The program of research reported here was guided by both epidemiology and developmental theory. Many prevention programs fail because they are not based on a thorough understanding of: (i) the risk characteristics of the targeted population; (ii) the nature and developmental pathways of the negative outcomes they intend to prevent and the positive outcomes they intend to promote; and (iii) the mechanisms, based on sound theory and evidence, through which they expect their programs to produce behavioral change (Olds & Kitzman, 1993). This paper describes the empirical and theoretical foundations upon which this program of research was founded; the design of the program itself; and the research designs, methods and findings from each trial. In the final section, we examine the policy implications of the findings and describe an emerging effort to disseminate the program model and study the process of dissemination itself. We start with its grounding in epidemiology and developmental research.

II. DEVELOPMENTAL EPIDEMIOLOGY

This program of research has been firmly grounded in epidemiology and theories of child development. Sheppard Kellam has referred to the integration of these disciplines in guiding prevention science as developmental epidemiology (Kellam and Werthammer-Larsson, 1986). On reflection, the current program of research was guided by this integration of epidemiology and a developmental perspective on maternal and child functioning.

In planning the original Elmira trial, we noted that although the problems identified above cut across all segments of U.S. society, they were more common among children born to poor, teenage, and single parents. This observation led to our decision to focus recruitment on women bearing first children who were either teenaged, unmarried, or from low-income families, although any pregnant woman bearing a first child was accepted into the study in order to avoid creating a program stigmatized because it served only the poor. Given that the beneficial effects of the Elmira program (described below) were concentrated on women who were unmarried and from low-income families, we modified the sampling designs in Memphis and Denver to focus more exclusively on low-income women, the vast majority of whom were unmarried and teenaged. Each of the trials focused on women who had no previous live births because we reasoned that offering them services during the transition to parenthood would increase their receptivity to offers of help. Moreover, from the standpoint of a public health strategy, this approach held the promise of improving the life chances of subsequent children, because parents were hypothesized to have better skills for managing the demands of pregnancy and early care of the child after they had been helped with their first child. In addition, to the extent that the rates of rapid successive births were reduced, parents would be able to focus their caregiving resources on a smaller number of children.

Selected maternal and child health problems were identified as potential targets for a comprehensive, intensive preventive intervention. This original work was directed toward improving the outcomes of pregnancy, parents’ caregiving skills (and the corresponding health and development of the child), and the early life course of the mothers. More recently, we have begun to consider the potential influence of prenatal and infant preventive intervention to produce improvements in long-term outcomes such as women’s completed family size, economic self-sufficiency, rates of child abuse and neglect over the course of a family’s childrearing years, and maternal and youth involvement in the criminal justice system.

In designing the program, we reviewed the literature to determine behavioral and contextual conditions that were consistently correlated with the adverse maternal and child outcomes that we wished to affect. We analyzed the literature to determine the extent to which these variables were most likely to be causally related to the outcomes of interest and which were simply markers for maladaptive functioning. Those that were hypothesized to be causally related to the outcome of interest and that were potentially modifiable with social and behavioral interventions became primary candidates for targeted interventions to reduce the rates of adverse outcomes identified for prevention. Theory played an important role in helping us integrate the epidemiologic data into a coherent developmental framework regarding both the proximity of risk to adverse outcome as well as the developmental progression of maladaptive functioning. It is important to note that the epidemiologic evidence indicated that some of the problems targeted for prevention early in the program were also risks for problems targeted later. This is best illustrated by reference to Figure 1.

A. Modifiable Risks for Low Birthweight, Preterm Delivery, and Fetal Neurodevelopmental Impairment

Epidemiologic evidence on risks for low birthweight indicates that, in developed countries, prenatal exposure to tobacco is an established determinant of compromised fetal growth (Kramer, 1987) and, to a lesser extent, shortened length of gestation (Kramer, 1987). Similarly, prenatal tobacco exposure increases children’s likelihood of neurodevelopmental impairments associated with compromised intellectual functioning (Olds, Henderson, & Tatelbaum, 1994a, 1994b) and behavioral problems (Olds, 1997) such as Attention Deficit-Hyperactivity Disorder (Milberger, Biederman, Faraone, Chen, & Jones, 1996), and Conduct Disorder (Wakschlag, Lahey, Loeber, Green, Gordon et al., 1997). Similar risks are posed by prenatal exposure to alcohol (Streissguth, Sampson, Barr, Bookstein, & Olson, 1994), marijuana (Fried, Watkinson, & Dillon, 1987), and other illegal drugs, such as cocaine (Mayes, 1994). While the evidence on these risks was not as coherent at the start of this series of trials 20 years ago as it is today, we chose to promote a reduction in use of all of these substances as a precaution.

In addition to the risks posed by maternal use of these harmful substances during pregnancy, the epidemiologic evidence indicated that other prenatal behaviors, such as inadequate weight gain (Institute of Medicine, 1990), inadequate diet (Institute of Medicine, 1990), inadequate use of office-based prenatal care (Klein & Goldenberg, 1990), and unattended obstetric complications, such as genitourinary tract infections and hypertensive disorders (Klein & Goldenberg, 1990) increased the risk for low birthweight, preterm delivery, and compromised neurologic development.

Moreover, there was some suggestion that children with compromised neurodevelopmental functioning would be more difficult for their parents to nurture (Parke & Collmer, 1975). Newborns whose mothers smoked cigarettes during pregnancy display higher rates of neurobehavioral disturbance -- reduced habituation to a variety of stimuli, lower arousal, increased tremulousness, weaker suck, longer latency to suck, reduced autonomic regulation, reduced orientation to auditory stimuli, and cries with higher pitches and other altered characteristics suggestive of neurodevelopmental perturbations compared to offspring of women who did not smoke (Fried & Makin, 1987; Jacobson, Fein, Jacobson, Schwartz, & Dowler, 1984; Makin, Fried, & Watkinson, 1991; Nugent, Lester, Greene, Wieczorek-Deering & O’Mahony, 1996; Picone, Allen, Olsen & Ferris, 1982; Richardson, Day & Taylor, 1989; Streissguth, Barr & Marti 1994; Streissguth, Barr & Martin, 1983). These effects remain after control for the newborns’ exposure to other possible toxicants, such as alcohol and other adverse maternal behaviors (Olds, 1997). Such children are more likely to present caregiving challenges to their parents and to increase their risk for being abused or neglected (Parke & Collmer, 1975).

B. Modifiable Risks for Child Abuse and Neglect and Injuries to Children

In planning this study, we made an explicit inventory of risks for child abuse and neglect and chose to design the intervention in a way that would reduce those risks. The risks for child abuse and neglect were organized according to their levels of immediate proximity to parental behavior. At a proximal level, risk assessment focused on the mother’s psychological immaturity and mental health problems that affect parents’ feelings and competencies in caring for their infants (Newberger & White, 1990; Sameroff, 1980;) and their internal working models of self and relationships (Carlson & Sroufe, 1995; Main, Kaplan, & Cassidy, 1985;). Markers for immaturity and/or psychological disturbance included holding unrealistic expectations for infants’ development (Epstein, 1979), lack of responsiveness toward their newborns (Jones, Green & Krauss, 1980), limited verbal engagement with their babies (Osofsky & Osofsky, 1970), expressing little empathy for their infant (Feshbach, 1989), displaying either little capacity to cope with frustration (Spinetta & Rigler, 1972), and displaying apathetic or depressive interpersonal styles (Polansky, 1981).

At a more distal level, risks focused on those environmental conditions that would create stressful conditions in the household that would interfere with parents’ care of their children, such as unemployment (Gil, 1970), poor housing and household conditions (Quinton & Rutter, 1984b), or marital discord (Belsky, 1981), and isolation from supportive family members and friends (Bakan, 1971; Kempe, 1973). A history of punitive, rejecting, abusive, or neglectful caregiving on the parents’ own part was considered a risk factor if they had no corrective emotional experience (such as effective caregiving from another parent, successful therapy, or a healthy marriage) that would allow them to resolve adequately these experiences (Egeland, Jacovitz, & Papatola, 1984; Pianta, Egeland, & Erickson, 1990).

C. Modifiable Risks for Welfare Dependence and Compromised Maternal Life-Course Development

While many of the maternal and child problems described above occur commonly in our society, they are found more frequently among children born into families in which the parents are teenagers, unmarried, and poor, and especially among women who have rapid, successive, subsequent pregnancies. Proximal risks for rapid successive pregnancies include women’s having little sense of control over their life circumstances and contraceptive practices in particular (Brafford & Beck, 1991; Levinson, 1986; Heinrich, 1993) and limited visions for their own personal development in the areas of education and work (Musick, 1993). It should be noted that these risks are found more frequently in low-income populations, where the opportunities for advancement in education and work have been constrained by limited financial resources.

D. Modifiable Risks for Early-Onset Antisocial Behavior

More recently, we have analyzed risks for early-onset antisocial behavior (Olds, 1997; Olds et al., 1997;) and determined that the impact of the program of prenatal and infancy home visitation on maternal and child health early in the life cycle reduces important risks for this important problem. We consider our perspective on early-onset antisocial behavior an elaboration of Moffitt’s model of risks (that is, neuropsychological deficits and dysfunctional caregiving) by adding an explicit focus on maternal life-course -- that is, large family size, closely spaced children, parental criminal involvement, and welfare dependence.

Moffitt has hypothesized that antisocial behavior emerges through two different developmental pathways. One type appears very early in life (with signs emerging as early as the preschool years) and a second that appears in adolescence (DSM-IV, 1994; Moffitt, Caspi, Dickson, Silva, & Stanton, 1996). The severity and longitudinal course of these two types of disorder are substantially different, with childhood-onset conduct disorders being the more serious. In one longitudinal study of 535 males conducted in Dunedin, New Zealand, 13% had characteristics of childhood-onset conduct disorder, and 31% exhibited the characteristics of adolescent-onset conduct disorder (Moffitt et al., 1996). The investigators found that children with behaviors indicative of childhood-onset conduct disorder were substantially more likely as adolescents to become violent, to display antisocial personalities, to leave school early, and to have weaker bonds to their families than did children whose antisocial behavior began to appear in adolescence. The reader will notice that the domains of risk for early onset antisocial behavior are exactly those targeted by the program to improve maternal and child health early in the life cycle.

1. Neuropsychological Deficits

Children with childhood-onset conduct disorder are more likely to have neuropsychological deficits, as reflected in compromised motor functioning, attention deficits, hyperactivity, impulsivity, and impaired language and cognitive functioning (Moffitt, 1993b). Although most children with these problems do not grow up to become criminals, subtle neurological deficits can increase children's susceptibility to other adverse environmental influences, such as harsh and rejecting parenting and rejection by peers, that can further increase their risk for later delinquency and crime (Moffitt et al, 1996).

While some of these childhood neuropsychological deficits probably have genetic origins, evidence is accumulating that a sizable portion can be traced to poor prenatal health conditions that compromise the development of the fetal nervous system (see, for example, Olds, 1997). Many of these neuropsychological deficits may be prevented by helping pregnant women (i) reduce their use of alcohol, illegal drugs, and cigarettes; (ii) improve their prenatal diet; and (iii) identify and obtain prompt treatment for emerging obstetric problems, such as genitourinary tract infections and hypertensive disorders.

