Predictors of pre-adolescent depression and suicidal ideation

Canadian Journal of Behavioural Science, Apr 1997 by Crocker, Alison D, Hakim-Larson, Julie

Abstract Harter and her associates have developed a model that examines perceived competence and social support as predictors of depression and suicidal ideation in adolescents (Harter, Marold, & Whitesell, 1992). This study applies a subset of concepts from Harter's model to a group of 9 to 12 year old boys and girls and their parents.

Using separate standard multiple regression analyses for child, mother, and father informants, support was found for perceived competence and social support as predictors of depression and suicidal ideation during pre-adolescence. Discussion focuses on how depression is manifest at different developmental levels. Further research using school-based assessment and intervention is needed to address those aspects of depression that are continuous and those that are discontinuous over time. In the last few decades, suicide rates have shown an alarming increase among older children and adolescents (McDowell & Stillion, 1994). Pre-adolescents who are suicidal have been found to be likely to repeat suicidal behaviours, especially if they have also suffered from depression (Pfeffer, 1994). Although completed suicide is less frequent among children than it is among adolescents and adults, suicidal ideation in children is more common than is typically recognized (see Marciano & Kazdin, 1994). These findings suggest that one possible way to prevent adolescent suicide is to be alert to the warning signs in younger pre-adolescent children. Although it is now recognized that all age groups share many of the more well-known symptoms of depression and suicidal ideation, some symptoms may differ across the life span. Suicidal acts tend to be especially impulsive and driven by proximal circumstances in children (McDowell & Stillion, 1994). In contrast to adolescents or adults, children may exhibit an irritable rather than a depressed mood, or they may fail to make expected weight gains rather than demonstrate a loss of appetite (Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV); American Psychiatric Association, 1994). Associated features include low self-worth or self-deprecatory ideation (e.g., Harter & Nowakowski, 1987). Self-worth refers to feelings of self-acceptance, self-liking, and self-respect (Rosenberg, 1986). Judgments about the self's competencies in domains important to the self are believed to contribute to self-worth, as does perceived support from significant others (e.g., Harter, 1986). Generally, children and adolescents who report low self-worth also report feeling sad (e.g., Renouf & Harter, 1990; Rosenberg, 1986). Self-worth, depressed affect, and feelings of hopelessness comprise what Harter, Marold, and Whitesell (1992) have called the "depression composite". The depression composite is viewed as mediating the relation between the child's competence and social supports and his or her suicidal ideation. To assess the constructs involved, Harter developed the Self-Perception Profile for Children (Harter, 1985a), the Social Support Scale for Children and Adolescents (Hatter, 1985b), and the Dimensions of Depression Profile for Children and Adolescents (Harter & Nowakowski, 1987). Using path-analysis with data from an adolescent school-based sample (12 to 15 years of age), Harter et al. (1992) examined several potential models. The model with the best-fit to their data confirmed the mediational role of the depression composite in adolescents (Harter & Marold, 1991; Hatter et al., 1992). The adolescent's perception of competence in self-concept domains that are often important to the developing self formed two clusters: (1) athletics, social acceptance, and physical appearance, and (2) scholastic competence and behavioural conduct. For each adolescent, only those domains judged to be important to the self-concept or important to parents were included in the analyses. The two self-concept cluster s predicted both parent support and peer support. These two sources of support as well as the two self-concept clusters also directly predicted the depression Composite. In addition, parent support, one of the two self-concept clusters (scholastic competence and behavioural conduct), and the depression composite directly predicted suicidal ideation. This model, although created on a data set with adolescents, has implications for pre-adolescent children as it provides a framework for the identification of features that may be associated with their depression and suicidal ideation. While child self-report is necessary to tap the internal perceptions of self-worth and depression, the use of the child's caregivers as informants is also important since many of the relevant features are readily observable by others (Robinson, Gather, & Hilsman, 1995). Thus, in the present study, data were gathered from the child and his or her caregiver(s). In addition to Hatter's measure of depression, converging information was collected from the more widely used Children's Depression Inventory (Kovacs, 1985). Both Harter's measures and the Children's Depression Inventory were minimally modified to enable completion by both the child and his or her caregiver(s). Additional concepts were addressed in Harter's model for adolescents that were not considered appropriate for the present study due to the developmental level of the sample and the complexity of using multiple informants. That is, ratings of hopelessness were omitted from the depression composite because they were