You are on page 1of 12

Prevention Science, Vol. 6, No. 4, December 2005 ( C 2005) DOI: 10.


Parent Recruitment and Retention in a Universal Prevention Program for Child Behavior and Emotional Problems: Barriers to Research and Program Participation
Nina Heinrichs,1,2 Heike Bertram,1 Annett Kuschel,1 and Kurt Hahlweg1
Published online: 2 August 2005

Despite the potential of parent training as a prevention and behavioral family intervention strategy, there are a number of important issues related to implementation (e.g., recruitment and retention of families). This paper presents recruitment and retention data from families enrolling in a randomized controlled universal prevention trial for child behavior problems conducted in Germany. The recruitment rate averaged 31% (general project participation), with families of lower socioeconomic status (SES) participating at a lower rate. Project-declining families most often reported intrusion of privacy as their primary concern. In contrast, once parents were enrolled in the project, participation among those randomized to the parent training group averaged 77% (program/intervention participation); nonparticipation was mostly due to logistical issues. Parents accepting the offer of parent training were more likely to report child behavior problems than did declining parents. Although parents from more disadvantaged areas had a lower overall level of participation in the project once recruited, parents with children having higher levels of behavior problems indeed were more likely to participate in the intervention. Different recruitment methods may be required to engage high-risk families from socioeconomically disadvantaged areas to further improve community-level impact on child mental health.
KEY WORDS: implementation; recruitment; triple p; child behavior problems.

Increasing rates of serious youth violence in recent years has led to a surge of research exploring risk and protective factors in children that may increase or decrease the likelihood of such a developmental pathway (Ingoldsby & Shaw, 2002; Loeber et al., 2001; Rutter et al., 1998; Rutter, 1999). Family risk factors, such as poor parenting, family conict and marriage breakdown also strongly inuence childrens development (Forehand et al., 1997; Olson et al., 2002). Moreover, it has been repeatedly shown that children at high-risk for emotional and behav-

Department of Clinical Psychology, Psychotherapy and Assessment, Technical University of Braunschweig. 2 Correspondence should be directed to Nina Heinrichs, Department of Clinical Psychology, Psychotherapy and Assessment, Technical University of Braunschweig, Spielmannstr. 12a, 38106 Braunschweig, Germany; e-mail:

ioral problems, as well as academic difculties and social competence decits, often come from economically disadvantaged homes (e.g., Bradley & Corwyn, 2002; Hart & Risley, 1995; Leventhal & Brooks-Gunn, 2000). Socioeconomic disadvantage in the family (McLoyd, 1998) and neighborhood (Boyle & Lipman, 2002) context, such as poverty, unemployment, low education and single parenthood, impacts child development, and similar risk factors (e.g., low income, social isolation, living in remote areas or in inner city areas, marital problems, parental depression, single parenthood and child behavioral problems) also characterize families with low access to mental health care (Snell-Johns et al., 2004). Thus, these factors put families not only at risk for child behavior problems, but also increase barriers to accessing effective help. Additionally, motivating families to participate in prevention efforts has also been difcult, particularly for family-based interventions (e.g., Prinz et al., 2001).


2005 Society for Prevention Research

Helping parents change their parenting skills is one promising method to positively inuence childrens developmental pathways. Parenting training has consistently been shown to improve parenting practices and reduce conduct problems in children (Greenberg et al., 2001; Webster-Stratton & Taylor, 2001) across treatment modalities (e.g., Sanders et al., 2000; Webster-Stratton, 1990). Compared to the acknowledged positive effects of these interventions, less is known about the mechanisms of action (Hinshaw, 2003). Enhanced parenting, especially improved parental discipline, however, is one of the few well-documented mediators for change in childrens problem behavior (Hinshaw, 2003). However, the potential impact of parent training is currently limited, primarily because of two factors: (a) evidence-based parenting programs are not readily accessible to families, neither in Germany nor in other countries across the world, and (b) the recruitment and retention of families in these parenting programs has been very difcult. The rates of families recruited for familyfocused preventions are typically very low (e.g., Spoth & Redmond, 2000). In addition, only a few published studies even describe the recruitment procedure, enrollment and retention in sufcient detail. We dene the recruitment rate as the percentage of study participants from the population of potentially eligible participants, for example, the number of eligible children in a school, kindergarten or community. A population may be exactly counted (e.g., a program introduced in an entire community or school) or estimated if an exact count is impossible (e.g., a program targeting children who show early signs of behavioral problems). An important distinction must be made between this population-based recruitment rate and the recruitment rate of those families who enroll in the project (e.g., provide informed consent) relative to the sample of interested/contacted participants (sample-based recruitment rate). For example, a number of studies report with how many potential participants they had contacted and how many of those subsequently participated. In contrast to the rst recruitment rate, however, this rate is not placed in the context of the respective population. Finally, attendance rates refer to the number of training sessions attended, usually in an intervention group. The signicance of these rates has been noted repeatedly, though without differentiating between population-