2. Dysfunctional Care of the Child

Abused and neglected children are at increased risk for persistent child behavior problems, academic failure, chronic delinquency, adult criminal behavior, antisocial personality disorder, and especially violent crime (Maxfield & Widom, 1996). Nevertheless, despite the risk posed by child abuse and neglect, the majority of such children do not become delinquent, criminal, or violent (Widom, 1989b). Moreover, we do not know why some abused and neglected children develop antisocial behavior, and others do not. It may have to do with some maltreated children's development of a belief that the world is a hostile place and a corresponding accumulation of experiences that channel such children into environmental contexts where they are increasingly exposed to criminogenic influences (a topic that we address more completely below).

Some researchers have reasoned that poor parenting practices fail to instill within the child the capacity for impulse regulation and empathy, increasing the risk for adolescent criminal behavior (Gottfredson & Hirschi, 1990). Particular attention has been given to abused and neglected children's difficulty in regulating emotions such as anger and aggression (Rogosch, Cicchetti, & Aber, 1995; Shields, Cicchetti, & Ryan, 1994). A number of investigators have noted that children who are abused develop a tendency to distrust others' motivations because they have found the world to be hostile and tend to attribute hostile motives to others' neutral behavior. This tendency leads children to confront and, at times, to attack others, as if to strike first before they are harmed themselves.

Another area of emotion regulation that has strong ties to behavior in later life is the development of empathy (Robinson, Zahn-Waxler, & Emde, 1994). When children are able to respond empathically, they are at lower risk for the development of antisocial behavior both during early periods of development and later in life (Eisenberg & Mussen, 1989). Abused and neglected children are less likely to be empathic.

3. Compromised Maternal Life Course

Women's own personal development is associated with whether their children will develop antisocial behavior. In a longitudinal study of adolescent parents in Baltimore, for example, young women with recent welfare experience were more likely to report that their children had been expelled from school and had engaged in a variety of antisocial and delinquent behaviors than were their low-income, nonwelfare counterparts (Furstenberg, Brooks-Gunn, & Morgan, 1987). Being unmarried, not having graduated from high school, and having three or more children also increased the likelihood of these reported behavioral problems. A Danish longitudinal study of 4000 males and their families found that poor social circumstances (reflected by mothers who were young, unmarried, and of low socioeconomic status, and poor conditions in the home) increased the risk for boys' violent behavior at age 18 (Raine, Brennan, & Mednick, 1994). Moreover, a study of 10th graders indicated that increased family size led to reduced parental influence and greater peer influence on both girls' and boys' development of antisocial behavior and delinquency (Tygart, 1991).

As with child maltreatment and antisocial behavior, the mechanisms linking maternal life course to children's antisocial behaviors are not well understood. Tygart's (1991) findings, for example, point to the role that parental monitoring may play in linking family size with antisocial behavior: The larger one's family, the more difficulties parents have supervising their children. This lack of supervision puts children at risk for poor academic outcomes which, in turn, are associated with antisocial behavior among white male 7th to 12th graders (Hirschi, 1994). In addition, poor parental monitoring, coupled with the likelihood that families with few economic resources tend to live in crime-ridden neighborhoods, may further increase children's exposure to negative peer influences (Dishion, Capaldi, Spracklen, & Li, 1995).

4. Interrelations Among these Domains of Functioning

Thus far, consideration has been given to how the three general domains of risk are individually related to the development of antisocial behavior. However, these risk factors probably combine developmentally to increase the risk for childhood-onset conduct disorder.

Recent evidence from the Danish longitudinal study cited above emphasizes the volatility of combining neurodevelopmental impairment on the part of the child and dysfunctional caregiving early in the life cycle (Raine et al., 1994). The combination of these two conditions produced a multiplicative increase in the children's risk for later violence. The subgroup of individuals at highest risk for violent behavior at age 18 were those who had both birth trauma (a marker for neurodevelopmental impairment) and parental rejection in the child's first year of life. These males had rates of violent behavior that were two to six times that of males who had neither risk factor or only one.

Although the pathogenic mechanisms by which violent behavior develops is unclear, one plausible explanation is that neurodevelopmental impairment due to birth complications compromises the child's ability to cope emotionally and intellectually with interpersonally traumatic experiences, such as parental rejection. Early rejection and/or violence may lead children with compromised emotional and cognitive functioning to overgeneralize their early-relationship experiences to other, later relationships and to be especially likely to attribute hostile intent or rejection to the neutral behavior of others.

Hostile attributions can have devastating consequences when viewed developmentally. Children prone to act aggressively are more likely to be rejected by peers. Peer rejection, in turn, is associated with academic, social, and behavioral maladjustment in middle school, major risk factors for the development of conduct disorder (Coie, Lochman, Terry, & Hyman, 1992; Dodge, Coie, Pettit, & Price, 1990; Kupersmidt & Coie, 1990).

The importance of cumulative risk is also emphasized in other studies (e.g., Kupersmidt, Burchinal, & Patterson, 1995; Loeber, 1990; Yoshikawa, 1994). Increases in dysfunctional caregiving are more frequently occurring when parents experience financial difficulties (Conger, Conger, Elder, Lorenz, Simons et al., 1992, 1993) and have larger families (Hirschi, 1994). In such cases, children's risks for antisocial behavior are further increased by the nesting of risk conditions that can ensnare them in deviant developmental trajectories (Felner, Brand, DuBois, Adan, Mulhall et al., 1995; Hirschi, 1994; Moffitt, 1993a, 1993b). Although it is clear that the co-occurrence of factors multiplies the risk for conduct problems, it is not yet clear how the combination of conditions increases the risk for violent and persistent antisocial behavior (Moffitt, 1993a; Raine et al., 1994).

Figure 1 provides a framework for integrating our thinking about how these diverse influences converge in producing childhood-onset conduct disorder and how this program of prenatal and early childhood home visitation by nurses reduces its risks. It is important to note that this is simply a hypothetical model; while evidence supports each of the links shown by arrows in the figure, we are unable to say with certainty that these associations operate in causal fashions.

Subtle damage to the developing fetal nervous system may interfere with children's capacity to respond effectively to their parents' efforts to care for them, resulting in patterns of interaction characterized by frustration and anger that interfere with the development of secure attachment (Olds, 1990; Rodning, Beckwith, & Howard, 1989; Sanson, Smart, Prior, & Oberklaid, 1993). Even if children with neurodevelopmental impairment do not "cause" their parents to be frustrated and angry, they are more likely to have parents who provide inconsistent discipline and who may be impatient and irritable themselves, possibly because of genetic links or because of shared environmental or behavioral contexts, such as substance abuse (Moffitt, 1993b). These parenting practices can lead to cycles of interaction in which the child's problems with emotional and behavioral regulation precipitate child abuse or neglect, which further intensifies the child's emotional and behavioral dysregulation.

The combination of these factors can also contribute to children's development of dysfunctional patterns of interaction with peers and teachers (Bierman & Wargo, 1995; Coie & Jacobs, 1993). Increased economic difficulties and parent depression can lead to a lack of nurturant and involved parenting, which, in turn, is associated with negative peer relations (Conger et al., 1992, 1993). Furthermore, children with behavioral dysregulation are likely to be labeled as deviant in the classroom environment (Bierman & Wargo, 1995; Coie, Terry, Lenox, Lochman, & Hyman, 1995; Dishion, French & Patterson, 1995; Eder, 1983). Eder (1982) found that first graders from lower socioeconomic backgrounds, as compared to their counterparts from higher SES backgrounds, were placed more often in lower-level reading groups. These groups were more disruptive and less facilitative of learning than high-level groups (Eder, 1981), exacerbating students' academic vulnerabilities and making them even more susceptible to negative peer influences (Dishion et al., 1995).

An analysis of the risks for prenatal health, caregiving, and life-course problems we wished to affect was the first step in the design of the program. That effort clarified the focuses for the intervention’s activities. The next step was to devise clinical and psychoeducational activities that held promise for bringing about enduring adaptive behavioral change during pregnancy and the early years of the child’s life. This required an ongoing examination of theories of human development, motivation, and behavior in light of accumulating research evidence and clinical experience (Olds et al., 1997). In the next sections, we describe the intervention and examine its theoretical underpinnings and the processes through which the program is thought to produce its effects.

III. THEORETICAL FOUNDATIONS

The program has been grounded in theories of human ecology (Bronfenbrenner, 1979, 1992), self-efficacy (Bandura, 1977), and human attachment (Bowlby, 1969). The earliest formulations of the program gave greatest emphasis to human ecology, but as the program has evolved, it has been grounded more explicitly in theories of self-efficacy and attachment.

A. Human Ecology Theory

The original formulation of this program was based in large part on Bronfenbrenner's theory of human ecology (Bronfenbrenner, 1979). Human ecology theory emphasizes the importance of social contexts as influences on human development. Parents' care of their infants, from this perspective, is influenced by characteristics of their families, social networks, neighborhoods, communities, and cultures, and interrelations among these structures. Bronfenbrenner's original theoretical framework has been elaborated more recently (with greater attention to individual influences) in a person-process-context model of research on human development (Bronfenbrenner, 1992).

The person elements of the model are reflected in the program components that have to do with behavioral and psychological characteristics of the parent and child. In the formulation of the theoretical foundations of this program, parents, and especially mothers, are considered both developing persons and the primary focus of the preventive intervention. Particular attention is focused on parents' progressive mastery of their roles as parents and as adults responsible for their own health and economic self-sufficiency. This program emphasizes parent development because parents' behavior constitutes the most powerful and potentially alterable influence on the developing child, particularly given parents' control over their children's prenatal environment, their face-to-face interaction with their children postnatally, and their influence on the family's home environment.

The concept of process encompasses parents' interaction with their environment as well as the intrapsychic changes that characterize their mastery of their roles as parents and providers. Three aspects of process emphasized here relate to individuals' functioning: (i) program processes (e.g., the ways in which the visitors work with parents to strengthen parents' competencies); (ii) processes that take place within parents (i.e., the influence of their psychological resources -- developmental histories, mental health, and coping styles -- on behavioral adaptation); and (iii) parents' interaction with their children, other family members, friends, and health and human service providers. For the sake of simplicity, the discussion of these processes has been integrated below into the person (parent) part of the model.

The focus on parents elaborated here is not intended to minimize the role that contextual factors such as economic conditions, cultural patterns, racism, and sexism play in shaping the opportunities that parents are afforded (Olds, 1980). Most of those features of the environment, however, are outside of the influence of preventive interventions provided through health and human service systems. Certain contexts, nevertheless, are affected by parents' adaptive competencies. It is these features of the environment that the current program attempts to affect, primarily by enhancing parents' social skills. The aspects of context that most concern us have to do with informal and formal sources of support for the family, characteristics of communities that can support or undermine the functioning of the program and families, the impact of going to school or participating in work on family life, as well as cultural conditions that need to be taken into consideration in the design and conduct of the program.

Similarly, this perspective is not inconsistent with an approach that recognizes that some individual characteristics that lead individuals toward certain social contexts have genetic influences (Plomin, Reiss, Hetherington, & Howe, 1994). The most important question is to what extent individual behavior and adaptive functioning can be improved when both structural features of the society and genetic propensities tend to increase the risk for maladaptive functioning.

