Conditional pseudolikelihood methods for clustered ordinal, multinomial, or count outcomes with complex survey data

Abstract

In order to adjust individual-level covariate effects for confounding due to unmeasured neighborhood characteristics, we have recently developed conditional pseudolikelihood methods to estimate the parameters of a proportional odds model for clustered ordinal outcomes with complex survey data. The methods require sampling design joint probabilities for each within-neighborhood pair. In the present article, we develop a similar methodology for a baseline category logit model for clustered multinomial outcomes and for a loglinear model for clustered count outcomes. All of the estimators and asymptotic sampling distributions we present can be conveniently computed using standard logistic regression software for complex survey data, such as sas proc surveylogistic. We demonstrate validity of the methods theoretically and also empirically by using simulations. We apply the new method for clustered multinomial outcomes to data from the 2008 Florida Behavioral Risk Factor Surveillance System survey in order to investigate disparities in frequency of dental cleaning both unadjusted and adjusted for confounding by neighborhood.

Adjusting for confounding by neighborhood using generalized linear mixed models and complex survey data

“…When investigating health disparities, it can be of interest to explore whether adjustment for unmeasured socioeconomic factors at the neighborhood level can account for, or even reverse, an unadjusted difference. Recently, we have been investigating racial and ethnic disparities in dental preventive care using complex survey data from the 2008 Florida Behavioral Risk Factor Surveillance System (BRFSS) survey [1,2]. We have needed to develop new statistical methods to simultaneously overcome two hurdles.…”

Neighborhood Contributions to Racial and Ethnic Disparities in Obesity Among New York City Adults

“…The rationale for the second set of analyses was that (1) a specific neighborhood effect can be obtained only by means of conventional multilevel modeling, and (2) estimates from multilevel modeling are likely to be biased when complex sample design is taken into account. 11,12 Even if results from the second set of analyses cannot be generalized to all NYC adults, they allow us to examine the magnitude of an unweighted contribution of each neighborhood factor to individual-level racial and ethnic disparities in obesity.…”

“…Then we compared the estimated odds ratios (ORs) for race/ethnicity with those estimated by using the conditional pseudolikelihood method for complex survey data, which enabled us to fully account for measured and unmeasured neighborhood confounding as well as complex sample design. 11,12 The purpose of this first set of analyses was to estimate the entire contribution of the neighborhood factors to racial/ethnic obesity disparities. In the conditional pseudolikelihood method, individuals living in the same neighborhood were treated as matched samples, requiring the use of the conditional pseudolikelihood.…”

“…11,12 Furthermore, it controls for full confounding by neighborhood, including both observed and unobserved neighborhood-level factors. 11,12 We applied this method to obtain unbiased estimates of obesity disparities by race/ethnicity in NYC after adequately accounting for neighborhood confounding and complex sample design.…”

“…11,12 Furthermore, it controls for full confounding by neighborhood, including both observed and unobserved neighborhood-level factors. 11,12 We applied this method to obtain unbiased estimates of obesity disparities by race/ethnicity in NYC after adequately accounting for neighborhood confounding and complex sample design. In addition, to address the limitations in current research, we estimated the total neighborhood effect on racial/ethnic disparities in obesity and assessed the extent to which this effect was attributed to each of 3 neighborhood characteristics (neighborhood walkability, residential segregation, and neighborhood poverty).…”