Heinrichs, Bertram, Kuschel, and Hahlweg based and sample-based recruitment rates (Prinz et al., 2001; Snell-Johns et al., 2004; for an exception see Dumka et al., 1997). A review of over 40 frequently cited prevention studies demonstrated that population-based recruitment rates differ by type of prevention (universal, selective or indicated) and target population (child, parent or teacher).3 For example, though universal prevention with parent training has not often been done, one study using a universal-like approach recruited 38% of the parent population (Bronstein et al., 1998). In contrast, programs focusing solely on the child often nd rates in the range of 6697% (Grossman et al., 1997; Furr-Holden et al., 2004, respectively). Selective prevention programs focusing on parents show population-based recruitment rates between 49% (Mason et al., 2003) and 70% (Dumka et al., 1997). Further important variables may be the intervention demands (e.g., number of sessions), the recruitment setting (e.g., low SES factors) or recruitment method. Attendance rates for parent trainings demonstrate that it is common for approximately 50% of the recruited parents to participate in only half (or fewer) of the parent sessions, particularly if there are more than seven parenting sessions (e.g., Barrera et al., 2002; Charlebois et al., 2001). In general, attendance seems to decrease over the length of the program (Reid et al., 1999). The majority of programs have not reported population-based recruitment rates at all, particularly if they chose an indicated prevention approach. Unfortunately, few studies have paid theoretical and empirical attention to recruitment and retention despite its signicance. One exception is a group of researchers at the Iowa State University working on the Project Family. Richard Spoth and Cleve Redmond (Spoth & Redmond, 1993, 1994, 1995, 2000) and others (Spoth et al., 1996, 2000) have attempted to understand what makes families decide to accept or decline participation in two universal prevention programs in economically stressed school

We reviewed more than 40 frequently cited universal, selective and indicated prevention trials for the introduction which we summarized in a Table. It is not a systematic review but it demonstrates how recruitment and attendance rates may be inuenced by the type of prevention (universal, selective, indicated), targeted population characteristics (such as the childs developmental stage or the program target), intervention demands (i.e., how many sessions to commit to), recruitment setting (such as low SES factors) and recruitment method. This Table is available as a PDF document from the rst author (

Parent Recruitment and Retention districts. Spoth and Redmond (1995) developed and tested a two-component model of family engagement in which health beliefs, including perceived susceptibility of the child to behavior problems and perceived program benets and disadvantages, directly impact parents inclination to enroll in a program. Family context factors, such as number of children, household income, past parenting resource use, educational attainment, and level of child behavior problems, comprise the secondary components. Spoth et al. (1996) explored barriers to participation by asking parents about their reasons for refusing to participate in the overall projects assessment procedures (11 assessment-related barriers) or in the prevention program (17 program-related barriers). Fifty-two percent reported scheduling conicts and 42% mentioned being videotaped as assessmentrelated barriers. Program-related barriers were similar, with the major focus on logistical demands put upon families in order to participate. In addition, privacy issues (e.g., not wanting to be videotaped) were particularly important for families of lower SES. In the present study, we were interested in (a) investigating the generalizability of these assessment- and program-related barriers under different program circumstances, and (b) extending our understanding of the factors associated with parent participation in a parent training program to prevent child behavior problems. Below, we elaborate on the similarities and differences between Spoth et al. (1996) and the present study, as this information may be benecial in evaluating the reported results.

In contrast to the Spoth et al. (1996) study, the present study involves an approach in which preschool (and thus family) eligibility status was independent of any risk factors for child behavior or emotional problems (no active attempts were made to have a specic proportion of lower income households). Children were younger (36 years) than in Spoth et al.s study (6th graders) and from a different country. In addition, the telephone procedures for assessing barriers were not feasible in Germany because of strict data protection laws enforced by the youth welfare department. We were not allowed access to any information about non-participants (including phone numbers or addresses); therefore, we assessed barriers through a voluntary questionnaire distributed through the preschools to all families. Further, though both studies differentiated between project assessment and prevention program participation, completion of step one was necessary to move on to step two in this study; participation in each step was independent in Spoth et al. (1996). Finally, while Spoth and colleagues used a three-point scale to assess barriers (yes, somewhat, no), we assessed the presence of barriers dichotomously (yes/no) to make completion of the questionnaire as easy as possible. We expected 50% project participation. This was based on a previous questionnaire study that did not offer participation incentives in the same preschools in which 50% of all parents responded. To detect a clinically meaningful reduction in prevalence of targeted behavior problems with this intervention, we needed 56 families assuming a large effect size (e.g., Sanders et al., 2000) for children with elevated scores on child behavior problem instruments. In comparison, 272 families were needed if the effect size was reduced to 0.4 (for example because more families participate in a trial where no parenting or child problems need to be present to be eligible for the trial). It was hypothesized that the overall recruitment rate would be rather low, with families from lower SES neighborhoods being less represented in the sample than those from higher SES neighborhoods. Based on Spoth et al.s (1996) work, we assumed that time and scheduling factors would be important as barriers for participation for both project and program participation. In contrast, disinclination to be videotaped and perceived invasion of privacy were anticipated to hinder only initial participation in the project.

Similarities and Differences Between Spoth et al. (1996) and the Present Study The questionnaires used by Spoth et al. were translated and only slightly modied to reect differences in the current study. Overall, the same assessment- and program-related barriers were measured. In addition, sample units were randomly assigned to conditions in both studies and families did not know about their assignment prior to their decision to participate or decline participation in the project. Families in both studies were paid for project-related assessments. Furthermore, both studies assessed basic sociodemographic information in addition to participation barriers and differentiated between assessment- and program-related barriers.

