Introduction
Depression causes a significant burden to individuals and brings high costs for society (Sobocki et al.
2006) and almost 30 % of all people experience a major depressive episode in their lives (Kessler et al.
2012). As symptoms of depression in adolescence strongly predict depressive episodes in adulthood (Pine et al.
1999), depression rates strongly increase in adolescence (Hankin et al.
1998), and 13 % of the 13- to 17-year olds reported to have experienced a depressive episode (Kessler et al.
2012), depression prevention programs have been developed to decrease the incidence of depression in youth. Most of these prevention programs are based on cognitive theories and teach adolescents skills to change a negative cognitive style into a more realistic, helpful cognitive style (Gillham et al.
2007; Kindt et al.
2014; Pössel et al.
2004). Modification of a maladaptive cognitive style should thus prevent or restrain the increase of depressive symptoms.
Yet, even though research on cognitive theory-based depression prevention programs for adolescents from community samples shows encouraging results, the effects are small and inconsistent across studies (Merry et al.
2012; Stice et al.
2009). This could imply that these prevention programs do not sufficiently impact or change a negative cognitive style in young adolescents. However, another reason for why cognitive theory-based prevention programs might not show the desired effects could be that a negative cognitive style does not have a major impact on depressive symptoms, or that depressive symptoms drive negative cognitive styles and not vice versa. Increased knowledge about the prospective links between a negative cognitive style and depressive symptoms in early adolescence is essential to understand why cognitive theory-based depression prevention programs only show small effects. Therefore, the aim of this study was to examine the longitudinal, bidirectional relationships between a negative cognitive style and depressive symptoms.
Meta-analytic reviews clearly showed that a negative cognitive style and depressive symptoms are cross-sectionally associated in children and adolescents [effect size = .41 in Gladstone and Kaslow (
1995); average r = .35, average Z = 4.29,
ps < .0001 in Joiner and Wagner (
1995)]. After childhood, these associations become stronger when youth develop a more stable cognitive style (Cole et al.
2008) and show strong improvements in reasoning and abstract thinking (Steinberg
2005). Previous longitudinal studies have tried to reveal the temporal order of the bidirectional associations between a negative cognitive style and depressive symptoms in adolescents, but so far, findings are mixed. While a previous study showed a negative cognitive style to precede the increase of depressive symptoms, and not vice versa (Kindt et al.
2015), other research showed the opposite effect, namely, that depressive symptoms predicted a negative cognitive style (Hankin et al.
2001; Johnson and Miller
1990; LaGrange et al.
2011; McCarty et al.
2007; Timbremont and Braet
2006). Yet, again other research showed bidirectional effects, with depressive symptoms and a negative cognitive style mutually influencing each other over time (Calvete
2011; Calvete et al.
2013; Stewart et al.
2004).
There may be two explanations for these contradictory findings—other than differences in design, samples and measurements. The relationship between a negative cognitive style and depressive symptoms could be explained by another variable, or the relationship is spurious due to a confounding variable and disappears after controlling for the confounding variable. Previous longitudinal research has revealed that time-varying factors partly account for the variance of both depressive symptoms and a negative cognitive style (LaGrange et al.
2011).
Negative life events could be such a time-varying variable and might act as a confounding variable and thus alter the link between depressive symptoms and a negative cognitive style. Negative life events are found to be associated with higher levels of depressive symptoms (Hammen
2005) and a more negative cognitive style (Calvete
2011; Hankin
2008; Hankin et al.
2001). Although cognitive theories hypothesize that a negative cognitive style is a cause for depressive symptoms, with negative life events acting as a moderator (Abramson et al.
1989; Beck et al.
1979), a meta-analytic review on this topic, that included studies with two assessment waves (e.g., Abela and Payne
2003; Abela and Seligman
2000; Hankin et al.
2001; Nolen-Hoeksema et al.
1992; Southall and Roberts
2002), showed that the overall interaction effect of a negative cognitive style and negative life events on depressive symptoms was small (partial correlation of 0.22; Lakdawalla et al.
2007). In addition, recent empirical research showed that the interaction between a negative cognitive style and negative life events did not predict depression until late adolescence (Braet et al.
2013). Moreover, these cognitive theories do not explain bidirectional relationships that have been found between negative life events, negative cognitive style and depressive symptoms. A recent study revealed that each variable predicted the others: depressive symptoms predicted a negative cognitive style and negative life events, a negative cognitive style predicted depressive symptoms and negative life events, and negative life events predicted depressive symptoms and a negative cognitive style (Calvete et al.
2013).