One of the central hypotheses of ecological theory is that the capacity of the parent-child relationship to function effectively as a context for development depends on the existence and nature of other relationships that the parent may have. The parent-child relationship is enhanced as a context for development to the extent that each of these other relationships involves mutual positive feelings and that the other parties are supportive of the developmental activities carried on in the parent-child relationship. Conversely, the developmental potential of the parent-child relationship is impaired to the extent that each of the other relationships in which the parent is involved consists of mutual antagonism or interference with the developmental activities carried on in the parent-child relationship (Bronfenbrenner, 1979, p. 77).

1. Program Implications

Human ecology theory played an important role in identifying which families would be enrolled in the study and when. We chose to work with women who had no previous live births and thus were undergoing a major role change that Bronfenbrenner calls an ecological transition. We began the program during pregnancy and the early years of the child's life with women who had no previous live births because these women are in the process of assuming the parental role. In providing support to young people prior to and while they were learning about being parents, we reasoned that the visitors would enhance their influence on parents' enduring orientation to their roles as parents and providers. The skills and resources that parents develop around the care of the first child would presumably also carry over to later children. Also, to the extent that the program was successful in helping parents plan for their futures (including planning subsequent pregnancies), we thought that parents might have fewer unintended children, thereby easing some of the subsequent challenges of caring for the first child.

Human ecology theory also focused the home visitors' attention on the systematic evaluation and enhancement of the material and social environment of the family. Indeed, it was because of our conviction that these social and material contexts of the family were so important that we chose to deliver the services in the home, where the nurses could evaluate, first hand, the family environment in which the parents and children were living. The visitors assess and promote informal social support (individuals within the family and friend network who can serve as reliable sources of material and emotional support for the mother in her efforts to care for her children), and families' use of formal community services.

Human ecologists would hypothesize that women's capacity to improve their health-related behaviors is influenced by their levels of informal support for change. Women's efforts to reduce cigarette smoking during pregnancy, for example, are affected by the extent to which individuals close to them believe that smoking is bad for pregnant women and the fetus and the extent that they actively support women's efforts to quit. Consequently, the visitors encourage mothers during pregnancy to invite other family members and friends to the visits in an effort to enhance friends' and family members' support of the mothers' efforts to improve their health-related behaviors and to prepare for labor, delivery, and early care of the child.

The involvement of other family members, friends, and mothers' partners is especially important in helping women practice contraception, finish their education, and find work. For discussions of family planning and contraception, the nurses make every effort to conduct some of those visits when mothers' partners are present. In addition, returning to school after delivery or finding work requires finding appropriate care for the child. For low-income families, this usually means that the mother must find someone in the household or network of friends who might be able to provide reliable and safe care for the baby. The nurses help mothers identify safe and nurturant care within their network of family members and friends and, if none can be found, help them find appropriate subsidized center-based care. To the extent that the visitors have been successful in helping women to complete their education and participate in the work force, they have altered the ecology of the family by placing additional demands upon other family members and friends. Moreover, in spending more time in educational or work settings, women are integrated into social contexts where there are greater pressures to conform to societal expectations. These activities thus change the ecology of the family in fundamental ways.

Human ecology theory also focuses the visitors' attention on the identification of family stressors and needed health and human services. The visitors assess families' needs and then systematically help them make use of their health-care providers and obtain other needed services in an attempt to reduce the situational stressors that many low-income families encounter. Families are helped to obtain services such as Medicaid, Aid to Families with Dependent Children (AFDC), subsidized housing, help with family counseling, nutritional supplementation, substance-abuse counseling, and assistance finding clothing and furniture.

After the baby is born, the visitors continue to inform mothers and other family members about the availability of formal community services and provide mothers with the skills to use those services more effectively. As during pregnancy, the visitors communicate with the children's physicians and their office staff in order to reinforce the medical staff's recommendations in the home and to enable the medical staff to provide more informed and sensitive care in the office. Parents are taught to observe their children's indicators of health and illness, to use thermometers, and to call the physician's office with signs of their children's illnesses. The expectation is that this approach will increase the appropriate use and decrease the inappropriate use of emergency rooms.

As the program model was transferred from Elmira, New York (where it served a primarily European-American population), to Memphis and Denver, it was reviewed from the standpoint of its congruence with the cultural beliefs of the African-American and Mexican-American families that it increasingly served. This work was facilitated by the creation of community advisory committees that reviewed the protocols. The reassuring message in both Memphis and Denver was that the protocols were essentially culturally competent. This sanctioning of the program was based in part on its inclusion of other family members and friends in the program and its creation of racially and ethnically diverse teams of visitors and supervisors.

2. Limitations of Human Ecology Theory

Compared to other developmental theories, Bronfenbrenner's framework provides a more extended and elaborated conception of the environment. The original formulation of the theory, however, tended to treat the immediate settings in which children and families find themselves as shaped by cultural and structural characteristics of the society, with little consideration given to the role that adults (in particular, parents) can play in selecting and shaping the settings in which they find themselves. While many investigators today reason that the personal characteristics that influence individuals’ selection and shaping of their contexts have genetic origins, we have focused this program of research on determining the extent to which and the means by which those choices and adaptive behaviors can be supported in ways that promote more effective health and development of the mother and child by focusing on individual characteristics that are centrally involved in adaptive behavioral change.

Consequently, self-efficacy and attachment theories were integrated into the model to provide a broader conception of the parent–setting relationship. The integration of these theories allows a conceptualization of development that encompasses truly reciprocal relationships in which settings, children, and other adults influence parental behavior and in which parents simultaneously select and shape their settings and interpersonal relationships.

B. Self-Efficacy Theory

Self-efficacy theory provides a useful framework for promoting women's health-related behavior during pregnancy, their care of their children, and their own personal development. According to Bandura (1977), differences in motivation, behavior, and persistence in efforts to change a wide range of social behaviors are a function of individuals' beliefs about the connection between their efforts and desired results. According to this view, cognitive processes play a central role in the acquisition and retention of new behavior patterns. In self-efficacy theory, Bandura (1977) distinguishes efficacy expectations from outcome expectations. Outcome expectations are individuals' estimates that a given behavior will lead to a given outcome. Efficacy expectations are individuals' beliefs that they can successfully carry out the behavior required to produce the outcome. Efficacy expectations affect both the initiation and persistence of coping behavior. Individuals' perceptions of self-efficacy can influence their choice of activities and settings and can determine how much effort they put forth in the face of obstacles.

1. Program Implications

While self-efficacy theory played a role in the design of the Elmira program through an emphasis on helping women set small achievable objectives for themselves that would strengthen their confidence in their capacity for behavioral change, it was not emphasized explicitly as a theoretical foundation in Elmira to the same degree as it was in Memphis and Denver. The increased focus on self-efficacy in the later trials grew out of our observation that several of the most important program effects in Elmira (discussed below), such as the reduction in child maltreatment and emergency-room encounters for injuries were concentrated among women who at registration had little sense of control over their life circumstances (Olds, Henderson, Chamberlin, & Tatelbaum, 1986). We hypothesized that the promotion of self-efficacy in the Elmira program played a central role in enabling at-risk women to reduce their prenatal cigarette smoking, rates of subsequent pregnancy, and rates of unemployment (Olds, Henderson, Tatelbaum, & Chamberlin, 1986a, 1988), given that the nurses used these methods in helping women manage these aspects of their lives. We reasoned that the nurses' emphasis on helping women gain control over specific life circumstances such as these promoted women's generalized self-efficacy.

As a result of these observations, in the Memphis and Denver trials, the visitors were trained explicitly in self-efficacy theory and its applications, and the program protocols were written in a way that distinguished efficacy expectations from outcome expectations. For instance, women may acknowledge that smoking is harmful for themselves and their babies (an outcome expectation) but not believe that they will be able to quit (an efficacy expectation). Distinguishing these two aspects of the problem helps in the specification of smoking reduction efforts and other individualized interventions.

Much of the educational content of the program was focused on helping women understand what is known (or thought about) the influence of particular behaviors on the health and growth of the fetus, on women's own health, and on the subsequent health and development of the child. The educational program represented an effort to bring women's outcome expectations into alignment with the best evidence available.

Improvements in individuals' behavior depends upon their confidence in their ability to change. According to Bandura, helping services like those carried out in the current program achieve their primary effect by creating and strengthening the individual's expectation of personal efficacy. Self-efficacy theory has a number of direct implications for the methods that the home-visitors used to promote mothers' healthy behavior, optimal caregiving, family planning, and economic self-sufficiency.

First, because the power-of-efficacy information is greater if it is based on the individual's personal accomplishments than if it derives from vicarious experiences and verbal persuasion (Bandura, 1977), the home visitors emphasize methods of enhancing self-efficacy that rely on women actually carrying out parts of the desired behavior. Verbal persuasion methods are used, of course, but whenever possible, they serve as guides and reinforcers for behaviors that the women already have enacted. Women who already display some adequate prenatal behaviors are encouraged for what they are doing well. Similarly, the visitors reinforce caregiving behaviors that are close to the goals of the program, such as the sensitive identification of and response to the child's cries or the removal of safety hazards in the home environment. This identification of family strengths helps build mothers' and other family members' confidence in their roles as parents and provides incentives for their acquiring new caregiving skills.

Second, the visitors employ methods of behavioral and problem analysis that emphasize the establishment of realistic goals and behavioral objectives in which the chances for successful performance are increased. The same principles apply whether the individual is trying to quit drinking, correct her diet, or improve her relationship with her boyfriend. Because perceptions of self-efficacy predict coping and self-regulatory behavior, the home visitors periodically ask women about their beliefs concerning their abilities to manage all types of problems related to the overall goals of the program or to the concerns of the women themselves. This information is used to help the nurses focus their efforts on creating opportunities for women to accomplish small, achievable objectives related to particular goals. As a result of these observations, visitors in the Memphis program developed a series of questionnaires used clinically to assess women's and other family members' beliefs (outcome and efficacy expectations) about their health-related behavior, their care of their children, and their life course. These assessments now provide the nurses with a basis upon which to begin their work with mothers and other family members.

Our articulation of self-efficacy in the program protocols has evolved over each of the three trials. In the Memphis trial, we augmented the emphasis on setting small, realistic objectives with a program of goal-setting and problem-solving (Haley, 1991; Wasik, Bryant, Ramey, & Sparling, 1992). The theory of self-efficacy was built into the training program more formally, and we began teaching the problem-solving method (defining the problem, generating sets of possible solutions, trying certain solutions, and evaluating the results) as a general approach to coping (Haley, 1991; Wasik et al., 1992). In addition, assessments of efficacy and outcome expectations with respect to critical behaviors were added to the formal test of program effects. In the Denver trial, the program model has been further refined with solution-focused methods that emphasize the competence of family members and that focus on parents' successes (O'Brien & Baca, 1997).

2. Limitations of Self-Efficacy Theory

While self-efficacy theory provides powerful insights into human motivation and behavior, it is limited in several respects. First, it is primarily a cognitive-behavioral theory. It attends to the emotional life of the mother and other family members only through the impact of behavior on women's beliefs or expectations, which, in turn, affect emotions. Many people have experienced multiple adversities in the form of overly harsh parenting, rejection, or neglect that often contribute to a sense of worthlessness, depression, and cynicism about relationships. Self-efficacy gives inadequate attention to methods of helping parents cope with these features of their personal history or the impact of those early experiences on their care of their children. We have augmented the theoretical underpinnings of the program regarding these social and emotional issues with attachment theory (discussed below).