METHOD General Study Design The present investigation was embedded in a randomized examination of a universal prevention efcacy study which targeted families with children aged 36 years. In an intervention/control group design, the parent training Triple P was introduced to families randomized to the experimental group; the control group was not offered training and was naturally observed for the course of the study. Triple P is a multilevel parenting and family support strategy developed in Australia by Sanders and colleagues (e.g., Sanders, 1999). It provides ve levels of intervention of increasing strength and specicity. Level 1 targets all families interested in information about parenting and promoting their childs development. Poster, yer, brochures, or media campaigns are strategies used on this level. On levels 2 and 3, Triple P targets families with a specic concern about their childs development. Strategies used are booklets, tip sheets or videos and on the third level also some active skills practice. One level further ahead, parents wanting more intensive training are addressed with Standard Triple P, which teaches positive parenting skills and generalization enhancement strategies. Finally, on level 5 behavioral family intervention is provided for families with already existing child behavior problems not resolved by any of the previous levels. Among the key features of Triple P is the exibility of delivery format (group, individual, selfdirected), the multidisciplinary approach (media, primary care, psychologists), and the self-regulation framework (parents are supported to enhance their self-sufciency and self-efcacy). In terms of design, Triple P is the only known parenting and family support strategy that attempts to create an optimal system of clinical, indicated and universal prevention strategies under the roof of a unied theory, as recommended by Offord and colleagues (1998) to lower the burden of suffering. Families were assigned to the experimental or control group based on preschool afliation. Preschools were randomized to the two conditions after being matched according to the social structure of their respective neighborhoods. Social structure was determined by an objective kindergarten social structure index (OKS) derived in collaboration with the Census Bureau that consists of four indices: rate of unemployment, number of families on welfare, number of immigrants and quality of housing

Heinrichs, Bertram, Kuschel, and Hahlweg in the particular neighborhood (Base, 1995). While, for most readers, the rst three indices likely contain some face validity as indicators of social structure, the number of immigrants may not. This rate was used only as an indirect indicator for social structural problems of the area. Many immigrants and refugees in Germany are forced to live on welfare because they do not immediately receive the needed permission for employment upon arrival. Without this permission, they are unable to seek or nd work and must be nancially supported by the state. Furthermore, immigrants in Germany usually live in the same neighborhoods together, decreasing chances for integration with Germans and enhancing social problems within this area. Thus, even though the number of immigrants enters the OKS equation, it is used to more generally mirror the social structural difculties encountered by these groups. The parent training for the experimental condition consisted of four weekly group sessions of 2 h each, and four optional 15-min phone contacts made on a weekly basis. In dual-parent families, both parents were invited to participate in program sessions. However, since children did not attend the trainings, dual-parent families usually left one parent with the child(ren) while the other attended the session. Attendance by one parent at one program session was sufcient to be considered program participation. Assessments were conducted prior to beginning the parent training, after completing the program, and one and two years after pre-assessment.

Recruitment Procedure We rst contacted all potentially eligible preschools in the city of Braunschweig (N = 33). Project staff members were present at preschool teacher meetings and explained the project to preschools. Twenty-three preschools (70%) expressed interest in participating in the project. Interested and non-interested preschools did not signicantly differ in OKS [2 (2) = .95, p > .62, V (Cramers statistic) = .17]. Seventeen of these interested preschools were then randomly selected to participate in the project (due to lack of project manpower) and assigned to one of the two conditions. The project was then presented to the families. Repeated project announcements at preschools were organized, parentteacher meetings about the project were held, and posters about the project were pinned on preschool halls. Finally, two staff members

Parent Recruitment and Retention were present at any special preschool activities (e.g., summer festivals). Families received information about the course of the project, the study conditions (developmental/control versus prevention program/experimental), home visit procedure, and nancial reimbursement. Interested families could enroll at any time through their preschool. It is important to recognize that participants could decline participation at two steps in the recruitment process. First, the project was described to the families who then decided whether to participate, recognizing that they might be assigned to either of the two conditions. Second, once they learned of their assignment, families in the experimental condition could elect to participate in or decline the parent training. We dene the rst step as project participation and the second step as program participation. Recruitment of families occurred in two waves (wave 1: 05/2001-09/2001; wave 2: 02/2002-04/2002). Inclusion criteria were the childs age (2.66.0 years) and parents German language ability. The nal eligible population for the project included 915 families. The entire recruitment process is presented graphically in Fig. 1 and can be best described as a two-step process (step 1: project participation and step 2: program participation).

follow-up assessments, 10 Euro for the reduced postassessment with self-report instruments only), and parents received feedback about the results of the assessments.

Teacher Ratings Each preschool teacher team was asked to rate each family in their group on a number of sociodemographic variables (age of parents: <25, 2535, 3645, >45; family status: living in a partnership, one-parent household; parent occupation: workers, employers, public servant, self-employed, other; number of people in the family; estimated social status of the family: low, middle, high; and family on-welfare: yes, no). The teachers evaluated all children in their preschool with this instrument (N = 1153). The total sample involved both participants (n = 282) and non-participants (n = 871). This sample of project non-participants, however, includes true non-participants (n = 633) as well as families who did not meet inclusion criteria (n = 238; see Fig. 1). These ratings were completed anonymously for nonparticipants because of youth welfare department restrictions. Thus, we were unable to analyze these two groups separately. We collected data about all families through the teachers instead of parents in order to gather information about the population of preschool families. However, it is questionable how valid the ratings of teachers are. Although they have up to daily contacts with the primary caregiver of the child (e.g., at dropping off and pick-up times, at parent meetings, for organization of excursions, etc.), it is unknown if these teacher ratings proved to be an adequately reliable way to secure data for the population of eligible participants. Since participating families were assigned a code number (and non-participating families were not), and we collected sociodemographic information from participating families during the initial home visit, we compared the quality of the ratings provided by the teachers for those families with the information provided by the families themselves. Unfortunately, we did not assess all relevant variables in the same way. We report results only for variables assessed equally in both samples. Spearman-rho correlations ranged from .60 (age of parent) to .91 (number of family members in the household) with most of the variables yielding correlations equal to or above .74. The amount of exact agreement was 96, 90, 89,

