Trajectories of Birth Family Contact in Domestic Adoptions

Abstract

Emotional distance regulation theory (Broderick, 1993; Grotevant, 2009) guided this examination of the changes in family structure and process in adoptive kinship networks experiencing different arrangements of contact between birth and adoptive family members. Group-based trajectory modeling was used to reveal four trajectories of post-adoption contact experienced between adoptive and birth family members in adoptive kinship networks of same-race, domestic infant adoptions. Data were drawn from the Minnesota Texas Adoption Research Project, a study of 190 adoptive families and 169 birth mothers followed across four longitudinal waves (middle childhood, adolescence, emerging adulthood, young adulthood). Three aspects of the birth family-adoptive family relationship measured at four times were used to create the groups: frequency of contact between the adopted person and birth mother, satisfaction of the adopted person with the openness arrangements, and number of adoptive and birth family members involved in the contact. Four trajectory groups emerged: no contact (41.6% of sample), stopped contact (13.7%), limited contact (26.3%), and extended contact (18.4%). Group membership was validated by coders who matched interview transcripts with group descriptions at levels significantly above chance. Knowledge of trajectories will assist professionals providing post-adoption services.

Adoption practices in the United States have been evolving over the past four decades, with a gradual shift from secrecy and closed adoptions (in which the child’s adoptive and birth family members have no identifying information about each other) to open adoptions, in which direct and sometimes frequent contact takes place (Grotevant, 2012; Grotevant & McDermott, 2014). At the time openness practices were being introduced in the 1980s, some considered the practice experimental and potentially dangerous for children (e.g., Kraft, Palombo, Mitchell, Woods, Schmidt, & Tucker, 1985). However, no empirical research was then available to evaluate the competing views of advocates of openness vs. those desiring to retain the status quo of closed adoptions.

The movement toward openness in adoption represented a major shift in thinking about the meaning of adoption. In the era of closed adoptions, children were thought to be “subtracted” from their family of birth and “added” to their family of adoption; a clean break between families was considered to be best for all parties (Carp, 1998). Open adoption implies a shift in thinking, acknowledging that adoption creates an adoptive kinship network (AKN: Grotevant & McRoy, 1998), in which the child connects his or her extended families of birth and rearing. This model recognizes that neither adopted children nor their birth parents forget about one another; they may remain physically and/or psychologically present to one another in varying degrees over their lifetimes (Fravel, McRoy, & Grotevant, 2000).

In the practice of domestic infant adoptions, some form of contact between adoptive and birth family members has become quite common and is almost expected (Siegel & Smith, 2012). Many adopted persons and birth parents are positively inclined toward contact, because it alleviates the secrecy of closed adoptions and allows each others’ questions and concerns to be addressed directly. Nevertheless, the paths that AKN members take in the development of their relationships are quite varied.

In this study, emotional distance regulation theory (Broderick, 1993; Grotevant, 2009) guided the examination of the changes in family structure and process in adoptive kinship networks experiencing different openness arrangements. The theory recognizes that the adults (initially adoptive parents and birth parents) bring their own developmental histories, expectations about relationships, expectations about adoption, and relationship skills to the creation of the new AKN. Early phases of the development of the AKN involve negotiation among the participants in terms of their tolerance for separation and connection, as they create a new family structure for which there is no shared cultural template.

If contact between adoptive and birth family members occurs, initial decisions about stopping, continuing, or expanding contact are based on the adults’ evaluations of the importance of the relationships for the child and their prognosis for its success; in other words, their satisfaction with the openness arrangements that they have developed up to that point. Over time, as individuals within the family change and as the child matures and asserts his or her own wishes for contact, relationships are once again renegotiated. The entrance and exit of significant family members (e.g., marriage of a birth mother, divorce of adoptive parents, birth of a grandchild) can also evoke changes in the family’s dynamics. Important inflection points occur in middle childhood, as the child’s understanding of adoption matures (Brodzinsky, 2011); adolescence, as the youth explores and begins to consolidate a sense of adoptive identity (Grotevant & Von Korff, 2011); emerging adulthood, as the adopted young adult leaves home and makes independent decisions about contact with birth relatives (Farr, Grant-Marsney, Musante, Grotevant, & Wrobel, 2014); and young adulthood, as the adopted person’s family of orientation typically expands with the addition of committed relationships and children.

Emotional distance regulation theory, highlighting the dynamic nature of relationships, suggests that different adoptive kinship networks will follow different developmental pathways over time, as a function of several factors central in this investigation: a) degree of contact that the adopted person and his or her birth mother have with one another (frequency); b) the adopted person’s appraisal of that contact (satisfaction); and c) the members of the AKN who are in contact with one another (cross-network participation). Although each family is unique in some ways, there should be a relatively parsimonious number of prototypical pathways that networks follow. This study employs empirical methods to identify and describe such subgroups of adoptive kinship networks that have evolved through processes of emotional distance regulation.

