American Society for Peripheral Nerve
ASPN Home ASPN Home Past & Future Meetings Past & Future Meetings

Back to 2023 Posters


Identifying the Predictors of Cold Sensitivity Severity in Patients with Carpal Tunnel Syndrome using Artificial Intelligence
Moaath Saggaf, MD1; Christine B Novak, PT, PhD2; Dimitri J Anastakis, MD3; Brian M Feldman, MD, MSc?
1University of Toronto, Toronto, ON, Canada; 2Toronto Western Hospital Hand Program, Hospital for Sick Children, University of Toronto, Toronto, ON, Canada; 3Division of Plastic and Reconstructive Surgery, University of Toronto, Toronto, ON, Canada

Introduction: Several associations of cold sensitivity in carpal tunnel syndrome (CTS) were sporadically reported in the literature. Machine learning methods can identify the factors that are associated with cold sensitivity. The aim of the study was to identify the predictors of cold sensitivity severity in CTS patients using machine learning methods.
Methods: We used data from a prospective cohort study for patients with CTS. LASSO regularization was applied to identify the clinical features associated with cold sensitivity severity as measured by the cold intolerance symptom severity (CISS) questionnaire. We validated the model internally using 5-fold cross-validation. The model was optimized based on the minimum lambda value from the training subset during cross-validation. The analysis was repeated 100 iterations to ensure the robustness of the results.
Results: The analysis was conducted on 106 patients with CTS. The mean age of the participants was 56.2 years (SD=12.7), and the majority of the participants were female patients (n= 71, 67.0%). Age, medical comorbidities, smoking status, history of steroid injection and a previous trial of splinting were not associated with an increased cold sensitivity severity. CTS severity had the strongest clinical association with cold sensitivity severity (mean beta = 8.6, SD=0.04). Female sex was also associated with increased cold sensitivity severity (mean beta = 1.8, SD=0.6). Of the cold-induced symptoms, cold-induced pain and numbness had the highest effect size for cold sensitivity (mean beta = 10.7 for both symptoms, SD= 0.4 and 0.3, respectively).
Conclusion: The machine learning model identified that CTS severity and female sex were associated with increased cold sensitivity severity. Cold-induced pain and numbness can be used to screen for cold sensitivity in CTS patients.


Back to 2023 Posters