Instrumentation Assessment of Families The project-related assessments for each family consisted of a battery of self-report questionnaires (N = 11, including the Child Behavior Checklist for ages 1.55; Achenbach & Rescorla, 2000) and diagnoses checklists (N = 2). Two project staff members also visited each home for approximately 2 3 h to complete an interview, child developmental test (Kaufman Assessment Battery for Children, KABC, Kaufman & Kaufman, 2001) and videotaped parentchild interaction task. The self-report measures were completed by both parents in dual-parent families, whereas the interview and the parentchild interaction were usually completed by the mother (97%). The multi-method assessment is modeled on other large prevention trial studies, such as Fast Track (Conduct Problems Prevention Research Group, 1999a, 1999b) and Sanders et al. (2000). Compensation for time and effort for the assessments was provided (50 Euro for pre-, and 12-month


Heinrichs, Bertram, Kuschel, and Hahlweg

Fig. 1. Course of recruitment in the Project Zukunft Familie (Future Family).

74 and 67% for family status, on welfare, number of family members in the household, age and occupation of parents. Thus, the estimation error ranges from 33 to 4%.

Family Non-Participation Survey Finally, non-participants were asked to complete two survey instruments (modied from Spoth

Parent Recruitment and Retention et al., 1996). One consisted of 15 barriers to project participation (assessment-related barriers) and the other consisted of 16 barriers to program participation (program-related barriers). Additional sociodemographic information was also obtained. The surveys for project decliners were distributed by the preschool teachers and returned anonymously. Only non-participants who were initially eligible for the study (n = 633) received the survey. Each responding family that declined to participate in the project returned one survey, yet anonymity precluded knowledge of the gender of the completing parent. Surveys regarding program non-participation were sent directly from project staff to the families and were mostly completed by mothers (92%).

df = 2, p < .001. Subsequent analyses showed that only the recruitment rate in high SES neighborhoods differed from participation rates in low or moderate SES neighborhoods (high-middle: 2 = 17.9, df = 1, p < .001; middle-low: 2 = 1.5, df = 1, p > .21). Given that participants are nested within preschools, we also analyzed whether there were signicant differences in participation rates across preschools that were not accounted for by the social structure index. In the low OKS preschool sample, one preschool signicantly differed from the ve other preschools with an outlier recruitment rate of 58% (recruitment range of the other ve preschools: 929%). In the moderate OKS preschool sample, no differences between the ve preschools occurred (recruitment rates 1937%). In the high OKS preschool sample, recruitment ranged from 27 to 64% with a bimodal distribution: three preschools ranged from 27 to 37% and the other three preschools ranged between 50 and 64%. All preschools with higher or lower recruitment rates, respectively, bordered neighborhoods with moderate OKS.

RESULTS Statistical Analyses First, we analyzed the recruitment rate and predicted project participation with a logistic regression. For this analysis, families from all participating preschools (N = 17) were included. Second, we analyzed barriers to participation in the project. Finally, we predicted program participation with a logistic regression and analyzed barriers to program participation. Only families who had been offered the program were included in these analyses (N = 186). Recruitment Rate: Project The total population consisted of 915 eligible participants. Across waves, 282 families (31%) enrolled in the project. The second wave replicated the results from the rst wave not only in the recruitment rate (see Fig. 1) but in all other results reported below, demonstrating considerable consistency across recruitment waves. We only report the results for the total sample. The social structure index of the preschools neighborhood (OKS) was inversely related to the participation of families in the project; in preschools located in neighborhoods with a high or moderate number of social problems (low or middle OKS), only 69 (23%) out of 298 families and 95 (27%) out of 347 families participated, respectively. In contrast, in neighborhoods with few social problems (high OKS), 118 (44%) out of 270 families participated. These recruitment rates differ signicantly, 2 = 31.2,

Differences Between Project Participants and Non-Participants A logistic regression predicting project participation and including a number of sociodemographic variables (single parent, occupation, social status of family and of the neighborhood) and parent/family characteristics (number of family members in the household, age of parents in years) showed that being in a single-parent home made the parent 1.56 times (condence interval, CI = 1.052.32) more likely to participate in the project when controlling for occupation, social status on a family and neighborhood level, family members in the household and parents age. Parents from a low SES family or neighborhood were 0.27 (CI = .14.51) and 0.49 (CI = .34 .72) times less likely, respectively, to participate in the project when considering these other sociodemographic variables, and those from middle SES families or neighborhoods were 0.50 (CI = .34.73) and 0.69 (CI = .49.97) times less likely, respectively. Parents occupation was signicantly related to participation, but this relationship became non-signicant when adjusting for the other predictors. Number of family members in the household and parents age were not signicant predictors.

