This study examined the geographic variation of infant mortality using zip codes as the unit of analysis in order to facilitate site location and site specific health services for community health centers and/or mobile health units. The relationship between infant mortality and the place of residence by zip code was analyzed using factors related to residential deprivation and selected maternal-infant factors. The design of this study was non-experimental retrospective correlational using a mix of ecological and individual variables. The sample chosen included all births and associated infant deaths from 1988 through 1992 in 22 of the 64 parishes in Louisiana. The 22 sample parishes contained 192 zip codes. Variable data was extracted from linked birth and death files and 1990 Census tapes. Chi-square was used to test variation of infant mortality among and within parishes by zip codes. Correlation and multivariate regression analysis were used to examine the association of six maternal/infant factors and five residential deprivation factors, separately and then combined, with infant mortality. A geographic information system was also used to formulate thematic maps for presentation of data. Results indicated that the mother's marital status was the most significant predictor of infant mortality; being married was associated with decreased infant mortality. Identification of significant predictors varied on an individual basis. The results also indicate that infant mortality does vary significantly within each parish. Overall use of this model in order to locate healthcare services or in predicting infant mortality by zip code is questionable, but may be useful in those parishes that have large populations and exhibit greater internal variability and segregation in general
