The Longitudinal Stability and Dynamics of Group Membership in the Dual-Factor Model of Mental Health

Abstract

French

This study examined the longitudinal stability and dynamics of group membership within the Greenspoon and Sakflofske’s dual-factor model of mental health. This expanded model incorporates information about subjective well-being (SWB), in addition to psychopathological symptoms, to better identify the mental health status and current functioning of youths. Participants included 730 seventh- and eighth-grade U.S. middle school students. Four groups of students were identified: flourishing (i.e., high SWB and low psychopathology), vulnerable (i.e., low SWB and low psychopathology), symptomatic but content (i.e., high SWB and high psychopathology), and troubled (i.e., low SWB and high psychopathology). Analyses of stability of group membership revealed varying levels of stability, with flourishing students demonstrating the highest stability. Analyses also revealed that social support variables yielded meaningful predictions of changes in group membership. Implications for professional practice are discussed.

Keywords

positive mental health adolescents dual-factor model subjective well-being

Psychologists and educators have traditionally relied on a unidimensional, deficit-focused approach to practice, in which mental health is conceptualized as the absence of psychopathological symptoms (Keyes, 2005). As a result, professional practices are organized around identification and amelioration of psychological symptoms with little regard for positive psychological factors (e.g., character strengths, subjective well-being) that may constitute or promote optimal levels of mental health. Positive psychologists have argued that the absence of symptoms is not an adequate definition of mental health; rather, it must include the presence of positive psychological characteristics (Seligman & Csikszentmihalyi, 2000).

Subjective well-being (SWB) is a key construct in positive psychology comprised of three major components: life satisfaction, positive affect, and negative affect (Diener, 1984). Life satisfaction is defined as a person’s cognitive appraisal of the positivity of her or his life overall based on her or his unique standards. Positive affect refers to the frequency of positive emotions (e.g., joy, excitement, affection) experienced by an individual. Conversely, negative affect represents how often an individual experiences negative emotions (e.g., anxiety, sadness). Research has demonstrated the utility of SWB assessment as an indicator and determinant of positive aspects of mental health (Proctor, Linley, & Maltby, 2009)

Research has supported positive and negative well-being as distinct constructs (Huppert, 2009). Based on such distinctions, a multidimensional, dual-factor model (DFM) of mental health was proposed by Greenspoon and Saklofske (2001) to integrate assessments of SWB and psychopathology. From a sample of 407 Canadian elementary students, Greenspoon and Saklofske (2001) identified four distinct groups varying in mental health. Using the nomenclature of Antaramian, Huebner, Hills, and Valois (2010), the first group was identified as having positive mental health, with high SWB scores and low psychopathology scores. A second group was characterized as troubled children, with low SWB and high psychopathology. The third group, labelled as vulnerable, demonstrated low levels of psychopathology with low SWB. The fourth group, labelled symptomatic but content, reported high SWB, despite high psychopathology. The latter two groups do not neatly fit into the traditional, unidimensional model of mental health; in fact, Groups 3 and 4 would not have been identified via traditional assessments. Greenspoon and Saklofske (2001) further identified a variety of psychosocial variables that differentiated the groups. Additionally, Lyons, Huebner, Hills, and Shinkareva (2012) also found that parent social support was a significant determinant of group membership in a sample of U.S. secondary students.

Suldo and Shaffer (2008) subsequently evaluated the DFM by classifying U.S. middle school students into the four groups and investigating differences in educational, social, and health-related presumed “outcomes.” Most notably, students with the highest mental health had higher reading achievement, academic self-perceptions, greater motivation and valuing of school, fewer social problems and school absences, and better parent and peer relationships than vulnerable students (i.e., low SWB, low psychopathology).

Antaramian and colleagues (2010) further investigated the DFM. As expected, middle school (U.S.) students with positive mental health consistently demonstrated higher school grades and higher levels of behavioural, cognitive, and emotional engagement, when compared to vulnerable students. Also, the two groups with high SWB had significantly more social support (i.e., parent, teacher, and peer) for learning compared to the groups with low SWB, despite comparable levels of psychopathology.

The aforementioned research suggests that the DFM provides a more comprehensive understanding of students’ well-being relative to traditional unidimensional models. To date, research has been limited to cross-sectional designs. Although useful in early research, cross-sectional approaches do not account for temporal processes or causal mechanisms. Thus this study used a short-term longitudinal design to assess the stability and predictors of DFM group membership. Students were categorized into one of four DFM groups across two different time periods, 5 months apart. Drawn from Lent’s (2004) integrative model of SWB, four social support variables (i.e., teacher–student relationships, family support for learning, and peer support for learning, seeking social support) were evaluated as predictors of group membership. Hence, this study further investigated the utility of the DFM in youth by determining (a) the longitudinal stability of group membership and (b) whether social support variables predicted changes in group membership.