The second limitation is that self-efficacy attends to environmental influences in a cursory way. People can give up because they do not believe that they can do what is required, but they also can give up because they expect that their efforts will meet with punitiveness, resistance, or unresponsiveness. While Bandura acknowledges that adversity and intractable environmental conditions are important factors in the development of individuals' sense of futility (Bandura, 1982), the structure of those environmental forces is not the subject of Bandura's theory. In other words, individuals' feelings of helplessness and futility are not simply intrapsychic phenomena, but are connected to environmental contexts that provide limited opportunities and that fail to nurture individuals' growth and well-being. The structure of those environmental influences is the primary subject of human ecology theory.

Finally, although Bandura (1982) discusses self-efficacy in terms of groups, communities, and nations, the focus of the theory tends to be on the individual. In this sense, the theory may be less relevant for cultural groups that place greater emphasis on group accomplishments (or survival), such as kin networks, families, and communities.

C. Attachment Theory

Historically, this program owes much to Bowlby's theory of attachment (Bowlby, 1969). Attachment theory posits that human beings (and other primates) have evolved a repertoire of behaviors that promote interaction between caregivers and their infants (such as crying, clinging, smiling, signaling) and that these behaviors tend to keep specific caregivers in proximity to defenseless youngsters, thus promoting their survival, especially in emergencies. Humans (as well as many other species) are biologically predisposed to seek proximity to specific caregivers in times of stress, illness, or fatigue in order to promote survival. This organization of behavior directed toward the caregiver is called "attachment".

A growing body of evidence indicates that caregivers' levels of responsivity to their children can be traced to caregivers' own childrearing histories and attachment-related experiences (Main, Kaplan, & Cassidy, 1985). Caregivers' attachment-related experiences are thought to be encoded in "internal working models" of self and others that create styles of emotional communication and relationships that either buffer the individual in times of stress or that lead to maladaptive patterns of affect regulation and create feelings of worthlessness (Carlson & Sroufe, 1995). Differences in internal working models, according to attachment theorists, have enormous implications for mothers' capacities for developing sensitive and responsive relationships, especially with their own children.

1. Program Implications

Attachment theory has affected the design of the home-visitation programs in three fundamental ways. The first has to do with its emphasis on the visitors' developing an empathic relationship with the mother and other family members, where possible. The second has to do with the emphasis of the program on helping mothers and other caregivers review their own childrearing histories and make decisions about how they wish to care for their children in light of the way they were parented. And the third has to do with its explicit promotion of sensitive, responsive, and engaged caregiving in the early years of the child's life.

A fundamental element of the program has been the visitors' close, therapeutic alliances with the mother and other family members, beginning during pregnancy. The establishment of such a relationship, consisting of empathy and respect, was expected to help modify a woman's internal working models of herself and her relationships (most importantly her developing relationship with her child).

It is important for the visitors to know about women's childrearing histories and their internal working models of relationships because, without intervention, destructive models are likely to undermine the quality of care that parents provide to their own children. By assessing women's beliefs and attitudes toward their children's behavior during pregnancy, the visitors were able to help women and other caregivers develop more accurate conceptions about their infant's motivations and methods of communicating.

Program protocols have been designed to present systematically how infants communicate, giving special attention to nonverbal cues, crying behavior, and colic, and how parents can meet their infants' and toddlers' emotional needs. An emphasis on mothers and other caregivers correctly reading and responding to the infant's cues begins during pregnancy and continues through the end of the program.

In order to promote sensitive and responsive caregiving, increasingly comprehensive parent-infant curricula were incorporated into the program in each of the three trials. For example, in the Elmira program, all of the nurses were trained in the Brazelton newborn examination (Brazelton, 1973), and they were provided teaching materials to promote sensitive, responsive care on the part of parents. The nurses in the Elmira program, however, felt that the primarily didactic nature of the parent-child curriculum failed to provide them with the kind of guidance they needed to promote emotionally responsive caregiving. We realized that we had too few activities incorporated into the program to promote parents' sense of success in interacting with their children. In the Memphis program, the number of standardized materials employed to promote sensitive and responsive caregiving was expanded to include activities such as Barnard's Keys to Caregiving program, her NCAST feeding scale (Barnard, 1979), and an adaptation of Sparling's Partners for Learning program (Sparling & Lewis, 1981). In the Denver program, a curriculum has been incorporated explicitly to promote parents' emotional availability and joy in interacting with their children. Known as the Partnership in Parenting Education (PIPE), the program was designed originally for adolescents in classroom settings (Dolezol & Butterfield, 1994), but has been adapted for home visitors in the Denver trial. Like Partners for Learning, it uses recommended activities for caregivers and children, but promotes interaction with structured interactive guidance. An additional differences is its explicit focus on shared positive emotions to enhance early development. While to date we have only preliminary staff feedback supporting the value of this component of the program, as we have reflected on the development and shortcomings of the home-visitation program to date, we are increasingly convinced that the emphasis on the emotional features of the relationship is fundamental.

2. Limitations of Attachment Theory

While attachment theory provides a rich set of insights into the origins of dysfunctional caregiving and possible preventive interventions focused on parent-visitor and parent-child relationships, it gives scant attention to the role of individual differences in infants as independent influences on parental behavior, and it provides inadequate attention to issues of parental motivation for change in caregiving. Moreover, it minimizes the importance of the current social and material environment in which the family is functioning as influences on parents' capacities to care for their children. For more systematic treatments of these issues, we turned to self-efficacy and human ecology theories.

D. Summary and Conclusions: Theoretical Foundations

In the visitors' efforts to help women improve the outcomes of pregnancy, child health and development, and maternal life course, they have been equipped with a theory-driven program design and visit protocol that guides their efforts to help women improve their health-related behaviors, their care of their children, their planning of subsequent pregnancies, and participation in the work force. These adaptive skills focus on both parents’ health, caregiving, and life-course behaviors and their learning to summon family and community support to improve the material and social contexts in which they live.

IV. PROGRAM DESIGN

A. Frequency of Visitation

The frequency of home visits changes with the stages of pregnancy and can be adapted to the mother’s needs. Mothers are enrolled through the third trimester of pregnancy. Visits are scheduled once a week during the first month after enrollment, which assists the new mother and the home visitor to establish a trusted relationship. Thereafter, visits are scheduled every other week until the birth of the baby. Nurses again visit weekly for 6 weeks after the baby is born, helping the new mother and newborn adjust. From the child’s 2nd to 21st postnatal month, visits are scheduled twice a week. From the 21st to 24th postnatal month, visits are scheduled once a month, as the nurse and mother work to bring their relationship to a close and assure that the mother is connected with any services she will continue to need for support. Each visit lasts approximately 90 min.

B. Nurses as Home Visitors

This program model calls for nurses to be the home visitors. We have chosen nurses because of their formal training regarding women’s and children’s health and because of their competence in managing the types of complex clinical situations often presented by at-risk families. We have hypothesized that the nurses’ ability to effectively address mothers’ and family members’ concerns about complications of pregnancy, the physiologic and anatomic changes of pregnancy, labor, and delivery, and the physical health of the infant provide nurses with credibility in the eyes of the family that increase their influence. Moreover, through their ability to teach mothers and other family members to identify emerging health problems and to use the health-care system to address those problems, the nurses increase the clinical influence of the program through the early detection and treatment of disorders.

While the content, theory, and clinical methods of the program have now been thoroughly specified, these materials cannot provide unequivocal guidance in all situations, as qualitative studies of the implementation of this program have demonstrated (Kitzman, Yoos, Cole, Korfmacher, and Hanks, 1997; Kitzman, Cole, Yoos, and Olds, 1997). In the highly complex situations often encountered in this program, the nurses must rely upon their competent clinical-decision making skills and excellent supervision in order to maintain an effective working relationship with the mother (and her family) and to simultaneously accomplish the goals of the program (Kitzman, Yoos, Korfmacher, et al., 1997; Kitzman, Cole, Yoos, et al., 1997).

While evidence from published randomized trials indicates that programs that employ nurses who work intensively with families employing comprehensive program models are more likely to achieve their goals (Olds and Kitzman, 1990; 1993), the reason that programs that have employed paraprofessionals have failed is not clear. Is it because of inadequate program models or because of paraprofessionals’ limited training? As noted above, this is the issue addressed in the Denver trial.

C. Outline of Program Content

Nurse home visitors follow detailed visit-by-visit program protocols that focus on five domains of functioning: personal health, environmental health, maternal role development, maternal life-course development, and family and friend support. The content of the protocols is organized developmentally to reflect those challenges that women are likely to confront at different stages of pregnancy and during the first 2 years of the child’s life. Within each of the five domains, specific assessments are made of maternal, child, and family functioning, and specific educational content and psychosocial interventions are prescribed, depending upon the nature and degree of vulnerability revealed in the assessment.

While the predominant population served by the program is low-income, unmarried women, the women’s husbands or boyfriends and their own mothers are especially encouraged to participate in the home visits because they often play decisive roles in determining the extent to which women will improve their health habits, finish their education, find work, secure appropriate child care, and address the needs of the child. In order to facilitate the involvement of friends and family members, nurses schedule weekend and evening visits to accommodate their work schedules.

During home visits, nurses carry out three major activities: (i) promoting adaptive change in behavior that affects the outcomes of pregnancy, the health and development of the child, and maternal life course; (ii) helping women build supportive relationships with family members and friends; and (iii) linking family members with other health and human services. In carrying out these activities, emphasis is placed on the importance of building on parents’ strengths and promoting parental competence and control over life circumstances.

1. Adaptive Behavior Change

a. Prenatal Behavioral Objectives. The nurse’s activities during pregnancy vary considerably among families because women enroll at various stages of gestation and because their knowledge, motivation, and ability to assimilate material differs, affecting the amount of time nurses must spend on any one topic. The major behavioral objectives include:

helping women improve their diets and monitor their weight gain;

helping women eliminate their use of cigarettes, alcohol and drugs;

teaching parents to identify the signs of pregnancy complications and learn how to use the health-care system to address those problems before they become more serious;

encouraging regular rest, appropriate exercise, and good personal hygiene related to obstetrical health;

preparing parents for labor and delivery;

preparing parents for early care of the newborn;

encouraging appropriate use of the health care system; and

encouraging mothers to make plans regarding subsequent pregnancies, returning to school, and finding employment.

b. Infancy and Early Childhood Behavioral Objectives. As during pregnancy, women’s learning needs and ability to assimilate educational materials varies considerably. The curriculum is organized so that nurses are able to cover issues of common concern to all primiparous mothers, while simultaneously responding to individual needs and differences. Through this phase of the program, the nurses assess the mother and infant with respect to the behavioral objectives outlined below, recognizing that the norms within these categories change as both the infant and mother mature. The major objectives include:

improving parent’s understanding of the infant’s temperament;

promoting the physical care of the child;

promoting the behavioral and emotional regulation of the child;

improving the safety of the home environment

helping mothers adapt to changing roles

encouraging mothers to further clarify their plans for returning to school, finding work and bearing additional children

helping women make concrete plans for completing their educations

helping women search for, secure, and retain a job

identifying safe and reliable child care

employing a reliable method of contraception

2. Enhancing Informal Support

The second major activity of the nurses during home visits is to enhance the informal support available to the women during pregnancy, birth, and the first 2 years after delivery. The nurses assess the quality of the women’s relationships with their husbands, boyfriends, mothers, friends, and other family members by asking the mother about these individuals and by observing their interactions. The nurses determine the extent to which inadequate support is due to the mother simply having no one to turn to versus her inability to use the support available to her. The nurses also attempt to predict the likelihood that new ideas introduced by the program will create or intensify hostilities among members of the support network or between the mother and the primary support person. These are the kinds of situations that require skilled clinical decision making on the part of the nurse that needs to be supported with competent supervision.

