Development, reliability, and validity assessment of a numerical algorithm to detect centralization phenomenon and directional preference among spinal pain patients. (Preprint)

BACKGROUND Affecting millions of people, spinal pain is the leading cause of years lived with disability worldwide since the nineties. Centralization phenomenon (CP) and directional preference (DP) are common features in spinal pain patients, indicating a good clinical prognosis. CP is defined as a rapid and lasting migration of distal pain from the limb to the center of the spine after repeated movement tests. DP is a broader concept in which both the migration and decrease of pain are considered. While their detection has still been based on clinician decision, digital tool would provide objective measurements. OBJECTIVE We developed and assessed the reliability and validity of a new algorithm to detect CP and DP among spinal pain patients using quantitative a pain mapping software (PRISMAP). METHODS We designed a two-phase retrospective, cross-sectional, double-blinded diagnostic accuracy study. In Phase 1, using PRISMAP, we recorded and analyzed pain variations before and after a CP-focused physiotherapy session with specialized physiotherapy (PT). PTs classified patients as CP+ or CP-. We developed an algorithm to model changes in pain topography and intensity, identifying patients relative to PT classification. In Phase 2, PTs conducted four pain mappings. The initial two maps depicted a three-day overall patient pain profile, from which reliability and agreement were calculated using Intraclass Coefficient Correlation (ICC), Standard Error of Measurement (SEM), Coefficient of Variation (CV), and Bland-Altman analysis (BA). PTs then documented t-time pain mapping pre- and post-repeated movement test, classifying patients into CP-/CP+ and DP-/DP+. Validity parameters (sensitivity, specificity, positive and negative likelihood ratios (LR+/LR-)) were calculated from the latter two maps, using PT classification as the standard reference. RESULTS Twelve patients were included in Phase 1 and 49 in Phase 2. The algorithm demonstrated good reliability (ICC=0.993 [95%CI 0.988–0.996], SEM=0.211, CV=12.2%, and bias error with the BA of -0.041 representing 2.4% of the sample mean). Validity for CP was 92.0% [95%CI 73.7–99.02], 79.2% [95%CI 57.8–92.9], 4.42 [95%CI 2.01–9.71], 0.1 [95%CI 0.03–0.39] for sensitivity, specificity, LR+, and LR-. Validity for DP was 81.3% [95%CI 63.56–92.79], 88.2% [95%CI 63.6–98.5], 6.91 [1.86–25.66], and 0.21 [95%CI 0.1–0.45] for sensitivity, specificity, LR+, and LR-. CONCLUSIONS The mathematical modeling of CP and DP is reliable and valid. This approach may enhance patient selection for future studies and serve as a clinical aid for practitioners.