A-MUPS score to differentiate patients with somatic symptom disorder from those with medical disease for complaints of non-acute pain.

PURPOSE

To develop a clinical score to discriminate patients with somatic symptom disorder (SSD) from those with medical disease (MD) for complaints of non-acute pain.

METHODS

We retrospectively examined the clinical records of consecutive patients with pain for a duration of ≥1 month in our department from April 2003 to March 2015. We divided the subjects according to the diagnoses of definite SSD (as diagnosed and tracked by psychiatrists in our hospital), probable SSD (without evaluation by psychiatrists in our hospital), matched MD (randomly matched two patients by age, sex, and pain location for each definite SSD patient), unmatched MD, other mental disease, or functional somatic syndrome (FSS). We investigated eight clinical factors for definite SSD and matched MD, and developed a diagnostic score to identify SSD. We subsequently validated the model with cases of probable SSD and unmatched MD.

RESULTS

The number of patients with definite SSD, probable SSD, matched MD, unmatched MD, other mental disease, and FSS was 104 (3.5%), 214 (7.3%), 197 (6.7%), 742 (25%), 708 (24%), and 978 (33%), respectively. In a conditional logistic regression analysis, the following five factors were included as independent predictors of SSD: Analgesics ineffective, Mental disorder history, Unclear provocative/palliative factors, Persistence without cessation, and Stress feelings/episodes (A-MUPS). The area under the receiver operating characteristic curve (AUC) of the model was 0.900 (95% CI: 0.864-0.937, <0.001), and the McFadden's pseudo--squared was 0.709. For internal validation, the AUC between probable SSD and unmatched MD was 0.930 (95% CI: 0.910-0.950, <0.001). The prevalence and the likelihood ratio of SSD increased as the score increased.

CONCLUSION

The A-MUPS score was useful for discriminating patients with SSD from those with MD for complaints of non-acute pain, although external validation and refinement should be needed.