The Minnesota Texas Adoption Research Project (MTARP: Grotevant & McRoy, 1998; Grotevant, McRoy, Wrobel, & Ayers-Lopez, 2013) was launched in the mid-1980s to examine developmental outcomes for children growing up in adoptive families varying in levels of contact, to examine how adoptive kinship networks managed contact over time, and to examine adjustment outcomes for adoptive parents and birth mothers (e.g., Henney, Ayers-Lopez, McRoy, & Grotevant, 2007). Data were collected from adoptive parents, their adopted children, and the children’s birth mothers. Even by the time of the first wave of data collection (when the children averaged 7.8 years of age), some adoptive parents reported that openness arrangements had changed since the placement. Some adoptions that started as closed had become open, and some that had had initial contact had stopped (Grotevant & McRoy, 1998). It was clear that child outcome predictions made from snapshots of contact at any one point in time would be limited, and that more complete information would be provided by examining trajectories of contact over time.

Adoptive and birth family members maintained contact through varied means, including the exchange of letters, pictures, emails, gifts, face-to-face meetings, and extended visits. By Wave 4 (young adulthood), there was evidence of increased use of technology-mediated communication, such as texting, Skype and FaceTime calls, and the use of social media platforms such as Facebook (Grotevant & McRoy, 2016).

In MTARP, emerging adult adoptees with greater contact were more satisfied with that contact, which was consistent with findings from earlier developmental time points (Farr, Grant-Marsney, Musante, Grotevant, & Wrobel, 2014). Adoptive families who were more satisfied with their openness arrangements also tended to have adoptees with fewer externalizing problems in adolescence and emerging adulthood (Grotevant et al., 2011). Nevertheless, the frequency of contact, participants in the contact, and feelings about the contact may vary dramatically across families.

Earlier reports from MTARP examined changes from waves 1 to 2 (middle childhood to adolescence; e.g., Dunbar et al., 2006) and from waves 2 to 3 (adolescence to emerging adulthood; e.g., Wrobel, Grotevant, Samek, & Von Korff, 2013). For example, most adoptive families (71.8%) remained within the same level of openness from wave 1 to wave 2. Among those who changed, roughly equal percentages increased in openness level (14.7%) and decreased (13.6%) (Grotevant, McRoy, & Ayers-Lopez, 2004). However, these single lag changes do not tell the full story of contact dynamics. A method was needed a) that would take into account patterns of change in contact across multiple waves of data and b) that would provide a case-centered approach to reflect the heterogeneity in contact pathways AKNs had experienced across time.

Group-Based Trajectory Modeling

Group-based trajectory modeling (GBTM: Nagin, 2005; Nagin & Odgers, 2010, 2012) provides an empirical method for identifying trajectory groups, typically groups of individuals who follow similar pathways over time (Nagin & Odgers, 2012). The method has four capabilities: 1) empirically identifying (rather than assuming) distinctive groups of trajectories; 2) estimating the proportion of the population comprising the various groups; 3) relating group membership to other characteristics of individual participants; and 4) examining predictors and outcomes relating to trajectory group membership (Nagin & Jones, 2012). Another strength of the method is that it can handle missing data well (Nagin, 2005; Nagin & Odgers, 2010), as missing data inevitably are found in longitudinal studies. The method has been used to study diverse topics, including depressive symptoms across adolescence (Duchesne & Ratelle, 2014); employment patterns of young adults (van der Geest, Bijleveld, Blokland, & Nagin, 2016); sociocultural adjustment of international students over their first year at university (Hirai, Frazier, & Syed, 2015); and change in posttraumatic stress symptoms among children who had been exposed to violence (Miller-Graff & Howell, 2015). It has also been used in the family literature, as seen in a recent critical review examining trajectories of marital quality (Proulx, Ermer, & Kanter, 2017).

In the present study, the units being followed over time are adoptive kinship networks rather than individuals; this is the first study to examine change over time in trajectories at the level of the adoptive kinship network. Trajectory groups yielded by the analysis, then, may be understood as clusters of individual cases (i.e., adoptive kinship networks) that follow similar contact pathways over time. In most studies using GBTM, trajectories are identified from individual variables (such as depressive symptoms or marital satisfaction). In the present study, trajectories are identified from the simultaneous consideration of three variables that characterize distinctive features of contact (see below) measured across time.

Characterizing Trajectories of Birth Family Contact

Prior reports from MTARP have examined links between contact and various psychosocial outcomes for children, adolescents, and young adults, especially with regard to identity (e.g., Von Korff & Grotevant, 2011) and psychosocial adjustment (e.g., Grotevant, Rueter, Von Korff, & Gonzalez, 2011). However, these investigations raised two new questions: a) Are there distinctive pathways or trajectories of contact that subsets of adoptive kinship networks have followed over time? and b) In addition to characterizing trajectories for each type of contact over time, can trajectories be developed that take into account multiple aspects of contact? Other reports have conducted more detailed analyses of the numerous aspects of contact, such as type of contact (e.g., pictures, letters, in person visits, etc.); frequency of contact; satisfaction with openness arrangements; people involved in the contact; duration of the contact; and so on (e.g., Grotevant et al., 2007). Taken together, this body of work suggested that three nonredundant variables would be most important to consider: a) frequency of the adopted person’s contact with his or her birth mother, b) the adopted person’s satisfaction with contact arrangements, and c) the number of adoptive kinship network members participating in cross-network contact.

