OBJECTIVE. The goal was to test the hypothesis that participation in a community-based home-visiting program is associated with a decreased risk of infant death.
METHODS. A retrospective, case-control design was used to compare the risk of infant death among participants in Cincinnati's Every Child Succeeds program and control subjects matched for gestational age at birth, previous pregnancy loss, marital status, and maternal age. The likelihood of infant death, adjusted for level of prenatal care, maternal smoking, maternal education, race, and age, was determined with multivariate logistic regression. The interaction between race and program participation and the effect of home visiting on the risk of preterm birth were explored.
RESULTS. Infants whose families did not receive home visiting (n = 4995) were 2.5 times more likely to die in infancy compared with infants whose families received home visiting (n = 1665). Black infants were at least as likely to benefit from home visiting as were nonblack infants. No effect of program participation on the risk of preterm birth was observed.
CONCLUSION. The current study is consistent with the hypothesis that intensive home visiting reduces the risk of infant death.
The American Academy of Pediatrics encourages pediatricians to “support referral of high-risk parents to home-visitation programs as early as possible, ideally before or at the time of the prenatal visit to the pediatrician.”1 Prenatal and infancy home visitation seeks to optimize pregnancy outcomes and child development through family education, training, and social support. Programs often target mothers and children at greater risk for adverse outcomes.
In a series of controlled trials, Olds et al2,3 found that home visiting by nurses reduced smoking during pregnancy, decreased preterm birth rates for smokers, increased birth weights among adolescent mothers, and decreased rates of child abuse and accidental injuries in children. Studying the effects of home visitation on infant mortality rates requires a large sample size, which may be available only in larger programs implemented in community settings. Examination of such programs is valuable in its own right, given the challenges of implementing programs originally developed in the setting of a controlled trial and given the need to better understand the effects of larger programs taken to scale.4
The purpose of the current study was to examine the impact of a large-scale, community-based, home visitation program. We tested the hypothesis that participation in greater Cincinnati's Every Child Succeeds (ECS) program is associated with a decreased risk of infant death.
DESCRIPTION OF ECS PROGRAM
ECS is a community-based, home visitation program for first-time mothers and their children. ECS uses 2 national models of home visitation, namely, Nurse-Family Partnership (NFP)4 and Healthy Families America (HFA).5 Eligible mothers have ≥1 of 4 risk characteristics, that is, unmarried, inadequate income (up to 300% of poverty level, receipt of Medicaid, or reported concerns about finances), <18 years of age, or suboptimal prenatal care. Women are enrolled either during pregnancy (before 28 weeks for NFP) or before their child reaches 3 months of age (HFA only). Regular home visits are provided by social workers, child development specialists or related professionals (82%), trained nurses (12%), or paraprofessionals (6%). Home visits are made until the child reaches 2 years (NFP) or 3 years (HFA) of age, starting with weekly or more-frequent visits and tapering to fewer visits as the child ages. The goals of home visitation, as provided by ECS, are (1) to improve pregnancy outcomes through nutrition education and substance use reduction, (2) to support parents in providing children with a safe, nurturing, and stimulating home environment, (3) to optimize child health and development, (4) to link families to health care and other needed services, and (5) to promote economic self-sufficiency.
STUDY DESIGN AND METHODS
The study and use of Ohio birth certificates were approved by the Cincinnati Children's Hospital Medical Center and the Ohio Department of Health institutional review boards. A retrospective, case-control design was used to examine the impact of home visiting on the likelihood of death before 1 year of age among infants enrolled in the ECS program. Study infants were limited to those residing in the ECS service area, including infants born to mothers residing in 1 of 4 southwestern Ohio counties and born between January 1, 2000, and December 31, 2001, and infants born to mothers residing in 1 of 3 northern Kentucky counties and born between January 1, 2000, and December 31, 2002. Infant deaths were limited to individuals who were residing in Ohio or Kentucky at the time of death.
