Depression and mortality: Artifact of measurement and analysis?

Appleton, K., Woodside, J.V., Yarnell, J.W.G., Kee, F., Evans, A., Patterson, C.C., Arveiler, D., Haas, B., Amouyel, P., Montaye, M., Ferrieres, J., Ruidavets, J.B., Bingham, A. and Ducimetiere, P., 2013. Depression and mortality: Artifact of measurement and analysis? Journal of Affective Disorders, 151 (2), 632 - 638 .

Full text available as:

Appleton JAD 2014.pdf - Accepted Version


DOI: 10.1016/j.jad.2013.07.010


Background Previous research demonstrates various associations between depression, cardiovascular disease (CVD) incidence and mortality, possibly as a result of the different methodologies used to measure depression and analyse relationships. This analysis investigated the association between depression, CVD incidence (CVDI) and mortality from CVD (MCVD), smoking related conditions (MSRC), and all causes (MALL), in a sample data set, where depression was measured using items from a validated questionnaire and using items derived from the factor analysis of a larger questionnaire, and analyses were conducted based on continuous data and grouped data. Methods Data from the PRIME Study (N=9798 men) on depression and 10-year CVD incidence and mortality were analysed using Cox proportional hazards models. Results Using continuous data, both measures of depression resulted in the emergence of positive associations between depression and mortality (MCVD, MSRC, MALL). Using grouped data, however, associations between a validated measure of depression and MCVD, and between a measure of depression derived from factor analysis and all measures of mortality were lost. Limitations Low levels of depression, low numbers of individuals with high depression and low numbers of outcome events may limit these analyses, but levels are usual for the population studied. Conclusions These data demonstrate a possible association between depression and mortality but detecting this association is dependent on the measurement used and method of analysis. Different findings based on methodology present clear problems for the elucidation and determination of relationships. The differences here argue for the use of validated scales where possible and suggest against over-reduction via factor analysis and grouping. CrownCopyright © 2013PublishedbyElsevierB.V.Allrightsreserved.

Item Type:Article
Group:Faculty of Science and Technology
ID Code:21349
Deposited By: Unnamed user with email symplectic@symplectic
Deposited On:15 Jul 2014 09:18
Last Modified:10 Sep 2014 14:57


Downloads per month over past year

More statistics for this item...
Repository Staff Only -