Key Takeaways
Extended Takeaways
- The Brazilian Portuguese BSL-23 supported a stable 1-factor structure across 3 random subsamples, with factor 1 explaining 56.6%, 56.8%, and 55.5% of the variance; clinicians and researchers can interpret the total score as a coherent index of borderline symptom severity rather than as separate subscales.
- Reliability was consistently excellent across subsamples, with Cronbach α and McDonald ω both equal to 0.95 (95% CI, 0.95–0.96), and no item-rest correlation below 0.40, supporting repeated use when a brief self-report measure is needed.
- In confirmatory factor analysis, model fit improved after allowing correlated residuals for items 02 and 08 (r = 0.62, P < .001) and items 01 and 03 (r = 0.53, P < .001), suggesting some symptom pairs may cluster clinically even though the overall construct remains unidimensional.
- Gender comparisons appear usable in Brazilian samples: despite statistically significant item-level differential functioning on several items, the latent gender effect was small (λ = −0.07, P < .001) and the mean-group difference was also small (Cohen d = −0.18).
- The scale showed very strong separation between lower- and higher-severity groups, with Welch t1336.8 = 74.94, P < .001, Cohen d = 6.54; 95% CI, 6.17–6.91, and mean scores of 32.31 (SD = 4.28) versus 88.54 (SD = 11.33).
- Concurrent validity patterns suggest the BSL-23 is most aligned with emotional dysregulation and depressive distress, showing stronger correlations with DASS-21 depression (r=0.79, P<.001), DASS-21 stress (r=0.70, P<.001), and DERS total score (r=0.69, P<.001) than with impulsivity measures.
