Infant ID deaths occurring in the United States were examined using the 2008 and 2009 Period Linked Birth/Infant Death public use data files from the National Center for Health Statistics, Centers for Disease Control and Prevention.^14,15 Infant ID deaths were defined as singleton infants (<1 year of age) born in the United States to U.S. resident mothers with an ID International Classification of Diseases, 10^th Revision, (ICD-10) code listed as the underlying cause of death (UCOD) on the death record. ¹⁶ The ID ICD-10 codes used in this study were adapted from a previous classification of ID deaths using ICD-9 codes.¹⁷ Results are reported among all infant ID deaths unless the neonatal or postneonatal periods are specified.

In 2008 and 2009, 1.3% and 1.4%, respectively, of all infant death records could not be linked to the corresponding birth certificate.^14,15 To account for the unlinked death records, the weighted number of infant ID deaths was estimated using weights provided by the National Center for Health Statistics.^12–15 Maternal and infant characteristics that were comparable between the 1989 and 2003 birth certificate revisions were examined and included sex, birth weight (very low birth weight [VLBW, <1500 grams], moderately low birth weight [1500–2499 grams], LBW [<2500 grams] and normal birth weight [NBW, ≥2500 grams]), gestational age (<37 weeks and ≥37 weeks), 5-minute Apgar score (<7 and ≥7), maternal age (≤19, 20–24, 25–29, and ≥30 years), maternal marital status, maternal race (white, black, American Indian/Alaska Native [AI/AN], and Asian/Pacific Islander [A/PI]), maternal Hispanic origin, live birth order (first, second, third, and fourth or more), and number of prenatal visits (0, 1–2, 3–5, and ≥6).^14,15 Some characteristics had missing data and therefore did not sum to the total number of infant ID deaths; percentages calculated exclude missing data. Average annual ID infant mortality rates (IMRs) were calculated overall and by each characteristic as the weighted number of deaths per 100,000 live births of the corresponding group during 2008–2009. In order to compare age periods at death—since the neonatal (<28 days of age) period is shorter than the postneonatal period—neonatal and postneonatal IMRs were calculated as the weighted number of deaths in the age period per 100,000 person-years. Person-years were calculated from the total number of person-days that contributed to each age period. Poisson regression was used to calculate risk ratios (RRs) and 95% confidence intervals (CIs) within each characteristic.

The proportion of unweighted neonatal and postneonatal ID deaths and age at death in months were examined overall and by birth weight group. The median age at death in weeks and the interquartile range (IQR) were calculated overall and by birth weight group; the Wilcoxon rank-sum test¹⁸ was used to statistically compare age at death in weeks between LBW and NBW infants.

The proportion of weighted infant deaths due to ID was calculated. The most frequently listed ID UCODs on the death certificate were examined overall and by birth weight group, and the proportion of neonatal deaths among LBW infants with a most frequently listed ID death was calculated. The number and percent of ID death records with bacterial sepsis of newborn (ICD-10 code P36) as the UCOD or with prematurity (P07) listed anywhere on the death record were examined overall and by birth weight group. The proportion of neonatal deaths among all deaths due to bacterial sepsis was calculated.

A retrospective case-control study was conducted to determine infant and maternal characteristics associated with infant mortality due to IDs among LBW and NBW infants. The cases for each birth weight group were the unweighted infant ID deaths. Controls for each birth weight group were selected from singleton infants of the corresponding birth weight group who survived to the end of their birth year and were born in the United States to U.S. resident mothers using a simple random sample to obtain a 1:1 ratio of cases to controls.

Univariate analysis was conducted for each of the birth weight groups using logistic regression.¹⁹ Odds ratios (ORs) and 95% CIs were calculated for sex, 5-minute Apgar score, maternal age (years), race, maternal marital status, Hispanic origin, and live birth order. Significant characteristics (P<0.10) in univariate logistic regression, along with the interaction of these characteristics with maternal race, were included as potential predictors in the initial LBW and NBW multivariable logistic regression models. The interaction terms were included to further examine maternal race in relationship with these characteristics. Hierarchical backward elimination was then used to determine significant characteristics and interactions within the multivariable models for each birth weight group. Characteristics and interaction terms were retained in the LBW and NBW multivariable models at the P<0.05 significance level. Adjusted ORs and 95% CIs were calculated for the significant characteristics within each of the final LBW and NBW multivariable regression models.

