Obesity, Shame, and Depression in School-Aged Children: A Population-Based Study

Study Population

The results are based on records from the SALVe-2004, which is a survey that is regularly distributed by the County Council of Vestmanland to monitor the psychosocial health of the adolescent population of the county. During the spring of 2004, 3505 individuals in Vestmanland attended the mandatory ninth grade (15-year-olds) of the primary school (grundskolan), and 3073 individuals attended the second grade (17-year-olds) of college (gymnasium). These 6578 individuals comprised the target population. To be included in the study, individuals had to answer the questionnaire and give a reasonable estimation of their length and weight. (Individuals reporting measures of height and weight yielding a BMI outside the interval of 12–42.9 were not included in the study.) In the group of 15-year-olds, 2679 individuals (1357 boys and 1322 girls) fulfilled these criteria. In the group of 17-year-olds, 2024 individuals (1033 boys and 991 girls) fulfilled these criteria. These 4703 individuals comprised 71% of the target population.

For the SALVe-2004, no analysis of missing cases was performed. However, such an analysis is available for a similar survey performed in 2001 (SALVe-2001). This analysis followed up a subsample of 365 nonresponders. The results showed that nonresponders had significantly (P = .002) higher BMIs (22.1 kg/m²) than those responding to the survey (mean: 21.5 kg/m²). (Depression scores and questions regarding shame were not asked in the SALVe 2001.)

Estimations of BMI and BMI Groupings

Height and body mass were based on self-report. The BMI was calculated as body mass divided by height squared (kg/m²). Results from a previous, similar study from the same county suggests that self-reported BMI is relatively reliable (correlation of 0.88–0.98 with objective measures) among similar age groups.¹³

The international age- and gender-specific BMI cutoff points for children developed by the Childhood Obesity Working Group of the International Obesity Task Force¹⁴ were used to define subjects as normal weight, overweight (preobese), or obese. These cutoff points were derived from the data of a number of large international cross-sectional growth-study surveys. Cutoff points were determined through centile curves that at 18 years were drawn through the widely accepted cutoff points, 25 and 30 kg/m², for adult overweight and obesity, respectively.

Depressive Symptoms and Depression

We used a self-rating scale (Depression Self-Rating Scales [DSRS]¹⁵) of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (A-criterion) for major depression, with a reported sensitivity of 96.1% and specificity of 59.4% for major depression. The depressive-symptom index (N = 5066; mean: 2.91; SD: 2.50) was calculated as a summation of symptoms reported on the scale. Each symptom was counted only once. (Thus, for instance, on the third criterion of significant weight reduction or addition, the participants could only score once even if they have had experienced both weight loss and weight gain in the previous 2 weeks.)

If ≥5 symptoms were present and at least 1 of the 2 general criteria was fulfilled, the individual was classified as “depressed.”

Shame and Psychosocial and Economic Status

In this study the shame index was built through a summation index of the following questions: Have you during the latest period of 3 months experienced that someone

1. treated you in a degrading manner?

2. made fun of you in front of others?

3. questioned your sense of honor?

4. talked about you in a degrading manner?

5. ignored you or behaved as if you did not exist?

Answers were given on a 5-point scale: 1, no never; 2, rather seldom; 3, sometimes; 4, rather often; and 5, almost always. These 5 items had a Chronbach's α value of .845. In a factor analysis, principal component analysis, method varimax, the 5 items had an eigenvalue of 3.192, explaining 63.85% of the variance in 1 component with factor scores from .704 to .830. The principal component analysis index and the summation index were highly correlated (r = 0.999; P < .00001). Because of this, only the summation index was used in additional analyses.

Those 25% who reported the lowest number of shame experiences were categorized as the low-shame group. Those who reported such experiences between 26% and 75% of the time formed an intermediate group, and the remaining individuals were placed in the high-shame group.

A family-economy index (N = 5066; mean: 3.47; SD: 1.27) was measured through questions of whether the family owned a computer; had Internet access; owned a house in the country for leisure activities; owned a boat big enough to sleep in; owned a caravan; or used to go for vacations in foreign countries and go alpine skiing on vacations. These questions could be answered with “yes” or “no.” Every “yes” answer was counted as 1 point in the socioeconomic summation index (range: 0–7). The survey also included questions regarding parental separation (whether the parents of the subjects were separated or living together) and parental employment (whether their parents worked or were unemployed). Adolescents who had at least 1 unemployed parent was coded as “unemployed parent.”

Statistical Analysis

We used χ² tests for association between BMI grouping and depression according to DSM-IV and analysis of variance and post hoc Scheffe's test for BMI groupings and relations to the depression index (DSRS) and the shame index. To measure the relation of employed versus unemployed parents and parental separation in relation to BMI, we used independent sample t test. The relation between BMI, family-economy index, shame index, and depression index was calculated with Pearson's r. Additionally, we used multiple regression to test our model of independent variables and their relation to the variation of the depression index. Finally, we used binary logistic regression to test our model in relation to depression according to DSM-IV.

Multiple-Regression Model of Predictors of Depressive Symptoms

In our multiple-regression model of depression we found that the relation between BMI and depression disappeared when controlling for parental employment, family situation, family economy, and shaming experiences (Table 2). We also found that female gender had a moderate relation to depressive symptoms. The socioeconomic variables of family constellation (separated parents) and parental employment (parental unemployment) significantly predicted major depression. However, the effect sizes were low. Family economy did not significantly contribute to the model. Shaming experiences accounted for 14% of the variation in depressive symptoms. The total model had an adjusted R² of 22.8%.

Binary Logistic Regression of Predictors of Depression

The result for BMI grouping and its association to depression analyzed in a logistic analysis shows that there was a significantly higher risk for obese adolescents to experience depression compared with normal-weight (reference category) and preobese adolescents (Table 3).

Almost the same pattern as in the multiple-regression model was found in the binary logistic model. The relation of BMI grouping to depression according to DSM-IV was nonsignificant when adjusted for socioeconomic variables and shaming experiences. From Table 4 we can observe that adolescents from separated families or unemployed parents have a higher risk for being depressed, and those adolescents from families with better socioeconomic circumstances have a significantly lower risk. However, those adolescents belonged to the high-shame group had an increased risk of >11 times (odds ratio) for being depressed, compared with those from the low-shame group, when all independent variables in the model were adjusted for. The model explained 21% of the self-reported major depressions (Nagelkerke R²) in the study population.