The categorisation of the Short Warwick Edinburgh Mental Wellbeing Scale scores: an exploration from an...

There has been a notable increase in the utilisation of SWEMWBS as a measure of mental well-being globally. To enhance its interpretability for both healthcare professionals and laypeople, categorising SWEMWBS scores is considered beneficial. Two approaches have been recommended, yet they have not been thoroughly investigated. This study aimed to explore the categorisation of the scores from an epidemiological perspective.

Adopting PHQ-9 results, as suggested by the original scale author, to be the benchmarking comparator and employing the epidemiological approach, the concordance between SWEMWBS and PHQ-9 was examined using data from a health survey. The scales were categorised following the recommended cutoffs suggested by the authors. An additional cutoff was generated from the Nonparametric Receiver Operating Characteristic (ROC) Analysis and verified using the multiclass ROC analysis. The agreement indicators, including the sensitivity, specificity, Positive Predictive Value, Negative Predictive Value, Likelihood Ratio Positive, Likelihood Ratio Negative, and Receiver Operating Characteristic-Area Under the Curve (ROC AUC), were calculated.

The categorisation of SWEMWBS scores by benchmarking yielded the highest sensitivity, but the smallest specificity with 86.1% (95% C.I. = 84.1%-87.9%) and 56.6% (95% C.I. = 49.1%- 63.9%) respectively. Categorisation using the mean and SD approach resulted in a sensitivity of 81.3% (95% C.I. = 79.1%- 83.4%) and a specificity of 68.1% (95% C.I. = 60.8%-74.8%). In contrast, categorisation using the ROC analysis approach provided a sensitivity of 76.5% (95% C.I. = 73.8%-79.0%) and a specificity of 77.5% (95% C.I. = 70.7%-83.3%). The ROC AUC values were moderately low with the largest being 0.769 (95% C.I. = 0.737-0.802).

The concordance of the Chinese version of the SWEMWBS has been examined using PHQ-9 as the benchmarking comparator. The results indicate moderate sensitivity, specificity, LR+, and LR- values.