Psychometric investigation of Liebowitz Social Anxiety Scale

Objective To assess the reliability and validity of the Chinese version of Liebowitz Social Anxiety Scale (LSAS) and its applicability in Chinese social anxiety disorder. Methods A multi-center study, including 167 cases of social anxiety disordor(SAD) and 587 cases of non-SAD, was carried out. LSAS, Hamilton Anxiety Scale (HAMA), the Severity Index (SI) and Global Improvement (GI) of Clinical Global Impression (CGI) were all used. The SAD group was scored by LSAS, HAMA and SI before treatment and after 8 weeks of treatment and by LSAS after 1 week of treatment. LSAS was administered by both clinician-scoring and self-scoring. The non-SAD group was measured with LSAS by clinician-scoring and one third of them also did self-scoring. Results 1 The internal consistency of the LSAS was quite good with item-total score if item deleted, with the correlation being 0.32-0.97 and the 4 subscale α above 0.9. 2 The rsults of scoring after 1 and 8 weeks of treatmen showed good correlation (P0.01). 3 The average score of total scale, total fear and total avoidance in SAD group were all higher than that of non-SAD group with statistical significance. 4 The scores of total fear and total avoidance were closely correlated with each other and with total scale in both clinician-rating and self-rating. 5 Good parallel validity and treatment validity were showed by the positive correlation of LSAS total score and HAMA, CGI-SI (r0.5, P0.01) and negative correlation of LSAS total score reduction before and after treatment and CGI-GI. 6Clinician-scoring and self-scoring were closely correlated (r0.89) but the average of self-scoring was higher than clinician-scoreing in SAD group (P0.05). 7 Cut-off of 38 in diagnosing SAD on the LSAS total score showed a sensitivity of 83.0% (85.5% for self-rating) and a specificity of 81.3%. Conclusion These findings support the use of the Chinese version of LSAS for social anxiety disorders among Chinese with good reliability and validity. It could be used for self-scoring. Cut-off of 38 of the total score represents the best balance of sensitivity and specificity.