Reasons for Non-Participation in the Project ( "Assessment-Related Barriers")4 Of 633 non-participating families, 253 (40%) reported what they considered the main project barriers. Ten returned survey instruments were uninterpretable. Results showed that most families did not participate because they were afraid of someone intruding into their privacy (e.g., being videotaped: 29%, having someone come into their homes: 25%, etc.). Since only 40% of all non-participating families returned the anonymous survey instrument, one could argue that the sample might not be representative. However, comparison of these responders with the total population of non-responders rated by teachers (N = 871) yielded no signicant differences in sociodemographic variables (alpha adjusted to < .017), such as age (<25, 2535, 3645, >45: 2 = 1.8, df = 3, p > .60), family status (single or in a relationship: 2 = .15, df = 1, p > .69), or number of family members living in household (2, 3, 4, 5, or more than 5: 2 = 9.8, df = 3, p = .02).

Heinrichs, Bertram, Kuschel, and Hahlweg

Table 1 Proportion of Children in the Elevated Range of the Child Behavior Checklist (CBCL) and of At-Risk Families in Recruited Sample (N = 280) Risk denition Proportion of families (%)

Clinical range (CBCL > 90th percentile) Internalizing 9.8 Externalizing 6.2 Total score 8.0 Elevated CBCL scores (CBCL > 85th percentile) Internalizing 18.1 Externalizing 14.5 Total score 14.5 Family adversity factors 90.4 Low risk (maximum of 1 risk factor) High risk (2 risk factors) 9.6 Note. CBCL data from mothers only, N = 4 missings with n = 1 single-parent father.

Differences Between Program Participants and Non-Participants A second regression predicting program participation included social status of the neighborhood, income, parent education, single parent, child gender, and child internalizing and externalizing problem behavior as predictors. Being in a single-parent home made the parent one-fourth as likely to participate in the program (OR = .25, CI = .10.62) after controlling for the other variables. Furthermore, being from a low SES neighborhood made the parent approximately one-third (OR = .31, CI = .13.75) as likely to participate in the program. In contrast, more externalizing child behavior made a parent 1.06 times more likely to participate in the program (CI = 1.03 1.12). Parent education, income, internalizing problem behavior and the childs sex were not signicant predictors. Reasons for Non-Participation in the Prevention ( "Program-Related Barriers")4 Of the 43 families who chose not to participate in the parent training, 23 (54%) returned the questionnaire about barriers to participation. The most frequent barriers to program participation were scheduling conicts and time demands (57%). DISCUSSION

Recruitment of Families Reporting Child Behavior Problems Table 1 reports the proportion of children in the clinically elevated and at-risk range of child behavior problems. The present sample size of N = 280 entails between 28 and 50 high-risk children, depending on the denition used. Retention Rate: Participation in the Prevention Program and Attendance Overall, the average program participation rate was 77%. In the rst and second wave, 69 (out of 90; 76.6%) and 75 (out of 96; 78.1%) joined the parent training, respectively. Most of these families participated in three or four group sessions (89%). Seventyeight percent took advantage of the opportunity of a voluntary and individual phone contact with their group facilitator at least once.

Item-by-item comparisons between Spoth et al. (1996) and our sample on reasons for non-participation in the project and the program, respectively, are also available as a PDF document.

The overall recruitment rate for this project (31%) reects an important issue to consider in

Parent Recruitment and Retention prevention efcacy studies focusing on parents: only one-third of parents agreed to be involved in the rst place. This nding was robust, as it was replicated in the independent second recruitment wave. As pointed out by others (Guyll et al., in press), this result poses challenges for interpreting effects, as well as for disseminating such prevention programs. Response and non-participation rates are important for two reasons: (a) the generalizability of the ndings is greatly inuenced by the participation rate, and (b) they may point to the need for altering recruitment and subsequent assessment strategies. Participation in the present study (project participation) depended signicantly on the social structure index of the preschools. In addition, project participants had a higher family SES. Despite this inuence of socioeconomic status (SES) on a neighborhood- and family-level, it is also important to realize that some variance in recruitment rates across sites within the same social status level occurred. Preschools demonstrating different recruitment rates than the rest of their OKS siblings were usually located in the outskirts of the respective geographical area. Thus, their catchment areas may have overlapped with neighboring, more moderate OKS preschools. Although more disadvantaged communities produce more children with mental health problems compared to less disadvantaged communities, most children in each SES level do not develop mental health problems. In addition, the majority of cases of both internalizing and externalizing child behavior problems do not reside in the most impoverished neighborhoods even though they are proportionately overrepresented. Thus, despite a universal parenting strategy having greater difculties in engaging the most disadvantaged neighborhoods such a strategy still has a signicant population-level impact on the prevalence of behavioral and emotional problems in the larger community because the majority of children who actually develop mental health problems are not from the lowest SES groups but from middle class homes. Offord and colleagues (1998) provide an intriguing example for this focus on public health impact: In the Ontario Child Health Study (Offord et al., 1987), children living in families with an income level below $10,000 were at high risk for behavioral and emotional problems (36% qualify for caseness). However, only 7% of children belonged to this family-income category. Thus, socially disadvantaged children accounted for only 14.5% of the population of behaviorally or emotionally disturbed

