During the last decades, a number of different relationship educational programs for couples have been developed, with the aims of strengthening marital and similar intimate relationships and preventing relationship deterioration and dissolution. Several of these programs are based on empirical research regarding predictors of relationship quality and stability. The most widely used and best documented program is the Prevention and Relationship Education Program (PREP) (Markman, Stanley, & Blumberg, 2010). This program focuses on teaching communication and conflict management skills and on helping couples foster emotional safety, protect and preserve positive connections, and deepen commitment.
Numerous studies have assessed the possible effects of PREP and similar programs, and according to meta-analyses, they seem to have moderate effects on communication style and relationship satisfaction, with average effect sizes in the range of .40 to .45 (Blanchard, Hawkins, Baldwin, & Fawcett, 2009; Hawkins, Blanchard, Baldwin, & Fawcett, 2008). A few studies have also assessed the extent to which couple relationship education may have any effect on divorce rate, and in both randomized studies (Markman, Rhoades, Stanley, & Peterson, 2013; Stanley, Allen, Markman, Rhoades, & Prentice, 2010) and quasi-experimental studies (Hahlweg, Markman, Thurmaier, Engl, & Eckert, 1998; Markman, Renick, Floyd, Stanley, & Clements, 1993), clinically significant effects have been found.
Although there is ample evidence of positive effects stemming from relationship education programs for couples in general, and from PREP in particular, such programs may not be equally effective for all couples since participants have varying levels of risk of distress. Some start at high risk for negative outcomes and dissolution whereas others attend preventive programs despite low levels of risk. Nevertheless, all couples typically receive the same intervention, and most programs do not screen or select participants for program participation based on risk (Wadsworth & Markman, 2012). Hence, Halford, Markman, and Stanley (2008) have suggested that couples probably would benefit from different types of interventions based on their risk of relationship problems. At the very least, evaluation studies of relationship education programs for couples should track outcomes for high- versus low-risk couples. However, few such studies exist.
In a recent study of three different relationship education programs, including PREP, the effects of program participation varied as a function of risk (Williamson et al., 2015). High-risk couples with low levels of commitment and relationship satisfaction experienced more positive effects than low-risk couples. However, couples with acute concerns at baseline, including physical aggression and alcohol use, benefitted less from the intervention than couples without these concerns. Halford, Sanders, and Behrens (2001) compared a self-help version of PREP with a control group and divided participants into a high- and low-risk group, respectively, based on parental divorce and physical aggression between parents during participants’ childhood and adolescence. These are family of origin characteristics that are associated with increased risk of experiencing relationship distress and dissolution (Van Widenfelt, Hosman, Schaap, & Van der Staak, 1996). The findings indicated beneficial effects from the program for the high-risk couples only. This finding was partly replicated in a more recent study with a slightly different relationship education program in which these family background characteristics were examined in combination with other kinds of risk factors, such as low educational level and aggression in the current relationship (Petch, Halford, Creedy, & Gamble, 2012). Here, high-risk women receiving relationship education for couples reported higher relationship satisfaction, whereas high-risk men showed a trend of higher relationship satisfaction. Furthermore, in a study among young and unmarried couples, the effect of a relationship education program was stronger the more socially and economically disadvantaged the participants were (Amato, 2014). However, a recent meta-analytic study of relationship education programs for lower-income couples (Hawkins & Erickson, 2015) revealed considerably smaller effect sizes (in the range between .06 and .09) than the observed effects in meta-analytic studies of relationship education programs in general (Blanchard et al., 2009; Hawkins et al., 2008).
A similar contradiction is apparent in a recent study of PREP (Markman et al., 2013). Participants with a history of physical aggression in the current relationship before marriage and before participating in a PREP intervention benefitted tentatively more from the program, albeit this moderation effect was only marginally significant. In contrast, parental divorce and physical aggression in the family of origin were not related to the program’s effect. Moreover, in a study of a Dutch version of PREP for couples in which at least one of the partners had experienced a parental divorce, participation in the program did not protect against a decline in relationship functioning. Rather, intervention couples reported a greater number of negative psychological symptoms and higher dissatisfaction with life in general than controls (Van Widenfelt et al., 1996). Thus, more research is needed to understand whether the effects of relationship education programs for couples are moderated by various risk factors for relationship problems. Moreover, Wadsworth and Markman (2012) have called for more research on any factors that may be related to the effects of relationship education programs for couples.
In the present study, risk was measured in terms of parental divorce in the family of origin and any previous experiences with professional help and support for relational problems in the current couple relationship. The latter is considered a more proximal and direct risk factor than the former and thus a more likely moderator of couple relationship education outcomes. Besides these risks for distress or dissolution, men and women might be affected by relationship education programs for couples in different ways. However, evaluation studies of such programs have seldom reported any gender differences in the effect of the intervention. This tendency may seem surprising since gender differences are frequently observed within the realm of couple relationships. For example, women are generally less satisfied than men with their couple relationship (Tai, Baxter, & Hewitt, 2014), women are more likely than men to initiate dissolution of the couple relationship (Amato & Previti, 2003; Hewitt, 2009), and women tend to respond more than men to variability in relationship quality (Kiecolt-Glaser & Newton, 2001; Waite & Gallagher, 2000). Thus, this study explored whether PREP is effective in women only or whether sizable effects occur in both genders.
