External validation of prognostic models for recovery in patients with neck pain.
Roel W Wingbermühle, Martijn W Heymans, Emiel van Trijffel and 3 others
PMID 34301471WHAT IT FOUND
Two prognostic models for predicting neck pain recovery failed external validation.
They could not accurately distinguish between patients who would recover and those who would remain disabled. Clinicians should not use these specific models to guide patient care or trial stratification.
Key findings
01The models demonstrated poor discriminative performance, with AUC values ranging from 0.43 to 0.54, indicating they could not reliably distinguish between patients who recovered and those who did not.
02Calibration was poor for both models, with significant deviations from the ideal intercept of 0 and slope of 1, meaning predicted risks did not match observed outcomes.
03The authors conclude that clinical use of these models cannot be advocated and that no useful models are currently available for predicting neck pain outcomes in primary care.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The trauma subset for the Amodel had too few events (24 recovered, 9 with moderate/severe disability) for stable estimates, leading to wide confidence intervals. Substantial missing data required multiple imputation, though sensitivity analyses supported the main findings. Some predictor variables from the original models (e.g., EuroQoL, hyperarousal subscale) were missing or proxied in the validation cohort. The validation cohort included patients with non-traumatic neck pain, whereas the original Amodel was derived from whiplash patients, introducing potential case-mix differences.
Declared interests
The authors declare no conflicts of interest.
The easy way to misread this
Do not interpret the poor performance as evidence that prognosis cannot be predicted at all. It means these specific models failed. The authors suggest that individual prognostic factors like baseline pain and disability may still be useful, but the models combining them were not accurate.