considered more appropriate for older children (Harter, 1989). Ratings of the importance of the domains to the child were omitted since disagreement among the multiple informants as to whether or not a domain was or wasn't important to a child would be difficult to resolve in an interpretable manner. Using both parent and child informants, the purpose of the present study was to examine whether perceptions of competence and social support in pre-adolescent children would predict their feelings of depression and suicidal ideation as would be expected on the basis of prior research with adolescents. METHOD Participants One hundred and three children in grades 4, 5, and 6, 96 female caregivers, and 43 male caregivers were recruited from regular education classes in eight Roman Catholic elementary schools in southwestern Ontario (50% of the schools contacted). Recruitment occurred via the principals who were given letters for distribution to students through their teachers. Using this method, 11% of the eligible families participated. While the need for both parent and child participation may have prohibited some from volunteering, another possibility is that depression and suicidal ideation in children is minimized by many adults and thus some are not convinced of the potential benefit of early identification. It may be that only the most concerned parents with time to complete the measures were willing to participate. Four families were dropped because of incomplete data. Seven families had two children who participated but only one child from each was included in the analyses. These exclusions resulted in the following participants: 92 female caregivers, 39 male caregivers, and 92 children (49 female, 43 male) in grades 4 (18 boys, 16 girls), 5 (16 boys; 16 girls), and 6 (9 boys; 17 girls). Ninety-eight percent of the female caregivers and 90% of the male caregivers were biological parents; the remaining caregivers were either adoptive, step- or grand-parents. Most respondents were Caucasian (90% female, 92% male) with Hispanic, African-Canadian, and other ethnicities comprising the remainder. Twenty-six percent of the sample reported professional/business-related occupations, 35% reported skilled/semi-skilled occupations, 4% reported unskilled occupations, and 35% provided no information (Hollingshead, 1975). MEASURES Self-Perception Profile for Children (SPPC). The SPPC (Harter, 1985a) assesses children's perceptions of competence in the domains of Athletic Competence, Social Acceptance, Physical Appearance, Scholastic Competence, and Behavioral Conduct. The subscale Global Self-Worth measures the extent to which children like themselves and feel happy about the way they are leading their life. Each subscale consists of six items; item scores range from one (perceptions of inadequacy) to four (perceptions of competency). Psychometric data for previous samples of school children indicated acceptable internal consistency (range from .73 to .82) and factor analysis revealed a five factor structure corresponding to the five domain-specific subscales of the SPPC (Harter, 1985a). Harter et al. (1992) have also defined two composite scores by summing the scores of Athletic Competence, Social Acceptance, and Physical Appearance (ASP), and the scores of Scholastic Competence and Behavioral Conduct (SBC). In the present study, the internal consistency of the subscales was judged acceptable (range .78 to .89) and the range of mean scores obtained (M = 2.84 to 3.11) was comparable to that of Harter (1985a) (M = 2.82 to 3.04, N = 167). Social Support Scale for Children and Adolescents (SSSC). The SSSC (Hatter, 1985b) assesses children's perceptions of support from significant others and includes: Parent, Classmate, Close Friend, and Teacher Support. Scale composition and scoring follows the format for the SPPC. For grades 3-6, acceptable internal consistency estimates were reported for the SSSC (range .72 to .82) and factor analysis revealed a three factor solution: Parent, Teacher, and Classmate/Close Friend (Harter, 1985b). Thus, the values for Classmate and Close Friend Support (CS) were summed. In the present study, internal consistency was judged to be acceptable (range .79 to .87) and the range of mean scores obtained (M = 3.07 to 3.51) was comparable to that reported by Harter (1985b) (M = 2.94 to 3.42, N = 330). Only CS and Parent Support (PS) were used in the present study. Dimensions of Depression Profile for Children and Adolescents (DDPC). The DDPC (Harter & Nowakowski, 1987) assesses the following domains: Mood/Affect, Self-Blame, Self-Worth, Energy/Interest, and Suicidal Ideation. Scale composition and scoring follows the format for the SPPC and SSSC. Acceptable internal consistency has been reported for the DDPC scales for grades 3-6 (range .72 to .86); factor analysis revealed a four factor solution, with Mood/Affect and Self-Worth defining a single factor (Harter & Nowakowski, 1987). Thus, the Depression Composite (DC) can be computed by summing the values for Self-Worth and Mood/Affect. The other three subscales represent separate factors. Validity indices indicate adequate convergent, construct, and discriminant validity (Harter & Nowakowski, 1987). In the present study, internal reliability coefficients were comparable to those reported by the test author and were judged to be acceptable (range .75 to .90). Excluding Suicidal Ideation which was not reported by the test authors for elementary school children, the range of mean scores obtained (M = 2.70 to 3.18) was comparable to that obtained by Harter and Nowakowski (1987) (M = 2.59 