Assessment and validation of prognostic models for poor functional recovery 12 months after whiplash injury: A multicentre inception cohort study

• Published in: Pain
• Main Authors: Sterling, Michele, Hendrikz, Joan, Kenardy, Justin, Kristjansson, Eythor, Dumas, Jean-Pierre, Niere, Ken, Cote, Julie, deSerres, Sophie, Rivest, Karine, Jull, Gwendolen
• Format: Article in Journal/Newspaper
• Language: English
• Published: Elsevier 2012
• Subjects: Physiotherapy; Canada; Brisbane; Iceland
• Online Access: http://hdl.handle.net/10072/60105; https://doi.org/10.1016/j.pain.2012.05.004

Uncertainty surrounds prognostic factors after whiplash injury. Previously we identified a prognostic model for 6-month pain-related disability in a cohort of 80 participants with acute whiplash. Predictors included initial disability, older age, decreased cold pain thresholds, decreased neck rotation movement, posttraumatic stress symptoms and decreased sympathetic vasoconstriction. The objective of this study was to externally validate this model. In a multicentre inception cohort study, 286 participants with acute whiplash (I, II or III) were assessed at <3 weeks and 12 months after injury. The Neck Disability Index (NDI) was the outcome. Observed and predicted NDI scores were generated using the published equation of the original model. Model discrimination between participants with no or mild disability from those with moderate to severe disability was examined by receiver operating characteristic curves. Initial NDI and cold pain threshold predicted current observed 12-month NDI scores (r2 = 0.50, 95% confidence interval 0.42 to 0.58). There was a significant site effect, and the estimated marginal mean ᠓E of 12-month NDI for Iceland (27.6 ᠱ.79%) was higher than the other 3 sites (Melbourne 11.2 ᠵ.03%, Canada 16.4 ᠲ.36%, Brisbane 16.8 ᠱ.17%). After adjusting for site, age and Impact of Events Scale scores regained significance (r2 = 0.56, 95% confidence interval 0.48 to 0.64). The tested model was not precise in predicting NDI as a continuous variable. However, it found good accuracy to discriminate participants with moderate to severe disability at 12 months (area under the receiver operating characteristic curve 0.89 [95% confidence interval 0.84-0.94], P < .001) which is clinically useful. No Full Text