300 likes | 404 Views
Infant Race at Birth and Death as Recorded on Minnesota Vital Records, 1990-1999. Wendy L. Hellerstedt, MPH, PhD Associate Professor School of Public Health Division of Epidemiology Maternal and Child Health Program. Introduction: Race identification.
E N D
Infant Race at Birth and Deathas Recorded on Minnesota Vital Records, 1990-1999 Wendy L. Hellerstedt, MPH, PhD Associate Professor School of Public Health Division of Epidemiology Maternal and Child Health Program
Introduction: Race identification • Race is often thought of as a social--not a genetic--factor • Self-identified race is acceptable in public health research • Some studies have shown that race “changes” for the same individual from survey to survey--sometimes depending on mode or type of survey administration
Race on birth and death records • Twin Cities: one of the highest rates of births to parents of different races • Census 2000, MN: 1.7% checked more than one race (in U.S., 2.4%) • Racial misidentification on vital records could be related to multiple race identity, clerical error, or other biases related to self-report • Racial identification is important because federal and state funds are directed at health indices that disproportionately affect groups defined by race or ethnicity
Coding race on birth certificates • Since 1989, NCHS started tabulating U.S. birth data primarily by race of mother because: • increased incidence of births to parents from different races: 5.3% in 2000 • large proportion of births with no data about father: 14% in 2000 • If race of mother is not known, father’s race is used, if known. Race of mother was not known for 0.5% births in 2000
Race and infant mortality • Infant mortality rates in the U.S.--and within states--are at least twice higher among African-Americans and American Indians compared with whites • Hahn, et al. (JAMA, 1992) examined U.S. birth-death records and found inconsistencies in infant race at birth and death low for whites (1.2%) and greatest for races other than white or black (43%). Their work preceded the NCHS rule that defined infant race as mother’s race.
Purpose • Examine infant race at birth--using the NCHS coding rules for identification of births by maternal race--and race as coded on the death certificate • Examine the maternal and paternal characteristics of infants who have different races at birth and death
Methods • Minnesota linked birth-death files, 1990-1999 • Birth certificates available for >99% all resident births • 4302 deaths to infants who died within first 365 days of life • Birth certificates included maternal and paternal race, age, and education. Also includes some prenatal health data for mother • Missing data were recorded as such and kept in analyses
Definitions of race at birth • Maternal race • If maternal race is missing, paternal race • If both races missing, code infant race as missing • Note: MN (and U.S.) don’t report “infant” race; data are reported by maternal race (and paternal race if maternal data missing).
Infant race at death • Reported by next of kin to medical examiner or coroner (all infant deaths autopsied) • Minnesota does not impute missing race data
Analysis • Descriptive • Multivariate, stepwise logistic regression to determine correlates of inconsistent race coding
Products of analyses • Sample characteristics • Differences between race coding at birth and death • Correlates of differences
Maternal and paternal age of Minnesota infants who died, 1990-1999, n = 4302 • 78.7% records had maternal and paternal data for age • 63.6% of mothers and fathers were within 5 years of age • 15.1% of the fathers were 5 years or more older than mother
Maternal and paternal adequacy of education for Minnesota infants who died, 1990-1999, n = 4302 • Adequacy of education defined as at least high school completion for adults. For adolescents, an algorithm employing age and years completed used. • 81% of fathers had adequate educations • 64% of mothers had adequate educations
Maternal and paternal race for Minnesota infants who died, 1990-1999, n = 4302 • 73% same • 4% different • 23% missing data • 66% parents both white • 7% both not white
Of the 3440 whites at birth: 94.8% white at death 2.4% black at death 0.7% AI at death 0.3% Asian at death 1.8% missing Of the 3362 whites at death: 97% white at birth <1% each black, AI, or Asian 1% missing at birth Race coding of white infants
Of the 435 blacks at birth: 95.2% black at death 2.8% white at death < 1% each AI, Asian 1.4% missing Of the 513 blacks at death: 80.7% black at birth 16.0% white at birth <1% each AI or Asian at birth 2.0%missing Race coding of black infants
Of the 180 American Indians at birth: 81.7% Am Ind at death 14.4% white at death 2.0% black 0.5% Asian 1.7% missing Of the 181 American Indians at death: 81.2% Am Ind at birth 13.3% white at birth 1.1% black at birth 1.1% Asian at birth 3.3% missing Race coding of American Indian infants
Of the 183 Asians at birth: 85.3% Asian at death 9.8% white at death 2.2% black at death 1.1% AI at death 1.6% missing Of the 171 Asians at death: 91.2% Asian at birth 6.4% white at birth 0.5% each black and AI at birth 1.2% missing Race coding of Asian infants
Variables associated (Adjusted odds ratios and 95% C.I.s) with different infant races at birth and death • Race of mother (compared with white): • Black: OR = 0.10 (0.05, 0.2) • Race of father (compared with white): • Black: OR = 49.8 (29.2, 85.3) • American Indian: OR = 15.1 (8.1, 27.1) • Asian: OR = 8.6 (4.8, 14.7) • Missing: OR = 22.1 (12.9, 38.7)
Variables associated (Adjusted odds ratios and 95% C.I.s) with different infant races at birth and death • Age of mother (compared with older than 30 yrs): • 15-19 yrs: OR = 0.7 (0.5, 0.9) • Age of father (compared with older than 30 yrs): • 20-24 yrs: OR = 1.4 (0.9, 2.2)
Variables associated (Adjusted odds ratios and 95% C.I.s) with different infant races at birth and death • Race of parents (compared with same): • Different: OR = 10.5 (7.2, 13.6) • Race missing, one or both: OR = 5.0 (3.2, 7.6) • Education of parents (compared with both adequate): • Education missing, one or both: OR = 0.6 (0.3, 0.9)
Variables not associated with racial coding inconsistency • Low birthweight • Maternal tobacco use • Adequate prenatal care • Compared with white: American Indian mother and Asian mother • Discordance in parental education • Discordance in parental age
So what? • Depending on whether race is coded by mother or at death will have different estimates of race-specific correlates of death
Correlates of death for black infants • If use maternal race, 7.7% of low birthweight infants, 5.1% of preterm infants, and 1.9% of infants born to teen mothers died • If use race at death, 12.0% of low birthweight infants, 5.7% of preterm infants, and 2.3% of infants born to teen mothers died
Correlates of death for American Indian infants • If use maternal race, 9.8% of low birthweight and 4.2% of preterm infants died • If use race at death, 10.5% of low birthweight and 5.1% of preterm infants died
Summary • There were 18% more blacks at death than birth, 7% fewer Asians, 1% more American Indians, and 2% fewer whites • While inconsistency was found in both directions, in general, more infants were identified as non-white at death than they were at birth
Numbers at birth and death vs. inconsistent coding of an individual • 82% of American Indians and 85% Asians were coded the same at birth and death • 95% blacks and 95% whites were coded the same on both records • Note: In Minnesota Census 2000, 1/3 of American Indians also listed another race--most common among children
Conclusions and Limitations • It is not clear which certificate is “valid” for race • Tip of the iceberg: only 1-2% of infants die, thus we don’t know how stable the racial coding is for the majority who survived the first year of life • Inconsistencies on birth and death records provide more evidence that classifying race is difficult • Will allowing multiple race coding on vital records help or hinder enumerating race? • Would a more inclusive definition of social disparities, encompassing various measures of class, race, ethnicity, income, and geography, ultimately be more useful in serving disadvantaged individuals?