INFANT MORTALITY AND LOW BIRTH WEIGHT RATES COMPARED TO EXPECTED RATES BY COUNTY FOR FLORIDA 1999 AND 2000

By: Daniel Thompson, M.P.H.*, Melanie Simmons, M.S.*, Carol Graham, Ph.D.*

August 14, 2001

*Florida State University Center for Prevention and Early Intervention Policy
Consultant to Florida Department of Health

This document is also available in the Portable Document Format (PDF). If you have trouble viewing or printing the HTML version you may want to try the PDF version. Note: you may need to download Adobe Acrobat Reader to view or print the PDF version.

Introduction

Infant mortality and birth weight statistics are used extensively in public health. These statistics are especially useful because of their relevance as maternal and child health indicators and because of their ease of availability. These data are also virtually 100 percent complete since they are recorded for every birth and death that occurs in the state.

The purpose of this analysis is to identify geographic areas in the state where low birth weight (LBW) rates and infant mortality (IM) rates are statistically significantly higher than would be expected considering the unique demographics of each area. These areas should then be the focus of further, more detailed analyses to determine the reasons for the high rates and to develop intervention strategies for improving the outcomes.

IM and LBW rates vary in relation to the demographic characteristics of each of the 67 counties in Florida, and the variation in rates across the counties is due in part to the unique demographic characteristics of the county populations. In this analysis, adjustments are made to account for the differences in demographic characteristics. The adjusted statistics can then be compared across counties independently of the differences in demographic characteristics.

IM and LBW rates also reflect random variation. In this analysis, statistical methods are used to separate the random variation from the non-random variation, so rates that are significantly high are most likely a result of non-random influences. Likewise, rates that are higher than expected, but not significantly high, are likely to be the result of random variation and are said to be within the range of normal variation.

Methods

The data used in this analysis were extracted from the birth records for residents of Florida born in calendar years 1999 and 2000. Births were classified as LBW if the birth weight on the birth record was in the range 1 to 2499 grams. Three demographic variables were used in this analysis-mother's race, marital status and education. These are recorded on the birth record, and for the purposes of this analysis, two categories were used for each variable. Mother's race was classified as black or non-black, marital status was classified as married or not married, and mother's education was classified as 12th grade or higher completed or less than 12th grade completed. The three variables were then used to classify the births into eight mutually exclusive categories. Birth records with unknown values for any of the three variables were placed in a ninth category. There were roughly 1100 birth records in the ninth category for each year, or about one half of 1 percent of the resident births. The nine categories are as follows:

Using this classification, the category specific rates were calculated from the statewide totals, and these rates were used with the births in each county to calculate the expected LBW births and infant deaths. In this way the county expected statistics are adjusted for the three demographic characteristics and then used to calculate the adjusted rates. The term for this adjustment technique is indirect adjustment.

For example, if a county existed where all the births were in category 1, then the expected statistics for the county would be the same as the statewide statistics for category 1. Another county might have had births that were all in category 8. For this county, the expected statistics would be the same as the statewide statistics for category 8. These two hypothetical counties would have different expected statistics because they have populations with different demographic characteristics. If both counties had actual rates equal to the expected rates, they would be considered equal regarding the rates. Stated differently, both counties are doing equally well at preventing IM and LBW, considering their different demographic characteristics.

Results

The results of this analysis are shown in tables 1 through 4 and maps 1 through 4. In these tables, actual statistics are compared to expected statistics. The expected statistics are adjusted for the demographic characteristics in each county, as described in the methods section above. The maps display the results of the statistical tests for significance. Counties where the actual statistics are significantly higher or lower are shaded, as indicated by the legend on the maps.

Discussion

This analysis should be considered a preliminary step in the continuing endeavor to reduce risk of low birth weight and infant death in Florida. The rationale is to use the results of this analysis to focus further analysis and efforts on the areas where the risks are significantly high. Since adjustments were used to account for the differing demographic composition in each county, further analysis would focus on other factors such as smoking rates and mother's age at birth. The process becomes much more complicated at this point, and a separate analysis should be done for each area of concern.