children. In fact, Offord and colleagues (1998) reviewed the trade-offs among intervention and indicated versus universal prevention in their impact on lowering the burden of suffering in children and families. Among the cardinal advantages of universal programs are that small effects on an individual level may translate into large effects in the population and the de-stigmatization of enrollment. However, the ip-op side is that participants (and health care providers) may be less motivated and therefore hard to recruit and retain in a program. Offord and colleagues (1998) offered some advice in the decision process of what level or type of program may be most appropriate under particular circumstances; however, the ideal healthcare paradigm seems to be a multilevel system in which all levels of intervention are integrated and realized within a community. Evaluating the present recruitment rate is challenging because few studies exist for comparison. Most studies either introduce universal prevention programs that target the child through school classes or offer a program to families who are at-risk or already exhibit difculties (selective or indicated prevention). Recruiting families into these kinds of programs is different from recruitment for a universal prevention parenting program. In the rst case, the prevention program can be built into the regular curricula and may thus elicit more parental cooperation, especially when the intervention is primarily directed at the children. In the second case, the child or the family as a whole already suffers from some problems. According to the theoretical model of Spoth and Redmond (1995), higher levels of reported child behavior problems and lower levels of household income indirectly inuence inclination to enroll (see Offord et al., 1998; Spoth et al., 1996). It is not surprising that selective/indicative prevention parenting programs therefore report higher enrollment rates than universal prevention programs where no such difculties necessarily exist. Other studies attempting to introduce a parent-focused universal training (and reporting sufcient recruitment information for analysis) yielded similar results to the present study (e.g., Bronstein et al., 1998: 38%). School-based recruitment may be more feasible if the intervention is child-focused and does not require active involvement of parents. Between 66% (Grossman et al., 1997) and 97% (Furr-Holden et al., 2004) of parents provide their consent for their child to be involved in a school-based prevention program; however, the rate decreases considerably if parental involvement in a program is necessary.

Recruitment may therefore still be a major problem with school-based projects if active parental involvement is required. This is important because there is now good evidence arguing for the introduction of prevention trials early in a childs development (e.g., Metropolitan Area Conduct Research Group, 2002). The earlier the prevention programs are introduced in a childs development, the more important the parents are. Therefore, we believe recruiting parents into prevention programs will remain among the biggest challenges for prevention of child behavior problems. Analysis of the barriers to initial participation in this project demonstrated that many families who might otherwise choose to participate opt to forego participation in prevention research because of the emotional demands placed upon them (such as having an unknown psychologist entering their private residence, being videotaped). Overall, we obtained very similar reasons for non-participation as Spoth et al.s (1996) study. Although the frequency for each item may deviate slightly across these two studies (e.g., Item 1: I did not want to be videotaped: 29% versus 42% and Item 3 I did not want my family to be studied: 25% versus 14%), summarizing ndings across items of privacy concerns shows similar results.4 We thus conclude that similar barriers hinder families from participating in research projects in the USA and Germany, independent of the exact focus of prevention (e.g., prevention of adolescence substance use or prevention of behavior problems in childhood). The nding that the majority of nonparticipating families nd the multimodal, multimethod assessment burdensome and at times intrusive deserves special consideration. In neighborhoods with salient social problems, the resistance to outside involvement in family affairs was especially powerful. It could be that parents living in these neighborhoods are more suspicious and doubtful towards strangers who show (for subjectively no reason) interest in their family because they have had bad experiences in the past (e.g., strangers showing interest and then threatening to take away their child). In any case, investigators must strike a realistic balance between their needs for a thorough, detailed assessment and their desire to include a broader array of families. In fact, these results may also mean that we need to develop new methods of studying families that are valid and reliable, yet accommodate the expressed needs of these families.

Heinrichs, Bertram, Kuschel, and Hahlweg The recruitment rate for the prevention program, in contrast to that of the project, was 77%. In addition, more than 88% of families attended at least 75% of the sessions. Other prevention studies targeting parents usually report similar program recruitment rates but lower attendance rates. However, these studies typically offered parent support to families in which the child already showed some behavior problems. We believe that this difference in program participation is due to the length of the offered program: the more sessions that are required, the less likely parents will attend regularly. Alternatively, parents with a lack of interest in the present prevention program may have already decided at an earlier recruitment stage (i.e., in the rst stage) that they would not participate in the project. We also found a signicant difference between program participants and non-participants, with nonparticipants more often stemming from neighborhoods with a lower social structure index and reporting fewer child behavior problems. This may indicate that once families are recruited, those who need the parent training most (i.e., higher CBCL score) gets self-selected into the program. However, the present study design only allows conclusions about the 23% who initially agreed to participate in the project and later declined the offer of a parent training. Recruitment barriers may be different if all potential families from the initial recruitment stage knew that they would be in a parent training. In contrast to barriers to the initial project participation, parents felt most deterred from program participation by time conicts. This was the main reason across families and indicates that exible program delivery is required in order to facilitate parents participation (c.f., Spoth et al., 1996; Spoth &Redmond, 1994). Although it is the goal of the present paper to explicitly note and discuss the issue of initial recruitment rates, the low rates nevertheless affect condence in the results of a study; when the goal is to generalize about a whole group, yet about two-third of this solicited group refuses to participate in the project, generalizability of results is likely to be impaired. As we attempted to illustrate throughout this paper, we believe this problem of initial low recruitment is not specic to our study but a general issue that needs to be more explicitly addressed and discussed in publications of preventive interventions. When evaluating the present results, it is also necessary to consider the specic setting, the recruitment strategy (e.g., two-step procedure with rst step obligatory to move forward to step two, nancial