In general, nurses become involved in developing relationships with other family members and friends and in addressing their needs when the nurse assesses that these individuals play a direct role in affecting maternal and child functioning. During the course of home visits, insofar as possible, these individuals are encouraged to be sensitive to the mother’s needs and to help with household responsibilities, to accompany the woman to the hospital at the time of delivery, to be present for the birth, to aid in the subsequent care of the child, and to reinforce the advice of the nurses in their absence. They are encouraged to help her follow appropriate health behavior and health care practices without nagging or finding fault.

The mother’s husband or boyfriend, whether or not he is the father of the child, is included in the program as an important and highly influential figure in the child’s life. His parenting skills, contributions to family life, and support to the mother are seen as important resources. In some cases, the mother may be involved with men who are either abusive, neglectful, or engaged in illegal activities. In many of these cases, the nurse is able to serve as a support to the mother as she breaks away from these destructive relationships. In other cases, the mother may be determined to maintain contact with the man at any cost. It may not be prudent for nurses to intervene actively to discourage women’s involvement in these relationships, but by showing concern and respect for the women, nurses communicate their belief that the women do not have to accept poor treatment and, in the process, help women make decisions that are in their best interests.

The mother is viewed as the primary figure responsible for the health and well-being of the child. However, in some families, the grandmother is the individual most willing and able to provide for the child. For these families, the nurse directs her educational efforts regarding child care to the grandmother and tries to help the young mother articulate her own goals in life. Conflict between the mother and grandmother sometimes arises in families in which either both individuals are capable of and willing to assume responsibility for the child or neither is willing to assume responsibility for the child. The nurse’s role in such situations is to help resolve the conflict by encouraging both mother and grandmother to communicate openly about the issue.

3. Linkage With Formal Services

The nurses also attempt to reduce family stresses that they are unable to handle themselves by connecting families with formal health and human services. Beginning with the first home visit, the nurses systematically assess the extent to which the family’s basic survival needs are being met. Areas considered are income and basic shelter, food and medical care; reliable and adequate housing; and physical, mental or substance abuse problems that are unattended.

The nurses urge parents to keep prenatal and well-child care appointments and to call the physician’s office when a health problem arises, so that the office staff can help them make decisions as to whether or not sick or emergency-room visits are necessary. The nurses send reports of their observations regarding medical, social, and emotional conditions to both the obstetricians and pediatricians who provide the mothers’ and babies’ care. In this way, the physicians and office staff can provide more informed and sensitive care. Also, by communicating regularly with the mother’s and baby’s primary health care providers, the nurses can clarify and reinforce physicians’ recommendations in the home. When necessary, the nurses refer parents to other social services such as public assistance, Medicaid or food stamps; Planned Parenthood (for contraceptives); mental health or family counseling; legal aid; WIC; and educational services or job training.

D. Summary and Conclusions: Program Design

This program model is now both well-conceived and well-tested. Some questions, nevertheless, remain about several aspects of the program. While we have attempted to integrate theoretical perspectives with both individual and environmental emphases, the resulting framework and program has emphasized individual determinants of development to a greater degree than may be desirable, especially given the levels of deprivation and hostility that characterize many under-class, crime-ridden urban environments. To date, we know little about how this type of program might work in communities with vastly different social and economic resources for families than those found in Elmira, Memphis, and Denver. A series of secondary analyses of the Memphis and Denver data is being conducted to examine the extent to which neighborhoods with high rates of crime and poverty further condition the impact of the program.

Moreover, we know little about how differences in the levels of community services affect program process and outcome. To what extent are the effects of the program dependent upon the existing system of health and human services? To what extent might the program effects be enhanced in locales where a greater portion of the families' basic needs are assured? And to what extent would the effects of the program be diminished if the existing system were depleted further? Visitors depend upon linking families with other needed services but often find that those services are in scant supply, leading some experts to take the position that home visits are necessary but not sufficient for child health and development (Chamberlin, 1980; Weiss, 1993).

Finally, we acknowledge that the current model of preventive intervention is limited with respect to addressing certain individual characteristics of parents as well. The visitors in each of the programs have encountered parents with major mental illness, such as depression and schizophrenia. The problems posed with this population can be considerable. The model calls for linking mentally ill parents with treatment services in the community, of course, but such services often not readily available. Moreover, barriers frequently exist to individuals' use of those services that are available. Administrators of home visitation programs may choose to hire their own mental health professionals to ensure treatment of these parents, may intensify the search for effective mental-health services through better linkages with primary care providers, or may try to directly improve the availability of mental health services in their community.

In their application of this theoretical framework to the individual needs of pregnant women, parents of young children, and their families, the visitors have found that the theoretical and empirical foundations of the program often provide less specific guidance than they need in order to guide their work with unequivocal success (see Kitzman, Cole, Yoos, & Olds, 1997 and Kitzman, Yoos, Cole, Korfmacher, & Hanks, 1997, for expositions of the clinical challenges encountered in providing this service). Sound clinical work has required substantial doses of common sense, clinical insight, and excellent supervision. Nevertheless, the comprehensive program model produced by the integration of these theoretical perspectives has led to a program of prenatal and early childhood home visitation that is substantially more effective than programs that are built on more limited theoretical foundations or without any theoretical foundations at all (Olds, 1992; Olds & Kitzman, 1990, 1993). Moreover, the integration of these theoretical perspectives has laid the foundation for program and research activities that are far more sensible and useful than any one of these theoretical perspectives, clinical wisdom, or common sense alone.

V. Overview of Research Designs and Methods and Findings

In each of the three studies of the program described above, women were randomized to receive either home visitation services during pregnancy and the first two years of the children’s lives or comparison services. While the nature of the home-visitation services was essentially the same in each of the trials, the comparison services were slightly different. The designs and methods employed in each of the trials are outlined below.

A. Elmira Design and Methods

The first trial of the program was begun in 1977. In the original study, 400 women were recruited during pregnancy and followed through the child’s 15th birthday. In evaluating the results of the Elmira trial, it is important to note that this was an efficacy study. The investigators were intensely involved in monitoring the implementation of the program, and the same set of nurses worked with their families for the entire duration of the program. For these reasons, the results obtained in Elmira are probably an upper-bound estimate of what might be accomplished if a program like this were disseminated on a larger scale.

1. Context and Sample

The study was conducted in a small, semi-rural county of approximately 100,000 residents in the Appalachian region of New York State. At the time the study began, the local community was well served from the standpoint of both health and human services. In spite of this abundance of services, the community consistently exhibited the highest rates of reported and confirmed cases of child abuse and neglect in the state between 1972 and 1982 (NY State Dept. of Social Services, 1982). Moreover, the community was rated the worst Standard Metropolitan Statistical Area in the country in terms of its economic conditions (Boyer &

Savageau, 1981). Pregnant women were actively recruited for the study from offices of private obstetricians and a free antepartum clinic if, at intake, they had no previous live births, they were at less than 26 weeks of gestation, and they had any one of the following characteristics that predispose to infant health and developmental problems: (i) young age (<19 years); (ii) single parent status; and (iii) low socioeconomic status. As noted above, any woman who asked to participate was enrolled, regardless of her age, marital status, or income, if she had no previous live births. This approach avoided creating a program that was stigmatized as being exclusively for the poor and created sample heterogeneity, enabling us to determine if the effects of the program were greater for families at higher risk. We enrolled 400 women, 85% of whom were either low-income, unmarried, or teenaged; none had a previous live birth. Eighty-nine percent of the sample was Caucasian. We stratified the sample on a number of demographic factors and then randomly assigned participating women to one of four treatment groups.

Families in Treatment 1 (n=94) were provided sensory and developmental screening for the child at 12 and 24 months of age. Based upon these screenings, the children were referred for further clinical evaluation and treatment when needed. Families in Treatment 2 (n=90) were

provided the screening services offered those in Treatment 1 plus free transportation (employing a taxi-cab voucher system) for prenatal and well-child care through the child's 2nd birthday.

There were no differences between Treatments 1 and 2 in their use of prenatal and well-child care (both groups had a high rate of completed appointments). Therefore, these two groups were combined to form a single comparison group. Families in Treatment 3 (n=100) were provided

the screening and transportation services offered Treatment 2 but in addition were provided a nurse who visited them at home during pregnancy. Families in Treatment 4 (n=116) were

provided the same services as those in Treatment 3, except that the nurse continued to visit through the child’s 2nd birthday. For assessment of the prenatal phase of the program, Treatments 1 and 2 were combined and compared to the combination of Treatments 3 and 4.

Five registered nurses were hired through a non-profit private agency for this experimental program. Each nurse had a caseload of 20-25 families and received regular clinical supervision.

2. Assessment Procedures

The interviews and assessments for this trial were conducted at registration (before the 30th week of pregnancy) and at the 34th, 36th, 46th, and 48th month, and the 15th year of the children’s lives. At 34 and 36 months, staff members conducted interviews and observational assessments in the families’ homes. At 36 and 48 months of life, the children were brought to the project offices for standardized testing. At the 15th year, the mothers (or primary caregivers) were interviewed, and the children interviewed and tested. The rates of completed assessments were very high. For the 15th-year interview, for example, we completed assessments on 81% of those cases originally randomized and on 90% of those women for whom there was no miscarriage, still birth, death (infant, child, or maternal), or child adoption. There were no treatment differences in the rates of completed assessments at the 15-year follow-up. Except in a few cases in which women inadvertently disclosed their treatment assignments , all assessments were conducted without awareness of the women’s and children’s treatment assignment. These features of the design strengthen the causal inferences that can be made about the influence of the program on maternal and child outcomes.

During home visits, the interviewers completed the Caldwell and Bradley Home Inventory (Caldwell & Bradley, 1979). At the 34th and 46th month in-home observations, interviewers also observed the mother interacting with the child and completed a 7-point scale rating the mother’s warmth, control, and involvement. They also completed an observational checklist of safety hazards in the home and asked questions about the presence of poisonous substances and use of car seats and safety belts.

At 36 and 48 months, children were brought to the project offices for standardized testing by a school psychologist, who administered the Stanford-Binet Form L-M test of intelligence.

Children’s pediatric and hospital records were reviewed for the period spanning birth to 50 months of age, and Child Protective Service records were reviewed in New York State as well as in the 14 other states to which families had moved during the period between the child’s birth and his/her 15th birthday.

At the children’s 15th birthday, interviews were conducted with the mothers in their homes (or by telephone, for those who lived outside of the Elmira area). The children were interviewed and given short IQ tests in the study offices (or in their schools by school officials when they lived outside of the Elmira area).