Method

Participants

A total of 730 seventh and eighth graders (ages 11 to 15, 51% female) participated in a two-wave assessment of school climate and student well-being (T1, fall 2008; T2, spring 2009). These data have been used in prior research (e.g., Antaramian et al., 2010); however, these analyses are new. The students were from one middle school in the Southeastern United States. Of the participants, 27% were African American, 4% were Asian, 60% were White, 2% were Hispanic and 7% reported themselves as “other.” Additionally, 21% received free or reduced-cost lunch at school, suggesting lower socioeconomic status (SES).

Measures

Subjective well-being—life satisfaction

Life satisfaction was assessed using the Students’ Life Satisfaction Scale (SLSS; Huebner, 1991), a self-report scale composed of seven items that reflect general statements about one’s life (e.g., “My life is going well”). Students rated their agreement with each item on a 6-point scale, ranging from strongly disagree to strongly agree. The SLSS has been used in numerous studies, and research has provided evidence of its reliability and validity (Huebner & Hills, in press).

Subjective well-being—positive and negative affect

Positive and negative affect were assessed using the Positive Affect and Negative Affect Scale for Children (PANAS-C; Laurent et al., 1999). It is a 27-item scale comprised of two subscales: Positive Affect and Negative Affect. The Positive Affect subscale consists of 12 descriptors, such as happy, lively, and proud. The Negative Affect subscale includes 15 descriptor items, such as gloomy, mad, and afraid. Children respond to items using a 5-point scale from very little or not at all to extremely or all the time during the past few weeks. For the purposes of the larger survey from which these data were drawn, the instructions of the PANAS-C were modified to request students to report on their positive and negative affect in the school setting. The PANAS-C has been widely used in research, and there is evidence supporting its reliability and validity (Huebner & Hills, in press).

Psychopathology

Psychopathology was assessed using the Self-Report Coping Scale (SRCS; Causey & Dubow, 1992), a 34-item measure that assesses the use of five different coping strategies in a socially stressful situation. The Internalizing and Externalizing subscales were used in this study to measure psychopathological symptoms. The Internalizing subscale includes seven items, such as “Worry that others will think badly of me” and “Become so upset that I can’t talk to anyone.” The Externalizing subscale is composed of four items, such as “Curse out loud” and “Take it out on others because I feel sad or angry.” Students reported how often they engaged in each behaviour on a 5-point scale ranging from never to always. Studies indicate sound psychometric properties of the SRCS (Causey & Dubow, 1992; Roecker-Phelps, 2001).

Social support—seeking social support

An additional subscale of the SRCS was used to assess how often students’ rely on social support from teachers, family, and friends as a coping strategy. The Seeking Social Support subscale is composed of 8 items, such as “Tell a friend or family member what happened” and “Talk to somebody about how it made me feel.” Students reported how often they sought out social support in response to a stressful event by rating the items on a 5-point scale ranging from never to always.

Social support—environmental context

Students’ environmental context was measured using three subscales from the Student Engagement Instrument (SEI; Appleton, Christenson, Kim, & Reschly, 2006), a 35-item self-report scale. The Family Support for Learning subscale includes four items that assess family involvement in school (e.g., “When I have problems at school, my family/guardians are willing to help me”). The Teacher–Student Relationships subscale is composed of nine items that describe positive relationships with adults at school (e.g., “My teachers are there for me when I need them”). The Peer Support for Learning subscale includes six items reflecting good peer relations at school (e.g., “Other students at school care about me”). Responses to all three scales are indicated on a 4-point scale ranging from strongly agree to strongly disagree. Evidence of reliability and validity has been provided in several sources (e.g., Appleton et al., 2006; Reschly, Huebner, Appleton, & Antaramian, 2008).

Procedure

The study involved analyses of archival data collected in an in-house survey of school climate and students’ well-being. A total of 11 (1%) declined participation. Surveys were administered by teachers during homeroom in October 2008 and March 2009. Eighty-four participants were removed because of excessive missing data (> 25% of items per scale), yielding a final total sample of 730. Analyses of the relevant study variables revealed no statistically significant differences between the students removed from the database and the remaining sample.

Data Analysis

DFM classification

Consistent with past research, levels of SWB using Diener’s tripartite model (1984) were identified, including the three constructs of positive affect, negative affect, and life satisfaction. A SWB score was calculated based on the same formula used in previous research; that is, composite scores were calculated based on life satisfaction and positive and negative affect standardized scores (Antaramian et al., 2010; Suldo & Shaffer, 2008). Specifically, we standardized the SLSS and PANAS-C scores, summed the life satisfaction and positive affect scores, and subtracted negative affect scores.