As negative life events have thus been found to predict both a negative cognitive style and depressive symptoms (Barrocas and Hankin
2011), the prospective relationships between depressive symptoms and a negative cognitive style may be altered by the experienced negative life events. As we were specifically interested in the prospective relationship between a negative cognitive style and depressive symptoms, we tested whether this relationship remained meaningful when we included or excluded the role of negative life events in the models.
Life events are divided into independent and dependent life events. Independent life events occur without the influence of the person him- or herself (e.g., divorce of parents), while dependent life events refer to events to which the person has contributed, and include most interpersonal events (e.g., conflicts; Hammen
2005). Depressed individuals experience more dependent negative events than independent negative events, and dependent events are highly predictive of depressive symptoms and episodes in adults (Kendler et al.
1999) and adolescents (Auerbach et al.
2011; Shih et al.
2006). Moreover, dependent life events have also been found to be the result of a depressive episode both in a community sample (Patton et al.
2003) and clinically depressed children and adolescents (Rudolph and Hammen
1999; Rudolph et al.
2000). Because dependent life events are most strongly related to depression, we included these in our study.
The Current Study
As cognitive-behavioral depression prevention programs in community samples do not show the effects as hoped, teaching adolescents skills to change a negative cognitive style to prevent them from developing depressive symptoms may not be an effective approach. One explanation could be that a negative cognitive style does not precede the depressive symptoms in early adolescents. Although cross-sectional studies established that a negative cognitive style and depressive symptoms are related, few empirical studies used a longitudinal design to reveal the temporal ordering of those associations in adolescence. Empirical studies on the bidirectional relationship between depressive symptoms and a negative cognitive style have shown mixed results. Moreover, these studies have not always included dependent negative life events, while these events could alter or confound the relationship between a negative cognitive style and depressive symptoms. The present study contributes to the field by examining bidirectional relationships between a negative cognitive style and depressive symptoms and the role of dependent negative life events regarding these relationships. We used data that covered four assessments over an 18-month period and that were collected as part of an effectiveness study of a depression prevention program (Kindt et al.
2014).
The associations between a negative cognitive style and depressive symptoms were first analyzed without taking into account the potential role of dependent negative life events, and subsequently we considered dependent negative life events in the analytic models by using two different strategies. First, dependent negative life events were added as a variable to the basic model next to depressive symptoms and a negative cognitive style to analyze temporal relationships between those three variables. This way, we also examined the robustness of the relationship between a negative cognitive style and depressive symptoms when including the temporal influence of dependent negative life events. Second, we added dependent negative life events as a time-varying covariate to the basic model, to control for their confounding impact on the relationship between a negative cognitive style and depressive symptoms. We hypothesized that depressive symptoms are both preceding and following a negative cognitive style when dependent negative life events were not included. Further, with respect to the two additional analytic models, we hypothesized that the bidirectional associations between depressive symptoms and a negative cognitive style are less prominent when considering the role of dependent negative life events.
Discussion
Because high levels of depressive symptoms proceed the onset of a depressive disorder, which is the most serious mental health problem in adolescence (Birmaher et al.
2002), and depressive symptoms in adolescence are known to be predictive for major depressive episodes in adulthood (Pine et al.
1999), prevention of depressive symptoms in adolescence is of importance to improve mental health care. So far, the efforts have mainly been focused on cognitive-therapy based depression prevention in early adolescence, as from late childhood into middle adolescence, individuals develop a more stable cognitive style and are capable of abstract thinking (Cole et al.
2008; Steinberg
2005). Yet, cognitive-therapy based depression prevention programs in adolescents in a community setting have only showed small effects (Merry et al.
2012). Because research showed mixed findings on the relationships between depressive symptoms and a negative cognitive style in adolescents (Calvete
2011; Calvete et al.
2013; Hankin et al.
2001; Johnson and Miller
1990; Kindt et al.
2015; LaGrange et al.
2011; McCarty et al.
2007; Stewart et al.
2004; Timbremont and Braet
2006), we assumed that these programs may not have shown larger effects because the targeted negative cognitive style does not precede depressive symptoms in early adolescence. Hence, the first aim of the study was to test the bidirectional relationships between depressive symptoms and a negative cognitive style. Moreover, because of the strong links of both depressive symptoms and a negative cognitive style with dependent negative life events (Abramson et al.
1989; Auerbach et al.
2011; Beck et al.
1979), we also aimed to test whether the relationship between a negative cognitive style and depressive symptoms remained meaningful when considering dependent negative life events.