This paper has two goals. First, GBTM will be used to identify distinctive contact trajectories characterizing adoptive kinship networks over the course of the adopted person’s first three decades of life, from middle childhood into young adulthood. Second, the quantitatively-derived trajectory assignments will be validated through the use of data from extensive interviews that had been conducted with participants. Identification and validation of these contact trajectory groups will provide unprecedented clarity on how contact in adoption may evolve across time; this knowledge will be useful to adoption professionals as well as members of adoptive kinship networks.

Method

Participants

The study sample included adoptive families and birth mothers whose adoptions varied in post-adoption openness arrangements. We first describe the complete sample and then explain which data were used in the construction of the trajectories.

Participants were recruited from 35 adoption agencies across the United States to yield a demographically homogeneous sample that varied by openness arrangements. Families were included whose target adopted child was between the ages of 4 and 12 years at the time of the first data collection; the child had been adopted through a private agency before his or her first birthday; the adoption was not transracial, international, or special needs; and both adoptive parents were still married to each other. Simultaneously, birth mothers were recruited whose children had also been placed for adoption in infancy through a private domestic agency with adoptive parents of the same racial background as the baby (Grotevant & McRoy, 1998). The racial / ethnic makeup of this sample is representative of families who adopted through private agencies and expectant mothers who placed children for adoption in the late 1970s and early 1980s. Although transracial and international adoptions were becoming more common at that time, they were excluded from the study design because the primary goal of the study was to examine variations in contact between the child’s birth and adoptive family members in same-race adoptions. Contact in international adoptions was considered impossible at that time, and contact in transracial adoptions was extremely rare; the investigators felt that introducing sample variation in terms of experiences unique to transracially or internationally adoptive families would obscure conclusions that could be drawn about contact (see Jager, Putnick, & Bornstein, 2017, for further discussion of the scientific merits of homogeneous convenience samples).

At Wave 1 (1987 - 1992), participants in the larger study included 720 individuals: both parents in 190 adoptive families, one target adopted child in 171 of the families (90 males, 81 females; mean age = 7.8 years), and 169 birth mothers (mean age = 27.1 years). The Wave 1 sample included 82 adoptive kinship networks (AKNs) in which complete data were available for the adopted child, both adoptive parents, and their birth mother. In addition, data were available for 108 adoptive families for whom no birth mother data were available and for 87 birth mothers from whom no adoptive family data were available. Adopted individuals and their families were primarily White (93.2% non-Hispanic White/Caucasian, 2.1% Hispanic/Mexican American, 0.5% African American, 4.2% unknown); Protestant or Catholic (84%); and middle-class (mother’s education: M=15.1 years; father’s education: M=16.2 years). Nineteen of the families had children too young to participate or who refused to participate; however, parents provided data about these children. All children had been adopted as infants (mean age of placement = 4 weeks); none of them had experienced maltreatment prior to the adoption. Virtually all parents adopted because of infertility, and most birth mothers placed their children for adoption because they wanted them to be raised in two-parent families that could provide more opportunities than they felt they could provide.

Approximately 8 years later at Wave 2 (1996–2000), data were collected from 177 of the 190 adoptive families and 127 of the 169 birth mothers. Participating adopted adolescents (75 males, 81 females) averaged 15.7 years of age; adoptive parents were in their late 40s, and birth mothers were in their mid-30s. At Wave 3 (2005–2008), data were collected from 181 of the original 190 adoptive families. Adopted emerging adults (87 males, 82 females) averaged 25.0 years of age; adoptive parents were in their late 50s. At Wave 4 (2012–2014), adopted young adults (50 males, 64 females) averaged 31 years of age. For more information about the project sample, methods, and measures, see Grotevant and McRoy (1998), Grotevant, Perry, and McRoy (2005), and Grotevant, McRoy, Wrobel, & Ayers-Lopez (2013).

The trajectories were estimated using the full set of 190 adoptive families who participated in the study. Reports regarding contact were made by adoptive family members, prioritizing the views of the adopted person. Specifically, at Wave 1, variables were coded from interviews conducted with the adoptive mothers, reporting on their child’s contact with his or her birth mother. Since the adoptive mothers typically controlled the contact with the child’s birth relatives and some of the children were very young, it was felt that more reliable and valid information about contact at Wave 1 would come from her rather than from the child’s report. Interviews with the mothers covered a broad range of topics about the family’s adoption process, with significant attention to the family’s contact with their child’s birth relatives. At Waves 2 and 3, variables were coded from individual interviews with the adoptees. The interviews covered a number of adoption-related topics, with significant attention to contact with birth relatives. At Wave 4, specific questionnaire items administered to the young adult adoptees were used in order to code the relevant variables.

