Adjusting for confounding by neighborhood using a proportional odds model and complex survey data

Area of Science:

• Social Epidemiology
• Biostatistics
• Health Disparities Research

Background:

• Neighborhood environment is a key determinant of health behaviors and outcomes.
• Investigating health disparities requires accounting for unmeasured neighborhood factors.
• Existing methods for binary outcomes (conditional logistic regression) need generalization for ordinal outcomes and complex survey designs.

Purpose of the Study:

• To present a generalized statistical method for analyzing ordinal health outcomes in the presence of neighborhood-level confounding.
• To adapt conditional logistic regression for ordinal outcomes and complex sampling designs using a proportional odds model.
• To apply the method to assess racial/ethnic differences in dental preventative care, adjusting for neighborhood factors.

Main Methods:

• Developed a generalized conditional logistic regression approach based on a proportional odds model.
• Utilized SAS PROC SURVEYLOGISTIC for implementation, ensuring ease of use with standard software.
• Applied the method to the 2008 Florida Behavioral Risk Factor Surveillance System (BRFSS) data.

Main Results:

• The ordinal outcome, time since last dental cleaning, was analyzed.
• Individual-level confounders (gender, age, education, insurance) were adjusted for.
• Adjustment for neighborhood confounding (zip code) significantly altered the findings regarding racial/ethnic differences in dental care.

Conclusions:

• The generalized proportional odds model effectively adjusts for neighborhood confounding in social epidemiology.
• Neighborhood factors play a substantial role in explaining health disparities, as demonstrated in the dental care example.
• This method provides a valuable tool for researchers studying health disparities and the social determinants of health.