An intent-to-treat approach was used, in which all mothers enrolled during the study periods who had ≥1 home visit were included in the sample. ECS provided records of all 2308 eligible and enrolled women and their infants born during the study periods. Among potentially eligible ECS participants, 643 were excluded from the analysis data set because no matching birth certificate was identified (n = 615) or the birth occurred in another state (n = 28) (Fig 1). Because participants could be included only if the birth and death certificates for those specific infants could be identified, we used the electronic ECS database to compare selected characteristics that increase the risk of infant death between participants included in the study data set and excluded ECS participants. There were more married mothers among the excluded participants, but no other differences were observed (Table 1).
ECS records were linked to Ohio or Kentucky birth certificates by using common data fields. Because Ohio does not provide the mother's Social Security number on its electronic birth certificates, 4 fields in the ECS data set were potentially used for linkage with the appropriate Ohio birth certificate, namely, mother's name, mother's date of birth, child's name, and child's date of birth. Agreement in ≥2 fields was required for qualification as a match. For Kentucky residents, a hierarchical linking algorithm that used the mother's Social Security number in the first step was invoked. Because women might have had >1 birth during the study period, ECS records were linked to Kentucky births by using the Social Security number and birth year.
For linkage of Ohio cases to birth certificates, any combination of 2 or 3 fields was subjected to additional review. Additional variables were used for verification purposes. Some potential matches were rejected when, for instance, the birth certificate recorded a male birth but the ECS child was recorded as female. The number of previous live births is a birth certificate data item that was used as a check for finding the correct birth certificate for each ECS participant. Because ECS targets first-time mothers, a birth certificate identifying a second or higher-order birth was considered suspect. Whenever a potential match was rejected, the record was restored to both the ECS and state birth certificate data sets, so that it would be available for matching in subsequent combinations of fields.
For the linked ECS-Kentucky births, birth certificates were linked to electronic Kentucky or Ohio death certificates by using the birth certificate number that is included on the death certificate. If a child is born in Kentucky but dies elsewhere, then the birth certificate number field carries the number assigned by the birth state. For linked ECS-Ohio births, the electronic Ohio birth certificates were merged with an electronic linked birth-infant death file provided for infant deaths that occurred in 2000, 2001, or 2002. Because a linked birth-infant death file was not available for 2002 Ohio births, these could not be included in the analyses.
Three control subjects for each ECS study subject were selected randomly from among births to first-time mothers that occurred in the same time period to mothers residing in the ECS service area who were either <18 years of age or unmarried at the time of birth. Because ECS enrolls participants either during pregnancy or after birth, 2 different algorithms were used to select control subjects. Control subjects for infants enrolled during the pregnancy were selected randomly, without replacement, from among pregnancies that had not been delivered by the week of gestation of enrollment of the ECS subject. For example, if ECS enrolled a mother when the fetal gestational age was 26 weeks, then 3 control subjects were selected from among eligible infants not enrolled in ECS whose pregnancy duration was ≥26 weeks. For each ECS subject enrolled after birth, 3 control subjects were selected randomly, without replacement, by using 2 criteria, such that control subjects had both the same gestational age at birth and were alive at the same postmenstrual age as the postmenstrual age of enrollment of the index ECS subject. For ECS case subjects enrolled after birth because of poor prenatal care, the Kotelchuck index was used as an additional criterion for selecting control subjects. Therefore, control subjects for these ECS subjects were matched for gestational age at birth, postmenstrual age at ECS enrollment, and 1 of the 4 Kotelchuck index prenatal care categories (inadequate, intermediate, adequate, or adequate plus).6
The aforementioned selection of ECS and control subjects and all data analyses were performed by using SAS 9.1 (SAS Institute, Cary, NC). Adjusted comparisons of the likelihood of death before 1 year of age between ECS participants and control subjects were made by using multivariate logistic regression. Infants with trisomy 18, trisomy 13, anencephaly, or renal agenesis were not included in the study data set, because such anomalies are considered lethal and nonpreventable.
The primary independent variable was participation in ECS. Potential adjustment variables were selected from those available in the birth certificate file that are known to be associated with infant death, including mother's race (black versus not black, reflecting the race/ethnicity distribution for greater Cincinnati), maternal age <19 years, suboptimal prenatal care,6 single marital status at the time of birth (as reported by the mother), maternal smoking during the pregnancy, suboptimal maternal education (>1 year behind the expected grade level if <19 years of age or not a high school graduate if ≥19 years of age), fetal death in any previous pregnancy, and product of a multifetal gestation.
