45
On average, the case workers visited the boys in their homes twice a month, and also took the boys to sporting events and other community activities.
In addition, the intervention group boys participated in program-supported camping trips and summer camps. Well intended, the basic idea behind the intervention was to provide friendly, supportive counsel to high-risk boys and their families. Serious crimes were recorded after age seventeen and used as the outcome measure. At the end of the study period, no signifi cant differences were found between the treatment and control group boys (McCord 1992 ; Powers, Witmer, and Allport 1951 ).
However, fi ndings from a thirty-year follow-up suggested that the boys in the treatment group fared worse in adulthood and reported higher rates of violent crime and alcoholism than the control group boys (Dishion, McCord, and Poulin 1999 ; McCord 1992 ). The follow-up study concluded that supportive counseling was not effective, and suggested that aggregating high-risk youth may have a corrupting and deleterious effect (also called deviancy training, see Gifford-Smith et al . 2005 ).
Together with other early studies that focused on distal outcomes (e.g., Berleman, Seaberg, and Steinburn 1972 ; Glueck and Glueck 1950 ; Meyer, Borgatta, and Jones 1965 ), the Cambridge-Somerville study gave rise to a spate of editorials and reviews that were critical of social work, psychology, and other helping professions (e.g., Fischer 1973 ). These editorials and reviews tended to fuse professional affi liation with particu-lar interventions such as, in the case of the Cambridge-Somerville study, supportive counseling and case advocacy. The ensuing dialogue gave rise to renewed interest in social work research (Briar 1974 ; Hudson 1982 ).
Many schools of social work started PhD programs with emphases on research, and MSW training became more focused on evaluating practice (Hudson 1978 ). The development of greater research capability within the profession brought forth rich and impassioned methodological debate on epistemology and methodology (e.g., Harrison, Hudson, and Thyer 1992 ; Witkin 1991 ).
Intervention research emerged during this period of professional self-refl ection, intellectual turmoil, and methodological criticism. At the core, practitioners and researchers wanted to improve service outcomes and better understand how programs work. Both groups were frustrated
with evaluations that seemed to place too little emphasis on understanding the processes operating within interventions. Whether a program was determined to be effective or ineffective, it was usually unclear why this was the case. Evaluations that focused exclusively on outcomes came to be known as black box research because complex intervention processes could not be untangled. When a program was declared effective, all we knew was that a desirable social or health outcome was produced by the intervention. Although the program seemed to work, the data that were collected did not explain the mechanisms that produced the positive outcomes. The processes of the intervention remained as cryptic as a magic act because the researchers could not see into the black box.
Intervention research emerged with roots in both quantitative and qualitative research methods. The fi eld grew from the desire on the part of social work scholars to develop innovative programs and test them rigorously in controlled trials. It grew also from the desire to better understand why programs worked and, when they failed—as did the Cambridge-Somerville program—why they failed (Fraser 1994 ; Fraser, Taylor, Jackson, and O’Jack 1991 ).
With this heritage, intervention research centers both on program outcomes and on hypothesized change processes operating within inter-ventions. To maintain this dual focus, two kinds of conceptualizations underpin the design and development of interventions : problem theory and program theory . Problem theory has to do with understanding the biopsychosocial processes that produce social and health problems.
Typically, this involves considering both individual factors and environ-mental conditions. Although based on problem theory, program theory has to do with specifying and matching intervention methods to a range of proximal and distal outcomes. This matching process involves clarifying the causal logic of an intervention and describing how the intervention activities are expected to produce signifi cant effects.
This chapter focuses on the twin conceptualizations of problem theory and program theory. In the fi rst section, we discuss the identifi ca-tion of social and health problems and the specifi caca-tion of the risk and protective processes that give rise to problems. This perspective may be used to describe problems occurring at the individual, family, group,
organizational, societal, or other levels. The risk and protective perspective is rooted in ecological and systems theories, and draws from the rich literatures of many other disciplines and professions including biology, medicine, nursing, psychology, public health, and sociology. The second section discusses the design of an intervention based on a program theory. Program theories make explicit how an intervention is supposed to function. If a study shows that an intervention is effective, program theory should explain why—it should illuminate the black box.
