Colloquium

Department of Statistics

Statistics Colloquium
September 21 (Fri) 11 a.m.
Alavi Commons Room, 6625 Everett Tower

Making the Most of the Data Available to Answer Important Public Health Questions

Catherine Kothari
Senior Investigator, WMU School of Medicine
and graduate student, PhD Program in Interdisciplinary Health Sciences

Background and Objectives: Over the last several decades, as overall infant mortality has declined in the United States, the gap between Blacks and Whites has remained significant, with Black infants dying at 3.5 times the rate of White infants, right here in Kalamazoo County: 18.7 Black and 5.3 White infant deaths per 1,000 live births. This is a critical public health issue and studies which investigate predictors and effective interventions are much needed. It is common within public health research to try and draw causal and predictive conclusions when we only have access to quasi-experimental, observational data. This presentation will demonstrate three different methodological approaches to the issue of racial disparities in birth outcomes within Kalamazoo County. The focus of the first two analyses is whether participation in Healthy Start (HS), a home-visitation, case-management program is associated with improved birth outcomes for Black women. The third analysis examines the association of race and racial segregation with birth outcomes at the individual maternal level as well as at the neighborhood level.

Methods: All three study approaches are based upon secondary data analyses of countywide birth records databases. Supplementary information has been assembled from multiple sources and integrated into these datasets. The presentation will detail the data collection and database assembly process. In the first analysis, race and HS participation were predictors in two linear regression models for gestational age and birthweight. In the second analysis propensity-score matching was used to develop a comparison group and analyses, stratified by race, were completed using GLM repeated measures. In the final analysis, geographic information system (GIS) analysis was utilized to define racially segregated neighborhoods in Kalamazoo County after geo-coding birth records addresses. Logistic regression modeling was then completed at the individual level and linear regression at the tract level for each of the four birth outcome variables (death, VLBW, LBW and prematurity). Currently we are preparing to conduct multilevel modeling of this dataset.

Results: Interactively with colloquium participants, results from these analyses will be shared in this presentation for discussion of interpretations and conclusions. Also shared will be comments from journal reviewers and discussion of methods for addressing their concerns.

All statistics students are expected to attend.

Past colloquiums