Although congenital malformations are associated with infant death, no deaths among infants in the study data set involved infants with congenital malformations. Mothers with absent data on smoking were coded as smokers. This had little effect, in the regression models, on the coefficients for either the effect of ECS or the effect of smoking on the likelihood of infant death. A final set of variables to be included in the initial regression analyses was determined on the basis of the total number of observed deaths, the distribution of each independent variable among study infants, and the strength of the previously reported association between the variable and the likelihood of infant death.
For ECS case subjects enrolled before birth, ECS might have influenced the likelihood of receipt of prenatal care. To evaluate the relationship between ECS participation and adequacy of prenatal care, we compared the parameter estimate for ECS in the final model with the parameter estimate for ECS in the same model excluding the prenatal care variable.
Because infant mortality rates in the United States are greater among black infants, compared with white infants,7 we used a parsimonious, multivariate, logistic model to calculate predicted probabilities for comparisons of risk of infant death among black infants participating in ECS, black control subjects, nonblack infants participating in ECS, and nonblack control subjects. Because one possible benefit of prenatal enrollment would be to reduce the risk of preterm birth, we used the same independent variables in a linear regression model of the association between prenatal ECS enrollment and gestational age at birth, including as observations only the ECS subjects and associated control subjects enrolled before delivery.
Among 1665 ECS case subjects included in the analyses, 715 (43%) were enrolled in ECS before birth, 950 (57%) were enrolled after birth, 1386 (83%) received home-visiting services based on the HFA model, and 279 (17%) received home-visiting services based on the NFP model. The mean ± SD gestational ages of ECS subjects enrolled prenatally in the HFA and NFP subsections of the ECS program were 29 ± 8 and 20 ± 7 weeks, respectively. ECS participants who were enrolled before birth and survived to 1 year of age had a mean of 6.9 home visits before birth and 18.3 visits between birth and 1 year of age. ECS participants who were enrolled after birth and survived to 1 year of age had a mean of 14.3 home visits before the age of 1 year. Characteristics of infants enrolled in ECS before birth, infants enrolled after birth, and their respective control subjects are displayed in Table 2. These characteristics reflect efforts by ECS to target high-risk women having their first child, that is, adolescent, unmarried, low income, or suboptimal education. The rate of maternal smoking among study participants was ∼3 times that reported for the overall population of mothers in the United States.8
On the basis of the number of deaths in the study sample, the final model to test the effect of ECS on the likelihood of death included 7 independent variables. Three models are displayed in Table 3. Model 1 includes the 9 proposed independent variables.
Because there were few infants of multifetal gestations and because the coefficient for the ECS variable did not change when this variable was removed, the multifetal gestation variable was not included in model 2. Because there were few married women and because the coefficient for the ECS variable did not change when this variable was removed, neither multifetal gestation nor marital status was included in model 3. The association between enrollment in ECS and the likelihood of infant death was the same in all models tested. After controlling for race, prenatal care, maternal smoking, maternal education, and maternal age, enrollment in ECS was associated with a 60% decrease in the likelihood of infant death. When inadequate prenatal care was removed as an independent variable from model 3, the parameter estimate for ECS decreased slightly, from 0.41 (Table 3) to 0.38 (model not shown), consistent with the possibility that ECS increased the adequacy of prenatal care.
The majority of deaths among study infants involved the ECS subjects and their respective control subjects enrolled before birth (Table 2). We were unable to identify different predictors for infants enrolled in ECS before birth and those enrolled after birth. Table 4 displays the results of multivariate linear regression analysis evaluating the association between ECS enrollment and gestational age at birth for ECS subjects and control subjects identified before delivery. No influence of ECS enrollment on gestational age at birth was observed. The difference between the adjusted predicted probabilities of death for black infants participating in ECS and black control subjects tended to be larger than the corresponding difference for nonblack infants (Fig 2).
