Roel W Wingbermühle1, Emiel van Trijffel2, Paul M Nelissen3, Bart Koes4, Arianne P Verhagen4. 1. SOMT University of Physiotherapy, Amersfoort; Erasmus University Medical Centre, Rotterdam, The Netherlands. 2. SOMT University of Physiotherapy, Amersfoort; Vrije Universiteit Brussel, Brussels, Belgium. 3. SOMT University of Physiotherapy, Amersfoort. 4. Erasmus University Medical Centre, Rotterdam, The Netherlands.
Abstract
QUESTION: Which multivariable prognostic model(s) for recovery in people with neck pain can be used in primary care? DESIGN: Systematic review of studies evaluating multivariable prognostic models. PARTICIPANTS: People with non-specific neck pain presenting at primary care. DETERMINANTS: Baseline characteristics of the participants. OUTCOME MEASURES: Recovery measured as pain reduction, reduced disability, or perceived recovery at short-term and long-term follow-up. RESULTS: Fifty-three publications were included, of which 46 were derivation studies, four were validation studies, and three concerned combined studies. The derivation studies presented 99 multivariate models, all of which were at high risk of bias. Three externally validated models generated usable models in low risk of bias studies. One predicted recovery in non-specific neck pain, while two concerned participants with whiplash-associated disorders (WAD). Discriminative ability of the non-specific neck pain model was area under the curve (AUC) 0.65 (95% CI 0.59 to 0.71). For the first WAD model, discriminative ability was AUC 0.85 (95% CI 0.79 to 0.91). For the second WAD model, specificity was 99% (95% CI 93 to 100) and sensitivity was 44% (95% CI 23 to 65) for prediction of non-recovery, and 86% (95% CI 73 to 94) and 55% (95% CI 41 to 69) for prediction of recovery, respectively. Initial Neck Disability Index scores and age were identified as consistent prognostic factors in these three models. CONCLUSION: Three externally validated models were found to be usable and to have low risk of bias, of which two showed acceptable discriminative properties for predicting recovery in people with neck pain. These three models need further validation and evaluation of their clinical impact before their broad clinical use can be advocated. REGISTRATION: PROSPERO CRD42016042204. [Wingbermühle RW, van Trijffel E, Nelissen PM, Koes B, Verhagen AP (2018) Few promising multivariable prognostic models exist for recovery of people with non-specific neck pain in musculoskeletal primary care: a systematic review. Journal of Physiotherapy 64: 16-23].
QUESTION: Which multivariable prognostic model(s) for recovery in people with neck pain can be used in primary care? DESIGN: Systematic review of studies evaluating multivariable prognostic models. PARTICIPANTS: People with non-specific neck pain presenting at primary care. DETERMINANTS: Baseline characteristics of the participants. OUTCOME MEASURES: Recovery measured as pain reduction, reduced disability, or perceived recovery at short-term and long-term follow-up. RESULTS: Fifty-three publications were included, of which 46 were derivation studies, four were validation studies, and three concerned combined studies. The derivation studies presented 99 multivariate models, all of which were at high risk of bias. Three externally validated models generated usable models in low risk of bias studies. One predicted recovery in non-specific neck pain, while two concerned participants with whiplash-associated disorders (WAD). Discriminative ability of the non-specific neck pain model was area under the curve (AUC) 0.65 (95% CI 0.59 to 0.71). For the first WAD model, discriminative ability was AUC 0.85 (95% CI 0.79 to 0.91). For the second WAD model, specificity was 99% (95% CI 93 to 100) and sensitivity was 44% (95% CI 23 to 65) for prediction of non-recovery, and 86% (95% CI 73 to 94) and 55% (95% CI 41 to 69) for prediction of recovery, respectively. Initial Neck Disability Index scores and age were identified as consistent prognostic factors in these three models. CONCLUSION: Three externally validated models were found to be usable and to have low risk of bias, of which two showed acceptable discriminative properties for predicting recovery in people with neck pain. These three models need further validation and evaluation of their clinical impact before their broad clinical use can be advocated. REGISTRATION: PROSPERO CRD42016042204. [Wingbermühle RW, van Trijffel E, Nelissen PM, Koes B, Verhagen AP (2018) Few promising multivariable prognostic models exist for recovery of people with non-specific neck pain in musculoskeletal primary care: a systematic review. Journal of Physiotherapy 64: 16-23].
Authors: Lene Aasdahl; Fredrik Granviken; Ingebrigt Meisingset; Astrid Woodhouse; Kari Anne I Evensen; Ottar Vasseljen Journal: BMC Musculoskelet Disord Date: 2021-05-19 Impact factor: 2.362
Authors: Peter Kent; Carol Cancelliere; Eleanor Boyle; J David Cassidy; Alice Kongsted Journal: BMC Med Res Methodol Date: 2020-06-29 Impact factor: 4.615
Authors: Isabel R A Retel Helmrich; Ana Mikolić; David M Kent; Hester F Lingsma; Laure Wynants; Ewout W Steyerberg; David van Klaveren Journal: Diagn Progn Res Date: 2022-05-05
Authors: Rob A B Oostendorp; J W Hans Elvers; Emiel van Trijffel; Geert M Rutten; Gwendolyne G M Scholten-Peeters; Marcel Heijmans; Erik Hendriks; Emilia Mikolajewska; Margot De Kooning; Marjan Laekeman; Jo Nijs; Nathalie Roussel; Han Samwel Journal: Patient Prefer Adherence Date: 2020-03-02 Impact factor: 2.711