Background & Aims

Chronic pain is an important burden on patients and the health care system. The psychological factors should be screened and managed to prevent chronification. The development of mathematical models may augment the current understanding of pain.
The study aimed to mathematically analyze the relationship between psychological factors on pain chronification.

Methods

The study included 976 people with pain (headache) who were questioned and evaluated. In the research group, 31.6% (216) were men and 68.4% (467) were women, average age 37 ±13 years. The research group was divided into four categories: patients without pain – 126 p (18.4%), patients with rare pain – 200 p (29.3%), patients with frequent pain – 132 p (19.3 %) and patients with chronic pain – 225 p (32.9 %). Patients completed the SCL–90 (Symptom Rating Scale) and were analyzed according to categories. This questionnaire allows the analysis of psychological manifestations and physical signs. The distribution by age demonstrates that the groups are homogeneous by this character. The patients completed the PID personality inventory elaborated according to DSM-5 which allows the assessment of possible maladaptive or dysfunctional tendencies.

Results

In the study, 13 predictive models were developed. The comparative evaluation of the proposed models highlighted the best performance for the DL model (deep learning model), which demonstrated sensitivity (recall) at the level of 69.5%, with the f1 score (f1) being estimated at the level of 64.3%, the mentioned indicators being the parameters optimal for the current study design. The development of the model was carried out according to the principles of the development of artificial intelligence models, which means the division of the cohort into the components for training the model (80% of the total number), as well as the validation component of the model (20% of the number total). Model training was performed by cross-validation (5 layers), the optimal model being identified by the average performance of the formed layers. The optimal model was applied for training components, and estimated coefficients for the respective equations were applied for validation components.

Conclusions

The SHAP analysis highlighted the 20 basic parameters, also determining their polarity. It is important to mention that the effects presented are adjusted to the effects of the parameters included in the current study. In the study, the predictive models for pain chronification were developed and the potential psychiatric predictors were evaluated.

References

1. Pellicer-Valero OJ, Martín-Guerrero JD, Cigarán-Méndez MI, Écija-Gallardo C, Fernández-De-Las-Peñas C, Navarro-Pardo E. Mathematical Modeling for Neuropathic Pain: Bayesian Linear Regression and Self-Organizing Maps Applied to Carpal Tunnel Syndrome. [cited 2024 Jan 31]; Available from: www.mdpi.com/journal/symmetry
2. Ashley Lang V, rn Lundh T, Ortiz-Catalan M. Mathematical and Computational Models for Pain: A Systematic Review. [cited 2024 Jan 31]; Available from: https://academic.oup.com/painmedicine/article/22/12/2806/6288499

Presenting Author

Oxana Grosu

Poster Authors

Misic Octavian

MD, PhD

Lead Author

Misic Maximilain

MD

Lead Author

Grosu Oxana

MD

Diomid Gherman Institute of Neurology and Neurosurgery

Lead Author

Topics

  • Novel Experimental/Analytic Approaches/Tools