Background & Aims
Interdisciplinary pain rehabilitation is a promising approach for many patients suffering from chronic pain, but not all see improvement. A model predicting health outcomes after interdisciplinary pain rehabilitation could improve targeting of treatment. The aim of this study was to demonstrate a modelling approach capable of predicting change simultaneously in several health measures, after interdisciplinary pain rehabilitation programs, using predictors before treatment. The model should explicitly model estimation and prediction uncertainty. The model could be implemented as a support tool when selecting patients to be included in pain rehabilitation programs.
Methods
Five outcome measures for chronic pain patients were selected, in accordance with the literature: SF-36 mental and physical component summaries, pain intensity during the past week (NRS), and the HADS anxiety and depression subscales . The predictive model was based on multivariate Bayesian linear regression. The multivariate approach reduces estimation uncertainty in the outcomes. The Bayesian method allows for explicit modeling of prediction uncertainty. This as predictions are given as predictive distributions rather than point predictions of change. A posterior predictive distribution describes the probabilities of different outcomes for some patient according to the model, accounting for the uncertainty regarding model parameters.
A study sample of 8168 individuals from the Swedish Quality Registry for Pain Rehabilitation was used in estimation. Model predictors were selected by a stepwise procedure comparing model AIC across different sets of predictors.
Results
Several models with different amounts of covariates were created. The most important covariates were baseline levels for all outcomes, but predictive performance improved for models including more covariates. Posterior predictive distributions of outcomes were created for several subjects, demonstrating the model’s ability to generate interpretable output that takes uncertainty into account and to account for several outcome dimensions simultaneously.
Conclusions
This study demonstrates a multivariate modeling approach for predicting several important health outcomes after interdisciplinary pain rehabilitation. The model explicitly accounts for uncertainty in estimation and prediction and can produce a probability distribution for the health outcomes of a patient, in addition to a single predicted most likely value. Such a model could be an important tool in patient selection for interdisciplinary pain rehabilitation and for communicating with patients regarding likely outcomes of rehabilitation.
References
•Turk DC, Dworkin RH, Allen RR, Bellamy N, Brandenburg N, Carr DB, Cleeland C, Dionne R, Farrar JT, Galer BS, Hewitt DJ, Jadad AR, Katz NP, Kramer LD, Manning DC, McCormick CG, McDermott MP, McGrath P, Quessy S, Rappaport BA, Robinson JP, Royal MA, Simon L, Stauffer JW, Stein W, Tollett J, Witter J. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. Pain 2003;106:337–345.
•Kaiser U, Kopkow C, Deckert S, Neustadt K, Jacobi L, Cameron P, De Angelis V, Apfelbacher C, Arnold B, Birch J, Bjarnegård A, Christiansen S, C de C Williams A, Gossrau G, Heinks A, Hüppe M, Kiers H, Kleinert U, Martelletti P, McCracken L, de Meij N, Nagel B, Nijs J, Norda H, Singh JA, Spengler E, Terwee CB, Tugwell P, Vlaeyen JWS, Wandrey H, Neugebauer E, Sabatowski R, Schmitt J. Developing a core outcome domain set to assessing effectiveness of interdisciplinary multimodal pain therapy: the VAPAIN consensus statement on core outcome domains. Pain 2018;159:673–683.
•Björn Gerdle, Peter Molander, Gunilla Stenberg, Britt-Marie Stålnacke, and Paul Enthoven. “Weak outcome predictors of multimodal rehabilitation at one-year follow-up in patients with chronic pain—a practice based evidence study from two SQRP centres”. In: BMC Musculoskeletal Disorders 17.1 (Nov. 2016)
•Geoffrey Thompson. MixMatrix: Classification with Matrix Variate Normal and t Distributions. R package version 0.2.6. 2021.
•Andrew Gelman. Bayesian Data Analysis Third Edition 3rd Edition. Chapman Hall CRC Texts in Statistical Science. CRC Press, 2013. isbn: 9781439840955.
Presenting Author
Rode Grönkvist
Poster Authors
Rode Grönkvist, MSc
MSc
Inst of Medicine, Sahlgrenska Academy, Gothenburg University
Lead Author
Linda Vixner
PhD
University of Dalarna
Lead Author
Björn Äng
Dalarna University
Lead Author
Anna Grimby
School of Public Health and Community Medicine, Department of Med., Gothenburg Univ
Lead Author
Topics
- Evidence, Clinical Trials, Systematic Review, Guidelines, and Implementation Science