Zoran Bukumiric1, Vladan Starcevic2, Dejana Stanisavljevic3, Jelena Marinkovic3, Natasa Milic3, Slavica Djukic-Dejanovic4, Vladimir Janjic4, Aleksandar Corac5, Aleksandra Ilic5, Mirjana Kostic5, Ivan Nikolic3, Goran Trajkovic3. 1. Institute for Medical Statistics and Informatics, Faculty of Medicine, University of Belgrade, Serbia. Electronic address: zoran.bukumiric@mfub.bg.ac.rs. 2. University of Sydney, Sydney Medical School - Nepean, Discipline of Psychiatry, Sydney, Australia. 3. Institute for Medical Statistics and Informatics, Faculty of Medicine, University of Belgrade, Serbia. 4. Department of Psychiatry, Faculty of Medical Sciences, University of Kragujevac, Serbia. 5. Department of Public Health, Medical Faculty, University of Pristina, Kosovska Mitrovica, Serbia.
Abstract
BACKGROUND: Correlations between instruments measuring the same construct reflect their concurrent validity. Little is known about changes in correlations between such instruments employed in studies with repeated assessment. The aim of this meta-analysis was to examine the changes in correlations between depression instruments in the course of longitudinal studies. METHODS: A literature search was conducted using MEDLINE and PsycINFO for the period from 1960 to 2013. The total number of collected articles was 3723, of which 61 were included. Three meta-analyses were performed for the changes in correlations between each pair of the three depression scales: Hamilton Rating Scale for Depression (HAMD), Montgomery-Asberg Depression Rating Scale (MADRS) and Beck Depression Inventory (BDI). The effect size in these meta-analyses was obtained by the z-transformation of correlation coefficients. RESULTS: Correlations between depression scales increased over time in 52 studies. Significant changes in correlation coefficients were found for correlations between HAMD and BDI (p<0.001) and for correlations between HAMD and MADRS (p<0.001). An increase in correlations between the scales was associated with a decrease in depression scores and increase in their variability. LIMITATIONS: Univariable and multivariable meta-regression models were not obtained in all three meta-analyses because of the lack of data. CONCLUSIONS: A finding that correlations between depression instruments tended to increase over time has significant implications for assessment of the concurrent validity of these instruments. In longitudinal designs it is important to estimate correlations between depression scales over time because different thresholds for scale correlations indicate acceptable concurrent validity at different times.
BACKGROUND: Correlations between instruments measuring the same construct reflect their concurrent validity. Little is known about changes in correlations between such instruments employed in studies with repeated assessment. The aim of this meta-analysis was to examine the changes in correlations between depression instruments in the course of longitudinal studies. METHODS: A literature search was conducted using MEDLINE and PsycINFO for the period from 1960 to 2013. The total number of collected articles was 3723, of which 61 were included. Three meta-analyses were performed for the changes in correlations between each pair of the three depression scales: Hamilton Rating Scale for Depression (HAMD), Montgomery-Asberg Depression Rating Scale (MADRS) and Beck Depression Inventory (BDI). The effect size in these meta-analyses was obtained by the z-transformation of correlation coefficients. RESULTS: Correlations between depression scales increased over time in 52 studies. Significant changes in correlation coefficients were found for correlations between HAMD and BDI (p<0.001) and for correlations between HAMD and MADRS (p<0.001). An increase in correlations between the scales was associated with a decrease in depression scores and increase in their variability. LIMITATIONS: Univariable and multivariable meta-regression models were not obtained in all three meta-analyses because of the lack of data. CONCLUSIONS: A finding that correlations between depression instruments tended to increase over time has significant implications for assessment of the concurrent validity of these instruments. In longitudinal designs it is important to estimate correlations between depression scales over time because different thresholds for scale correlations indicate acceptable concurrent validity at different times.