Parent Recruitment and Retention reimbursement for assessment, no regular child care available) and the measurement requirements (e.g., long home visits, questionnaires as well as interaction observation). These specics may limit the generalizability of the ndings. In sum, the key issue is to determine what recruitment rates are adequate for prevention. In the absence of evidence-based guidelines, we suggest that the present recruitment effort was clearly effective in recruiting many suitable parents to participate in a parent training. Even with a skew favoring high SES areas, we still enrolled a number of families at risk. Moreover, if we were to apply our recruitment rate to other areas of health promotion (such as exercise, smoking cessation or nutritional approaches), the rate would be considered moderate to high. From a public health perspective, combining an intervention that is not only effective but also efcient, such as Triple P, with universal community outreach, as done in the present study, a 31% project recruitment rate seems promising, particularly if we consider the increased likelihood of families with more child behavior problems to participate in the training. Thus, a favorable self-selection occurs, with those families experiencing more problems also more often accepting help within the context of a non-stigmatizing parent training offer. Future recruitment research on universal prevention programs should focus on reasons for, and not only against, participation. If a child is ne and the parents do not have had any experience with conduct problems in their family why should they enroll in a program? It may be helpful to explore why so many parents still participate despite the demands on their time. However, we should also realize that experiences with conduct problems should not become the main argument for participation in parenting programs. Just as childbirth classes are useful for mothers with and without expected difculties, so can parenting programs be useful to families that do not yet experience signicant child behavioral problems. Parenting is something every parent has to go through (as child birth is), one of the differences however is that parenting is a lifelong task while child birth is usually over after a limited amount of time. Childbirth classes have been normalized and de-stigmatized as part of the typical parenting experience; perhaps normalizing parent training could lead to a different set of participation predictors, particularly if participation is linked to other family benets (e.g., universal health care or tax rebates).

The differences between the Spoth et al. (1996) and the present study invite the question as to whether different or similar outcomes should be expected. To our knowledge, there is no data available to convincingly argue either side. The present study adds to the literature by demonstrating fairly similar barriers across these two studies despite remarkable differences in design. This is a promising result because it is initial evidence for cross-national similarities in prevention participation and hints at the powerful accumulation of knowledge that may be possible around the world. Therefore, research in family recruitment may benet from further international collaboration to overcome these barriers and help families make use of prevention offers in order to better promote child mental health. ACKNOWLEDGMENTS This study was supported by the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation), HA 1400/14-1. We are very grateful to Don Baucom and Lynlee Tanner for their comments on earlier versions of this manuscript. REFERENCES
Achenbach, T. M., & Resorla, C. A. (2000). Manual for the ASEBA preschool forms and proles. Burlington: University of Vermont. Barrera, M., Jr., Biglan, A., Taylor, T. K., Gunn, B. K., Smolkowski, K., Black, C., et al. (2002). Early elementary school intervention to reduce conduct problems: A randomized trial with hispanic and non-hispanic children. Prevention Science, 3, 8394. Base, B. (1995). Die sozialr aumliche Gliederung der Stadt Braunschweig. Methodik und Durchfuhrung sozialgeographischer Analysen im st adtischen Wohnumfeld auf Grundlage des Zensus von 1987 [The social-ecological structure of the city of Brunswick. Method and procedure of socio-geographical analyses in urban living areas based upon the census of 1987]. Braunschweig. Dissertation, Technische Universitat Bronstein, P., Duncan, P., Clauson, J., Abrams, C. C., Yannett, N., Ginsburg, G., & Milne, M. (1998). Preventing middle school adjustment problems for children from lower income families: A program for aware parenting. Journal of Applied Developmental Psychology, 19, 129152. Boyle, M. H., & Lipman, E. L. (2002). Do places matter? Socioeconomic disadvantage and behavioral problems of children in Canada. Journal of Consulting and Clinical Psychology, 70, 378389. Bradley, R. H., & Corwyn, R. F. (2002). Socioeconomic status and child development. Annual Review of Psychology, 53, 371 399. Charlebois, P., Vitaro, F., Normandeau, S., & Rondeau, N. (2001). Predictors of persistence in a longitudinal preventive intervention program for young disruptive boys. Prevention Science, 2, 133143.

