BACKGROUND: Geographic variation may be an indicator of risk factors for birth defects. This study models the geographic distribution of three complex congenital heart defects (CHDs) in eastern Wisconsin, and evaluates effects of demographic census variables linked to geographic location. METHODS: Cases of Hypoplastic Left Heart Syndrome (HLHS), Tetralogy of Fallot (TOF) and d-Transposition of the Great Arteries (d-TGAs) born between1995 and 2004 were identified from three medical centers serving eastern Wisconsin. Case diagnoses were assigned by a pediatric cardiologist using echocardiographic records. Births by ZIP code were obtained from the State of Wisconsin. ZIP Code demographic variables were derived from 2000 census data. Numbers of cardiac defects by ZIP code were modeled using cluster analysis and Poisson generalized additive models (GAMs) for spatial coordinates including all and white only cases (excluding trisomies). GAM analyses were repeated adjusting for census variables. RESULTS: Four hundred forty-eight cases were ascertained. A significant south-to-north spatial gradient for HLHS, TOF, and combined CHDs, but not d-TGAs was identified. This gradient remained significant when census variables were included in the model for the full sample. In the analysis excluding non-white cases, findings were the same for TOF, combined CHDs, and d-TGAs. However, the geographic gradient for HLHS was not significant in the adjusted model. CONCLUSIONS: A south-to-north gradient was apparent for two of three complex CHDs in eastern Wisconsin. For white cases, demographic variation seems to explain some of this spatial gradient in HLHS.