The sample of Brazilian street-involved youth of this study were recruited in different settings. Specifically, we briefly described data collection spaces. Institutional shelter (or "homeless shelter") is a protective measure that consists of the removal of the child or adolescent from his / her family of origin, when some (or several) situations of violation of rights occur, such as when the child/teenager is a victim of neglect, physical or sexual violence. In this case, the child/adolescent —from a court order—starts living in an institution that should be as similar as possible to a residence, for a period that can extend for a maximum of two years ¹ . At the shelter, the children/adolescents perform all activities necessary for their survival and their routine (food, sleep, hygiene, school, leisure, etc.).

Differently, in open institutions young people have the freedom to walk through these spaces, for example, using them only a day per week or for three days in different shifts, depending on the rules of the institution and the young person's desire. In this case, no court order has required the child/adolescent to remain in that space. However, in the case of non-compliance with rules and regulations, young people are deprived of the use of these services, usually for 72 hours. These spaces are like pass houses, where young people find protection, and can be sent to an institutional shelter, to their families or returning to the streets. Regarding the third data collection space, the street, there was only a small portion of the sample (from Fortaleza) that was properly in the streets, sleeping in porches, bus terminals and frequenting the city center and beaches, a reality that explained the change in the profile of this population.

In Time 1, 113 youth who had to live on the street for at least six months of the first data collection were recruited in institutions shelter (80%), open institutions serving street-involved youth (17%) and street (3%). The participants were recruited in Fortaleza (n = 45, 39.8%), Salvador (n = 40, 35.4%), and Porto Alegre (n = 28, 24.8%). There were no significant differences between cities for sex and age. The following shows the number of respondents in each data collect Time: 81 (72%) young people were interviewed in T2 and 68 (61%) in T3. The analytical sample of this article consisted of 104 young people who completed all measures related to SWB (PA, NA, and LS) in Time 1. These young ones ranged from 9 to 18 years (M = 14.22 years, SD = 2.4) and the majority (82%) were males and (91%) non-white.

Participant’s Form. A set of 11 questions for sociodemographic categorization of participants based on the Life Experience Interview (Raffaelli, Koller, & Morais, 2007) was created. In this article, we analyzed the items that dealt with the sociodemographic data: age, sex, relations with the institution (a type of institution attended), street and family (contact intensity - weak, medium and strong).

Stressful Events (Number and Impact). It is a checklist of 22 events that could have occurred during the last six months (e.g., relocation, return home, death threat, and the death of a friend or family member). The frequency of occurrence of the events was used to calculate the number of events. For those events occurred, participants rated each one as impact (5 items; 1- nothing stressful to 5 - totally stressful). The average of the total impact is achieved by dividing the sum of the assigned impact values by the events experienced by the total number of events with a valid impact.

Health-Risk Behaviors. Formed from a combination of three indicators (suicidality, sexual risk, and drug use). For its construction, the following steps were conducted: 1) the transformation of each adjustment indicator into a standardized score (z-score); and 2) the sum of the normalized scores of all indicators (suicide rate, sexual risk, and drug use).

Suicidality Index. Computed from items on suicide ideation and attempted suicide, being ranked in 0 (no ideation and attempts), 1 (thought about suicide but had no attempt), 2 (attempted once) and 3 (attempted two or more times).

Sexual Risk Index. Composed of the sum of four indicators (already had sex, sexual intercourse before the age of 13, two or more partners in the last 6 months, no condom at the last sexual intercourse). Values ranged from 0 (no risk) to 4 (presence of all sexual risk behaviors).

Drugs Use. Drug use was measured by the frequency of consumption of seven drugs (licit and illicit: alcohol, cigarettes, marijuana, cocaine, crack and others) in the last month, with values from 0 (did not consume) to 7 (consumed all types drugs).

Positive and Negative Affect. The measurement of affect was obtained by the Positive and Negative Affect Schedule for Children (PANAS-C, Watson, Clark, & Tellegen, 1988; Laurent et al., 1999) adapted by Raffaelli et al. (2007). It is a two-factor scale (Positive Affect, 17 items, alpha = 0.86; Negative Affect, 17 items, alpha = 0.89) of 34 items classified in five points (1 - not at all to 5 - very much). Feeling participatory, determined, strong and courageous are examples of PA. While feeling sad, humiliated, worried, and nervous are examples of items of NA.

Life Satisfaction. Composed of five items (1 - strongly disagree to 5 - strongly agree) of the Life Satisfaction Scale (Diener, Emmons, Larsen, & Griffin, 1985) adapted by Koller et al. (1996) (alpha = 0.72). One example is the following questions: "Is your life close to your ideal?" and "Are you satisfied with your life?”.

