Familial aggregation of a disease is important for studying the genetic etiology of a disease. A useful measure of family aggregation is recurrence risk (RR), the conditional probability of disease given a family member is diseased. For SRS of families with varying sizes, quadratic exponential models (QEM) have been used to estimate RR. Household health surveys with (family) network sampling in which sampled individuals report their disease status and disease statuses of specified relatives, are useful for estimating disease prevalence and RR of disease. This paper extends QEM’s for RR’s to household network sample surveys with complex sample designs involving sample weighting and cluster sampling of observations. In addition, inferential methods are developed for comparing RR’s and related parameters in the QEM among multiple family relationships and across family-level covariates (e.g., race) after adjustment for individual-level covariates (e.g., age) using propensity score weighting. Significance level and power of hypotheses tests based on derived Wald and Quasi-score tests are evaluated using simulations. Our methods are illustrated using the 1976 NHIS diabetes data set.