Procedures

At Waves 1 and 2, data were collected from adoptive families during visits to their homes, located across the United States. The data collection sessions typically lasted 4–5 hours and included individual interviews with the adoptive mother, adoptive father, and adopted child or adolescent; administration of several questionnaires; a couples interview with the adoptive parents (Wave 1); and a family interaction task (Wave 2). For the few cases in which home visits were not possible, interviews were completed over the phone and questionnaires were administered through the mail. At Wave 3 (emerging adulthood), interviews were conducted with adoptees through a secure synchronous online interview system, and questionnaires were administered through a secure online system. For the few cases in which online data collection was not possible, questionnaires were mailed to the participant. At Wave 4 (young adulthood), questionnaires were administered online through the secure Qualtrics Survey platform. Modest stipends were paid to adoptees for participation at Waves 3 and 4. All procedures were reviewed and approved by the relevant Institutional Review Boards at the time of data collection.

Variables

Variables for this analysis were coded from interview transcripts, described above. All transcripts were coded by the principal investigators or students who received extensive training and supervision. Coders were initially trained to a criterion of .80 agreement before they could begin coding independently. All transcripts were coded independently by two coders, who subsequently met to compare ratings and resolve disagreements. Three variables were used to construct the contact trajectories. Frequency of contact was defined as the frequency of current contact (within approximately the past year) between the adoptee and his or her birth mother. The following scale was used: 0 = no current contact, 1 = rarely (less than once / year), 2 = occasionally (once or twice / year), 3 = often (3–11 times / year, or once / year for an extended visit); 4 = frequently (once / month or more, or more than one extended visit / year). Frequency was coded from the adoptive mother’s interview at W1, the adoptee’s interview at W2 and W3, and the adoptee’s responses to specific questionnaire items at W4. Contact was coded only if the child was involved, although contact among the adults without the child was rare. The adoptee’s satisfaction with contact with his or her birth mother was coded on a 5-point scale: 0 = very dissatisfied, 1 = dissatisfied, 2 = neutral, or rough balance between satisfied and dissatisfied; 3 = satisfied; 4 = very satisfied. Satisfaction was coded from the adoptive mother’s interview at W1, although coders looked for mothers’ descriptions of their child’s satisfaction. The adoptee’s interview at W2 and W3, and the adoptee’s responses to specific questionnaire items at W4 were used to code their satisfaction with birth mother contact at those time points. Participation in cross-network contact was coded in terms of the total number of people in the adoptive family who had any contact with person(s) in the adoptee’s birth family plus the total number of people in the birth family who had any contact with person(s) in the adoptee’s adoptive family, within the year prior to the interview. Specifically, cross-network contact of the following AKN members was recorded as yes or no, and then summed: adopted person, adoptive mother, adoptive father, adoptive family sibling, adoptive family extended relatives, birth mother, birth mother’s spouse / partner, birth father, birth father’s spouse / partner, birth siblings, birth grandparents, birth family extended relatives. For group categories (e.g., adoptive family extended relatives, birth grandparents), the number of people involved was tallied, if a number was provided, or estimated, if a number was not. This variable was coded from the adoptive mother’s interview at W1, since the children were too young to provide accurate reports of such contact. Adoptees’ interview responses were coded at W2 and W3. Directly comparable data were not collected at W4; thus, W4 responses for this variable were not included in the analysis.

Analysis Plan

Group-based trajectory modeling is a semi-parametric method which is used to identify trajectories of one’s outcome variable or variables of interest (Nagin & Tremblay, 1999). GBTM is a specialized application of finite mixture modeling where the aim of the analysis is to identify groupings of participants with unique trajectories; maximum likelihood estimation is used to create the model parameter estimates (Nagin, 2005). GBTM generates the likelihood that individuals are members of one of a specified k number of classes based on their intercepts and slopes; thus, cases that follow a similar growth pattern over time are assigned to the same trajectory group membership based on sharing the highest probability that they belong to the same group.

The current study utilized the STATA procedure “traj” (Nagin & Jones, 2012) in order to estimate group-based trajectory models (GBTM) predicated on the three contact constructs of interest: frequency of contact with the birth mother, adoptee satisfaction with contact with the birth mother, and participation in cross-network contact. For the cross-network contact variable, only three waves of complete data were available, as participants were asked about cross-network contact at Wave 4 in a less comprehensive way that was not analogous to the previous three waves. Cases with missing data at specific time points were not excluded from the models, as GBTM estimates missing parameters utilizing full information maximum likelihood estimation and is quite robust to missing data (Nagin, 2005; Nagin & Odgers, 2010).

The vast majority of families (n = 180, 94.7%) had fully complete data for at least two of the four time points under study; of the remaining 5.3% (n = 10), five families had partial data at only one wave, three had partial data at two waves, and two had partial data at three waves. Most families had complete data for at least three waves (n = 138, 72.6%), but a minority had complete data across all four waves (n = 63, 33.2%); there was considerable attrition at the fourth wave of data collection, so this figure is not surprising. Additionally, this attrition partially motivated the deliberate choice not to include the fourth wave of data in the cross-network contact variable. The sample was considerably smaller at that wave, and questions about the size of the adoptive kinship network (AKN) were not asked in the same manner. Therefore, our cross-network contact variable was intentionally limited to only three waves of data for these analyses.

