Agreement between 'time-blind' and 'time-non-blind' assessments of depressive symptomatology

In a 4-week study, two methods were used simultaneously in the assessment of depressive symptomatology with videotaped structured clinical interviews: a 'time-non-blind' (TNB) method (chronological order, observer aware of the previous duration of drug treatment) and a 'time-blind' (TB) method (no chronological order, rater unaware of the previous duration of treatment). Sixty newly admitted depressed inpatients with Montgomery-Asberg Depression Rating Scale scores higher than 20 were assessed before (D0), after 10 days (D10) and after 28 days (D28) of antidepressant treatment. Agreement between TNB and TB methods on the Montgomery-Asberg Depression Rating Scale, measured by intra-class correlation coefficients, was good at D0 (0.68), excellent at D10 (0.81) and D28 (0.86), but not significantly different between D0, D10 and D28. The statistical method of Bland and Altman (1986) was also used to evaluate the degree of agreement. Results of this second analysis were in accordance with the intra-class correlation coefficient results, and showed significantly (P < 0.05) higher D0-D28 and D10-D28 intrasubject changes with the TB method, which were largely accounted for by some particular items (inner tension, pessimism, lassitude). With the Clinical Global Impression-Severity score, the Bland and Altman method failed to show significant differences between the two methods, and compared with the Montgomery-Asberg Depression Rating Scale, intra-class correlation coefficients were lower with larger confidence intervals, suggesting that global ratings are less reliable than itemized symptom ratings.