The composition of the sample performed three stages. In the first stage, the potential recruitment places identified included social assistance institutions that provided services to street-involved youth. In the second stage, places identified as regular meeting places for young people were selected for the recruitment of participants. In the third stage, the research team was inserted in these spaces, beginning the ecological engagement with six-month duration. This approach privileged the insertion of the researcher into the research context, enabling the researcher-participant interaction regularly of time (Koller, Morais, & Paludo, 2016). Participants were eligible if they had current or recent street experience (e.g., sleeping, working, or “hanging out” on the street) or engage in illicit or risky street-based activities (e.g., drug trafficking, sexual exploitation). After selection and acceptance, the participants completed a set of measures in three times (T1, T2, and T3) with an interval of six months between each one. The instruments were applied, on average, in two individual 35-minute meetings, conducted by the research assistants duly trained in the joint study protocol.

We conducted a hierarchical cluster analysis to identify groups of street-involved youth with different SWB profiles at Time 1. The Ward method with squared Euclidean distance as a measure of proximity was used. The score for each variable was standardized (z-score), before the analysis, to minimize the effects of different scales between variables. The selection of the clusters based on an inspection of the dendrogram branches and the coherence of the resulting clusters through the analysis of variance, being verified the existence of significant variability between the clusters in relation to the components of SWB. For the longitudinal characterization of the clusters, descriptive statistics (mean, standard deviation, frequency, and percentage) and analysis of comparison (ANOVA) between clusters were carried out, considering the differentiation of the groups in the other variables investigated over time (T1, T2, and T3). The data were analyzed using the Statistical Package for Social Sciences - SPSS (version 19), considering a level of significance of p<0.05.

The Ethics Committee from the institution of origin of the authors (Protocol 2011023) approved the study, and we followed all procedures necessary for the research with youth in situations of social vulnerability, suitably.

The cluster analysis performed with the three components of the SWB (LS, PA, and NA) indicated three distinct SWB profiles of young people in the street situation. As depicted in Figure 1, clusters include an average SWB (n = 56; 49.6%) characterized by moderate LS, PA, and NA. A positive SWB (n = 21; 18.6%) defined by high LS and PA and low NA; and a negative SWB (n = 27; 23.9%) distinguished by low LS and PA and moderate NA. The values of SWB variables correspond to normalized means and are described in Table 2.

Figure 1 Figure 1 Clusters formed according to different levels of SWB (LS, PA, and NA)

As for the sociodemographic characterization, as shown in Table 1, clusters do not differ significantly in gender. On the topic of the age, the follow-up test (LSD) verified that the positive SWB is composed of younger children/adolescents than the negative SWB. Regarding the relationship with the streets, all clusters had medium to strong contact intensity, and there are no significant differences between the groups. The clusters differed significantly for the variables’ intensity of contact with family [F (2) = 1.14, p <0.05, Cohen’s d = 1.20] and type of institution (open institution) [F (2) = 3.6, p <0.05, Cohen’s d = 0.68]. The post hoc analysis (LSD) identified that positive SWB was made up of young people with stronger family ties than the other groups. Participants who composed the average SWB and positive SWB clusters used the open institutions more frequently than those of the negative SWB.

Table 1

Subscribers reflect differences between groups based on ANOVA (p < 0.05).

Table 2 presents the descriptive analysis and comparisons between groups for SWB variables, stressful events and health-risk behaviors in Times 1, 2 and 3. In T1, besides the clusters differ in relation to SWB as previously described in the identification of the profiles, there were significant differences for the health-risk behaviors [F (2) = 6.24, p < 0.01, Cohen’s d = 0.96] and sexual risk [F (2) = 8.73, p < 0.001, Cohen’s d = 1.08]. Follow-up tests (LSD) indicated that the positive SWB group included participants with lower levels of health-risk behaviors and sexual risk than the other groups, as well as suicidality lower than the negative SWB group.

At Time 2, clusters had significant differences for PA [F (2) = 12.65, p < 0.001, Cohen’s d = 1.56]. Analyzes with the LSD test indicated that the positive SWB group had higher mean PA than the other groups, as well as higher LS levels and lower NA levels than the average SWB group. In T3, there were significant differences between the clusters in the PA [F (2) = 5.32, p <0.01, Cohen’s d = 1.89], LS [F (2) = 4.62, p <0.01, Cohen’s d = 1.12] and health-risk behaviors [F (2) = 3.18, p < 0.05, Cohen’s d = 0.88]. The post hoc (LSD) test indicated that the negative SWB group had lower LS levels than the other groups, as well as higher health-risk behaviors and the impact of stressful events than positive SWB. This cluster still stood out for lower means of health-risk behaviors than the average SWB cluster. The different SWB profiles did not differ significantly in terms of the number of stressors and drug use at any time (T1, T2, and T3).

Table 2 Table 2 Means, Standard Deviations and SWB Cluster Comparisons, Stressful Events (Number and Impact) and Health-Risk Behaviors Variables ¹

Means of the variables based on z-score. The value of N for some variables may have variation due to missing. Subscribers reflect differences between groups based on ANOVA (p < 0.05). SE = Stressful Events.