Participants were classified as having either moderate-to-high SWB or low SWB based on widely used interpretation guidelines (Reynolds & Kamphaus, 2004). Students with a T score of 40 or lower were considered to have low SWB, whereas those with a T score higher than 40 were classified as having moderate-to-high SWB.

To determine DFM status, participants were first classified according to their level of psychopathology, measured by SCRS scores. Again, based on commonly used guidelines (Reynolds & Kamphaus, 2004), participants were classified as having high psychopathology if they had a T score of 60 or higher on Internalizing, Externalizing, or both scales. Students DFM’ group affiliation was then determined by pairing dichotomized levels of SWB and psychopathology using these decision points.

Results

Descriptive Analysis and Intercorrelations

Descriptive statistics are reported in Table 1. Mean scores on the SWB variables were comparable to those found in prior studies (Haranin, Huebner, & Suldo, 2007; Reschly et al., 2008). Means on the internalizing and externalizing variables were also consistent with those from prior research (Causey & Dubow, 1992).

Table 1.

Descriptive Statistics, Intercorrelations, and Coefficient Alphas at Time 1 and Time 2

Time 1	1	2	3	4	5	6	7	8	9
1. Life satisfaction	0.82***
2. Positive affect	0.40***	0.90***
3. Negative affect	−0.42***	−0.25***	0.90***
4. Externalizing	−0.32***	−0.10***	0.33***	0.71***
5. Internalizing	−0.27***	−0.13***	0.48***	0.48***	0.74***
6. Family support for learning	0.48***	0.35***	−0.27***	−0.26***	−0.08*	0.81***
7. Teacher–student relationships	0.37***	0.34***	−0.23***	−0.29***	−0.15***	0.49***	0.91***
8. Peer support for learning	0.42***	0.49***	−0.32***	−0.15***	−0.18***	0.41***	0.45***	0.84***
9. Seeking social support	0.31***	0.37***	−0.08*	−0.10***	0.13***	0.38***	0.41***	0.35***	0.85***
M	4.46	3.86	1.89	2.26	2.56	3.56	2.92	3.29	2.98
SD	1.01	0.81	0.74	0.97	0.81	0.55	0.67	0.58	0.89
Time 2	1	2	3	4	5	6	7	8	9
1. Life satisfaction	0.87***
2. Positive affect	0.42***	0.92***
3. Negative affect	−0.47***	−0.23***	0.90***
4. Externalizing	−0.29***	−0.14***	0.36***	0.75***
5. Internalizing	−0.29***	−0.16***	0.48***	0.53***	0.79***
6. Family support for learning	0.55***	0.43***	−0.35***	−0.27***	−0.20***	0.86***
7. Teacher–student relationships	0.49***	0.42***	−0.35***	−0.23***	−0.18***	0.58***	0.92***
8. Peer support for learning	0.42***	0.51***	−0.34***	−0.07*	−0.16***	0.47***	0.55***	0.88***
9. Seeking social support	0.33***	0.42***	−0.12***	−0.01	0.10***	0.44***	0.49***	0.44***	0.87***
M	4.30	3.88	1.82	2.33	2.56	3.52	2.97	3.31	3.01
SD	0.58	0.82	0.69	1.00	0.85	0.60	0.70	0.61	0.92

Note: N = 730. Bold = coefficient alpha.

p < .05. ***p < .005.

Descriptive statistics were used to examine distributional properties. Skew and kurtosis were evaluated using rules of thumb provided by Tabachnick and Fidell (2007). Specifically, absolute values of skew below 2 and kurtosis below 3 were not regarded as severe deviations. Based on these criteria, one outcome variable (i.e., family support for learning) was leptokurtic (K = 3.4). Because this variable was only slightly leptokurtic, and because the statistical models used to analyze the data are generally robust to distributional violations, the variables were not transformed to preserve integrity of interpretation.

Coefficient alphas were subsequently calculated; all exceeded .70 (Table 1). Test-retest coefficients were also calculated and were as follows: global life satisfaction (.64), positive affect (.57), negative affect (.49), externalizing behaviour (.59), internalizing behaviour (.49), family support for learning (.60), teacher–student relationships (.62), peer support for learning (.59), and seeking social support (.59).

DFM Classification

Participants were classified into DFM groups based on levels of SWB and psychopathology. These groups will be referred to using Suldo and Shaffer’s (2008) nomenclature with the exception of the “complete mental health” group, which will be referred to as “flourishing.” Tables 2 and 3 show the distributions of participants. The flourishing group had average to high levels of SWB and low levels of psychopathological symptoms. With 466 participants, the flourishing group was the largest of the groups, comprising 64% of the sample. The vulnerable group included 59 adolescents who not only had low levels of psychopathology but also had low SWB; they comprised 8% of the sample. Approximately 20% of the participants were classified into the third group, labelled as symptomatic but content. These adolescents had average-to-high SWB; in spite of also having high psychopathology. Finally, 57 adolescents were in the troubled group, comprising 8% of the sample.