Our findings consistently demonstrated that, in a sample of young adolescents with depressive symptoms in the normative range, a negative cognitive style did not predict depressive symptoms. In addition, when we did not consider the role of dependent negative life events in the analyses, the opposite direction was revealed: depressive symptoms predicted a negative cognitive style. Yet, when dependent negative life events were added as a variable next to depressive symptoms and negative cognitive style, the prospective relationship between depressive symptoms and a negative cognitive style did not maintain consistently over time. Furthermore, the prospective association of depressive symptoms on negative cognitive style disappeared when we controlled for the presence of dependent negative life events. These models including dependent negative life events were invariant for gender and intervention condition. Our findings indicate that (a) a negative cognitive style does not seem directly predictive of depressive symptoms, and (b) a longitudinal relationship between depressive symptoms and a negative cognitive style is hard to interpret when dependent negative life events are not taken into account. As a negative cognitive style did not precede and predict depressive symptoms, we may need to revisit the idea of preventing depression in a community sample of early adolescents by changing a negative cognitive style.
Previous longitudinal research among community samples of adolescents that included three or more measurements of depressive symptoms and a negative cognitive style, but did not include dependent negative life events, showed contradicting findings (LaGrange et al.
2011; Kindt et al.
2015). One study found depressive symptoms to be predictive of a negative cognitive style (LaGrange et al.
2011), while another found a negative cognitive style to be predictive of depressive symptoms (Kindt et al.
2015). The variations in findings between these studies and the current study might partly be explained by differences in questionnaires used for assessing depression and a negative cognitive style, or participant characteristics such as the age range. As important, both studies (LaGrange et al.
2011; Kindt et al.
2015) differed from the current study because dependent negative life events were not measured and could thus not be added to the analyses. When dependent negative life events would have been controlled for in those studies, as we did in the current study, the predictive relationships between depressive symptoms and a negative cognitive style may also have disappeared.
Previous studies that are most similar to ours also used cross-lagged panel designs, the same questionnaire to measure negative cognitive style, and a questionnaire for dependent negative life events, however, they did not control for the dependent life events (Calvete
2011; Calvete et al.
2013). With regard to the relationship between depressive symptoms and negative cognitive style, they found reciprocal associations (Calvete et al.
2013) or a more mixed pattern depending on which subscale of the cognitive style questionnaire was analyzed (Calvete
2011). Why different results were found could be due to stronger associations in the current study between dependent negative life events on the one hand, and depressive symptoms and negative cognitive style on the other hand compared to Calvete and colleagues’ studies. Between depressive symptoms and dependent negative life events, the Pearson correlations ranged between .56 and .67 in our study versus between .30 and .40 (Calvete
2011), and between .29 and .36 (Calvete et al.
2013). Between negative cognitive style and dependent negative life events the Pearson correlations ranged between .38 and .48 in the current study, and ranged between .10 and .27 and between .22 and .28 in Calvete’s studies (respectively Calvete
2011; Calvete et al.
2013). Note that differences can only be concluded with precaution, since the Pearson correlations were based on different sample sizes and different questionnaires for depressive symptoms and dependent negative life events.
The different associations may also reflect different sample characteristics such as cultural differences between Spain (Calvete
2011; Calvete et al.
2013) and The Netherlands, or between a mainly native sample (Calvete
2011; Calvete et al.
2013) versus a sample with a high proportion of immigrants, which may be more vulnerable to develop depressive symptoms (Siegel et al.
1998). Although we included adolescents from schools with a high proportion of pupils living in low-income areas who may be at higher risk to develop depressive symptoms during adolescence, we have labeled our sample as a community sample as it shows normative depressive symptoms that are similar to a Dutch universal sample (Tak et al.
2014).
Nevertheless, because of the higher cross-sectional correlations between depressive symptoms, negative cognitive style and dependent life events in the current study, controlling for the dependent negative life events had a higher restrictive role on the association between depressive symptoms and a negative cognitive style than it would have had in the studies of Calvete and colleagues (Calvete
2011; Calvete et al.
2013). As such, the question remains whether controlling for dependent negative life events in those studies would also have led to the same conclusions as presented in the current study, that is, that depressive symptoms and a negative cognitive style are prospectively unrelated when controlled for dependent negative life events.
Implications
If replicated, this study has important implications for the perspective of depression prevention programs in samples of adolescents with normative depressive symptoms. Most depression prevention programs for community samples of young adolescents are based on principles from cognitive behavioral therapy and target a negative cognitive style (e.g., Gillham et al.
2007; Kindt et al.
2014; Pössel et al.
2004; Stice et al.
2009). Even though the impact of those programs is encouraging, it is not considered satisfying (Merry et al.