B. Elmira Results

1. Prenatal Results

We found that during pregnancy, nurse-visited women improved the quality of their diets to a greater extent, and those identified as smokers smoked 25% fewer cigarettes by the end of the pregnancy, than did their counterparts in the comparison group. By the end of pregnancy, nurse-visited women had fewer kidney infections, experienced greater informal social support, and made better use of formal community services. Among women who smoked, those who were nurse-visited had 75% fewer preterm deliveries, and among very young adolescents (aged 14-16), those who were nurse-visited had babies who were nearly 400 grams heavier, than their counterparts assigned to the comparison group (Olds et al., 1986a).

2. Caregiving and Child Development Results

After delivery, 19% of the poor, unmarried teens in the comparison group abused or neglected their children during the first 2 years after delivery as opposed to 4% of the poor, unmarried teens visited by a nurse (Olds, Henderson, Chamberlin & Tatelbaum, 1986b). This result was corroborated by independent measures of mothers’ interactions with their children, observations of conditions in the home using the HOME scale, medical records, and children’s developmental status at 12 and 24 months of age. The impact of the program on child maltreatment was further moderated by women’s sense of control (or mastery) over their life circumstances when they registered in the program during pregnancy (see Figure 2). For poor, unmarried teenagers, as their sense of control declined, the rates of child maltreatment increased substantially in the comparison group but not in the nurse-visited group. We see the same pattern of results for emergency room encounters (both overall and for injuries and ingestions) during the 2nd year of the children’s lives for the sample as a whole (see Figure 3). (As indicated below, because of this pattern of results in the Elmira trial, maternal psychological resources were hypothesized to moderate program impact in our Memphis replication.) The concentration of effects in the 2nd year of the child’s life makes sense given the dramatic increase in injuries at that time, when children become more mobile and the rates of injuries increase.

Overall, the children of nurse-visited women were less likely to receive emergency room treatment and to visit either a physician or emergency room for injuries and ingestions from their 12th to 48th month of life than were their comparison group counterparts (Olds, Henderson, & Kitzman, 1994; Olds et al., 1986b).

It is important to note that the impact of the program on state-verified cases of child abuse and neglect was attenuated during the 2-year period following the end of the program -- probably due to increased surveillance for child abuse and neglect in the nurse-visited group resulting from the nurses being mandated to report suspected maltreatment and their linking families with needed community services, where their parenting needs were more completely assessed by other service providers. Less serious forms of child maltreatment were thus likely to be detected among the nurse-visited families. (Olds et al., 1994, Olds et al., 1995). While the rate of substantiated cases of child abuse and neglect was equivalent for the nurse-visited and comparison women during the 2-year period following the end of the program, an examination of the living conditions and emergency-department encounters for the "maltreated" children showed that those who were visited by nurses were substantially less serious. Nurse-visited "maltreated" children lived in homes that were more conducive to children’s intellectual and socioemotional development, as indicated by higher HOME scores; the homes of nurse-visited "maltreated" children were substantially safer; and the children themselves had far fewer emergency-room encounters and physician visits in which injuries were detected. We have interpreted these differences as a reflection of greater surveillance for child abuse and neglect in the nurse-visited conditions, leading to more frequent identification of less serious forms of child abuse and neglect in the nurse-visited condition (Olds et al., 1995).

This interpretation has been reinforced with results from the 15-year follow-up (Olds et al., 1997). During the 15-year period after delivery of their first child, in contrast to women in the comparison group, those who were visited by nurses during pregnancy and infancy were identified as perpetrators of child abuse and neglect in 0.21 versus 0.46 verified reports, p=.0006. This effect was greater for women who were unmarried and from low-SES households at registration (p=.0002). The effect of the program on number of verified reports was especially strong for the 4- to 15-year period after the birth of the child — that is, for the period not assessed in previous reports.

3. Prenatal Tobacco Exposure, Prenatal Home Visitation, and Mental Development in the First 4 Years of the Child’s Life

Children born to women who smoked a moderate to heavy amount when they registered in the program during pregnancy and who received prenatal home visitation had significantly higher IQ scores at 3 and 4 years of age than their counterparts in the comparison group (Olds et al., 1994a, 1994b). As shown in Figure 4, control-group children born to women who smoked 10 or more cigarettes per day during pregnancy had mental development scores that declined over the first 4 years of the child’s life, in contrast to their counterparts in the comparison group whose mothers did not smoke during pregnancy (Olds et al, 1996a) (Figure 4 represents the effects of tobacco exposure after statistical adjustment for a host of background characteristics that distinguished moderate-to-heavy smokers from non- or light smokers. The apparent adverse influence of prenatal tobacco exposure was twice as large before these statistical adjustments.) In the nurse-visited condition, children born to women who smoked 10 or more cigarettes at registration during pregnancy had mental development scores in infancy, toddlerhood, and the preschool period that were the same as those who did not smoke at all or who smoked only a few cigarettes per day (Olds et al., 1994b). These beneficial effects of prenatal home visitation held for the group visited only during pregnancy and were not explained by differences in measured aspects of the postnatal environment, given that the estimates of both the impact of prenatal tobacco exposure and the impact of the program controlled for qualities of the home environment measured at annual intervals over the first 4 years of the child’s life. These findings have led us to focus substantially greater attention to the role that an improvement in prenatal health-related behaviors can play in reducing neurodevelopmental impairment on the part of the children.

4. Maternal Life Course

During the 4 year period after delivery of the first child, among low income, unmarried women, the rate of subsequent pregnancy was reduced by 42%, and the number of months that nurse-visited women participated in the work force was increased by 83% (Olds, Henderson, Tatelbaum et al., 1988). By the first-born’s 2nd year of life, the rate of subsequent pregnancy was reduced by 33%. Much of the impact of the program on work force participation among the adolescent portion of the sample did not occur until the 2-year period after the program ended, when the teens were old enough to obtain jobs (Olds et al., 1988). This has important implications for the interpretation of the findings for the first 2 years of life in the Memphis study.

The 15-year follow-up study provides even more compelling results. During the 15-year period after delivery of their first child, unmarried women from low socioeconomic (SES) households at enrollment who were visited by nurses during pregnancy and infancy, in contrast to those in the comparison group, had 1.1 versus 1.6 subsequent births (p=.02), 65 versus 37 months between the birth of their first and 2nd children (p=.001), 60 versus 90 months on welfare (p=.005), 0.41 versus 0.73 behavioral impairments due to substance abuse (p=.03), and 0.18 versus 0.58 arrests by self-report (p=.0009). New York State records revealed that they had 0.16 versus 0.90 arrests (p=.00002; Olds et al., 1997).

5. Cost Analysis

We examined the impact of the program on families’ use and the corresponding cost of other government services (Olds et al., 1993). In 1980 dollars, the program cost $3,173 for 2� years of intervention. We conceived of government savings as the difference in government spending for these other services between the group that received postnatal home visitation and the comparison group. Savings also were expressed in 1980 dollars and were adjusted using a 3% discount rate. By the time the children were 4 years of age, low-income families who were visited by a nurse during pregnancy and through the 2nd year of life cost the government $3,313 less than their counterparts in the comparison group. Thus, when focused on low-income families, the investment in the service was recovered with a dividend of about $180 within 2 years after the program ended (Olds et al., 1993). Figure 5 shows that this $3,300 cost savings to government for low-income families was largely due to reductions in expenditure for AFDC and for food stamps. More recently, the Rand Corporation has conducted an economic evaluation of the program that extends the estimate of cost savings through the children’s entire lifetime and that considers savings to society as well as government (Karoly, Everingham, Hoube, Kilburn, Rydell et al., in press). While there are no net savings to government or society for serving low-risk families, the savings to government and society for serving high-risk families (i.e., those in which the mother is low-income and unmarried) are substantial.

C. Memphis Design and Methods

The Memphis trial was designed to determine if the effects of the Elmira program could be replicated with a large sample of low-income African-American women, children, and their families living in a major urban area and when the program was conducted through an existing health department. Unlike the Elmira trial, the Memphis replication study was an effectiveness study. Efficacy trials test interventions under optimal conditions, whereas effectiveness trials test interventions in contexts that are closer to real-life conditions. In the Memphis trial, the investigators were less involved in the administration of the program. Moreover, the study was conducted during a nursing shortage, which led to fairly high rates of staff turn-over because nurses could earn substantially more in competing hospitals than they could earn as home-visitors in this grant-funded program through the local health department. Given that these kinds of factors are likely to buffet the program if it were administered as an on-going program in new sites, it is particularly useful to examine the Memphis replication in relation to the Elmira trial.

1. Context and Sample

The program was conducted through the Memphis/Shelby County Health Department. From June 1990 through August 1991, 1,139 low-income women who were less than 29 weeks of gestation were recruited from the obstetrical clinic at the Regional Medical Center in Memphis. Women were recruited if they had no previous live births, no specific chronic illnesses thought to contribute to fetal growth retardation or preterm delivery, and at least two of the following sociodemographic risk conditions: (i) unmarried, (ii) less than 12 years of education, (iii) unemployed. At registration, 92% were African-American, 97% were unmarried, 65% were aged 18 or younger. 85% came from households with incomes at or below the federal poverty guidelines, and 9% smoked cigarettes.

2. Randomization and Treatment Conditions

Slightly different assignment ratios and treatment allocation schemes were used during three different time frames over the 15-month period of sample recruitment. This procedure was employed to accommodate shifting expectations about completed sample size (due to the competition with other studies that sampled the same population) and to manage the relatively large number of women enrolled during the first 2 months of the study when only 10 of 12 project nurses had been hired. Consequently, Treatment 1 was added to the design during the 2nd and 3rd allocation periods to reduce the number of families assigned to the nurse-visited

conditions (Treatments 3 and 4) and to reduce workload stress among the nurses. After

completion of informed consent and baseline interviews, identifying information on the participants was transmitted electronically to Rochester, New York, and entered into a computer program that randomized the assignment of women to one of four groups.

Women in Treatment 1 (n = 166) were provided free round-trip taxi-cab transportation for scheduled prenatal care appointments; they did not receive any postpartum services or child developmental assessments/screening. Women in Treatment 2 (n = 515) were provided the free transportation for scheduled prenatal care appointments plus developmental screening and referral services for the child at 6, 12, and 24 months of age. Women in Treatment 3 (n = 230) were provided the free transportation and screening offered Treatment 2 plus intensive nurse home-visitation services during pregnancy, one postpartum visit in the hospital before discharge, and one postpartum visit in the home. Women in Treatment 4 (n = 228) were provided the same services as those in Treatment 3; in addition, they continued to be visited by nurses through the child's 2nd birthday.

For the evaluation of the prenatal phase of the program, Treatments 1 and 2 were combined to form a single comparison group, which was contrasted with combined Treatments 3 and 4, a group that was nurse-visited during pregnancy. For the postnatal phase of the study, Treatment 2 was contrasted with Treatment 4.

3. Assessment Procedures

Assessments were carried out with participating women by research staff members at the time of registration (prior to their assignment to treatments) and again at the 28th and 36th weeks of pregnancy, and the 6th, 12th, and 24th months postpartum. Obstetrical and newborn records were abstracted directly and verified against an on-line perinatal database from the University of Tennessee, which contains information on many of the outcomes of interest for this study. All data were gathered by staff members who were unaware of the women’s treatment assignment.