Developing a Problem Theory
Though micro- to macro-social in character, interventions in social work share a common focus on enhancing human well-being and helping to meet basic human needs (National Association of Social Workers, 2007 ). Interventions usually center on signifi cant social problems such as hunger, mental illness, family violence, or child maltreatment. However, a problem focus does not mean that we subscribe to a pathology perspective. Indeed, many interventions are comprised of activities designed to strengthen protective factors , which are also called assets or strengths. Protective factors operate to disrupt the infl uence of risk factors (Fraser 2004 ). For example, having a supportive and involved spouse may promote a patient’s recovery from a heart attack or other serious illness. Living in a neighborhood where adults monitor children may reduce gang activity-related injuries. These factors function protectively—they reduce vulnerability in the presence of risk. To design and develop an effective intervention, we must clearly specify the problem and the mechanisms that produce or suppress it. These mechanisms are often combinations of risk and protective factors. It is not uncommon for an intervention to concomitantly build strengths (i.e., promote protection) and reduce risk.
Problem theory is a portrayal of the individual and environmental factors—both risk inducing and risk suppressing (i.e., protective)—that give rise to a problem or that sustain a problem over time. We use problem theory to identify leverage points for intervention. In defi ning a problem
clearly, we are often able to work backward to identify these leverage points, and to discover the risk and protective factors that may be mal-leable in intervention. Defi ning the problem is the fi rst step in building the causal logic of an intervention.
What Is the Problem?
Problems are often easier to identify at the individual level. A teenager has no home, and therefore the problem is homelessness. But is home-lessness the only problem? What caused the homehome-lessness? If the root cause is mental illness, homelessness may be a manifestation of an untreated serious mental disorder. Perhaps untreated mental disorders should be the stated problem. If the teen was living on the street, does the problem include drug use or prostitution? Does the problem also include HIV exposure or other serious physical ailments? Even at the individual level, problems are usually complicated. Designing an intervention requires making a strategic decision about where to start. In this case, you might pick homelessness as a starting-point. If you can resolve the homelessness (i.e., the problem of the greatest urgency), you may be able to address the other problems.
Problems can and should be conceptualized at a variety of levels.
Indeed, individuals and their families are always embedded in larger systems that defi ne the parameters of services and resources. Returning to the example of homelessness, living on the street may be an unintended consequence of federal or state decisions to limit spending on mental health care for low-income families. Or homelessness could be a distal function of private insurer decisions to limit mental health-care coverage for insured families. Alternatively, it may be a function of the inability of local law enforcement to protect a young person from sexual exploitation in her home or, if drug abuse is involved, the dearth of adequate residential drug treatment programs for adolescents. The policy context creates environmental conditions and service resources that relate to the prevalence of social and health problems.
Acknowledging the policy context, we usually begin to design an intervention by estimating the prevalence and incidence of a problem.
Prevalence is the proportion of a population that has a problem at a given point in time. Incidence refers to the proportion of new cases in a population within a defi ned period. Incidence is usually expressed as a rate, such as the number of new cases in a year divided by the total population. Incidence can be thought of as the chance that someone within the population will develop a particular problem within a defi ned period.
In contrast, prevalence is expressed as a simple proportion, that is, the percentage of people within a population who experience a problem.
Prevalence data are often available from federal and state agencies.
The Centers for Disease Control and Prevention (CDC) maintains a Youth Risk Behavior Surveillance System that reports national- and state-level prevalences for fi ghting, victimization, drug use, obesity, and other adolescent problems (CDC 2007d). The CDC also maintains a Behavioral Risk Factor Surveillance System for adults. It collects state-specifi c information on asthma, diabetes, health-care access, alcohol use, hypertension, obesity, cancer screening, nutrition, physical activity, tobacco use, and other health problems (CDC 2007a). Similarly, crime data are available from the Federal Bureau of Investigation’s (FBI) Uniform Crime Reporting system (FBI 2007), and the National Institutes of Health (NIH) publishes prevalence and incidence data on a wide variety of topics, such as suicide and mental illness (e.g., National Institute on Mental Health 2007a , 2007b ). These data are useful in describing the dimensions of a problem, including differential risk based on gender, income, race/ethnicity, and sexual orientation. In short, these public data resources may help you make the case for developing a new intervention.