Any program effects might also be moderated by level of education. Although previous research on couple relationship education has taken participants’ income level into consideration (Amato, 2014; Hawkins & Erickson, 2015), the moderating effects of education level have seldom been addressed. However, numerous studies have observed an association between education level and divorce rate and couple relationship dissolution (Matysiak, Styrc, & Vignoli, 2014). Thus, participants’ level of education might moderate the effects of couple relationship education.
Moreover, few studies of relationship education programs for couples have assessed any moderating effect of marital status. Such moderation appears likely given the numerous studies showing that married couples have generally higher relationship satisfaction and stability than cohabitating couples (Dush, Cohan, & Amato, 2003; Stafford, Kline, & Rankin, 2004). This finding may reflect a tendency for cohabitating couples to be less committed than married couples (Mortensen, Torsheim, Melkevik, & Thuen, 2012; Stanley, Rhoades, & Markman, 2006). Thus, being married as opposed to cohabiting could possibly moderate the outcomes of relationship education programs for couples.
The number of children could also possibly act as a moderator, although to our knowledge this assumption has not been assessed in previous studies. However, the relationship may be less demanding for couples with only one child or two children compared with more children, and any such differences might influence how the couples respond to relationship education. Besides, the number of children may be an indicator of the duration of the relationship, which could possibly also influence how relationship education is experienced.
Finally, both partners are not necessarily equally interested in participating in a couple relationship education program. Participants might experience a relationship education program differently, depending on whether they, their partner, or both decided to take part in it. To our knowledge, this has never been assessed as a possible moderator of couple relationship education programs.
Thus, the present study assessed the extent to which distal and proximal risk-factors, in terms of parental divorce in the family of origin, and any previous experiences with professional help and support for relational problems in the current couple relationship, could moderate any effects of the PREP program. In addition, this study explored whether beneficial effects of PREP could be observed in both genders. Finally, moderating effects of participants’ education level, marital status, number of children, and initiator-status with regard to participating on a PREP program, were assessed. Findings will provide insights into the efficiency of PREP, and particularly whether it suits various kinds of participants equally well, or if some groups benefit more than others.
The Danish Ministry of Social Affairs offers participation in PREP for an income-dependent symbolic fee (e.g., 73 USD at an annual household income up to 88,000 USD; 147 USD at an income up to 147,000 USD) to a limited number of couples with children per year. Besides the requirement of having children, no admission criteria was applied other than that couples had to actively apply for participation. Admitted couples were informed that their governmentally sponsored participation in PREP required each partner to fill out a questionnaire on the Internet both before and after the program, which was supposed to take approximately 15 minutes each time. Besides the participation in PREP, couples did not receive any further compensation for contributing to data collections. Neither, however, faced those who failed to fill out the questionnaire any sanctions, as data collections were completely anonymous.
This study is based on data from PREP participants during the period from January 2012 to August 2013. A total of 889 individuals participated in pretest, of which 476 were female and 412 male (1 did not indicate the gender). All participants had children, all but one participant from their present relationships, and 12% in addition from previous relationships. In total, 92% of participants had between two and four children (two: 28%, three: 46%, four: 18%). One percent had a single child.
The average year of birth was 1974 (SD = 6.70, Mdn = 1975). Thus, most participants were in their 30’s or 40’s at the time of data collections. The majority of participants were either public (35.1%) or private employees (33.0%), or students (10.9%). The remainder were unemployed or retired (7.1%), self-employed (6.7%), on maternity leave (3.8%), “other,” or did not indicate their professional status (3.4%). About three of ten participants (29.2%) indicated an annual income in the range 3-400,000 DKK (44-59,000 USD) which was the most frequently endorsed category. Four of five participants (77.7%) ranged between 100,000 and 500,000 DKK (15-73,000 USD). The vast majority (91.4%) of participants was born in Denmark.
As described in more detail below, relatively stable demographic markers were used to identify longitudinal cases. Out of 889 participants at pretest, 52 (5.8%) did not provide complete demographic information and 40 (4.5%) left information that was identical to another participant. Thus, 797 participants at pretest (89.7%) were uniquely identifiable cases. The respective numbers at posttest were 589 participants of which 42 (7.1%) left incomplete demographic information and 11 (1.9%) identical information to another participant, leaving 536 (91.0%) unique cases. Of the 797 unique cases at pretest, 304 (38.1%) could be identified as longitudinal thanks to identical demographic information at both time points. Note that the low retention rate of 38.1% is not only due to drop-out but also due to difficulties in identifying longitudinal participants. Thus, assuming a change of zero in the intention-to-treat analyses, if no posttest data could be identified, probably underestimates the actual PREP effect.
Besides the variables used as potential moderators of PREP effect described below, participants indicated their (a) year of birth, (b) gender, (c) postal code, (d) country of birth with the categories Denmark and foreign-born, and (e) annual income, in steps of 100,000 DKK (approximately 15,000 USD).
Potential moderator variables of PREP success.