At intake, women were interviewed to determine their socioeconomic conditions, mental health, personality characteristics, and child-rearing beliefs. Women also completed brief assessments of their intellectual functioning, mental health, and sense of mastery/self-efficacy (this assessment served as a basis for indexing their sense of mastery).

At 6 months postpartum, mothers and children were assessed in the study offices. Mothers were interviewed to assess their rates of breast-feeding and beliefs associated with child abuse and neglect. Mothers and children were observed while the mothers taught their children a developmentally challenging task. The mothers’ behaviors were rated to characterize their sensitivity, responsiveness, and quality of teaching. The infants’ behaviors were coded and aggregated to characterize their responsiveness to and clarity of communication with their mothers.

At 12 and 24 months postpartum, mothers again completed the standardized interviews given earlier. They and their children were also observed using the teaching-interaction procedure. During the home visits, the educational and socioemotional properties of the home environments were assessed. Children’s medical records were also reviewed with a focus on hospitalizations, emergency-room visits, and outpatient encounters in which injuries and ingestions were detected. The dates and types of immunizations were also recorded. Finally, data were extracted from Tennessee Department of Human Services records to ascertain women’s and their first-born children’s use of AFDC during the period from the child’s birth through his/her 2nd birthday.

A follow-up study is currently under way to determine the long-term influence of the program on the maternal life course and child development when children complete kindergarten (at approximately age 6).

D. Memphis Results

The treatment conditions were essentially equivalent at all stages of randomization, and assessments were conducted on a large portion of the women originally assigned to treatment conditions. For example, office-based assessments were completed at 24 months postpartum on 96% of the cases for which there was no fetal or child death.

1. Prenatal Findings

There were no treatment main effects on birthweight, low birthweight, length of gestation, spontaneous preterm delivery, indicated preterm delivery, or Apgar scores. Nevertheless, by the 36th week of pregnancy, nurse-visited women were more likely to use other community services than women in the control group (p= .01). They also were more likely to be working (p=.06), an effect that was particularly strong among women who were not in school when they were randomized (14% vs. 8%, p=.045, and 8% vs. 2% (p=.01), at the 28th and 36th weeks, respectively). There were no program effects on women’s use of standard prenatal care or obstetrical emergency services after registration in the study, but nurse-visited women who were in school at the time of registration had twice as many predelivery hospitalizations as their counterparts in the comparison condition (0.18 versus 0.09, p=.003). This difference was not explained by any coherent pattern of diagnoses associated with those hospitalizations.

In contrast to women in the comparison group, nurse-visited women had fewer yeast infections after randomization and fewer instances of Pregnancy-Induced Hypertension (PIH) (p = .05 and p= .02, respectively). In contrast to those in the comparison group, women with PIH who received a nurse home visitor had mean arterial blood pressures during labor that were 3.5 points lower (p = .05), although there were no treatment differences in birthweight, length of gestation, or Apgar scores for women with PIH.

2. Dysfunctional Caregiving and Child Development

During their first 2 years, nurse-visited children had fewer health-care encounters in which injuries and ingestions were detected than children in the comparison condition (p=.05), an effect that was accounted for primarily by a reduction in outpatient encounters (p=.02). Nurse-visited children also were hospitalized for fewer days with injuries and/or ingestions than children in the comparison condition (p=.0002). As shown in Figures 6 and 7, these program effects on both total health-care encounters and number of days hospitalized with injuries and ingestions were greater for children born to women with few psychological resources (.41 versus .67, p=.003, and .02 versus .26, p=.0002, respectively). Note the similarity in the present pattern of results with child abuse and neglect and emergency-department visits (see Figures 2 and 3) in the Elmira study.

An explanation for this difference in number of days children were hospitalized with injuries can again be found in the nature of their problems. In general, nurse-visited children were hospitalized at older ages and for substantially less serious reasons. The three nurse-visited children who were hospitalized with injuries and ingestions were admitted when they were 12 months of age (and, thus, were mobile), while six (43%) of the 14 comparison children were hospitalized when they were 6 months of age (and, thus, immobile). Eight (57%) of the 14 comparison-group hospitalizations involved either fractures and/or head trauma, while none of the nurse-visited hospitalizations did. Two of the three nurse-visited children were hospitalized with ingestions. Nurse-visited mothers reported that they at least attempted breast-feeding more frequently than did women in the comparison group (p=.006), although there were no differences in duration of breast-feeding. By the 24th month of the child’s life, in contrast to their comparison-group counterparts, nurse-visited women held fewer beliefs about child-rearing associated with child abuse and neglect—lack of empathy, belief in physical punishment, unrealistic expectations for infants (p=.003). Moreover, the homes of nurse-visited women were rated on the HOME scale as more conducive to children's development (p=.003). There was no program effect on maternal teaching behavior, but children born to nurse-visited mothers with low levels of psychological resources were observed to be more communicative and responsive toward their mothers than their comparison-group counterparts (17.9 versus 17.2; p=.03). There were no program effects on the children’s use of well-child care, immunization status, mental development, or reported behavioral problems.

Children born to mothers with limited psychological resources were observed to be more responsive to their mothers and to communicate their needs more clearly than children born to low-resource mothers in the comparison group. We have interpreted the children’s behavior as a reflection of the enduring nature of their relationships with their mothers, with more responsive children indicating relationships in which their mothers were more sensitive and responsive to their children’s needs and less intrusive and hostile (Kitzman et al., 1997). There were no program effects on the children’s rates of immunization, mental development, or reported behavioral problems. The failure to affect immunization rates calls for a closer examination of the clinical protocols employed in promoting well-child care as the program is tested in new settings. Immunization rates approaching 100% could easily be realized simply by including immunization among the nurses’ responsibilities.

3. Maternal Life Course

At the 24th month of the first child’s life, nurse-visited women reported 23% fewer second pregnancies and 32% fewer subsequent live births that women in the comparison group. The program-control difference in subsequent live births was limited to women with high levels of psychological resources, for whom the rates were 14% versus 31%, respectively. Nurse-visited women and their first-born children relied upon AFDC for fewer months during the 2nd year of the child’s life than comparison-group women and their children. There were no program effects on reported educational achievement or length of employment. The program was able to help those women with fewer mental health symptoms, higher IQs, and more active

coping styles become less dependent upon welfare, but it was unable to help women with fewer psychological resources (Kitzman et al., 1997).

E. Comment on Elmira and Memphis Results

This program of prenatal and early childhood home visitation by nurses achieved two of its most important goals -- the reduction in dysfunctional care of children and the improvement of maternal life course, but its impact on a third goal -- the improvement of pregnancy outcomes (in particular, the reduction of preterm delivery and low birthweight) -- was equivocal.

In the Elmira trial, the program produced the anticipated reduction in cigarette smoking, improvement in diet, and increases in women’s use of needed social services and informal social support. There was an increase in the birthweight of infants born to women who were very young (i.e., less than 17 years of age at registration) and a reduction in the rates of preterm delivery from 10% to 2% among women identified as smokers (those who smoked five or more cigarettes per day at registration). It is important to note that 55% of the Caucasian women in the Elmira trial smoked cigarettes during pregnancy.

This impact on preterm delivery and birthweight among young adolescents and women identified as smokers was not replicated in the Memphis trial, although the program did produce the anticipated effects on women’s use of other human services and on the rates of Pregnancy Induced Hypertension (PIH). The absence of corresponding effects on the rates of preterm delivery among smokers in Memphis is probably a reflection of the very low rates of cigarette smoking among African-Americans. Nine percent of the Memphis sample smoked cigarettes overall, and only 7% of the African-Americans in the sample smoked. Moreover, the rates of heavy smoking among African-American pregnant women, in Memphis as elsewhere, is especially low compared to their Caucasian counterparts. Reproductive-tract infections (another major risk for preterm delivery), on the other hand, were much higher among African-Americans. The program did reduce their rate of Pregnancy-Induced Hypertension (PIH), however, an effect that was predicted by the pattern of results found for the Caucasians in Elmira. The prevalence of PIH among Caucasians in the Elmira trial was too low, however, to be statistically significant. We had predicted a corresponding pattern of results among African- Americans in Memphis because the prevalence of PIH among primiparous African-American women is very high (American College of OB/GYN, 1996).

This lack of correspondence between the results of the two trials emphasizes the importance of basing preventive interventions on sound epidemiologic evidence -- that is, a clear understanding of the modifiable risks for the disorder that one wishes to prevent. In this case, the pattern of risks was quite different for Caucasians in Central New York State than for African-Americans in Memphis. While the program can reduce cigarette smoking, it is more of a challenge to affect reproductive-tract infections, such as Gardnerella vaginalis and Ureaplasma urealyticum, given that many of these infections begin prior to pregnancy, are relatively asymptomatic, and are not easily detected outside of office-based medical settings after pregnancy has already progressed (Goldenberg, and Andrews, 1996).

There is some suggestion in the Elmira trial that the program may have reduced the rates of neurodevelopmental impairment associated with cigarette smoking during pregnancy (Olds, 1997; Olds et al., 1994a, 1994b). Given the simultaneous impact of the program on the rates of dysfunctional care and compromised maternal life-course, the program has reduced major risks for early-onset conduct disorder (Moffitt et al., 1996; Olds, 1997b; Olds et al., 1997).

The impact of the program on the rates of dysfunctional caregiving was substantially replicated. Recall that the beneficial effects of the program in Elmira on dysfunctional care (reflected in rates of state-verified cases of child abuse and neglect and on health-care encounters in which injuries were detected) were concentrated on women who were unmarried and from low-SES households. Most were teenagers. Corresponding effects were found in Memphis (where 97% of the sample was unmarried, all were from low-SES families, and more than two-thirds were less than 19 years of age) for health-care encounters in which injures were detected, for observations of the home environments, and for parents’ reports of caregiving and childrearing beliefs. The beneficial effects of the program on caregiving were concentrated among women with lower levels of psychological resources at the time of registration. For example, in contrast to children of the comparison group, children of nurse-visited mothers in Memphis who had few psychological resources were observed to be more responsive and communicative toward their mothers. Infant-attachment research suggests that toddlers’ behavior toward their mothers reveals the extent to which their mothers are sensitive and responsive rather than hostile, intrusive, or neglectful toward them, with toddlers’ behavior being a better indication of the quality of the parent-child relationship over time than currently observed behaviors of parents (Ainsworth et al., 1978; Sroufe & Carlson, 1995).

It is important to note in this regard that the program was designed to decrease risks posed by limited intellectual functioning, mental health, and a limited sense of mastery on the part of caregivers (Olds et al., 1997). The program provided a detailed educational protocol related to maternal and child health and was designed to help parents understand and respond appropriately to their children’s unique temperament and communicative style using educational materials adapted to the intellectual levels of the mothers. Given that limited knowledge and capacity to anticipate children’s needs can compromise parents’ ability to care for their children, it is reasonable that this program would reduce risks imparted by limited intellectual functioning even in the absence of an effect on maternal intellectual functioning.

