Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Predicting Clinician- and Patient-Assessed Pain Intensity from Wrist-Worn Physiological Signals in Chronic Spine Pain Patients: A Multi-Rater Machine Learning Study
ABSTRACT
Background:
Traditional pain assessment relies on a unidimensional "gold standard" of patient self-report, which is often unavailable in critical or non-verbal populations. While machine learning has shown promise in decoding physiological signals for pain using the patient’s self-report of pain as the ground truth, existing models do not account for situations in which the patient’s self-report of pain is unavailable.
Objective:
This study developed an automated, multi-rater pain assessment model using wrist-worn physiological signals to predict pain intensity scores from patients, nurses, and physicians.
Methods:
We recruited 23 chronic spine pain patients undergoing medial branch nerve block procedures. Physiological data (electrodermal activity (EDA), heart rate (HR), temperature (TEMP), interbeat interval (IBI), and blood volume pulse (BVP)) were captured via Empatica E4 and EmbracePlus sensors. Nine machine learning architectures were evaluated using leave-one-subject-out cross-validation (LOSO-CV). Manifold visualizations (PCA, t-SNE, UMAP) were utilized to interrogate latent data structures, and Spearman correlation analyses were conducted to examine rater-specific physiological association patterns.
Results:
The Ensemble (Random Forest + XGBoost) model emerged as the top performer, achieving an R2 of 0.940 and a mean absolute error (MAE) of 0.220. SHapley Additive exPlanations (SHAP) analysis revealed that nearly 70% of the model’s predictive weight was derived from temperature and electrodermal features. Correlation analysis revealed rater-specific physiological association patterns, with EDA showing the most consistently positive directional tendency across all three raters.
Conclusions:
Wearable physiological modeling may help characterize rater-specific pain assessment patterns and identify differences between patient self-report and clinician-assessed pain. Future work is needed using more raters to better understand how healthcare workers formulate their pain assessment with access to patients’ physiological data.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.