Participants’ parents’ relationship stability. Participants answered a question whether their parents (a) had been married throughout their lives, (b) had been cohabitating throughout life, or (c-f) separated during the respondent’s early childhood, middle childhood, adolescence, or adulthood, respectively. This variable was dichotomized representing whether the parents remained together (i.e., alternatives a and b) or had separated at one time in live (c-f).
Higher education. Participants assessed the length of their education on a six-point scale ranging from grade 9/10 to long higher education (5-6 years). They also could indicate other. As only few participants chose this possibility, this category was excluded from analyses including length of higher education. The remaining categories were collapsed into three categories of (a) up to two years of higher education, (b) three-four years, and (c) five-six years of higher education.
Marriage status. Participants indicated whether they were married, cohabitating, or living separately. As few participants were living apart, the last two categories were collapsed into unmarried participants.
Previous help for relationship problems. Participants answered a dichotomous question whether or not they had received professional help for their relationship before.
Who took initiative to participate in PREP. Participants were asked who took initiative to participate in the course, with the three possible answers (a) I did it myself, (b) my partner did it, and (c) it was a joint decision.
Number of children. Participants were asked “How many children do you have in your present relationship” and “How many children do you have from previous relationships”. The possible answers to both questions ranged from zero to “7+”. Each of these questions were used individually for identifying longitudinal cases. In order to assess the possible moderating effect of number of children, the two questions were summed and dichotomized into up to two children and three or more children, respectively.
Couple relationship measures.
Dyadic Adjustment Scale-4 (DAS-4). Sabourin, Valois, and Lussier (2005) developed a 4-item version of Spanier’s (1976) Dyadic Adjustment Scale that despite its shortness measures general relationship satisfaction reliably across a large range of satisfaction levels. Salari, Wells, and Sarkadi’s (2014) Swedish version of this scale was translated into the Danish language. As these two languages are highly similar, a translation-back translation procedure was deemed unnecessary. Participants answered three of the items on a 6-point scale ranging from all the time (0) to never (5) and the fourth item on a 7-point Likert scale ranging from extremely unhappy (0) to perfect (6). Despite the different ranges, descriptive statistics of the four items were similar. Thus, the four items were summed without re-scaling them, as in previous studies (e.g., Salari et al., 2014). For this scale as well as for the scales described below, items were reversed if necessary so that high values always represent positive couple relationships. Cronbach’s alphas were .76 and .79 at pre- and posttest, respectively.
Communication. Eleven items tapping communication difficulties addressed by PREP, were developed for this study by Center for Family Development (n.d.). These items represent negative attributions (e.g., “When we have problems that require a joint solution, it seems as if we are each our own team”), secrecy and avoidance of discussions (e.g., “I do not share what I think and feel with my partner”), and conflict escalation and verbal aggression (e.g., “I say offensive words about or scold my partner”). The frequencies of these communication difficulties were assessed on a 3-point Likert scale ranging from never or almost never (0) to this happens often (2). Cronbach’s alphas were .84 and .85, respectively, at the two occasions of measurement. The whole scale was reversed so that high levels represent the absence of these communication difficulties.
Danger signals. Three items representing negative escalation of conflicts, invalidation of the partner’s thoughts and feelings, and negative interpretations of his or her verbal and non-verbal behaviors were taken from Stanley, Markman, and Whitton’s (2002) Negative Interaction scale. In order to improve precision of measurement, the original 3-point scale was replaced by a 5-point Likert scale, ranging from agree completely (0) to disagree completely (4). Cronbach’s alphas were .75 and .73, respectively, at the two occasions of measurement.
Problem solving. Four items addressing rational, cooperative, and self-regulated problem-solving were developed for this study by the Center for Family Development (n.d.). The items tapped into arriving at a shared problem definition, brainstorming possible solutions and evaluating them, and re-assessing the problem in cases where it has not been solved. An example item is: “If a solution to a problem does not work, we go back and discuss why.” The participants filled out the same Likert scale as for Danger Signals. Cronbach’s alphas reached .85 and .84 before and after participation in PREP, respectively.
About two weeks before PREP and three months after completion of it, the Center for Family Development asked participants to fill out short questionnaires online. Participants were informed that they did not have to leave any identifying information and that the questionnaires were completely anonymous and analyzed not with respect to individual persons but in order to evaluate and improve the program.
PREP was conducted at several sites spread over the whole country, either on one weekend or distributed over three sessions during a three- or four-week period. During a total of 15 hours, PREP groups discussed the topics of expectations and roles, communication, danger signals, problem-solving, forgiveness, sexuality and sensuality, values, friendship, coziness and fun, and engagement and commitment. Focus was laid on decreasing risk factors associated with relationship dissolution, such as dysfunctional communication and increasing protective factors such as problem-solving skills. The program includes videos, role play exercises from the teachers, and training with the participant’s own partner. Although PREP was provided by a number of private and public organizations, it was manual based, and the course leaders had received a five-day training session at the Center for Family Development, which also administered admission to the program and the program evaluation.
