[Gender-sensitive epidemiological data analysis: methodological aspects and empirical outcomes. Illustrated by a health reporting example].

Geschlechtergerechte epidemiologische Datenanalyse: Methodische Aspekte und empirische Befunde. Dargestellt an einem Beispiel aus der Gesundheitsberichterstattung.

Publisher: Springer Country of Publication: Germany NLM ID: 101181368 Publication Model: Print Cited Medium: Print ISSN: 1436-9990 (Print) Linking ISSN: 14369990 NLM ISO Abbreviation: Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz Subsets: MEDLINE

Original Publication: Berlin : Springer,

Data Interpretation, Statistical* ; Epidemiologic Methods* ; Health Surveys* ; Prejudice*; Data Collection/*statistics & numerical data; Adult ; Aged ; Cross-Sectional Studies ; Decision Trees ; Empiricism ; Female ; Germany ; Guidelines as Topic ; Humans ; Male ; Middle Aged ; Models, Statistical ; Risk Factors ; Sex Factors ; Smoking/epidemiology

In Germany gender-sensitive approaches are part of guidelines for good epidemiological practice as well as health reporting. They are increasingly claimed to realize the gender mainstreaming strategy in research funding by the federation and federal states. This paper focuses on methodological aspects of data analysis, as an empirical data example of which serves the health report of Bremen, a population-based cross-sectional study. Health reporting requires analysis and reporting methods that are able to discover sex/gender issues of questions, on the one hand, and consider how results can adequately be communicated, on the other hand. The core question is: Which consequences do a different inclusion of the category sex in different statistical analyses for identification of potential target groups have on the results? As evaluation methods logistic regressions as well as a two-stage procedure were exploratively conducted. This procedure combines graphical models with CHAID decision trees and allows for visualising complex results. Both methods are analysed by stratification as well as adjusted by sex/gender and compared with each other. As a result, only stratified analyses are able to detect differences between the sexes and within the sex/gender groups as long as one cannot resort to previous knowledge. Adjusted analyses can detect sex/gender differences only if interaction terms have been included in the model. Results are discussed from a statistical-epidemiological perspective as well as in the context of health reporting. As a conclusion, the question, if a statistical method is gender-sensitive, can only be answered by having concrete research questions and known conditions. Often, an appropriate statistic procedure can be chosen after conducting a separate analysis for women and men. Future gender studies deserve innovative study designs as well as conceptual distinctiveness with regard to the biological and the sociocultural elements of the category sex/gender.

Date Created: 20080111 Date Completed: 20080409 Latest Revision: 20160707

20240829

10.1007/s00103-008-0415-y

18185965