In 2002, the US infant mortality rate ranked 24th worst among nations reporting to the World Health Organization.9 The health status of US residents overall and the health status of many US subgroups have improved remarkably in nearly all measures in recent decades. Many individuals enjoy better health, but large disparities persist among racial/ethnic, socioeconomic, and geographic subgroups. It is these disparities that, in very large part, explain the persistently low US ranking in international comparisons. For example, if the US infant mortality rate for black infants (14.4 deaths for every 1000 live births in 2002) were similar to that for white infants (5.8 deaths for every 1000 live births), then the US ranking among reporting United Nations member countries would increase from 24th to ∼7th. We report findings that support an association between intensive home visiting for high-risk, first-time, pregnant mothers and reduced likelihood of infant death. Nonparticipants were 2.5 times more likely to die in infancy, compared with those enrolled in ECS. These findings expand our understanding of the impact of home visitation and should guide future interventions designed to decrease infant mortality rates.
Causes of infant death can be categorized as possibly preventable or not preventable. The much higher infant mortality rate for US black infants, compared with white infants, suggests that some black deaths may be preventable. In the current study, the differences between black participants in ECS and nonparticipants tended to be much larger than differences observed for nonblack infants. Our analyses showed that adequacy of prenatal care had the strongest association with the likelihood of infant death. Black mothers enrolled in ECS before birth were more likely to receive adequate prenatal care, compared with control subjects (74% vs 60%), which might have been one of the determinants of the lower infant mortality rate seen in that group. Adequacy of prenatal care, as measured with the Kotelchuck index, includes both pregnancy month at initiation of prenatal care and number of prenatal visits. Because ECS could have no impact on initiation of prenatal care, we evaluated the study model by using month of initiation of prenatal care, rather than the entire Kotelchuck index, as a covariate. Therefore, any effect of the ECS program on the number of prenatal visits was embedded in the ECS variable. No change in the effect of ECS on the likelihood of infant death was observed (odds ratio: 0.43 vs 0.41).
We cannot make inferences about causation because of the retrospective, case-control design used in this study. However, it is encouraging that the largest association between ECS participation and reduced infant mortality rate was seen for black infants. It is not evident specifically how participation in ECS might result in reduced risk of infant death. The number of infant deaths in the current study does not allow careful analysis of age of death and cause of death. The largest contributors to infant death in the United States are preterm birth, complications of labor and delivery, congenital anomalies, sudden infant death syndrome (SIDS), infections, and injury. Except for congenital anomalies, home visiting theoretically could influence any of these death mechanisms. For women enrolled in ECS before delivery, improving the amount and content of prenatal care may decrease the risk of preterm birth, perinatal complications, and SIDS through maternal education and identification/amelioration of barriers to optimal care. For injuries (intentional and unintentional), infection, SIDS, and lasting complications of perinatal disease, home visiting may reduce the risk of death during infancy through similar mechanisms.
In well-designed randomized trials, home visiting has been shown to have important short-term and long-term benefits. In a randomized trial that compared provision of one half of indicated prenatal visits in the home by a nurse specialist and usual prenatal care for 173 high-risk pregnant women (94% black), the mean number of prenatal visits was increased and the home visit group had fewer preterm births and fewer combined fetal/infant deaths.10 Randomized trials of combined prenatal and infancy home visiting by nurses demonstrated decreases in reported and substantiated child abuse and neglect rates, fewer emergency department visits, fewer physician visits for treatment of accidents and poisonings, and healthier subsequent pregnancies.1,3,11 The average birth weight for mothers <17 years of age was 395 g greater than that for a comparison group with no home visits, and preterm delivery among smoking mothers was decreased by 75% with home visiting.2 In an uncontrolled study, home visiting among a cohort of high-risk women was associated with increased use of prenatal care.12 A prospective observational study suggested that pregnant, high-risk women who had 5 to 9 home visits during pregnancy had higher average birth weights, had fewer low birth weight infants, and were more likely to breastfeed.13 These studies suggest that home visiting, particularly during pregnancy, may help to reduce the risk of infant death by decreasing preterm birth, low birth weight, and child abuse/neglect. In the current study, home visiting was associated with better prenatal care, which was, in turn, associated strongly with decreased risk of infant death. Because less than one half of ECS families were enrolled before birth, we were unable to determine the extent to which prenatal home visits might have improved the adequacy of prenatal care.
