Data from national surveillance shows significant regional variation in the respiratory syncytial virus (RSV) season. To enable accurate recommendations for surveillance and prophylaxis at a local level, we examined this variation in detail and determined predictors for the duration of the RSV season. Statewide hospitalization data for the years (1996-2006) were summarized by postal zip code and linked to the corresponding Colorado Zip Code Tabulation Area (ZCTA). Demographic data from the 2000 Census and geographic variables created using geographic information systems were merged with hospitalization data, using the ZCTA as the common unit of analysis. Univariate and multivariate logistic regression analysis were used to determine predictors of the duration of the RSV season in a ZCTA. A 2-year quasi periodic (biennial) pattern was seen for RSV hospitalizations in Colorado. Significant variability was observed for the duration of the RSV season between ZCTA's and within the same ZCTA during 10 seasons. ZCTA's with longer season lengths tended to be urban crowded with larger families, with a higher proportion of children less than 5 years and Hispanics. The number of children younger than 5 years per room and per square km, and the urban/rural character of a ZCTA were found in logistic regression to be significant predictors of season length (P < 0.05). Demographic factors influence the duration of the RSV season. Indices of overcrowding and population density as well as being urban or rural can be used to stratify the duration of the RSV season, establish targeted surveillance, and tailor prophylaxis in an area.