Table 2.

Mental Health Status Groups Yielded From a Dual-Factor Model With Predictor Descriptive Statistics at Time 1

		Subjective well-being
Psychopathology		Average to high		Low
Low		Flourishing		Vulnerable
		N = 466		N = 59
		64%		8%
High		Symptomatic but content		Troubled
		N = 148		N = 57
		20%		8%
Flourishing	M	SD	Vulnerable	M	SD
Family support for learning	3.69	0.42	Family support for learning	3.00	0.80
Peer support for learning	3.41	0.50	Peer support for learning	2.75	0.69
Teacher–student relationships	3.10	0.60	Teacher–student relationships	2.43	0.73
Seeking social support	3.10	0.85	Seeking social support	2.44	0.79
Symptomatic but content	M	SD	Troubled	M	SD
Family support for learning	3.56	0.51	Family support for learning	3.12	0.63
Peer support for learning	3.30	0.53	Peer support for learning	2.80	0.63
Teacher–student relationships	2.78	0.66	Teacher–student relationships	2.50	0.63
Seeking social support	3.03	0.92	Seeking social support	2.64	0.90

Table 3.

Mental Health Status Groups Yielded From a Dual-Factor Model With Predictor Descriptive Statistics at Time 2

		Subjective well-being
Psychopathology		Average to high		Low
Low		Flourishing		Vulnerable
		N = 495		N = 67
		68%		9%
High		Symptomatic but content		Troubled
		N = 116		N = 52
		16%		7%
Flourishing	M	SD	Vulnerable	M	SD
Family support for learning	3.67	0.44	Family support for learning	2.90	0.69
Peer support for learning	3.40	0.53	Peer support for learning	2.76	0.69
Teacher–student relationships	3.10	0.64	Teacher–student relationships	2.34	0.64
Seeking social support	3.12	0.88	Seeking social support	2.51	0.76
Symptomatic but content	M	SD	Troubled	M	SD
Family support for learning	3.53	0.57	Family support for learning	2.87	0.93
Peer support for learning	3.50	0.49	Peer support for learning	2.80	0.83
Teacher–student relationships	3.02	0.66	Teacher–student relationships	2.30	0.67
Seeking social support	3.43	0.98	Seeking social support	2.61	0.85

Demographic characteristics of the groups and total sample were analyzed. Chi-square tests were conducted to determine if demographic variables were differentially represented in the four groups. Results indicated nonsignificant differences for gender, ethnicity, SES, grade, and age.

Stability and Dynamics of DFM Groups

To assess the stability of group membership, participants were reclassified according to DFM status 5 months later. To assess the dynamics of group membership, participants were further categorized into 12 additional subgroups of group change (Table 4). A total of 85% of the flourishing group continued their membership at Time 2, demonstrating the highest stability of the DFM groups. The vulnerable group demonstrated the lowest stability, with 29% remaining at Time 2. Most of these students moved to the flourishing group (46%). The symptomatic but content and troubled students demonstrated intermediate levels of stability, 42% and 47%, respectively.

Table 4.

Stability and Dynamics of DFM Group Affiliation

	Subjective well-being
Psychopathology	Average to high	Low
Low	Group 1 (64%, N = 466)	Group 2 (8%, N = 59)
	→ Group 1—85%, n = 394	→ Group 1—46%, n = 27
	→ Group 2—6%, n = 26	→ Group 2—29%, n = 17
	→ Group 3—9%, n = 39	→ Group 3—14%, n = 8
	→ Group 4—1%, n = 7	→ Group 4—12%, n = 7
High	Group 3 (20%, N = 148)	Group 4 (8%, N = 57)
	→ Group 1—43%, n = 64	→ Group 1—18%, n = 10
	→ Group 2—7%, n = 11	→ Group 2—23%, n = 13
	→ Group 3—42%, n = 62	→ Group 3—12%, n = 7
	→ Group 4—7%, n = 11	→ Group 4—47%, n = 27

Note: Bold = group stability across Time 1 and Time 2.

Influence of Support Variables on Group Change

To assess the predictive qualities of support variables on group change, multiple logistic regression models were used for each DFM group in which dichotomous outcome variables (i.e., no change; change from one category to another category/categories) were predicted by four continuous support variables (i.e., family support for learning; teacher–student relationships; peer support for learning; seeking social support). In addition to these models, multiple logistic bivariate models were used to accommodate for any loss of power due to moderate multicollinearity and limited sample sizes; a remedial approach suggested by Cohen, Cohen, West, and Aiken (2003).

