Predictive modeling toward refinement of behavior-based pain assessment in horses

• Published in: Applied animal behaviour science 2023-10, Vol.267, p.106059, Article 106059
• Main Authors: Trindade, Pedro Henrique Esteves, Barreto da Rocha, Paula, Driessen, Bernd, McDonnell, Sue M., Hopster, Klaus, Zarucco, Laura, Gozalo-Marcilla, Miguel, Hopster-Iversen, Charlotte, Rocha, Thamiris Kristine Gonzaga da, Taffarel, Marilda Onghero, Alonso, Bruna Bodini, Schauvliege, Stijn, Mello, João Fernando Serrajordia Rocha de, Luna, Stelio Pacca Loureiro
• Format: Article
• Language: English
• Subjects: Algorithm; animal behavior; Body language; confidence interval; horses; Logistic regression; orthopedics; pain; Pain assessment; regression analysis; surgery; tissues; Welfare
• ISSN: 0168-1591; 1872-9045
• DOI: 10.1016/j.applanim.2023.106059

After 25 years of studies on methodologies for behavioral assessment of equine pain, the Unesp-Botucatu Horse Acute Pain Scale (UHAPS) and the Orthopedic Composite Pain Scale (CPS) were recently considered suboptimal instruments to assess pain in hospitalized horses. However, the combination of the two instruments has never been examined. The objective was to investigate whether the merging, mining, and weighting of UHAPS and CPS behavioral items in a single instrument using a predictive model could improve the capacity to diagnose pain in horses. A previously video-collected behavioral database of 42 horses admitted to three different hospitals for orthopedic or soft tissue surgery was used. Multilevel binomial logistic regression models were used to merge, mine, and weight the behaviors of both instruments. The classification quality between the model and the instruments was compared by the area under the curve (AUC) and its 95% confidence interval. The short model containing 25% of the behaviors of the two instruments showed a higher AUC (98.64 [98.16 – 99.12]; p