Aim: The aim of this study was to update a previously developed clinical prediction model and to perform external validation of the updated model using a new sample from a different setting. The prediction model aimed to predict recovery from pain for patients with acute low back pain (LBP) who continue to have pain approximately 1-week after initially seeking care.
Methods: The study validation sample comprised 737 patients with acute LBP, with a pain score of ≥2/10, 1-week after initially seeking care and with duration of current episode of <4 weeks. The primary outcome measure was days to recovery from pain. Five predictor variables (duration of current episode, number of previous episodes, depression, pain intensity and pain intensity change over the first week) were included in the final model of the original development study. Three of these variables were categorized differently in the validation data set (duration of current episode, number of previous episodes, and depression). Hence, we needed to modify the coding for these three variables in the development dataset to derive an updated model. The performance of the updated model was assessed for discrimination and calibration in the validation sample.
Results: The performance of the updated prediction model was good and the discrimination (C-statistic = 0.76, 95% CI, 0.70-0.82) was equivalent to the original model. Applying the updated model coefficients and baseline survivor function to the validation data resulted in a C-statistic of 0.71 (95% CI, 0.63-0.78). The calibration for the validation sample was acceptable at 1-month. However, at 1-week the predicted proportions tended to overestimate the observed recovery proportions and at 3-months the predicted proportions tended to underestimate the observed recovery proportions.
Conclusions: The updated prediction model demonstrated reasonably good external validity and may be useful in practice, but further validation and impact studies should be conducted.