The first multiple logistic regression analysis determined which support variables predicted change in membership from flourishing at Time 1 to any of the other three groups at Time 2 (Table 5). One predictor (i.e., family support for learning) was significant (β = –.68, p < .05), and suggested that for every unit increase in family support for learning, the log odds of moving from the flourishing group decreased by .68. When bivariate regression models were used for individual predictors, teacher–student relationships, as well as family support for learning, were significant predictors of negative group change from the flourishing group (β = –.47, p < .05). The estimates can be converted to odds ratios by taking the natural log of the estimate. The odds ratios are interpreted as the corresponding odds of moving from the flourishing group to another group associated with a unit increase in the continuous predictor variable.

Table 5.

Predictors of Negative Change From the Flourishing Group to the Vulnerable, Symptomatic but Content or Troubled Group

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	1.89	1.12	1.69	0.09	6.60
Teacher–student relationships	−0.21	0.25	−0.83	0.41	0.81
Peer support for learning	−0.17	0.29	−0.56	0.55	0.84
Family support for learning	−0.68	0.31	−2.22	0.03*	0.51
Seeking social support (coping)	0.02	0.17	0.10	0.92	1.02
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	−0.47	0.21	−2.23	0.02*	0.63
Peer support for learning	−0.47	0.25	−1.93	0.05	0.62
Family support for learning	−0.84	0.26	−3.21	0.001***	0.43
Seeking social support (coping)	−0.19	0.15	−1.22	0.22	0.83

Note: N = 70.

p < .05. ***p < .005.

The second multiple logistic regression analysis determined which support variables predicted change in membership from vulnerable at Time 1 to flourishing at Time 2 (Table 6). In this model, none of the values was significant. When bivariate regression models were used for individual predictors, the variables of teacher–student relationships (β = −1.53, p < .05), peer support for learning (β = 1.07, p < .05), family support for learning (β = 1.31, p < .05), and seeking social support (β = 1.18, p < .05) were all significant predictors of group change.

Table 6.

Predictors of Positive Change From the Vulnerable Group to the Flourishing Group

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	−5.81	2.20	−2.65	0.01*	0.003
Teacher–student relationships	0.69	0.69	1.01	0.31	2.00
Peer support for learning	0.61	0.59	1.04	0.30	1.85
Family support for learning	0.69	0.65	1.068	0.29	1.99
Seeking social support (coping)	0.42	0.60	0.69	0.50	1.52
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	1.53	0.56	2.75	0.01*	4.60
Peer support for learning	1.07	0.48	2.21	0.03*	2.92
Family support for learning	1.31	0.53	2.48	0.01*	3.69
Seeking social support (coping)	1.18	0.49	2.42	0.02*	3.24

Note: N = 27.

p < .05. ***p < .005.

The third multiple logistic regression determined which support variables predicted a change in membership from vulnerable at Time 1 to symptomatic but content or troubled at Time 2 (Table 7). One predictor (i.e., teacher–student relationships) was significant (β = 2.34, p < .05), and suggested that for every unit increase in the teacher–student relationship variable at Time 1, the log odds of moving from the vulnerable group to the symptomatic but content group or the troubled group at Time 2 increased by 2.34. When bivariate regression models were used for individual predictors, teacher–student relationships (β = 2.08, p < .05) again was the only significant predictor.

Table 7.

Predictors of Negative Change From the Vulnerable Group to the Symptomatic but Content or Troubled Groups

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	−4.67	2.37	−1.97	0.05*	0.01
Teacher–student relationships	2.34	1.10	2.13	0.03*	10.37
Peer support for learning	0.28	0.70	0.40	0.70	1.32
Family support for learning	−0.23	0.64	−0.36	0.72	0.79
Seeking social support (coping)	−0.37	0.70	−0.53	0.60	0.70
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	2.08	0.83	2.51	0.01*	8.02
Peer support for learning	1.01	0.61	1.67	0.10	2.75
Family support for learning	0.31	0.45	0.69	0.49	1.36
Seeking social support (coping)	0.40	0.50	0.80	0.42	1.49

Note: N = 15.

p < .05. ***p < .005.

The fourth multiple logistic regression determined which support variables predicted a change in membership from symptomatic but content at Time 1 to vulnerable or flourishing at Time 2 (Table 8). In this model, teacher–student relationships (β = .82, p < .05) and seeking social support (β = –.61, p < .05) were significant predictors, suggesting that (a) for every unit increase in the teacher–student relationships variable assessed at Time 1, the log odds of moving from the symptomatic but content group to one of the two higher functioning groups at Time 2 increased by 0.82, and (b) for every unit increase in the seeking social support variable at Time 1, the log odds of moving from the symptomatic but content group to one of the two higher functioning groups at Time 2 decreased by .61. When bivariate regression models were used, only the variable of teacher–student relationships (β = .53, p < .05) was significant.