In order to comply with ethical requests of full anonymity, respondents did not provide names or other information that would make them identifiable. As a downside, it is impossible to know which respondents belonged to the same couple and even which responses were made by the same persons over time. To identify responses that likely came from the same person, in a first step, unique respondents in terms of year of birth, gender, postal code, marriage status, number of children in the current and from previous relationships respectively, country of birth, education level, and parents’ relationship stability were identified, using the original variables as described above, before collapsing categories. In a second step, unique cases from the two times of measurement were matched to arrive at longitudinal cases.
Two kinds of errors might have been caused by this procedure. First, actual longitudinal cases might have been missed because either participants were not unique on this set of nine demographic variables or they did not answer all of them. One or more of these variables may have changed over time as well, e.g., by moving into a different postal code area or by having another baby. Second, different persons might have erroneously been regarded as longitudinal cases if two different individuals matched on these nine variables at the two occasions of measurement, respectively. The first kind of error leads to an overestimation of attrition because people appear to have dropped out even though they have not. This tendency results in a loss of statistical power due to a reduced number of longitudinal cases and a reduced effect size in the intention-to-treat analyses in which dropped-out cases are assumed to have remained unaffected by PREP. The second kind of error, which probably is much less common, leads also to a reduced effect size because it introduces random variation at each occasion. Thus, these two errors lead rather to an underestimation of effect size and significance level and thus make the test of possible effects of PREP more conservative.
Besides this possible general underestimation of PREP effects, the analysis of moderating effects might be affected if participants with certain levels of the moderator variables were retained in the longitudinal sample more often than others. This might apply to four of the moderator variables because they were used for identifying longitudinal cases. Actually, the share of longitudinally matched unmarried cases (20.4%) was smaller than the respective percentage among the unidentified cases (35.2%, χ2(1) = 20.86, p < .001) at pretest. Similarly, the share of longitudinally matched participants with separated parents (33.6%) was smaller than the respective share among those who could not been matched over time (40.7%, χ2(1) = 4.26, p = .039). The shares of participants with various levels of education did not differ, however (χ2(2) = 4.51, ns), nor did the shares of those with few or many children, respectively (χ2(1) = 1.27, ns). Neither was there any difference in the moderator variables that were not used to identify longitudinal cases, i.e., previous psychological help (χ2(1) = .42, ns) and who in the couple took initiative to participate in PREP (χ2(2) = 3.84, ns). In summary, two groups of participants were retained less frequently than would have been expected, namely unmarried participants and participants whose parents had separated. Both of these groups were minorities from the beginning (see Table 2). Therefore, they should have been unique cases more often and thus should have been identified longitudinally more frequently than others. In conclusion, these risk groups were more prone to dropout from PREP as opposed to their loss being due to the method used to identify longitudinal cases. Even if the method of longitudinally matching participants did not produce any bias in the moderation analyses, the selective dropout might result in more positive estimates of PREP effects in the longitudinal analyses, especially among unmarried participants and those with separated parents, as compared to the intention-to-treat analyses.
Because it was impossible to know which partners belonged to the same couple, MANOVAs were performed separately for women and men, respectively, thereby avoiding the problem of dependency in the data. To the best of our knowledge, no homosexual couples participated in this study.
Table 1 summarizes the descriptive statistics of the four relationship measures and their inter-correlations. All measures were strongly associated. Participants who expressed satisfaction with their relationships tended also to report high communication quality, high skills in relationship problem-solving, and only a few danger signals in their relationships. Descriptively, women reported a somewhat poorer relationship quality on all four measures than men, especially before taking part in PREP.
Table 2 summarizes the frequencies of participants’ background characteristics that might indicate risk or protective factors and that might moderate the effectiveness of PREP. Most participants had some or extended higher education. However, about two in five participants had experienced the separation of their own parents or had sought help for problems in their own romantic relationship, respectively. One-third of the respondents were not married to their respective romantic partners. One-fourth indicated that it was the partner who had taken initiative to participate in PREP, whereas about equal shares of the remainder stated that they had taken initiative themselves or that it had been a joint initiative. As mentioned above, almost three-quarters of the sample had three or more children.
Effectivity of PREP
Multivariate repeated ANOVAs revealed significant changes in relationship quality from before to three months after participation in PREP (see Table 3). Conventionally, effects explaining about 25% of variance are called “large,” 9% “medium,” and 1% “small.” Similarly, Cohen’s d, which is a standardized mean difference, is called “large” if d reaches .8, “medium” at d = .5, and “small” if d = .3. Given these conventions, PREP had a large effect on women’s reports of their respective relationship quality and a medium to large effect on males’ reports. Women reported a large beneficial effect on communication in the relationship and medium to large improvements in general relationship satisfaction, problem-solving ability, and freedom of danger signals in the partner relation. As mentioned above, males appeared more satisfied with their relationships to begin with. Still, they reported medium-sized improvements in communication, problem-solving in the couple, and general relationship satisfaction. A reduction in experienced dangers to the relationship, however, failed to reach significance.