Understanding the dimensions of a problem is the fi rst step in designing an intervention because good prevalence or incidence data provide clues about who experiences the problem. However, demo-graphic data are primarily useful in calling attention to a problem and establishing the need for an intervention. To design an intervention, you need to understand how the problem develops, which includes understanding the risk and protective factors that produce the problem as well as the ways risk and protective factors may vary across populations.
Specifying Mediating Mechanisms
Mapping the interaction of risk and protective factors is akin to specify-ing the mechanisms that mediate social conditions and behavioral or health outcomes. Suppose that you are interested in developing an intervention to improve the social and emotional growth of children from low-income families. In trying to understand the problem (i.e., the social and emotional growth of children in low-income communities), we might develop a framework using the perspective of parenting as a crucial contributor to the growth of children, and poverty as a disorga-nizing infl uence on parenting (e.g., Gershoff, Aber, Raver, and Lennon 2007 ). The following example, in addition to other examples used in this chapter, are drawn from Gershoff and her colleagues (2007), who study the effects of material hardship on child development. A sequential argument, or risk chain, using this perspective might look like:
1. Poverty and material hardship create parental stress 2. Parental stress disorganizes parenting
3. Disorganized parenting affects a child’s social and emotional development.
Problem theory requires identifying targets for change by speculating on risk processes that produce social problems. The risk chain above of-fers many points for intervention. This risk process might be disrupted at any of these points using a variety of programs including those that re-duce poverty and material hardship, those aimed at decreasing parental stress or increasing coping skills, or those intended to alter parenting practices. Speculations that underpin putative risk chains, like the one above, are informed by the scientifi c evidence and theory. When the evi-dence is strong, these speculations may take the form of hypotheses.
Often, we are able to use path charts (see Figure 3.1 ) to create a graphic representation of the active pathways in a risk chain.
Figure 3.1 shows a structural equation model estimated by Gershoff et al . (2007) for the developmental outcomes of U.S. children entering kindergarten. The model includes the elements we outlined
for the sequential argument and adds a protective factor termed parent investment . This term is used to describe the amount of time parents spend with children, parental support for school and extracurricular activities, and, more generally, the academic richness of the home.
Notice that in this model, the pathogenic concept of disorganized parenting has been replaced with the alternative positive parenting behavior, which is regarded as a strength. As in the Gershoff et al . model, conceptual frameworks for social and health problems often contain both risk and protective factors.
To test this model, Gershoff and her colleagues collected data on a nationally representative sample of 21,255 children who entered 944 kindergarten programs in 1998. On the far right side, the fi gure shows distal developmental outcomes of child cognitive skills (i.e., academic achievement measured through vocabulary, math, reading, and general knowledge tests) and socioemotional competence (i.e., child behavior measured through teacher and parent ratings of the child’s social competence, self-regulation, internalizing problems, and externalizing problems). From left to right, the fi gure specifi es the putative risk process, including both risk and protective factors, and shows Gershoff et al .’s estimates of the Figure 3.1 Infl uence of family income and material hardship on child cognitive skills and socioemotional competence. Source: Gershoff et al . 2007, fi gure 3.
Reprinted with permission.
strength of relationships. The numbers associated with each pathway (represented by arrows) range from −1.0 to +1.0. Taken together, these coeffi cients portray the structure of the developmental correlates for child cognitive skills and socioemotional competence; this is one rationale for describing problem theory charts as structural (equation) models.
Using Problem Theory to Build an Intervention
At the start of an intervention research project, problem theory models can be used in two ways. First, prior research may point to pathways leading to a social or health problem of interest. The Gershoff et al . model contains two pathways: (1) a parent investment pathway that leads to child cognitive skills, and (2) a parent stress pathway that leads to child socioemotional competence. In the fi rst pathway, academic achievement appears to be infl uenced largely by a path running from family income to parent investment to child cognitive competence. This pathway is nearly independent of material hardship and parental stress. In the second path-way, child behavior appears to be infl uenced largely by a path running from family income to material hardship to parent stress to positive parenting to child socioemotional competence. This pathway appears to be independent of parental investment. These pathways specify mediating mechanisms for the effect of family income on the cognitive and socioe-motional skills of six-year-old children.
