Related Experiment Videos
Regression analysis of multiple-source longitudinal outcomes: a "Stirling County" depression study.
Constantine Daskalakis1, Nan M Laird, Jane M Murphy
1Biostatistics Section, Division of Clinical Pharmacology, Thomas Jefferson University, Philadelphia, PA 19107, USA. c_daskalakis@lac.jci.tju.edu
American Journal of Epidemiology
|January 5, 2002
Summary
This study introduces a multivariate logistic regression method for analyzing psychiatric disorder data from multiple sources and repeated assessments. The method was applied to depression data, revealing differences in sex ratios and low agreement between diagnostic schedules.
Area of Science:
- Epidemiology
- Psychiatric Research
- Biostatistics
Background:
- Epidemiologic studies of psychiatric disorders increasingly require multiple data sources and longitudinal assessments for diagnostic validity.
- Existing methods may not adequately handle complex data structures with repeated measures and missing outcomes.
Purpose of the Study:
- To present a general multivariate logistic regression method for simultaneous analysis of discrete outcomes from multiple sources and repeated assessments.
- To apply this method to analyze depression data from the Stirling County study, examining risk factors and diagnostic agreement.
Main Methods:
- Development and application of a multivariate logistic regression model.
- Analysis of data from 631 subjects assessed twice over 3-4 years using the DePression and AnXiety schedule (DPAX) and the Diagnostic Interview Schedule (DIS).
- Simultaneous analysis of risk factors and outcome agreement within a unified framework, accommodating missing data.
Main Results:
- The female:male ratio for depression differed between DPAX (0.8) and DIS (1.6).
- Education showed an inverse association with depression; time, age, and interviewer sex had null effects.
- Agreement between DPAX and DIS was low, but DPAX stability over time was significantly higher than DIS stability.
Conclusions:
- The proposed multivariate logistic regression method effectively analyzes complex longitudinal psychiatric data.
- The findings highlight significant differences in depression assessment between DPAX and DIS, with implications for diagnostic validity and reliability.
- The method's ability to handle missing data and analyze agreement enhances its utility in psychiatric epidemiology.