Table 8.

Predictors of Positive Change From the Symptomatic but Content Group to the Vulnerable or Flourishing Groups

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	−2.19	1.63	−1.35	0.18	0.11
Teacher–student relationships	0.82	0.33	2.51	0.01*	2.26
Peer support for learning	0.01	0.37	0.03	0.18	1.01
Family support for learning	0.54	0.41	1.33	0.18	1.72
Seeking social support (coping)	−0.61	0.24	−2.51	0.01*	0.55
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	0.53	0.27	1.98	0.05*	1.70
Peer support for learning	−0.04	0.32	−0.13	0.90	0.96
Family support for learning	0.57	0.36	1.60	0.11	1.77
Seeking social support (coping)	−0.26	0.19	−1.38	0.17	0.77

Note: N = 75.

p < .05. ***p < .005.

The fifth multiple logistic regression determined which support variables predicted a change in membership from symptomatic but content at Time 1 to troubled at Time 2 (Table 9). None of the values was significant.

Table 9.

Predictors of Negative Change From the Symptomatic but Content Group to the Troubled Group

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	−1.42	2.88	−0.50	0.62	0.24
Teacher–student relationships	−0.20	0.60	−0.33	0.74	0.82
Peer support for learning	0.83	0.77	1.08	0.28	2.29
Family support for learning	−0.53	0.70	−0.77	0.44	0.59
Seeking social support (coping)	−0.28	0.44	−0.64	0.52	0.75
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	−0.48	0.51	−0.95	0.34	0.62
Peer support for learning	0.43	0.73	0.60	0.55	1.54
Family support for learning	0.60	0.58	−1.04	0.30	0.55
Seeking social support (coping)	−0.35	0.36	−1.00	0.33	0.71

Note: N = 10.

p < .05. ***p < .005.

The sixth multiple logistic regression determined which support variables predicted a change in membership from troubled at Time 1 to any of the other three groups at Time 2 (Table 10). One predictor (i.e., family support for learning) was significant (β = −1.10, p < .05), and suggested that for every unit increase in the family support for learning variable at Time 1, the log odds of moving from the troubled group to another group at Time 2 decreased by −1.10. When bivariate regression models were used, none of the predictors was significant.

Table 10.

Predictors of Positive Changes From Troubled Group to Flourishing, Symptomatic but Content or Vulnerable Groups

Multiple logistic regression model
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
(Intercept)	0.76	1.79	0.43	0.67	2.15
Teacher–student relationships	0.29	0.58	0.50	0.62	1.33
Peer support for learning	0.55	0.51	1.10	0.28	1.74
Family support for learning	−1.10	0.56	−1.96	0.05*	0.33
Seeking social support (coping)	0.19	0.41	0.49	0.62	1.21
Bivariate regression models
Predictor	Estimate	Standard error	Z statistic	p value	Odds ratio
Teacher–student relationships	0.15	0.43	0.36	0.72	1.17
Peer support for learning	0.50	0.44	1.14	0.26	1.65
Family support for learning	−0.72	0.46	−1.60	0.12	0.49
Seeking social support (coping)	0.10	0.30	0.35	0.73	1.11

Note: N = 30.

p < .05. ***p < .005.

Discussion

The first aim of this study was to analyze the longitudinal stability of middle school students’ mental health status using Greenspoon and Saklofske’s (2001) DFM. Students were categorized into four mental health groups based on levels of SWB and psychopathology on two occasions, 5 months apart. The distribution across the four groups in this study was similar to previous research (Suldo & Shaffer, 2008). Findings also revealed that 85% of the students in the flourishing group remained in the flourishing group, 29% of the students in the vulnerable group remained in the vulnerable group, 42% of the students in symptomatic but content group remained in the symptomatic but content group, and 47% of the students in the troubled group remained in the troubled group. The vulnerable group demonstrated the lowest stability (29% remained at Time 2). Among the students with high psychopathology (troubled and symptomatic but content), those with high SWB were more likely to improve compared to those with low SWB, which provides support for SWB as a resiliency factor. This finding is consistent with that of a previous study, which found that SWB in the form of life satisfaction acted as a buffer against psychopathological behaviour (Suldo & Huebner, 2004). These results reveal further distinctions among the mental health groups and suggest that measures of SWB and psychopathology together better differentiate mental health than measures of psychopathology alone.