Two kinds of Cohen’s d are used in the literature. Some authors (e.g., Thomas & Zimmer-Gembeck, 2007) standardize the pre/post difference at the standard deviation of the population at pretest, thereby comparing the size of change with the “natural” variation of the variable of interest. Others (e.g., Nowak & Heinrichs, 2008) prefer to standardize at the standard deviation of change, thereby obtaining an indication of how large the change is in relation to the variability of the treatment effect. In other words, an indication is given of how homogeneously the effect occurs in the investigated sample. We believe that both kinds of information have their merits and thus both are summarized in Table 3. Generally, the effect size, if standardized at the variability of change, appeared somewhat larger than if standardized at the baseline variance. This tendency might indicate relatively homogeneous effects of participation in PREP.
Because those participants who provided data at both time points might be a positive selection of participants, we also conducted intention-to-treat analyses where missing data at Time 2 were replaced by the respective participant’s data from baseline. Such analysis gives a conservative estimate of the program effect, especially since subjects might have participated in—and benefitted from—the program without providing data at Time 2. Anyway, as Table 3 shows, all changes reported above were significant at p < .001 even in the intention-to-treat analyses.
Analyses of potential moderators
Participants’ parents’ relationship dissolution, having been treated for relationship problems before, low education, being unmarried, and having many children were hypothesized as being risk factors that might also moderate the effectivity of PREP. In addition, we suggested that any changes in relationship quality might vary depending on whether the respondent or his or her partner, or both, decided to participate in PREP, possibly reflecting various levels of motivation to participate.
As Table 4 (upper part) summarizes, all potential moderators except for participants’ parents’ relationship dissolution and number of children were associated with participants’ relationship quality, confirming their status as risk factors. More specifically, length of education predicted the experience of relationship danger signals, with women with fewer than two years of higher education reporting more danger signals than women with five or more years of higher education (and women with intermediate education scoring in between). For men, results differed somewhat. Men with little higher education reported fewer danger signals than men in the intermediate education category, with men with the highest education scoring in between. Participants’ marriage status was associated with all four measures of relationship quality. Married participants reported higher relationship satisfaction, better communication, fewer danger signals, and higher problem-solving capability compared to their cohabitating counterparts. The exceptions were female-reported problem-solving and male-reported relationship satisfaction, where this overall effect was not significant. In both males and females, previous relationship counseling was associated with poorer relationship quality. Finally, males who had made the decision to participate in PREP reported poorer relationship quality than males whose partners had decided to participate, with males reporting a joint decision scoring in between.
The abovementioned relatively large effects resulting from standardization at the variability of change suggested already that the effects of PREP might be relatively homogeneous. As the lower part of Table 4 shows, none of the six risk factors was a significant moderator of the overall effect of PREP or approached significance (p < .10). Because most F values were below 1, the moderating effects of the six risk factors were negligible, even descriptively.
In order to test whether one of the risk factor’s effects might have been masked by other correlated risk factors, an additional mixed MANOVA was conducted, with all risk factors entered simultaneously. Because estimations of the numerous higher-order interactions would have been unstable (Cohen, Cohen, West, & Aiken, 2003), only main effects and interactions with PREP effects were modeled. For one of the six potential moderating variables and only in women, the multivariate analysis approached significance. PREP tentatively affected women differently, depending on whether they had undergone relationship counseling before, F(4, 149) = 2.02, p = .095, η2 = .05. Univariate follow-up analyses revealed that women with previous relationship counseling benefitted more from PREP in terms of reduced perceived danger signals, F(1, 152) = 6.36, p = .013, η2 = .04, improved DAS-4 relationship measure, F(1, 152) = 5.60, p = .019, η2 = .04, and tentatively more improved communication, F(1, 152) = 3.87, p = .051, η2 = .02. The estimated marginal means of the PREP-by-previous-counseling interaction, net the other five risk factors and standardized by the pooled standard deviation before PREP, revealed effect sizes of d = 0.08, –0.21, and –0.39 for improvements in danger signals, DAS-4, and communication, respectively, in women without previous counseling compared to effect sizes of 0.47, –0.57, and –0.65, respectively, in their counterparts who had prior relationship counseling. That is, in all these facets of the relationship, both groups improved, but those women without previous counseling who had better couple relationships to begin with improved less than those with previous counseling, so that the two groups were more similar after PREP.
A methodological issue is the lack of power in the analysis of risk-factor-by-PREP interaction effects due to the moderate size of the longitudinal sample. The sample was comprised of 162 females and 136 males, respectively, split into two or three levels of the respective risk factor. Using the program G*Power 3.1.3, the power to detect interactions of medium size (f(V) = .25) at a significance level of .10, as applied above, proved satisfactory. It was 84.4% and 77.7% for women and men, respectively, when inspecting interactions with risk factors that had two levels (both parents’ and own marriage status, whether or not the couple had had counseling before, and the dichotomized number of children). It was 96.9% and 93.5% for women and men, respectively, when inspecting interactions with risk factors that had three levels (length of education, who decided to participate in PREP). The power to detect small interaction effects (f(V) = .10), however, was poor (22.8% to 32.7%).