Whereas GBTM is commonly utilized to create trajectories based on individual variables, the goal of the current study was to create a holistic picture of birth family contact over time that accounted for relations between different aspects of contact. Thus, it was decided that a multi-group trajectory model would be the primary focus of the study.

Trajectory Validation Process

Most studies using GBTM have no external means to validate their trajectory solutions, except through testing hypotheses about predictors of trajectories or outcomes of trajectory group membership. In this study, extensive interview and survey data provided by participants across waves permit validation of the trajectory solutions, by comparing descriptions of the trajectory groups yielded by the analysis to actual interview text of the participants whose cases were assigned to those groups.

In order to validate the group memberships generated by the group-based trajectory model, paragraph-length descriptions were first developed for each of the four groups, based on examination of the curves generated by the GBTM program. The set of curves for each group may be thought of as an “average” or “typical” case for that group. However, as with any type of latent trajectory modeling, there are variations within the sample around the group “average” or “typical” case. It is possible that no single case precisely follows the trajectory curves shown in the figure.

Next, a coding process was developed that utilized data from interviews with participants from the first three waves of the study and item responses from the fourth wave of the study. Thirty-two cases spread across the four groups were systematically selected for this validation process. Efforts were made to only include cases that had interview transcripts for the first three waves. In addition, no cases from the same adoption agency were selected within each group.

Summaries of these 32 cases were created. Each summary consisted of descriptions of the adoption kinship network’s openness arrangement at Waves 1 through 3. Descriptions also included quotes that pertained to frequency of contact with the birth mother, satisfaction with contact with the birth mother, and participation in cross-network contact, which were taken from the Wave 1 through Wave 3 interviews. As interviews were not conducted at Wave 4, the participants’ responses on items pertaining to frequency of current contact with the birth mother and satisfaction with birth mother contact were included in the descriptions, if available.

The 32 case summaries were then randomized and provided to three coders, who were asked to assign a trajectory group membership to each case based on how the case summary compared to descriptions of the four multi-trajectory groups. Coders included two current graduate students and one faculty member who were familiar with the overall MTARP study but were blind to the current study hypotheses and group assignments. Each coder’s results were then compared to each other and to the group memberships generated by the GBTM.

Results

Results are presented below in three sections: a) trajectory modeling results, b) descriptions of the trajectory groups, and c) validation of the trajectories.

Trajectories

Using the STATA procedure “traj,” a four-group multi-trajectory solution was sought in order to consider the combinatory effects of frequency of birth mother contact, adoptee satisfaction with birth mother contact, and cross-network contact. The “traj” procedure is able to model censored data (e.g., psychometric scales), count data, and binary data (Nagin, 2005), so any potential data to be modeled must be of one of those types. To estimate trajectories of frequency of birth mother contact, measured on a Likert scale, a censored normal (cnorm) model was fit to the data. A similar censored normal model was fit to the data pertaining to satisfaction with birth mother contact. For the cross-network contact variable, a zero-inflated Poisson (zip) model was fit, as this variable was a simple count of the number of cross-network contacts occurring at each wave; the zero-inflated Poisson model corrects for underestimation of occurrences of zero responses, which comprises at least 50% (Wave 1: 50.53%; Wave 2: 50.33%; Wave 3: 53.29%) of the cross-network contact responses across all waves. The number of groupings and trajectory shapes (i.e., linear, quadratic) were guided by significance of adding additional polynomial terms and the best fit of the model (i.e., AIC, BIC) as well as the interpretability of the resulting trajectories. Further, posterior probabilities of group membership were also considered when determining the appropriateness of the trajectories beyond fit statistics.

A four-group multi-trajectory model of frequency with birth mother contact, satisfaction with birth mother contact, and size of adoptive kinship network was identified (see Figure 1): a ‘no contact’ group, a ‘stopped contact’ group, a ‘limited contact’ group, and an ‘extended contact’ group were the resulting descriptors for each. Because group-based trajectory modeling is robust to missing data, all N = 190 cases were utilized in the analysis, and all were assigned to a trajectory grouping on the basis of probability of group membership. The overall model BIC (N = 190) = −2576.0, and AIC = −2517.55. Models testing three- and five-group multi-trajectory solutions were also considered and produced better (smaller) BIC (BIC₃ = −2543.80, BIC₅ = −2530.84) and AIC (AIC₃ = −2496.71, AIC₅ = −2457.78) fit statistics. As a strict model fitting exercise, a four-group solution would not be retained but was in this case. “For reasons of parsimony and comprehensibility, the fewer groups the better,” (Nagin, 2005), which would favor a three-group solution over a five-group solution.

Figure 1.

Profiles for the four multi-group trajectories, including frequency of contact, satisfaction with contact, and size of cross- network participation for each.