2012). Based on the present findings, we do not have evidence that a depression prevention program will be able to prevent depressive symptoms by changing a negative cognitive style, because negative cognitive style did not predict depressive symptoms over time. Yet, the merit of the impact of dependent negative life events on the development of both a negative cognitive style and depressive symptoms should be acknowledged. The occurrence of dependent negative life events was associated with depressive symptoms over time, hence, depression prevention programs aiming a reduction of dependent negative life events might be more effective in reducing depression levels than programs aiming a change in a negative cognitive style. As dependent negative life events cover mostly interpersonal events to which the person has contributed (Hammen
2005), prevention programs could focus on reducing dependent negative life events by social skills training to reduce conflicts with peers or parents. Also, anti-bullying projects, for a better school atmosphere, or programs teaching parents skills how to positively interact with their adolescent might reduce dependent negative life events in youth. Yet, many existing depression prevention programs already involve social problem solving skills (e.g., Gillham et al.
2007; Pössel et al.
2004), and a review showed that the specific content of depression prevention programs (e.g., reducing negative cognitions or focussing on problem-solving) was unrelated to the effect-sizes (Stice et al.
2009). Additional research on the depression prevention programs that reduce dependent negative life events could shed light on this field. As the design of the current study does not allow to draw conclusions on the causal interpretation of the relationship between dependent negative life events and depressive symptoms, additional research is first needed to reveal whether dependent negative life events cause an increase of depressive symptoms in young adolescents from a community sample.
Limitations
Several limitations provide opportunities for further research. First, we only used self-report questionnaires. Although self-reports receive criticism because of potential biases, it is a widely used strategy that is well suited for assessing human cognitive and emotional states when it fits in the theoretical context in which it is used (Haeffel and Howard
2010). Self-reports are a more accurate approach for assessing emotional states and cognitions compared to the use of parents or teachers as source for information (DiBartolo et al.
1998) and specifically for life events, findings suggest that self-reports and interviews may be equally viable methods (Wagner et al.
2006). Moreover, the use of self-reports enhanced comparison with previous research on depressive symptoms and a negative cognitive style that is also based on self-report measurements. Specifically for measuring parental psychopathology, a multi-informant strategy would have been more reliable and valid than questioning the adolescent whether their parents were ever treated by a psychiatrist. Assessing parental psychopathology the way we did probably excluded parents who had psychopathology but did not receive treatment, and also treated parents whose children did not know they received treatment.
Concerning the dependent negative life events measurements, we formulated three limitations. First, we only measured the frequency of dependent negative life events and did not include the perceived severity or impact of the events, which may have deflated the relationship between dependent negative life events and depressive symptoms (Cohen et al.
1983). Second, we found strong cross-sectional correlations between the variables dependent negative life events, depressive symptoms and a negative cognitive style. Adolescents may have reported dependent negative life events in line with their negative cognitive style and depressive symptoms, resulting in an inflated internal consistency of the dependent life events measure and an over-report of these dependent negative life events. More precisely, the dependent negative life events may be over-reported by those suffering from a negative cognitive style, as those individuals may excessively pay attention to negative aspects of a situation and thereby casting the whole situation in a negative context (Beck et al.
1979). In accordance with that, youth with clinical depressive symptoms are found to have an attentional bias toward negative interpersonal stimuli (Gotlib et al.
2004). Also, previous findings in adult samples suggested that a depressed mood (Shrout et al.
1989) and a negative cognitive style (Simons et al.
1993) influence over-reporting of negative life events on self-report scales. However, depressive symptoms and a negative cognitive style were not found to be associated with over-reporting of events in community children and adolescents compared to interviews (Wagner et al.
2006). Whether over-reporting has happened could be disentangled by the use of multi-informant strategies in future research, such as questionnaires or interviews with parents, peers or teachers. A third limitation is that we did not assess independent life events, which include major life events such as a divorce by parents, illness or death of family members. Although dependent life events are most important for depression research (Auerbach et al.
2011; Shih et al.
2006), additionally testing the effect of independent life events would contribute to the knowledge about the impact of negative life events on the development of depressive symptoms and a negative cognitive style.
Furthermore, although a large proportion of the immigrant adolescents had an ethnical background that is a minority group in the Netherlands (e.g., Moroccan, Turkish), this was not the case for a substantial proportion of the sample. As such, conclusions drawn about the immigrant adolescents cannot be generalized to minority groups in general. At last, the results of this study cannot be generalized to adolescents with clinical levels of depression. As meta-analyses have shown that indicated prevention efforts are more effective than universal prevention programs, even when the same program is used for a different target group (Merry et al.
2012; Stice et al.
2009), it may be that prospective relationships between a negative cognitive style and depressive symptoms are different among adolescents with elevated depressive symptoms.
Future research is encouraged to replicate the current study by following samples of youth over an extended period of time. As it was beyond the scope of the current study, further research could test existing theories such as the stress generation theory (Hammen
1991; Daley et al.
1997), postulating that negative life events are increased by depression, or scar theory (Lewinsohn et al.
1981), postulating that a negative cognitive style is left after a depressive episode.