LBPredict

This study aims at investigating possible causal links between measures of pain-related fear, whole body lifting strategies and LBP incidence using a supervised machine learning algorithm.

Fiche signalétique

  • Département responsable Santé
  • Institut Physiothérapie
  • Unité de recherche Biomécanique de la colonne vertébrale
  • Organisation d'encouragement Autres
  • Durée (prévue) 01.06.2021 - 31.07.2024
  • Responsable du projet Dr. Stefan Schmid
  • Direction du projet Dr. Stefan Schmid
  • Équipe du projet Dr. Stefan Schmid
    Christian Bangerter
    Michael L. Meier
  • Partenaire Universitätsklinik Balgrist
  • Mots-clés Low back pain, Object lifting, Spinal kinematics, Fear-avoidance beliefs, Artifical intelligence

Situation

Low back pain (LBP) is a major global health problem. To prevent LBP, health care professionals often promote that lifting should only be done with a straight back. Although this notion is widely accepted in general population, recent evidence has suggested, that lifting without bending the spine could even predispose individuals to back trouble. Indeed, the roles of pain-related fear and lifting behavior in the development of LBP remain unclear.

Approche

We will enroll 150 healthy and pain-free supermarket employees for this prospective observational cohort study. Pain-related fear and whole body movement strategies during object lifting will be assessed at baseline using self-reports (questionnaires) as well as a portable strain gauge-based measurement system and regular video recordings. For LBP incidence, we will conduct bi-weekly follow-up assessments over one year using an online questionnaire. In addition, we will assess LBP-related parameters such as duration and level of pain, disability and work absence for secondary analyses. Predictive modeling will be performed using a Support Vector Machine algorithm, which will be trained and tested using a randomly selected sub-sample of 100 of our 150 participants. To validate the model, we will conduct out-of-sample predictions using the data of the remaining 50 participants.