For this reason, we even inspected all interaction effects where the non-significant multivariate F values had at least exceeded 1. As the lower part of Table 4 summarizes, this was the case for previous relationship counseling in women (multivariate p = .127), length of men’s education (p = .226), and women’s parents’ relationship status (p = .390). If women reported previous counseling, PREP improvements were larger concerning DAS4, F(1, 160) = 4.55, p < .05, η2 = .03, d = 0.60, than if they did not have prior counseling, d = 0.30 (using the same overall before-treatment SD for calculation of both effect sizes to make them comparable). A similar result was obtained for the reduction of danger signals in the course of the PREP program, F(1, 160) = 5.90, p < .05, η2 = .04, d = –0.57 and –0.21 with or without, respectively, previous counseling. Because the initial problem levels were higher in participants with previous relationship counseling, a compensatory PREP effect in women at risk is implied, making them more similar to those women without pre-existing risk. Men with limited higher education of up to two years after graduation from high school did not improve (d = 0.00) their problem-solving ability in contrast to men at an intermediate or extensive level of higher education (d = 0.37 and 0.35, respectively), F(2, 132) = 3.14, p < .05, η2 = .05. None of the four relationship measures showed any interaction effect between PREP and parents’ relationship status in women. In summary, these additional more liberal analyses suggest that women at risk due to previous relationship counseling might benefit more from PREP, thus catching them up to some degree with women not at risk. In contrast, men at risk due to low education might not be able to improve their problem-solving ability with the help of PREP.
The aim of this study was to assess the extent to which any changes in couple relationship quality following a Danish version of the PREP program would 1) hold in both genders and 2) be moderated by distal and proximal risk factors, participants’ education level, marital status, and initiator-status. A precondition for the assessment of potential moderators is the existence of any effects of PREP to begin with, which then might be smaller or larger in various sub-groups. Both the pre/post comparisons of longitudinal cases and the intention-to-treat analyses revealed significant effects of PREP. More specifically, the findings indicated medium to large improvements in both relationship satisfaction and communication skills for couples. The magnitude of the changes was comparable to the effects observed in previous studies where PREP and other relationship education programs for couples have been evaluated by the use of quasi-experimental or experimental design (Blanchard et al., 2009; Hawkins et al., 2008). This tendency suggests that the Danish version of PREP is about as effective as previously evaluated versions of the program.
From a comparison of the two different ways of calculating effect sizes, relatively homogeneous PREP effects could be concluded. This homogeneity was corroborated by the formal test of moderation effects, which did not reveal any significant moderations, although the various risk factors and other studied variables had main effects on relationship quality. However, because the sample size limited the statistical power to detect small moderation effects, possible interactions between the risk factors and PREP were inspected more closely. Such scrutiny included even those instances where no overall moderation could be established across all outcome measures. This inspection revealed that women who might be at risk, as indicated by previous relationship counseling, caught up to some degree with their counterparts on relationship satisfaction and low levels of perceived danger signals. The implication then is that PREP might have a compensatory effect. This compensatory effect in women even approached significance in the multivariate analysis and extended tentatively on improved communication if the unique moderating effect of previous relationship counseling, was assessed. Turning to males, those without much higher education did not improve on one of the four relationship measures. However, they had relatively high scores already at baseline. Thus, there was less of a need to improve on problem-solving, or alternatively, a possible initial lack of insight into weaknesses in this respect may have been replaced by a more competent assessment of their problem-solving capability actually existing.
Thus, by and large, the Danish version of PREP appears to help couples from various backgrounds well, irrespective of any risk factors. This sparseness of moderation effects may seem at odds with previous findings where various kinds of risk-prone participants typically have responded somewhat differently to couple relationship programs as compared to other participants (Amato, 2014; Halford et al., 2001; Petch et al., 2012; Williamson et al., 2015). Therefore, the sparseness of differential PREP effects might to be blamed on methodological factors such as the measurement of risk factors, the absence of a control group, or the large attrition between measurement occasions. However, given that all risk factors except parental relationship dissolution had main effects on the participants’ romantic relationships, these methodological explanations appear unlikely. On the contrary, selective dropout among unmarried participants and participants with separated parents might have resulted in a positive bias of these sub-groups over time, such that successful participants remained and unsuccessful ones left the sample. Thus, the development of these two groups might, on average, actually be less positive than the present findings indicate. Still, in the context of a general absence of clear-cut moderation effects, risk factors do not appear to hamper the beneficial effects of participation in PREP. Moreover, PREP appeared as even more effective for some women at risk.
Some “risk factors” might also be ambiguous in nature. For example, experiences with professional help and support for relational problems in the current couple relationship could also indicate a willingness to seek professional help. In turn, this tendency might be considered a strength of the couple relationship. Any effects of having relationship problems might therefore have been balanced by the strength of willingness to seek help. Moreover, as also seen in the study by Markman and colleagues (2013), various risk factors failed to moderate or only marginally moderated the observed effects of PREP. Therefore, previous findings seem to be somewhat conflicting when it comes to any risk factor effects on the outcomes of couple relationship education. The present findings thus strengthen the notion that risk factors might affect relationship education programs such as PREP to a smaller extent than might be expected theoretically.