In addition to the statistical criteria, the three-, four-, and five-group solutions were reviewed in terms of interpretability and what was known about the typical contact arrangements in the sample. The three-group solution was not retained, because it obscured an important, theoretically meaningful difference in developmental trajectories for some of the variables; this was confirmed by plotting the model estimates for three- and four-group solutions. Specifically, the three-group solution combined families having fundamentally different contact experiences: those in which contact had stopped and those in which it was ongoing, albeit limited in scope. This difference is elaborated further in the descriptions that follow. The five-group solution was not retained, because two of the five groups would be small (less than 10% group membership). On balance, the four-group solution seemed to be the best fit, when statistical, conceptual, and sample criteria were considered together.

According to Nagin (2005), each trajectory group should meet the criteria of having a mean posterior probability of group membership greater than 70%, and this was true of the sample. Respectively, the ‘no contact’, ‘stopped contact’, ‘limited contact’, and ‘extended contact’ groups had mean posterior probabilities of 98.6%, 96.1%, 95.9%, and 94.0%. Across all cases, 91.6% had a posterior probability of group membership greater than 90%. There was no significant difference in gender by trajectory group (χ²(3) = 2.60, p = 0.46).

Trajectory Group Descriptions

Profiles of the four trajectory groups are depicted in Figure 1, and descriptions and quotes relevant to each trajectory are presented in Table 1. The quotes were taken from the 32 cases used to validate the trajectories. Each column in Figure 1 represents one of the trajectory groups, and the rows depict the three variables comprising each profile (frequency of contact on the top row, satisfaction with contact in the center row, participation in the network in the bottom row). The four groups are described as follows:

Table 1.

Quotes Illustrating Contact Patterns in Four Trajectory Groups

Trajectory Group	Context	Quotes
No Contact	Desire for Health and Medical History	“I would like to have a little more medical history. I’m just more curious more [than] anything, especially now that I have a child… I’m not quite sure I’m ready to involve her in my life with my child” (Adopted emerging adult, Wave 3) “Since I recently had my child I am now more curious about my birth family... I would like to at least meet my half siblings, I would like to know my family health history.” (Adopted emerging adult, Wave 3)
	Openness to Idea of Contact	“I’d just like to meet them, get to know them. And that would be about all. Let them know that I’m doing all right and that they can see me.” [8 years later] “I would extend to them [birth family] a relationship with them, to extend a helping hand” (Adopted person, Wave 2 and Wave 3) “I just kind of want to see her and meet her [her birth mother], but then I don’t. I do, but I don’t.” [8 years later] “I see it [searching] pessimistically, that it’s not going to be easy, and it’s not even worth the trouble. But it would be. I don’t know.” (Adopted person, Wave 2 and Wave 3)
Stopped Contact	Contact Fading Out or Ending	“The communication just stopped…I am guessing she [her birth mother] moved. Or she got too busy with three kids… I don't have time to think about where she is, or should I contact her” (Adopted emerging adult, Wave 3) “I try not to keep contact with her, my mom does, on and off I think, but I really don’t have any interest, so I don’t find out much.” (Adopted adolescent, Wave 2)
	Contact Not Working Out	“My [adoptive] mother felt like she [my birth mother] was intruding and I believe she felt threatened The only reason I am not pursuing a relationship with [my birth mother] right now is out of respect for the mother that raised me.” (Adopted emerging adult, Wave 3) “My folks said she [his birthmother] stopped sending letters because of something they did. They never told me why.” (Adopted emerging adult, Wave 3)
Limited Contact	Keeping Contact Limited	“I wish I had more contact with her [my birth mother] and her family, but I understand that they have their lives and I have my life.” (Adopted emerging adult, Wave 3) “I was extremely satisfied [when I was a high school sophomore] … I didn’t feel like anything was missing…. Then meeting her was a drastic change, and she instantly became a bigger part of my life…. Our relationship hasn’t strengthened all that much … I have just been sitting back on it a little bit.” (Adopted emerging adult, Wave 3)
	Challenges in Negotiating Increasing Contact	“She [birth mother] wishes we could do more stuff like spend a whole week together and we’d talk ... and stuff, instead of just one day, or something. I don’t know, ... it's hard to say if I’m pleasing her or not.” (Adopted adolescent, Wave 2) “She is trying too hard…I wanted to find her to say thank you for giving me life…I was glad to meet her but I am not ready to make her a part of the family.” She continues by conveying her frustration with unforeseen difficulties in managing contact: “I wanted to find her. And I had good reasons for it. I just wish that someone could have helped me to know what might possibly happen after I found her and helped me work through it.” (Adopted emerging adult, Wave 3)
Extended Contact	Birth Family as Extended Family	“We just sort of take it as a normal part of life. We don't get uptight or anxious. She's sort of an extended part of the family. We would treat her like you would any other family It's like a friend coming to visit.” (Adoptive mother, Wave 1) “I love spending time with my birth mothers [sic] family…we [the adopted young adult and his adoptive mother] talk about how great it is to be able to share in the lives of people who would seem to be strangers but in fact are family…to get together and be family first, but friends as well.” (Adopted emerging adult, Wave 3) “We have a Christmas party every year…and…we just count them as family…they’re just my family.” (Adopted adolescent, Wave 2)
	Relational Work for Extended Contact	“It’s just been, like such an awesome experience…and like hard, I guess, at times... You always have to, you know, consider everyone else’s feelings, too.” (Adopted adolescent, Wave 2)

