prevalent is chronic pain among people with cancer?
Pruinelli: Approximately 35% of patients with cancer will develop chronic pain, and the reported rate for people with breast cancer is even higher. It’s not just about the pain itself — it’s about how chronic pain affects quality of life and daily living.
Healio: How did you develop the AI model and how does it predict chronic pain?
Pruinelli: We used a deep learning model, which mimics our brain, and we used one that is able to account for time series data. We can take a sequence of data over time, not just a static measure, and combine that in the model. We extracted demographic information, diagnosis codes and survey data to predict risk factors that are more related to developing chronic pain in 3 years.
Healio: What did you find?
Pruinelli: The model accurately predicted chronic pain 82% of the time. We usually consider more than 80% very good. Imagine being able to tell a patient they have an 80% or higher chance of developing chronic pain and, more importantly, knowing how to prevent that? This could be a significant aid for clinicians and patients.
Healio: What are the potential implications of these findings?
Pruinelli: We now have more information to tailor treatment based on patients’ health status. We could apply this to people with other types of cancer, as well. We don’t necessarily need to be disease-specific, either. We can say a patient has cancer and a high chance of developing chronic pain, but we also know there are other factors that determine the likelihood of developing chronic pain. It’s not just the fact that a person has cancer, and our study showed that. We could apply this same modeling approach to noncancerous conditions, as well. We’d just have to change the parameters in such a way we would capture different factors based on the specific condition we want to address.
Healio: What is next in your research?
Pruinelli: One of the first things we want to do is to investigate whether this is replicable to a more diverse patient population. We want to evaluate how well our model performs when applied to a population that the original model didn’t see before. We also want to include genetic information, because we want to further understand how genetic factors play a role in cancer in the presence of other risk factors, as the ones we identified. It may be that genetic factors are correlated with other conditions that are related to development of chronic pain — not just with cancer — and that these other conditions are more easily treated or preventable than chronic pain.
Healio: Is there anything else you’d like to mention?
Pruinelli: First, I would like to thank Jung In Park, PhD, RN, for her leadership and commitment to improve the overall health of persons with breast cancer with studies like this one. We have a long history of collaboration, and we want to keep working together to implement this work to help clinicians to deliver better and patient-centric care. These models can help clinicians to summarize all information derived from care delivery and target better treatment strategies. This can be further used for shared decision-making with the patient, of course. Our ultimate goal is to pilot this model in the real world and evaluate how it will work, and how much chronic pain we can prevent over time and the impact on improving people’s overall health.
Reference:
For more information:
Lisiane Pruinelli, PhD, MSN, RN, FAMIA, can be reached at lisianepruinelli@ufl.edu.
Collapse
Healio Interviews
Click Here to Manage Email Alerts