Translating findings from randomized trials to the real world can be problematic.14 Carabin et al15 evaluated the effect of a statewide nurse home-visiting program on the risk of infant death by using Oklahoma birth and death certificates. Children of first-time mothers were compared with eligible nonparticipants. For high-risk pregnancies, participation was associated with decreased likelihood of birth at gestational age of <37 weeks and decreased infant mortality rates, with odds ratios very similar to those found in our study. Observed benefits of home visiting may depend on the qualifications of the home visitors, with larger effects being seen with nurses, compared with other types of home visitors.11 It is encouraging that use of home visitors with a variety of qualifications was associated with reduced infant mortality rates in the Cincinnati region. The current study has inadequate sample size for determination of the independent effects of the type of home visitor or the independent effects of the various ECS program components.
Confidence in inferences that can be made on the basis of retrospective case-control studies is limited. It is particularly difficult to address possible bias in the selection of control subjects. Inherent in the ECS enrollment process is the requirement that women initially agree to participate and the possibility that these women are inherently different from women who do not or cannot enroll. To help address this issue, we conducted a time-limited comparison of mothers who agreed to participate and those who refused to participate. This comparison of 625 mothers (agreed, n = 315; refused, n = 310) occurred over a 6-month period at birth hospitals in which all eligible mothers were approached and offered home visitation services. Mothers were compared with respect to 16 psychosocial risk factors (eg, social isolation, unemployment, and housing instability) used by many HFA programs. No significant differences between groups with respect to these risk factors or race were observed (P > .05), except that enrolled mothers were slightly younger than those who refused (18.7 vs 19.4 years).
Our study findings are consistent with the findings from randomized, controlled trials and suggest that home visiting reduces the risk of infant death. To identify and to characterize ECS participants, we used an administrative data set that was designed for management and evaluation of the ECS program, rather than scientifically rigorous research. We were unable to link a substantial portion (25%) of ECS participants to their respective birth and death certificates. It is possible that those ECS participants who were excluded from the study were systematically different from those who were included. However, we were unable to identify a rational explanation for such bias, and our comparison of risk characteristics for the included and excluded case subjects demonstrated no differences that would bias the study in a direction that favored ECS (Table 1). Although included ECS case subjects were less likely to be married (4% vs 17%), this would most likely bias the study in the direction of more-favorable outcomes among control infants. Moreover, marital status in our regression models had a weak association with the likelihood of infant death.
Infant death may be considered the “tip of the iceberg” in which the children of families at risk experience suboptimal care, poor health outcomes, and the possibility of lifelong disability; some die before their first birthday. The current study is consistent with the hypothesis that intensive home visiting reduces the risk of infant death.
This work was funded in part by the Children's Hospital Research Foundation (Cincinnati, OH) and the Maternal and Child Health Bureau (grant R40MC06632). ECS acknowledges the support of its founding partners: the United Way of Greater Cincinnati, Cincinnati-Hamilton County Community Action Agency, and Cincinnati Children's Hospital Medical Center.
- Accepted January 18, 2007.
- Address correspondence to Edward F. Donovan, MD, Child Policy Research Center, Cincinnati Children's Hospital Medical Center, 3333 Burnet Ave, MLC 7014, Cincinnati, OH 45229-3039. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵American Academy of Pediatrics, Council on Child and Adolescent Health. The role of home-visitation programs in improving health outcomes for children and families. Pediatrics.1998;101 :486– 489
- ↵Olds D, Henderson C, Tatelbaum R, Chamberlin R. Improving the delivery of prenatal care and outcomes of pregnancy: a randomized trial of nurse home visitation. Pediatrics.1986;77 :16– 28
- ↵Olds D, Henderson C, Chamberlin R, Tatelbaum R. Preventing child abuse and neglect: a randomized trial of nurse home visitation. Pediatrics.1986;78 :65– 78
- ↵World Health Organization. Annex Table 7: millenium development goals—selected health indicators in all WHO member states, 2000. In: The World Health Report: Changing History. Geneva, Switzerland: World Health Organization; 2004
- ↵Olds DL, Robinson J, Pettitt L, et al. Effects of home visits by paraprofessionals and by nurses: age 4 follow-up results of a randomized trial. Pediatrics.2004;114 :1560– 1568
- Copyright © 2007 by the American Academy of Pediatrics