Although consistent effects on the efficacy of PREP were revealed, PREP might not help men with low education improve their problem-solving ability, albeit PREP was effective otherwise even in this group. This potential weakness of the program should be acknowledged. Previous studies from other cultural contexts suggest that the effects of relationship education programs are contingent on various risk factors, which might imply restraint in offering any such programs as one size fits all. However, within a Danish context, there is limited empirical support for such a strong conclusion. Rather, it may seem more reasonable to consider PREP as a program with apparently favorable effects for broad groups of couples.
Four possible limitations need to be mentioned, however. First, it was not possible to identify which individuals participated in the same PREP groups or were educated by the same instructors. Thus, the analyses could not take this clustering into account. Second, the Danish context of this study is both a strength and a weakness. It is a strength because the bulk of findings stem from studies conducted in North America, and more knowledge is needed on the degree to which PREP and similar programs can be applied in a European context. On the other hand, the present findings do not necessarily generalize to other western regions such as North America. In a Danish context, PREP seems to be effective for all kinds of participants. Perhaps this latter tendency may reflect that Denmark is a more egalitarian and gender-equal society than the USA and most other countries in which studies of relationship education measures usually have taken place.
The third limitation is the composition of the sample consisting of couples in mid-adulthood who had, on average, almost three children together. That is, these were long-term and thus stable relationships but quite dissatisfied ones as the relationship measures revealed. This specificity of the sample needs to be taken into account when considering possible generalizations of the present findings.
Finally, the limited number of longitudinally identified participants implies an issue of statistical power when examining how the various risk factors might affect PREP’s beneficial effects. In this paper, the decision was made to minimize the risk for Type-II errors, that is, to miss possible moderation effects. For this reason, a 10% significance level was employed, and interactions where F values exceeded 1 were explored descriptively. Obviously, this decision came at the cost of an increased risk for Type-I errors. In other words, the compensatory effect in women who had prior counseling — which appears to be a proxy of dissatisfying couple relationships — and the, in part, smaller effect in men with low education may be due to sampling fluctuation and may not replicate in future studies. However, more important is that even these liberal analyses have not revealed any subgroups for which PREP would be of no use at all or even contraindicated. Thus, although some participants may have larger or broader benefits than others, depending on their background, the results speak for the general applicability of PREP in the studied population.
Our study suggests that programs such as PREP can be applied successfully across a wide range of participant backgrounds.
Amato, P. (2014). Does social and economic disadvantage moderate the effects of relationship education on unwed couples? An analysis of data from the 15-month Building Strong Families evaluation. Family Relations, 63, 343–355. doi:10.1111/fare.12069
Amato, P. R., & Previti, D. (2003). People’s reasons for divorcing: Gender, social class, the life course, and adjustment. Journal of Family Issues, 24, 602–626. doi:10.1177/0192513X03024005002
Blanchard, V. L., Hawkins, A. J., Baldwin, S. A., & Fawcett, E. B. (2009). Investigating the effects of marriage and relationship education on couples’ communication skills: A meta-analytic study. Journal of Family Psychology, 23, 203–214. doi:10.1037/a0015211
Center for Family Development, Copenhagen, Denmark. (n.d.). Udkast til spørgeskema [Questionnaire draft]. Unpublished manuscript.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). New York: Routledge.
Dush, C. M. K., Cohan, C. L., & Amato, P. R. (2003). The relationship between cohabitation and marital quality and stability: Change across cohorts? Journal of Marriage and Family, 65, 539–549. doi:10.1111/j.1741-3737.2003.00539.x
Hahlweg, K., Markman, H. J., Thurmaier, F., Engl, J., & Eckert, V. (1998). Prevention of marital distress: Results of a German prospective longitudinal study. Journal of Family Psychology, 12, 543–556. doi:10.1037/0893-3126.96.36.1993
Halford, W. K., Markman, H. J., & Stanley, S. M. (2008). Strengthening couplesʼ relationships with education: Social policy and public health perspectives. Journal of Family Psychology, 22, 497–505. doi:10.1037/a0012789
Halford, W. K., Sanders, M. R., & Behrens, B. C. (2001). Can skills training prevent relationship problems in at-risk couples? Four-year effects of a behavioral relationship education program. Journal of Family Psychology, 15, 750–768. doi:10.1037/0893-3188.8.131.520
Hawkins, A. J., Blanchard, V. L., Baldwin, S. A., & Fawcett, E. B. (2008). Does marriage and relationship education work? A meta-analytic study. Journal of Consulting and Clinical Psychology, 76, 723–734. doi:10.1037/a0012584
Hawkins, A. J., & Erickson, S. E. (2015). Is couple and relationship education effective for lower income participants? A meta-analytic study. Journal of Family Psychology, 29, 59–68. doi:10.1037/fam0000045
Hewitt, B. (2009). Which spouse initiates marital separation when there are children involved? Journal of Marriage and Family, 71, 362–372. doi:10.1111/j.1741-3737.2009.00603.x
Kiecolt-Glaser, J. K., & Newton, T. L. (2001). Marriage and health: His and hers. Psychological Bulletin, 127, 472–503. doi:10.1037/0033-2909.127.4.472
Markman, H. J., Renick, M. J., Floyd, F. J., Stanley, S. M., & Clements, M. (1993). Preventing marital distress through communication and conflict management training: A 4- and 5-year follow-up. Journal of Consulting and Clinical Psychology, 61, 70–77. doi:10.1037/0022-006X.61.1.70
Markman, H. J., Rhoades, G. K., Stanley, S. M., & Peterson, K. M. (2013). A randomized clinical trial of the effectiveness of premarital intervention: Moderators of divorce outcomes. Journal of Family Psychology, 27, 165–172. doi:10.1037/a0031134
Markman, H. J., Stanley, S. M., & Blumberg, S. L. (2010). Fighting for your marriage (3rd ed.). San Francisco, CA: Jossey- Bass.
