SEPTEMBER 6, 2024
A team of researchers has used data from the National Institutes of Health’s “All of Us” program to train a deep learning algorithm to identify breast cancer patients at high risk of developing chronic pain after treatment, according to recent research.
“This predictive approach enables earlier intervention and personalized pain management strategies, potentially improving patient outcomes by reducing the long-term burden of pain associated with breast cancer
A team of researchers has used data from the National Institutes of Health’s “All of Us” program to train a deep learning algorithm to identify breast cancer patients at high risk of developing chronic pain after treatment, according to recent research.
“This predictive approach enables earlier intervention and personalized pain management strategies, potentially improving patient outcomes by reducing the long-term burden of pain associated with breast cancer treatment,” study author Jung In Park, PhD, RN, an assistant professor at the Sue and Bill Gross School of Nursing at the University of California, in Irvine, told Pain Medicine News.
The final data set included 1,131 patients treated for breast cancer, with 199 (17.59%) samples representing patients with chronic pain and 932 (82.4%) representing those without chronic pain (J Nurs Scholarsh 2024 Jul 26. doi:10.1111/jnu.13009). Input variables for modeling were derived from demographic information, diagnosis codes prior to cancer diagnosis and survey data. Based on these inputs, the model predicted the occurrence of chronic pain following cancer diagnosis with an accuracy of 72.8%. The algorithm also identified diagnosis codes most likely to be associated with patients experiencing pain—or lack of pain—after breast cancer treatment.
A limitation of the study was the relatively homogeneous makeup of the participants, with White patients (n=997) accounting for 88.2% of the group. Black or African American individuals (n=84) and Asians (n=24) accounted for 7.5% and 2.1% of participants, respectively.
Park noted that the lack of racial diversity may limit the generalizability of the findings. He also said future research will focus on including more underrepresented populations when training the algorithm to enhance the study’s applicability across racial categories. Furthermore, newer models will incorporate more factors that influence pain, such as medications and genetic information, “for a more comprehensive analysis [of who is likely to suffer from pain after breast cancer therapy],” Park concluded.
—Myles Starr
Park reported no relevant financial disclosures.
