Marginal approach in analyzing sequential cross-sectional oral health survey data

Abstract

Objectives: To apply marginal approach in analyzing sequential cross-sectional survey data to adjust for the possible cluster effect of birth cohorts by generalized estimating equations (GEE). Methods: Data from 4330 respondents aged 15 years+ in 9 birth cohorts collected in 4 sequential cross-sectional surveys of non-institutionalized Danes from 1975-2005 were analyzed. The key study variables were edentulism (yes vs. no) and seeking dental care on an annual basis (ADC, regularly at least once a year in the preceding 5 years vs. not regularly every year). For the analysis of edentulism, only respondents aged 35 years+ were included. Survey year, age, gender, socio-economic status (SES) group and ADC were considered as the independent factors. For the analysis of ADC, survey year, age, gender, SES group and denture wearing were considered. To adjust for the cluster effect of birth cohorts, logistic regressions with an exchangeable correlation structure in GEE were adopted with the use of PROC GENMOD in the SAS software.

Results: Overall in the population, edentulousness decreased from 24% in 1975 to 3.8% in 2005 while proportion of people seeking dental care on an annual basis increased from 58.8% to 86.7%. All the five factors in each logistic regression were significant (P<0.05). Females, older age group, respondents in earlier survey years, not seeking ADC, and lower SES group were associated with higher probability of being edentulous (P<0.05). Females, respondents in higher SES group, in recent survey years, and with no denture or just lower denture were associated with higher probability of seeking ADC (P<0.05). Danes aged 15-24 years had significantly higher probability of seeking ADC than adults aged up to 54 years (P<0.05).

Conclusion: With the use of GEE, the cluster effect of birth cohorts that may exist in sequential cross-sectional oral health survey data could be appropriately adjusted.