An estimated 3,843 infant ID deaths occurred in the United States during 2008 and 2009, accounting for 8.3% of all infant deaths. Among these deaths, 2,665 (69.5%) were LBW infants and 1,170 (30.5%) were NBW infants (Table 1). More than 60% of the infant ID deaths occurred during the neonatal period; this percentage was higher for LBW infants (72.7%) compared to NBW infants (34.3%, P<0.01; Figure 1). The overall median age of ID death among infants was two weeks (IQR: 0–8 weeks). LBW infants had a lower age of death (median: 1 week, IQR: 0–4 weeks) than NBW infants (median: 7 weeks, IQR: 2–20 weeks; P<0.0001).

A large proportion of ID deaths were due to bacterial sepsis of newborn (28.3% overall, Table 2) with 96% of these deaths occurring in the neonatal period. The proportion of ID deaths due to bacterial sepsis of newborn was higher in LBW infants (34.1%) than NBW infants (15.3%; P<0.01). Bacterial sepsis of newborn, unspecified was the most frequently listed UCOD among LBW (27.5%) and NBW (10.8%) infants. Among LBW infants, the other most frequently listed UCODs were newborn affected by chorioamnionitis (27.0%); septicemia, unspecified (9.2%); infection specific to the perinatal period, unspecified (3.1%); and congenital pneumonia, unspecified (2.1%). The top five UCODs represent 68.9% of LBW infant deaths; 84.7% of these were for neonates. The other most frequently listed UCODs among NBW infants were pneumonia, unspecified (7.4%); bronchopneumonia, unspecified (6.9%); septicemia, unspecified (5.8%); and congenital herpes viral [herpes simplex] infection (5.6%). Overall, 2,294 (59.7%) of the infant ID deaths had prematurity (ICD-10 code P07) listed on their death record. Among LBW infant ID deaths, 84.8% had prematurity listed on their death record while 2.3% of NBW infant ID deaths had prematurity listed on their death record.

The overall ID IMR was 47.5 deaths per 100,000 live births (Table 1). The neonatal mortality rate was higher than the postneonatal mortality rate (RR: 19.7, 95% CI: 18.4–21.0). The IMR was higher for male infants, infants with a 5-minute Apgar score <7, infants with a gestational age <37 weeks, infants born to an unmarried mother, infants born to young mothers (<19 and 20–24 compared to 25–29 years), infants with a live birth order of first, third, or fourth or more compared to second, infants of black race or of AI/AN race compared to white race, and infants whose mothers had fewer prenatal visits (0, 1–2, and 3–5 compared to ≥6). The IMR for LBW infants was much higher (RR: 33.3, 95% CI: 31.1–35.7) than that for NBW infants and was highest for VLBW infants (RR: 160.3, 95% CI: 149.4–172.1). The IMR was lower for Hispanic infants and infants of A/PI race compared to white race.

The case-control study included infants whose 2008–2009 ID death record was linked to the corresponding birth certificate (2,633 LBW infants and 1,158 NBW; Table 3). In the univariate analyses, all characteristics were significantly associated with ID death in logistic regression for NBW infants, and all except for Hispanic origin were significant in logistic regression for LBW infants. Male sex, 5-minute Apgar score <7, young maternal age (≤19 and 20–24 years), black race, unmarried maternal marital status, and live birth order of fourth or more were positively associated with ID death in both LBW and NBW infants (Table 3). For NBW infants, AI/AN race was positively associated with ID death whereas having Hispanic origin and maternal age ≥30 years were negatively associated with ID death. For LBW infants, A/PI race was negatively associated with ID death compared to white race.

In the multivariable analyses, all characteristics that were included in the initial logistic regression models were significant except for maternal marital status in the model for LBW infants and maternal race in the model for NBW infants (Table 4). Maternal race did not interact with any of the characteristics included in the initial LBW and NBW multivariable regression models. Male sex, live birth order of fourth or more, young maternal age (<19 and 20–24 years), and low 5-minute Apgar score were positively associated with ID death in both LBW and NBW infants (Table 4); however, the effect of 5-minute Apgar score was greater for LBW infants than NBW infants. For LBW infants, black race was associated with increased odds of ID death. Among NBW infants, maternal age of ≥30 years, a live birth order of first, and Hispanic origin were associated with decreased odds of ID death whereas being born to an unmarried mother was associated with increased odds of ID death.