to 3.16, N = 176). In the present study, the mean was 3.41 for Suicidal Ideation. Children's Depression Inventory (CDI). The CDI (Kovacs, 1992), a derivative of the Beck Depression Inventory (Beck, 1967), is a 27 item self-report symptom-oriented scale designed for children and adolescents. Items sample behaviours associated with childhood depression (e.g., disturbed mood, hedonic capacity, self-evaluations, and vegetative functions). Total scores range from 0 to 54 in the direction of increasing severity (M = 10.5, SD = 7.3, N = 1266; Kovacs, 1992). Adequate internal consistency estimates (range .71 to .89) and test-retest reliability have been reported for psychiatric, pediatric medical, and Canadian school populations (Kovacs, 1985; 1992). Factor analysis revealed a single factor for the Canadian school group (Kovacs, 1985). The range of values reported by the children in the present study was from 0 to 51 (M = 9.72, SD = 10.5, N = 92). In the present study, the internal reliability coefficient obtained for the child informants was .94. Parent report measures. Parents completed a companion measure to the SPPC as developed by Harter (1985a). With permission from the test author, modifications were also made to SSSC and DDPC to create parent versions. For example, the measures were minimally altered by changing the item descriptors from "really true for me" in the child version to "really true for my child" in the parent version. For the present study, acceptable internal consistency was obtained on the parent measures. On the SPPC, alpha coefficients for mothers ranged from .64 to .81, and for fathers ranged from .67 to .87. For the SSSC, the range was from .71 to .91 for mothers, and from .70 to .80 for fathers. For the DDPC, the range was from .80 to .89 for mothers, and from .79 to .89 for fathers. For the CDI, permission was obtained from the test publisher to modify the wording of items. The phrase "My child" either replaced the pronoun "I" or began each statement on the CDI. For the present study, the internal consistency of the CDI for mothers was .86 and for fathers was .82. Procedure Permission was obtained from the school superintendent and each principal before recruitment letters were given to the principals for distribution through the teachers. Following parental consent, questionnaires were then mailed to participating parent's homes. Children did not participate in the study at school until parents had returned their completed questionnaires. After giving assent, children completed the questionnaires in small groups at school; instructions and items were read aloud and children were encouraged to read along silently at their desks. Standardized administration procedures as outlined in each test manual were followed. Parent and child measures were completed in a fixed order, with the exception of the two measures of childhood depression which were counterbalanced. Identification of Children at Potential Risk Although only screening measures were used, it was considered ethically appropriate to inform parents of children who might benefit from a complete mental health assessment (with child assent given apriori). While a cutoff score of 19 or greater has been used to identify potentially depressed children on the CDI (Stark, 1990), the liberal cut-off score of 15 was used because data collection for research purposes did not permit clarification of individual children's responses. Thirteen girls in grades 4, 5, and 6, and 12 boys in grades 4 and 5 were identified by self-report to be at possible risk using the cut-off score of 15. Thirteen children (14%) reported a score of 19 or higher. Using the cut-off score of 15, only four mothers and one father met the criterion. Comparison of child and parent reports revealed overlap between only one mother and her child, and one father and his child. During phone contacts with parents to inform them of the possibility of their children benefitting from mental health follow-up, 17 parents indicated that their child was already receiving some assistance or that they were aware that their child was experiencing difficulties and mentioned the antecedent (e.g., divorce, death of a family member, sibling conflict). These responses lend credence to the children's self-reports, but raise the question of why parents did not endorse more symptoms if they were aware of difficulties. RESULTS Pearson product moment correlations were computed between the CDI and Harter's measures for each informant. In support of convergent validity, the CDI was significantly correlated with all of the relevant composite measures (ASP, SBC, PS, CS, DC) and with Suicidal Ideation (SI) for both child and mother informants (range: r = -.37 to -.83, all p's .05). Intercorrelations of the composite variables derived from Harter's measures ranged from r = .39 to .77 for child informants (all p's .05), and ASP with SBC for father informants (r = .13, p > .05), all other intercorrelations between the composite variables were significant for mothers (range: r = .22 to .53, all p's .05), and DC (r = .11, p > .05) and PS (r = -.01, p > .05) for father-child, all other parent-child ratings were significantly related (range: r = .21 to .63, all p's .05). Preliminary Analyses Multivariate analyses were conducted to examine the effects of child grade, child gender, order of the two depression measures, and parent socio-economic status on the subscale scores by informant. Using Pillais' criterion, the MANOVAS revealed few differences. Significant multivariate effects for child gender were obtained for child (F(6,85) = 3.32, p