La réponse thérapeutique à une intervention multidisciplinaire pour les maux de dos chroniques : prédicteurs et algorithmes
|Advisor:||Gauthier, Janel G.; Lacouture, Yves|
|Abstract:||Multidisciplinary programs for chronic back pain patients, while returning about 60% of patients at work, represent major financial investments. The capacity to predict therapeutic outcomes may constitute an important asset to optimize treatment response and achieve cost reductions. This thesis has two objectives. The first objective was to complete a metaanalytic review of the predictors of treatment outcomes following multidisciplinary programs for low back pain. A total of 87 studies were selected, comprising 73 predictors and 773 effect sizes related to three outcomes : severity of pain and disability following program, and return to work. Age and education, receiving disability payments, medication and health services usage, positive coping strategies, anxious traits, severity of pain, disability and role limitation, aerobic capacity, physical endurance and muscular strength, low familial support, and psychological and physical demands related to employment were found to be significant predictors of outcomes. No evidence supported the predictive value of gender, medical diagnosis, radiological abnormalities, physical characteristics, and nature of work in tertiary prediction. The second objective was to study conceptually and empirically predictive performance of four algorithms coming from statistical and data mining areas, such as linear and logistic discriminant analysis, multivariate classification trees, and radial basis neural networks. Strengths and limitations of each technique are discussed using simulated data. Predictive performance with medical data are discussed based on five large comparison studies. Available data support the choice of multivariate decision trees for predictive engine in a medical setting. The selection of reliable and significant predictors of treatment response to multidisciplinary programs, coupled with the use of effective predictive tools, will allow health professionals to better tailor their interventions according to patient profiles and expected benefits, in order to maximize treatment response in a context of limited clinical and financial resources.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||24 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.