• Group 1 (No Contact. N = 79; 41.6% of sample; 44.3% female): Adopted persons in this group appeared to have no birth family contact across time; both frequency and cross-network contact appeared to be consistently zero or very low. Satisfaction with contact declined in a linear way across time, suggesting that the adoptees became less satisfied with no or minimal contact.

• Group 2 (Stopped Contact. N = 26; 13.7% of sample; 57.7% female): Adopted persons in this group had contact with their birth mothers initially (perhaps only for the first year or two after placement), but contact gradually decreased over time. This is mirrored in a declining level of cross-network contact. Satisfaction with contact declined from wave 1 to wave 2, and then remained at a moderate level.

• Group 3 (Limited Contact. N = 50; 26.3% of sample; 52.0% female): In this group, there was some level of initial contact for many families, and it generally increased over time. The level of cross-network contact also increased slightly, but remained relatively small in comparison to the level of contact in Group 4. Satisfaction with birth mother contact started high and decreased over time to wave 3.

• Group 4 (Extended Contact. N = 35; 18.4% of sample; 40.0% female): In this group, frequency of contact between the adopted person and birth mother started high and stayed high relative to the other groups, and the level of cross-network contact started out relatively high and grew over time relative to the other groups. Satisfaction with contact started high and stayed that way, with a slight dip at adolescence.

Validation of Trajectories

To validate the trajectories, we evaluated how many cases could be accurately assigned to trajectory groups by trained coders. Two of the coders accurately assigned 18 of the 32 cases (56.3%), whereas one coder accurately assigned 20 of the 32 cases (62.5%). Of the 32 cases, 15 were accurately assigned by all three coders (46.9%). Binomial tests were conducted to determine whether or not the levels of accurate group assignment exceeded the expected probability of chance. As there were four multi-group trajectories to which a case could be assigned, probability of chance was set to .25. Results of the one-tailed binomial tests indicated that the probability of accurately assigning 18 out of the 32 cases was significant higher than the expected probability of chance (p < .001), as was the probability of accurately assigning 20 out of the 32 cases (p < .001).

Discussion

This is the first paper to identify trajectories of change in contact between adoptive and birth family members across time at the level of the adoptive kinship network, from the time when the adoptee was in middle childhood to young adulthood. Group-based trajectory modeling revealed four distinctive groups: No Contact, Stopped Contact, Limited Contact, and Extended Contact. These groups were readily interpretable within the context of emotional distance regulation theory, which highlights ways in which the members of family systems navigate relationships over time and describes why some relationships thrive and expand while others fade or end. It is important to acknowledge, however, that descriptions of the group prototypes do not apply equally well to every case in the group; the prototypical group descriptions represent a “typical” or “average” case, but there is variation within each group.

Group 1 (No Contact) primarily consisted of kinship networks in which there had been no contact between adoptive and birth family members, reflected by the consistently low frequency of contact and number of members engaged in cross-network contact. It is possible that the decline in adoptees’ satisfaction with openness arrangements reflected some young adults’ desire for health and medical information (e.g., Wrobel & Grotevant, in press) as well as their openness to the idea of contact, especially in light of increasing societal acceptance of contact between birth and adoptive family members (Siegel & Smith, 2012).

Group 2 (Stopped Contact) was characterized by adoptive kinship networks in which there was some initial contact between the adoptive family and the child’s birth mother, but the contact gradually decreased over time and then stopped. This group includes some families who had time limited mediated contact, typically contact that had been planned for a limited time. Contact stopped for some families because someone lost interest in maintaining the contact, contact information was lost, or the initial contact “didn’t work out” for one of many possible reasons.

In Group 3 (Limited Contact), it appears that the adopted person and a small number of birth relatives have worked towards achieving a reasonably satisfying relationship by young adulthood. Contact in this group may mostly be with just a few people (perhaps the young adult’s birth mother and her immediate family) rather than the whole large extended family, and may have begun or increased dramatically later in the adoptee’s life. Some cited challenges in negotiating increased contact.

Finally, Group 4 (Extended Contact) appears to include adoptive kinship networks where there was a considerable amount of ongoing contact across the network. These appear to be the networks where the birth and adoptive families really engaged each other, spending considerable amounts of time together with multiple extended family members, including holidays and regular visits. The adoptees often talked about birth family members being part of their extended family; they also noted the relational work involved in maintaining contact over time.

