Heterogeneity in the Birth Cohort and Mixture Models

Comparisons of birth-weight-specific infant mortality indicate that low-birth-weight African American infants have lower mortality than low-birth-weight European American infants despite higher infant mortality overall—the ‘‘pediatric paradox.’’ One explanation is heterogeneity in birth weight. The shape of the birth weight distribution with its heavy tails also indicates heterogeneity. Analyses of African American and European American births suggest that birth cohorts consist of two heterogeneous subpopulations. A Guassian mixture model of two populations appears to fit the data very well. One subpopulation appears to account for the majority of births in a “normal” range (the primary subpopulation), whereas the other may consist of compromised births (the secondary subpopulation). The mixture model is shown in the figure below. The black line represents the total density. The red line is the density of the primary subpopulation and the blue line is the density of the secondary subpopulation. The rug plot displays the density of the actual data.


Using a mixture model also helps us to understand the differences in infant mortality between populations. Estimates of infant mortality indicate that the compromised subpopulation has higher overall mortality but lower birth-weight-specific mortality. We attribute lower birth-weight-specific infant mortality in the compromised subpopulation to higher rates of fetal loss in that population. Compared to European American birth cohorts, African American birth cohorts have (1) higher birth-weight-specific mortality in the normal subpopulation, (2) larger compromised subpopulations, and (3) lower birth-weight-specific mortality in the compromised subpopulation. Consequently, the pediatric paradox is attributable to greater rates of compromised pregnancies and higher fetal losses among African Americans.

  back to top