The second aim was to investigate whether social support variables predicted group changes. Among flourishing students, those students who had good relationships with teachers or received family support for learning were more likely to remain in the flourishing group. This finding is consistent with resiliency literature, which shows that mental health is nurtured by caring relationships with teachers and parents (Benard, 2006).

Among vulnerable students, those receiving greater social support for learning from family or peers and/or better teacher–student relationships at Time 1 were also more likely to move to the flourishing group. This suggests that positive social supports are especially beneficial for improving vulnerable students’ mental health. However, findings also showed that vulnerable students who had good teacher–student relationships were more likely to change to a group of lesser mental health at the 5-month follow up. This may be a spurious finding, or it may be an extraneous variable effect. For instance, past research has found that children who engage in informal help-seeking behaviour and have school-related problems are more likely to recognize a psychological problem and seek and receive professional help (John, Offord, Boyle, & Racine, 1995). Similarly, those vulnerable students who have good relationships with teachers may be more likely to seek help from their teachers when they recognize a problem; however, this is speculative at best. Follow-up assessments would be helpful to clarify the longer term outcomes for these students.

Among symptomatic but content students, those who had more positive relationships with their teachers were more likely to move to a higher functioning group. This finding is consistent with a prior variable-centered longitudinal study, which found that increasing teacher support for middle school students showed corresponding decreases in depressive symptoms and increases in self-esteem (Reddy, Rhodes, & Mulhall, 2003). Contrary to what might be expected, those who tended to seek out social support were less likely to make a positive change. This may be because students identified with a psychological disorder who engage in help-seeking behaviour are more likely to recognize their problems (Zwaanswijk, Verhaak, Bensing, Van der Ende, & Verhulst, 2003).

Among troubled students, those who received higher family support for learning were less likely to improve in mental health. This finding could suggest that these students do not benefit from family social support for learning, that family support for learning is insufficient as a protective factor in isolation, or that a statistical anomaly (e.g., suppression effect) has occurred. Again, future research is needed to clarify this finding.

Research Limitations and Future Directions

This study had several major limitations. First, all data were obtained from students from a single school, limiting generalizability. Second, all data were self-report, increasing method variance. Third, multiple logistic bivariate models were used to accommodate for small sample sizes in some groups. This analysis not only increases the likelihood of Type I error but also helps discern weaker signals for areas of future interest. Because this was an exploratory study, incurring a Type I error was preferred to a Type II error (Cohen et al., 2003). Still, results of the multiple logistic bivariate models should be interpreted cautiously.

This study also reflected strengths. First, this study used a two-wave longitudinal design to better identify predictors of mental health and to provide a novel understanding of the stability and dynamics of group membership. Second, this study identified four social support variables that were predictive of group change, illuminating areas of possible future research and practice applications.

However, future directions for research should be considered. One major recommendation is for longitudinal studies using larger samples across more frequent occasions. Such studies are needed to untangle the shorter term and longer term antecedents and consequences of group membership. A second major recommendation is that future researchers would likely benefit from including a broader array of individual difference and environmental variables to capture more fully the dynamics of changes in group membership for children of varying ages, gender, and backgrounds.

The findings from this study along with previous studies support the usefulness of more comprehensive psychological assessments, which incorporate positive psychology data, such as SWB. Mental health models (e.g., DFM), which conceptualize mental health as more than the absence of psychopathology, provide more nuanced information, leading to a more complete understanding of students’ functioning (e.g., identification of vulnerable vs. flourishing students) and ultimately should lead to more sophisticated efforts to prevent problems and promote optimal mental health.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The research was supported in part by a grant from the South Carolina Association of School Psychologists.

Bios

Ryan M. Kelly is an advanced doctoral student in the School Psychology Program at the University of South Carolina. His research interests include adolescent positive psychology, prevention, and adolescent psychotherapy.

Kimberly J. Hills, PhD, is clinical assistant professor in the Psychology Department at the University of South Carolina. Her major research interests and foci include the biopsychosocial factors associated with well-being in children and adolescents. Her interests also include teacher well-being, assessment of autism spectrum disorders, and school psychology training.

E. Scott Huebner, PhD, is professor and director of the School Psychology Program at the University of South Carolina. His research interests involve the conceptualization, measurement, and applications of positive psychological well-being constructs for school psychology practice. He is a fellow of Division 16 of the APA and the International Society for Quality of Life Studies.

Samuel D. McQuillin, PhD, is an assistant professor in the School Psychology Program at the University of Houston. His research interests include mentoring, positive youth development, and methodological and statistical applications in school psychology. He is a graduate of the University of South Carolina School Psychology Program.