The trajectory groups yielded by GBTM reveal ways in which the key openness variables (frequency, satisfaction, number of participants) are interdependent in different ways for different groups. In the ‘no contact’ and ‘extended contact’ groups, frequency of contact and number of participants are similar to each other (both close to zero in the ‘no contact’ group, and both high in the ‘extended contact’ group). However, in the ‘stopped contact’ and ‘limited contact’ groups, frequency of contact and number of participants vary. Although across the full sample there is a positive association between contact and satisfaction with contact (Grotevant, Rueter, Von Korff, & Gonzalez, 2011), satisfaction is seen to vary across the four trajectory groups. Thus, the data reported here reveal for the first time the interplay among frequency of contact, satisfaction with contact, and participation in the four distinctive groups. These groupings provide new insights not afforded by looking at the variables one at a time.

We acknowledge that adoptive persons, adoptive parents, and birth parents may have different perspectives on satisfaction and that these different perspectives were not captured in the current study. The decision to focus on adoptee’s satisfaction was due to their central role in the adoption triad. In addition, consideration of others’ satisfaction would introduce too many parameters into the model, and we believe that somehow combining satisfaction across the adoption triad would reduce meaningful variability and perhaps not represent any individual. Future examinations could and should examine birth mother or adoptive parent satisfaction as a group-level difference at the four time points.

Significance of Validating Group Memberships

The goal of the coding process was to validate the trajectory solution generated by the group-based trajectory model through the use of extensive interviews with the participants. Thus, the coding process was meant to show that the model generated and assigned group memberships in a meaningful way and not simply to copy the algorithm utilized by the model. Given this direction, it was decided that results from the coding process were to be compared to chance probability, as opposed to the strict reliability guidelines seen in traditional behavioral coding. Results from these comparisons provided evidence for the model’s ability to distill cases into four coherent groups that reflected the variability seen across participants.

To our knowledge, no existing studies have used qualitative interview data to validate group-based trajectory modeling, and validity has been primarily examined through examination of predictors and outcomes of trajectories (see, for example, Duchesne & Ratelle, 2014; Hirai, Frazier, & Syed, 2015; Miller-Graff & Howell, 2015; Proulx, Ermer, & Kanter, 2017; van der Geest, Bijleveld, Blokland, & Nagin, 2016).

Although coders successfully assigned cases well above chance probability, approximately 40% of cases were misclassified across coders. This percentage of misclassified cases and consistency in performance across coders suggest that there may be limitations in the group-based trajectory model’s ability to fit cases with multiple pieces of data into a small number of distinct groups, as well as human coders’ abilities to mimic the decision-making process of the model. In order to create the trajectories, the model was required to account for a maximum of eleven data points per case: four for frequency of contact, four for satisfaction with contact, and three for participation in cross-network contact. Thus, individual differences in cases, presented as variability across these eleven pieces of data, were compressed into four average trajectories. As expected, examination of the 32 cases indicated that many did not neatly fit into the groups generated by the model. Clearly, individual trajectories within each group are somewhat heterogeneous, especially when accounting for three different variables over three to four time-points spanning over 20 years. Thus, it was perhaps not surprising that coders’ abilities to classify cases correctly was moderate, albeit significantly higher than chance. Nevertheless, although each adoptive kinship network is unique in some ways, GBT modeling revealed four broad patterns found within the sample, albeit with variation within each pattern (as is found with any case-centered method that attempts to identify latent groups).

Strengths and Limitations

This study has a number of strengths. Findings were based on a large, nationwide longitudinal (over 30 years) sample of adoptive kinship networks varying in contact patterns. In the first study of this type, GBTM provided a systematic quantitative method for revealing distinctive groups varying in openness trajectories at the level of the adoptive kinship network. Generalizability from this type of adoption (domestic adoption of infants) to other forms of adoption (international, child welfare, kinship) is limited, but the present data may suggest important aspects of family relationships to consider in understanding more diverse families. Although the four waves of data provide important information about change, most waves are six to eight years apart in time, and some changes that occurred between points of measurement may not have been captured in the trajectory groupings.

Implications for Practice and Future Research

Identification of contact pathways will be useful for a variety of professionals working with families. Adoption agency staff working with prospective adopters and clinicians providing post-adoption services to adoptive and birth families will have an evidence base from which they can advise clients about the diversity of family configurations and, importantly, their dynamic nature. Knowing a family’s starting point in terms of frequency, satisfaction, and participation by itself is a poor predictor of the trajectory that that family might follow over time. The trajectories, in combination with emotional distance regulation theory, provide practitioners with language and concepts that family members can understand as they experience changes over the life course. The trajectories also normalize the expectation that there is not just one “correct” or “best” way to experience adoption; and that openness situations vary for many reasons, especially as a function of the cognitions and personalities of the individuals who comprise them. Finally, the trajectories provide family educators with information that can be used by school personnel, health providers, social workers, and other professionals who encounter adoptive kinship network members in the course of their daily work and yet may not be aware of the many forms that these complex family structures may manifest. Future analyses will permit examination of factors that predict trajectory group membership as well as the usefulness of these trajectories in predicting outcomes for adoptees, including psychological adjustment and well-being, adoptive identity, and close relationships.