While the program produced no effect on mothers’ mental health, it may have reduced psychological distress related to parents’ care of their children, which is affected by parents’ depression and sense of competence (Loyd & Abidin, 1985; Zuckerman & Beardslee, 1987). Moreover, by the end of the program at the children’s 2nd birthday, there were significant treatment effects in Memphis on women’s sense of mastery. Mastery is a general psychological attribute that reflects parents’ ability to cope effectively with a wide range of challenges; a deficit in sense of mastery has been associated with child abuse and neglect (Ellis & Milner, 1981). Thus, the program was designed to reduce risks posed by limited intellectual functioning, psychological distress, and a low sense of mastery. The evidence reported here indicates that it indeed diminished the influence of those risks.

Finally, the Elmira program produced dramatic effects on a host of maternal life-course outcomes from the birth of the first child to that child’s 15th birthday. Among women who were unmarried and from low-SES households at registration, those who were visited by nurses during pregnancy and infancy had fewer subsequent children, fewer months on welfare and food stamps, fewer behavioral impairments from use of alcohol and drugs, fewer arrests and convictions, and fewer days jailed during the 15-year period after birth of their first child. Moreover, the program reproduced the most important outcome with respect to maternal life-course in the Memphis replication -- a reduction in the rate of subsequent pregnancy. We should note that the beneficial effects of the program on life-course outcomes for teens in the Elmira trial were not reflected in increased rates of employment, greater educational achievements, or in reduced welfare dependence while the program was in operation (i.e., 2 years postpartum). It was reflected in the reduced rate of subsequent pregnancy, however, which positioned the teen mothers to eventually find work, become economically self-sufficient, and avoid substance abuse and criminal behavior (Olds et al., 1997).

F. Denver Design and Methods

The third trial under way in Denver compares paraprofessional home visitation with nurse home visitation as a means of improving the same maternal and child outcomes studied in the earlier trials, using essentially the same program model. This trial, known as Home Visitation 2000, is important in that many of the recommendations of policy groups call for the establishment of paraprofessional home visitation. The use of paraprofessionals has been recommended for a variety of practical and theoretical reasons. First, it is thought that paraprofessional home visitors would have less social distance from their clients compared to professional nurses and may for that reason, serve as uniquely powerful role models and more easily establish a trusted and empathic relationship with the women they serve. The role of the paraprofessional home visitor has also been discussed as a potential first career step for mothers leaving welfare dependence and entering the work force. Third, paraprofessional salaries tend to be substantially lower than those paid to professional nurses, and some cost savings are presumed when employing paraprofessionals as home visitors, thereby allowing a greater number of families to be served by a limited budget. The use of paraprofessionals as home visitors has been recommended, however, in the absence of solid scientific evidence that such an approach is effective in preventing maternal and child health problems. The evidence from past randomized trials of paraprofessional home visitation programs is not encouraging (Olds et al., 1993) and has been difficult to interpret because of problems with program design and/or implementation or flaws in study design. The evidence that does exist suggests that paraprofessional programs have not been particularly effective (Olds et al., 1993).

In addition to offering the first well-designed test of a prenatal and infancy paraprofessional home visitation program, the Denver trial will provide an opportunity to examine the impact of a model home-visitation program on low-income Hispanic families, in addition to low-income African-Americans and Caucasians. About 40% of the 736 families enrolled in the Denver study are Hispanic.

The 735 women recruited during pregnancy were randomly assigned to one of three treatment groups: Women in Treatment 1 (n = 255) received free sensory and developmental screening for their children at ages 1 and 2 years and referral to services if needed. Women in Treatment 2 (n = 235) received the developmental screening and referral plus nurse home visitation from pregnancy through the child’s 2nd year. Women in Treatment 3 (n = 245 ) received the developmental screening and referral plus paraprofessional home visitation from pregnancy through the child’s 2nd year.

Both nurses and paraprofessionals followed the same comprehensive, intensive program model. The program was augmented by including more recently developed methods of enhancing maternal-role functioning and by making the protocols more appropriate for the multi-ethnic populations in the Denver area and for both types of home visitor. The structure of the program is consistent with the program as it was implemented in Elmira and Memphis. Each visitor carries a caseload of 25 families from intake until the child is 2 years of age. The team of 10 nurses is supported by a supervisor and a secretary, while the paraprofessional team is supported by two supervisors and a secretary.

Nurses were required to have a BSN with experience in community or maternal and child health, and paraprofessionals were expected to have a high-school education, no college preparation in the helping professions, and strong "people skills". Preference for paraprofessionals was given to applicants who had been working for human service agencies. The home visitors selected were an ethnically diverse group representative of the population of women to be served, although the only matching between the race/ethnicity of the visitors and the mothers they served was in the case of monolingual Spanish mothers, who were matched with Spanish-speaking visitors.

The Denver trial is still being conducted, and we expect to be able to report the first results in 1998.

VI. Policy Implications and Program Dissemination

When the results of the Elmira trial were first reported in 1986, many program advocates contended that the program should be disseminated nationally. We, on the other hand, took the position that we needed to determine the extent to which the results from the Elmira program could be replicated in a major urban area with a minority sample and when the program was conducted under less than optimal conditions. This, of course, led to the Memphis replication study. At the time, we had no data on the extent to which the beneficial effects of the program would endure, especially given that the beneficial effects of many preventive interventions that begin during infancy eventually diminish after the program ends (e.g., McCarton, Brooks-Gunn, Wallace, Bauer, Bennett, et al., 1997).

While we decided to wait to determine the endurance and replicability of the Elmira findings, a number of major policy bodies used the data from early phases of the Elmira trial to promote a wide variety of home-visitation programs. The problem with the recommendations of these advisory bodies, however, was that the programs they recommended had little resemblance to the program tested in our randomized trial. The National Commission to Prevent Infant Mortality, for example, recommended that home-visitation services be made available to low-income pregnant women in an effort to improve their health-related behaviors, reduce the rates of low birthweight, and reduce the rates of infant mortality and morbidity. Even though the only results from a randomized trial cited to support this approach came from the Elmira study, the Infant Mortality Commission promoted the dissemination of the South Carolina Resource-Mothers program, a prenatal home-visitation program delivered by paraprofessionals with little resemblance to the program studied in Elmira, Memphis, and Denver.

An analogous process occurred with home-visitation services to prevent child abuse and neglect. The U.S. Advisory Board on Child Abuse and Neglect in 1991 identified child abuse and neglect as a national emergency and identified home-visitation services as the most promising method of preventing child maltreatment. This group also relied on the results of the Elmira trial to support their recommendation that home-visitation services be made available to parents of all newborns. The program that it promoted was Hawaii’s Healthy Start program, a program of paraprofessional home-visitation that begins in the newborn period. As with the Resource Mothers program, there was no evidence from our randomized trials to support the efficacy of this particular program, which again bore little resemblance to the program tested in the Elmira trial. As a result of the recommendations of these two groups, hundreds of home-visitation programs have now been spawned in the U.S. in the 1990s with little evidence supporting their potential efficacy.

At about the time these recommendations were made, we reviewed the randomized trials of home-visitation services for pregnant women and parents of young children (Olds & Kitzman, 1990, 1993). These reviews indicated that home-visitation services vary enormously and that those few programs that produced the largest and most broad-based effects were those that resembled the Elmira program (e.g., focused on at-risk families bearing first children, began during pregnancy, used nurses, followed comprehensive service strategies). Simply sharing one of these characteristics was insufficient. Most programs failed, including programs that had characteristics like the South Carolina Resource Mothers program and the Hawaii Healthy Start program.

These early attempts to apply the findings from the Elmira trial were unsuccessful because the National Commission to Prevent Infant Mortality and the U.S. Advisory Board on Child Abuse and Neglect did not have a nuanced understanding of the differences among home-visitation program models. Our analysis of the literature indicates that unless programs share the essential elements of the program tested in these trials, other programs are not likely to produce the same kinds of results.

Moreover, even when communities choose to develop programs based on models with good scientific evidence, all too often the programs are watered down and compromised in the process of being scaled up. We have recently begun work that addresses this problem.

In 1995, we were invited by the US Department of Justice to disseminate the program in several high-crime neighborhoods around the country. We accepted the invitation because the results from the Memphis replication trial and the Elmira follow-up study were promising. We intend to use the Justice Department initiative to learn more about what is required to develop the program in new communities while maintaining fidelity to its essential elements. Under the Justice Department initiative we are establishing the program in six communities in the country, including Los Angeles, Fresno, and Oakland California; Oklahoma City, Oklahoma; and Clearwater, Florida. A sixth site is still under development.

In this dissemination phase (which will soon expand to include 15-20 additional sites beyond the original Justice Department initiative), state and local governments are securing financial support for the program out of existing sources of funds, such as Temporary Assistance to Needy Families, Medicaid, child-abuse, and crime-prevention dollars. They are making this investment in part because the evidence indicates that the program will reduce future expenditures. This means that the cost of this program, which in 1997 dollars is about $7,000 per family for 2� years of service, can be shared by a variety of government agencies. This, in turn, reduces the strain on any one agency’s budget. While there are less expensive services, they are less expensive because they are less intensive and less comprehensive, and, as noted above, we find no evidence that less intensive services prevent child abuse or neglect, welfare dependence, or crime.

We wish to emphasize that we do not believe that we can disseminate this program on a large scale in a short period of time without compromising its effectiveness. We believe that it makes sense to develop a larger number of demonstration sites only once we have learned from our first set how to develop the program successfully in a variety of new contexts. In this next phase of this work, we are building in provisions for learning about the new implementation efforts so that we can disseminate the program to an even larger number of sites as quickly as is possible without losing program effectiveness.

In general, we believe that policies and practices for young children and their families must be based upon the best scientific evidence available. There is considerable enthusiasm these days about the promise of early preventive intervention programs that current evidence, unfortunately, cannot support. Public hope and confidence in the promise of such programs is a scarce commodity that we dare not squander on approaches that are not likely to work. As health and social welfare policy is redesigned in the near future, we believe that it makes sense to begin with programs that have been tested, replicated, and found to work.

Acknowledgments

The work reported here was made possible by support from many different sources. These include the Administration for Children and Families (90PD0215/01 and 90PJ0003), Biomedical Research Support (PHS S7RR05403-25), Bureau of Community Health Services, Maternal and Child Health Research Grants Division (MCR-360403-07-0), Carnegie Corporation (B-5492), Colorado Trust (93059), Commonwealth Fund (10443), David and Lucille Packard Foundation (95-1842), Ford Foundation (840-0545, 845-0031, and 875-0559), Maternal and Child Health, Department of Health and Human Services (MCJ-363378-01-0), National Center for Nursing Research (NR01-01691-05), National Institute of Mental Health (1-K05-MH01382-01 and 1-R01-MH49381-01A1), Pew Charitable Trusts (88-0211-000), Robert Wood Johnson Foundation (179-34, 5263, 6729, and 9677), US Department of Justice (95-DD-BX-0181), and the W. T. Grant Foundation (80072380, 84072380, 86108086, and 88124688).

We thank John Shannon for his support of the program and data gathering through Comprehensive Interdisciplinary Developmental Services, Elmira, New York; Robert Chamberlin for his contributions to the early phases of this research; Jackie Roberts, Liz Chilson, Lyn Scazafabo, Georgie McGrady, and Diane Farr for their home-visitation work with the Elmira families; Geraldine Smith, for her supervision of the nurses in Memphis; Jann Belton and Carol Ballard, for integrating the program into the Memphis/Shelby County Health Department; the many home visiting nurses in Memphis, and the participating families who have made this program of research possible.