Conduct Problems Prevention Research Group. (1999a). Initial impact of the Fast Track prevention trial for conduct problems: I. The high-risk sample. Journal of Consulting and Clinical Psychology, 67, 631647. Conduct Problems Prevention Research Group. (1999b). Initial impact of the fast track prevention trial for conduct problems: II. Classroom effects. Journal of Consulting and Clinical Psychology, 67, 648657. Dumka, L. E., Garza, C. A., Roosa, M. W., & Stoerzinger, D.H. (1997). Recruitment and retention of high-risk families into a preventive parent training intervention. The Journal of Primary Prevention, 18, 2539. Forehand, R., Miller, K., Dudra, R., & Chance, M. W. (1997). Role of parenting in adolescent deviant behavior: Replication across and within two ethnic group. Journal of Consulting and Clinical Psychology, 65, 10361041. Furr-Holden, C. D. M., Ialongo, N. S., Anthony, J. C., Petras, H., & Kellam, S. G. (2004). Developmentally inspired drug prevention: Middle school outcomes in a school-based randomized prevention trial. Drug and Alcohol Dependence, 73, 149 158. Greenberg, M. T., Domitrovich, C., & Bumbarger, B. (2001). The prevention of mental disorders in school-aged children: Current state of the eld. Prevention & Treatment, 4, Article 1. Retrieved from volume4/pre0040001.html Grossman, D. C., Neckerman, H. J., Koepsell, T. D., Liu, P., Asher, K. N., Beland, K., et al. (1997). Effectiveness of a violence prevention curriculum among children in elementary school. JAMA, 277, 16051611. Guyll, M., Spoth, R., & Redmond, C. (in press). The effects of incentives and research requirements on participation rates for a community-based preventive intervention research study. Journal of Primary Prevention. Ingoldsby, E. M., & Shaw, D. S. (2002). Neighborhood contextual factors and early-starting antisocial pathways. Clinical Child and Family Psychology Review, 5, 2155. Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children. Sydney: Paul H. Brooks. Hinshaw, S. P. (2003). Intervention research, theoretical mechanisms, and causal processes related to externalizing behavior patterns. Development and Psychopathology, 14, 789818. Kaufman, A. S., & Kaufman, N. L. (2001). Kaufman assessment battery for children (K-ABC). Deutschsprachige Fassung von P. Melchers & U. Preu. 5., korr. und erg. Auage. Leiden: Swets & Zeitlinger. Leventhal, T., & Brooks-Gunn, J. (2000). The neighborhood they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309337. Loeber, R., Farrington, D. P., Stouthamer-Loeber, M., Moftt, T. E., Caspi, A., & Lynam, D. (2001). Male mental health problems, psychopathy, and personality traits: Key ndings from the rst 14 years of the Pittsburgh youth study. Clinical Child and Family Psychology Review, 4, 273297. Mason, W. A., Kosterman, R., Hawkins, J. D., Haggerty, K. P., & Spoth, R. L. (2003). Reducing adolescentsgrowth in substance use and delinquency: Randomized trial effects of a parent-training prevention intervention. Prevention Science, 4, 203212. McLoyd, V. C. (1998). Socioeconomic disadvantage and child development. American Psychologist, 53, 185204. Metropolitan Area Conduct Research Group. (2002). A cognitiveecological approach to preventing aggression in urban settings: Initial outcomes for high-risk children. Journal of Consulting and Clinical Psychology, 70, 179194. Offord, D. R., Boyle, M. H., Szatmari, P., et al. (1987). Ontario Child Health Study: Six month prevalence of disorder and rates of service utilization. Archives of General Psychiatry, 44, 832836.

Heinrichs, Bertram, Kuschel, and Hahlweg

Offord, D. R., Kraemer, H. C., Kazdin, A. E., Jensen, P. S., & Harrington, R. (1998). Lowering the burden of suffering from child psychiatric disorder: Trade-offs among clinical targeted, and universal interventions. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 686694. Olson, S., Bates, J., Sandy, J., & Schilling, E. (2002). Early developmental precursors of impulsive and inattentive behavior: From infancy to middle childhood. Journal of Child Psychology and Psychiatry, 43, 435447. Prinz, R. J., Smith, E. P., Dumas, J. E., Laughlin, J. E., White, D. W., & Barron, R. (2001). Recruitment and retention of participants in prevention trials involving family-based interventions. American Journal of Preventive Medicine, 20, 31 36. Reid, J. B., Eddy, J. M., Fetrow, R. A., & Stoolmiller, M. (1999). Description and immediate impacts of a preventive intervention for conduct problems. American Journal of Community Psychology, 27, 483517. Rutter, M. L. (1999). Psychosocial adversity and child psychopathology. British Journal of Psychiatry, 174, 480493. Rutter, M. L., Giller, H., & Hagell, A. (1998). Antisocial behavior by young people. New York, NY: Cambridge University Press. Sanders, M. R. (1999). Triple P-Positive Parenting Program: Towards an empirically validated multilevel parenting and family support strategy for the prevention of behavior and emotional problems in children. Clinical Child and Family Psychology Review, 2, 7190. Sanders, M. R., Markie-Dadds, C., Tully, L. A., & Bor, W. (2000). The Triple P-Positive Parenting Program: A comparison of enhanced, standard, and self-directed behavioural family intervention for parents of children with early onset conduct problems. Journal of Consulting and Clinical Psychology, 68, 624640. Snell-Johns, J., Mendez, J. L., & Smith, B. H. (2004). Evidencebased solutions for overcoming access barriers, decreasing attrition, and promoting change with underserved families. Journal of Family Psychology, 18, 1935. Spoth, R., & Redmond, C. (1993). Study of participation barriers in family-focused prevention: Research issues and preliminary results. International Quarterly of Community Health Education, 13, 365388. Spoth, R., & Redmond, C. (1994). Effective recruitment of parents into family-focused prevention research: A comparison of two strategies. Psychology and Health: An International Journal, 9, 353370. Spoth, R., & Redmond, C. (1995). Parent motivation to enrolll in parenting skills programs: A model of family context and health belief predictors. Journal of Family Psychology, 9, 294310. Spoth, R., & Redmond, C. (2000). Research on family engagement in preventive interventions: Toward improved use of scientic ndings in primary prevention practice. Journal of Primary Prevention, 21, 267284. Spoth, R., Redmond, C., Hockaday, C., & Shin, C. (1996). Barriers to participation in family skills preventive interventions and their evaluations: A replication and extension. Family Relations, 45, 247254. Spoth, R., Redmond, C., & Shin, C. (2000). Modeling factors inuencing enrolllment in family-focused preventive intervention research. Prevention Science, 1, 213225. Webster-Stratton, C. (1990). Long-term follow-up of families with young conduct problem children: From preschool to grade school. Journal of Clinical Child Psychology, 19, 144 149. Webster-Stratton, C., & Taylor, T. (2001). Nipping early risk factors in the bud: Preventing substance abuse, delinquency, and violence in adolescence through interventions targeted at young children (08 years). Prevention Science, 2, 165 192.