Second, problem theory models identify leverage points. If you were interested in developing a kindergarten intervention to promote cognitive skills and reduce behavior problems, the pathways demonstrated in the Gershoff et al . study would give you an evidence base for two intervention strategies. To promote cognitive skills, you might develop a program to strengthen parent investment. Alternatively, to reduce problem behavior (strengthen children’s social and emotional skills), you might use these pathways as evidence to reduce material hardship and decrease parental stress. You might also address positive parenting behavior; however, based on the pathways fi ndings, you might not expect positive parenting changes to be sustained unless you also intervene to reduce parental stress and material hardship. By specifying the mediating mechanisms between economic conditions, such as family income, and developmental outcomes,
such as child cognitive skills, the research fi ndings from Gershoff et al . (2007) provide an evidence base for the design of an intervention.
To be useful in the design of an intervention, mediating mechanisms should include factors that are malleable. On the far left of the Gershoff et al . model are factors that contribute to family income, which include sociodemographic characteristics over which we have little control in an intervention (e.g., marital status, education, race/ethnicity, and family size). However, the factors in the middle of the model, such as parental investment, parental stress, and positive parenting behavior, are more easily infl uenced. Policy- or program-level interventions might affect family income or material hardship by expanding earned income tax credits (e.g., Okwuje and Johnson 2006 ); by creating individual develop-ment accounts or child savings accounts for low-income families (e.g., Schreiner et al . 2005 ); or by providing conditional cash transfers to low-income parents who make investments in the social, cognitive, and health needs of their children (e.g., Maluccio and Flores 2004 ). Likewise, parental stress might be reduced by organizational-level interventions such as the development of a school-based health clinic that offers positive parenting training among other family services (e.g., Allison et al . 2007 ).
Alternatively, parental stress could be addressed through an individual-level intervention such as the creation of a home-based visiting nurse program to provide support to new parents and to teach positive parenting skills (e.g., Olds et al . 2007 ). Structural models anchor program planning by specifying mediating mechanisms that are action points for the design and development of interventions.
Developing a Program Theory
As previously mentioned, the fi rst step in intervention design is the con-ceptualization of problem theory, which, in turn, forms the basis for a program theory. Described above, problem theory involves understand-ing the structure of a social or health problem. From an intervention perspective, structural models illuminate the mediating processes, and provide important clues for how and when to intervene. A good problem
theory is comprised of mediating constructs that may be changed through program or policy initiatives.
However, problem theories alone are not adequate for planning an intervention because they do not provide enough information. A second kind of conceptualization is needed, which specifi es the ways in which the intervention will change the mediating processes: this is called program theory.
Whether implicit or explicit, all interventions have an underlying program theory. A program theory is “the conception of what must be done to bring about the intended social benefi ts” (Rossi et al . 2003 , 134).
This underlying theory is a portrayal of the causal logic for an intervention.
In one picture, a program theory identifi es program targets (e.g., parental investment); core activities (e.g., skills training, conditional cash transfers);
change or intervention agents (e.g., social workers); and expected outcomes (e.g., academic achievement). Although there are many ways to portray the causal links of an intervention, we describe two frequently used methods: logic models and theories of change.
Logic Models: From Program Inputs to Distal Outcomes
Logic models show the connections between program objectives and inputs and distal outcomes. As Figure 3.2 shows, logic models usually specify an intervention process in terms of core program elements (i.e., objectives, inputs, activities); outputs (i.e., products of program activities); interme-diate outcomes (i.e., changes in mediators); and distal outcomes. Inputs are comprised of the resources needed to implement an intervention.
These might include staff, training, facilities, and equipment costs such as the purchase of treatment manuals or other program materials.
These might include staff, training, facilities, and equipment costs such as the purchase of treatment manuals or other program materials.