Matysiak, A., Styrc, M., & Vignoli, D. (2014). The educational gradient in marital disruption: A meta-analysis of European research findings. Population Studies, 68, 197–215. doi:10.1080/00324728.2013.856459
Mortensen, Ø., Torsheim, T., Melkevik, O., & Thuen, F. (2012). Adding a baby to the equation: Married and cohabiting women’s relationship satisfaction in the transition to parenthood. Family Process, 51, 122–139. doi:10.1111/j.1545-5300.2012.01384.x
Nowak, C., & Heinrichs, N. (2008). A comprehensive meta-analysis of Triple P-Positive Parenting Program using hierarchical linear modeling: Effectiveness and moderating variables. Clinical Child and Family Psychology Review, 11, 114–144. doi:10.1007/s10567-008-0033-0
Petch, J., Halford, W. K., Creedy, D. K., & Gamble, J. (2012). Couple relationship education at the transition to parenthood: A window of opportunity to reach high-risk couples. Family Process, 51, 498–511. doi:10.1111/j.1545-5300.2012.01420.x
Raffing, R. (2014). Parental divorce and relationship satisfaction: A longitudinal study of the PREP program in Denmark. Unpublished Bachelor’s thesis, Kristianstad University.
Sabourin, S., Valois, P., & Lussier, Y. (2005). Development and validation of a brief version of the dyadic adjustment scale with a nonparametric item analysis model. Psychological Assessment, 17, 15–27. doi:10.1037/1040-35184.108.40.206
Salari, R., Wells, M. B., & Sarkadi, A. (2014). Child behaviour problems, parenting behaviours and parental adjustment in mothers and fathers in Sweden. Scandinavian Journal of Public Health, 42, 547–553. doi:10.1177/1403494814541595
Spanier, G. B. (1976). Measuring dyadic adjustment: New scales for assessing the quality of marriage and similar dyads. Journal of Marriage and the Family, 38, 15–28. doi:10.2307/350547
Stafford, L., Kline, S. L., & Rankin, C. T. (2004). Married individuals, cohabiters, and cohabiters who marry: A longitudinal study of relational and individual well-being. Journal of Social and Personal Relationships, 21, 231–248. doi:10.1177/0265407504041385
Stanley, S. M., Allen, E. S., Markman, H. J., Rhoades, G. K., & Prentice, D. L. (2010). Decreasing divorce in U.S. Army couples: Results from a randomized controlled trial using PREP for Strong Bonds. Journal of Couple and Relationship Therapy, 9, 149–160. doi:10.1080/15332691003694901
Stanley, S. M., Markman, H. J., & Whitton, S. W. (2002). Communication, conflict, and commitment: Insights on the foundations of relationship success from a national survey. Family Process, 41, 659–675. doi:10.1111/j.1545-5300.2002.00659.x
Stanley, S. M., Rhoades, G. K., & Markman, H. J. (2006). Sliding versus deciding: Inertia and the premarital cohabitation effect. Family Relations, 55, 499–509. doi:10.1111/j.1741-3729.2006.00418.x
Tai, T., Baxter, J., & Hewitt, B. (2014). Do co-residence and intentions make a difference? Relationship satisfaction in married, cohabiting, and living apart together couples in four countries. Demographic Research, 31, 71–104. doi:10.4054/DemRes.2014.31.3
Thomas, R., & Zimmer-Gembeck, M. J. (2007). Behavioral outcomes of Parent-Child Interaction Therapy and Triple P-Positive Parenting Program: A review and meta-analysis. Journal of Abnormal Child Psychology, 35, 475–495. doi:10.1007/s10802-007-9104-9
Wadsworth, M. E., & Markman, H. J. (2012). Where’s the action? Understanding what works and why in relationship education. Behavior Therapy, 43, 99–112. doi:10.1016/j.beth.2011.01.006
Waite, L. J., & Gallagher, M. (2000). The case for marriage: Why married people are happier, healthier, and better off financially. New York: Broadway Books.
Van Widenfelt, B., Hosman, C., Schaap, C., & Van der Staak, C. (1996). The prevention of relationship distress for couples at risk: A controlled evaluation with nine-month and two-year follow-ups. Family Relations, 45, 156–165. doi:10.2307/585286
Williamson, H. C., Rogge, R. D., Cobb, R. J., Johnson, M. D., Lawrence, E., & Bradbury, T. N. (2015). Risk moderates the outcome of relationship education: A randomized controlled trial. Journal of Consulting and Clinical Psychology, 83, 617–629. doi:10.1037/a0038621