References

Antaramian

S. P.

Huebner

E. S.

Hills

K. J.

Valois

R. F.

(2010). A dual-factor model of mental health: Toward a more comprehensive understanding of youth functioning. American Journal of Orthopsychiatry, 80, 462-472. doi:10.1111/j.1939-0025.2010.01049.x

Appleton

J. J.

Christenson

S. L.

Kim

Reschly

A. L.

(2006). Measuring cognitive and psychological engagement: Validation of the Student Engagement Instrument. Journal of School Psychology, 44, 427-445.

Benard

(2006). Resliency: What we have learned. San Francisco, CA: WestEd.

Causey

D. L.

Dubow

E. F.

(1992). Development of a self-report coping measure for elementary school children. Journal of Clinical Child Psychology, 21, 47-59.

Cohen

West

S. G.

Aiken

L. S.

(2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum.

Diener

, (1984). Subjective well-being. Psychological Bulletin, 95, 42-575.

Greenspoon

P. J.

Saklofske

D. H.

(2001). Toward an integration of subjective well-being and psychopathology. Social Indicators Research, 54, 81-108.

Haranin

Huebner

E. S.

Suldo

S. M.

(2007). Predictive and incremental validity of global and domain-based adolescent life satisfaction reports. Journal of Psychoeducational Assessment, 25, 127-138.

Huebner

E. S.

(1991). Initial development of the Students’ Life Satisfaction Scale. School Psychology International, 12, 231-240.

10.

Huebner

E. S.

Hills

K. J.

(in press). Assessment of children’s subjective well-being. In Saklofske

Schwean

Reynolds

C. R.

(Eds.), Oxford handbook of psychological assessment. Oxford, UK: Oxford University Press.

11.

Huppert

F. A.

(2009). Psychological well-being: Evidence regarding its causes and consequences. Applied Psychology: Health and Well-Being, 1, 137-164.

12.

John

L. H.

Offord

D. R.

Boyle

M. H.

Racine

Y. A.

(1995). Factors predicting use of mental health and social services by children 6-16 years old: Findings from the Ontario Child Health Study. American Journal of Orthopsychiatry, 65, 76-86.

13.

Keyes

C. L. M.

(2005). Mental illness and/or mental health? Investigating axioms of the complete state model of health. Journal of Consulting and Clinical Psychology, 73, 539-548.

14.

Laurent

Catanzaro

S. J.

Joiner

T. E.

Rudolph

K. D.

Potter

K. I.

Lambert

. . . Gathright

(1999). A measure of positive and negative affect for children: Scale development and preliminary validation. Psychological Assessment, 11, 326-338.

15.

Lent

R. W.

(2004). Toward a unifying theoretical and practical perspective on well-being and psychosocial adjustment. Journal of Counseling Psychology, 51, 482-509.

16.

Lyons

M. D.

Huebner

E. S.

Hills

K. J.

Shinkareva

V. S.

(2012). The dual-factor model of mental health: Further study of the determinants of group differences. Canadian Journal of School Psychology, 27, 183-196.

17.

Proctor

Linley

P. P.

Maltby

(2009). Youth life satisfaction: A review of the literature. Journal of Happiness Studies, 10, 583-630.

18.

Reddy

Rhodes

J. E.

Mulhall

(2003). The influence of teacher support on student adjustment in the middle school years: A latent growth curve study. Development and Psychopathology, 15, 333-367.

19.

Reschly

A. L.

Huebner

E. S.

Appleton

J. J.

Antaramian

(2008). Engagement as flourishing: The contribution of positive emotions and coping to adolescents’ engagement at school and with learning. Psychology in the Schools, 45, 419-431.

20.

Reynolds

C. R.

Kamphaus

R. W.

(2004). Behavioral assessment system for children (2nd ed.). Circle Pines, MN: AGS.

21.

Roecker-Phelps

C.E.

(2001). Children’s responses to overt and relational aggression. Journal of Clinical Child Psychology, 30, 240-252.

22.

Seligman

M. E. P.

Csikszentmihalyi

(2000). Positive psychology: An introduction. American Psychologist, 55, 5-14.

23.

Suldo

S. M.

Huebner

E. S.

(2004). Does life satisfaction moderate the effects of stressful life events on psychopathological behavior during adolescence? School Psychology Quarterly, 19, 93-105.

24.

Suldo

S. M.

Shaffer

E. J.

(2008). Looking beyond psychopathology: The dual-factor model of mental health in youth. School Psychology Review, 37, 52-68.

25.

Tabachnick

B. G.

Fidell

L. S.

(2007). Using multivariate statistics (5th ed.). Boston, MA: Allyn and Bacon.

26.

Zwaanswijk

Verhaak

P. F.

Bensing

J. M.

Van der Ende

Verhulst

F. C.

(2003). Help seeking for emotional and behavioural problems in children and adolescents: A review of recent literature. European Child and Adolescent Psychiatry, 12, 153-161.