|
Activity Number:
|
90
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 9:00 AM to 10:50 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #301064 |
|
Title:
|
Statistical Models for the Influence of Death of Spouse on the Depression Status of the Elderly Twins
|
|
Author(s):
|
Maria Iachina*+ and Ivan Iachine
|
|
Companies:
|
University of Southern Denmark and University of Southern Denmark
|
|
Address:
|
Epidemiology, Institute of Public Health, Campusvej 55, Odense, International, DK-5230, Denmark
|
|
Keywords:
|
depression symptoms ; death of spouse ; missing data ; multiple imputation ; GLM ; longitudinal data
|
|
Abstract:
|
The Longitudinal Study of Aging Danish Twins included information on depression symptoms which were combined with data on death of spouse between consequent interviews. The aim was to study the short- and long-term influence of death of spouse on depression status taking into account the serial and intrapair twin correlation and incompleteness of depression data due to drop-outs. A logistic regression analysis indicated that observed depression score was a strong predictor of future drop-out, suggesting the inappropriateness of the complete case analysis. To deal with these issues we developed a marginal generalized linear model which related the mean depression changes with lagged death of spouse data. The incomplete data problem was handled using multiple imputation. Standard errors were obtained using a clustered version of the Huber-White sandwich estimator. The analysis shows a significant effect of the death of spouse on the depression status of the surviving person, the depression score increases by about 10%. The average depression score returns to normal in about 12 months for both males and females. Alternative likelihood-based approaches are also discussed.
|
