Trauma and severe stressors are highly prevalent among childrenReference McLaughlin, Koenen, Hill, Petukhova, Sampson and Zaslavsky1,Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 and are frequently associated with multiple psychiatric conditions.Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 However, the limited epidemiological research on incident psychiatric comorbidity in children exposed to trauma or severe stressors is currently subject to three limitations. First, research examining psychiatric morbidity following the experience of a potentially traumatic event has been dominated by a focus on post-traumatic stress disorder (PTSD).Reference McLaughlin, Koenen, Hill, Petukhova, Sampson and Zaslavsky1,Reference Gradus3 A substantial proportion of individuals who experience potentially traumatic events do not develop PTSD but may develop other stress-related disorders.Reference Gradus3 Within ICD-10, five stress-related diagnoses, including PTSD, are used for identifying post-trauma or severe stressor sequelae, depending on the type and duration of symptoms. Studies with adults indicate that individuals who develop PTSD following trauma have increased risk for other mental disorders compared with those exposed to trauma who do not develop PTSD.Reference Gradus, Antonsen, Svensson, Lash, Resick and Hansen4 Less is known about incident mental disorders in individuals who do not develop PTSD but instead develop other stress-related diagnoses. Second, there is no population-based study examining a range of stress disorders and incident psychopathology in children. Limited available evidence from studies of children who have experienced trauma suggests comparable levels of functional impairment across a range of symptom constellations.Reference Carrion, Weems, Ray and Reiss5–Reference Cohen and Scheeringa7 Also, consistent with findings for adult samples, studies with children suggest that stress disorders are often comorbid with other psychiatric disorders, most commonly depression, anxiety and behavioural disorders, and substance use disorders in older adolescents.Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 However, these findings are largely based on clinical treatment-seeking samples. Because psychiatric comorbidity is correlated with seeking treatment, clinical samples are not representative of the relationships between trauma or severe stressors, stress disorders and other psychiatric disorders in the general population.Reference Cohen and Cohen8 Third, available data from population-based studies with children and adolescents are also subject to biases due to their cross-sectional nature (e.g. reverse causation), utilisation of a subset of a population (e.g. selection bias), and limitations in age range (focusing mostly on adolescents) and type of diagnoses examined.Reference McLaughlin, Koenen, Hill, Petukhova, Sampson and Zaslavsky1,Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 Thus, longitudinal population-based data on children experiencing trauma or severe stressors and the course of incident psychiatric disorders can inform future intervention and prevention research. For instance, if stress disorders that require less severe or fewer symptoms (e.g. adjustment disorder) or a shorter duration of symptoms (e.g. acute stress disorder) for diagnosis increase the risk of incident disorders at rates comparable to more severe disorders (e.g. PTSD), then intervention or prevention research aimed at reducing the burden of incident comorbid psychiatric disorders would need to examine outcomes across a broad array of stress disorders. Accordingly, the aims of this study were to examine the association between all ICD-10 severe stress disorders and a range of incident depressive, anxiety and behavioural disorders in a population-based prospective longitudinal sample of children and adolescents. Additionally, given the well-established gender-related differences in stress reactions in adults, particularly following trauma exposure,Reference Gradus3 we examined effect modification of these associations by gender.
Method
Participants
Of the 1 698 654 Danish-born residents of Denmark between 6 and 15 years of age in the period between 1 January 1995 and 31 December 2011, this study included all children diagnosed with a severe stress disorder in either the Danish Psychiatric Central Research Register (DPCRR) or the Danish National Patient Registry (DNPR).Reference Schmidt, Schmidt, Adelborg, Sundbøll, Laugesen and Ehrenstein9 Given the developmentally distinct manifestations of traumatic stress reactions for very young children (0–5 years of age),Reference Cohen and Scheeringa7 we focus on school-aged (6–15 years) children and adolescents. We included adolescents through 15 years of age, as this marks the end of compulsory primary education in Denmark. Findings related to older adolescents and adults have been previously published.Reference Gradus, Antonsen, Svensson, Lash, Resick and Hansen4 Stress cohort members had to receive at least one incident stress disorder diagnosis during the study period. The year 1995 was chosen as the start of the study period because ICD-10 coding was implemented in 1994 and the inclusion of out-patient clinic visits in the psychiatric registry in Denmark was implemented in 1995. We excluded anyone with a diagnosis between 1 January 1994 and 31 December 1994 because persons who were diagnosed in 1994 may have been living with the disorder before receiving the diagnosis that year, making them prevalent cases. Prevalent cases may differ from incident cases with respect to disorder aetiology, clinical course and survival.Reference Rothman, Greenland and Lash10 For instance, prevalent cases are affected by factors associated with disorder maintenance and survival. Prevalent cases were therefore removed from the cohort and the current analyses to ensure that only newly occurring cases of stress disorders were examined. In total, 11 292 children received a first severe stress or adjustment disorder diagnosis between 1995 and 2011. A general-population comparison cohort of children was created of individually matched children who had not received a stress disorder diagnosis at the time that their matched stress cohort member was diagnosed (n = 56 460). Members of the comparison cohort were individually matched to stress cohort members by gender and age at the matched index date at a ratio of 5 to 1.Reference Heide-Jorgensen, Adelborg, Kahlert, Sorensen and Pedersen11,Reference Schmidt, Pedersen and Sorensen12 Of note, this cohort was designed such that if a member of the comparison cohort was diagnosed with a stress disorder after their match index date they were removed from the comparison cohort and designated to the appropriate stress disorder group, with five new non-stress disorder diagnosed comparison group members then selected to match that participant. Diagnostic validation processes have been the subject of prior studies and suggest that diagnoses (e.g. affective disorders, stress disorders) in the registries have high validity compared with computer-generated diagnoses or independent reassessment.Reference Svensson, Lash, Resick, Hansen and Gradus13
Predictor and outcome variables
We obtained data from national Danish medical longitudinal registers. The 10-digit Civil Registration number, a unique identifier assigned to all residents of Denmark, was used to retrieve and merge data for each individual for the creation of the cohort.
Predictors
The predictors in our analyses were five ICD-10 stress diagnoses: (a) acute stress reaction (diagnosed in the immediate aftermath of an event; ICD-10 code F43.0), (b) PTSD (diagnosed following a traumatic event and a specified period of symptom maintenance; ICD-10 code F43.1), (c) adjustment disorder (diagnosed following a stressful event and specified period of non-recovery; ICD-10 code F43.2), and two diagnoses – (d) other reactions to severe stress (ICD-10 code F43.8) and (e) reactions to severe stress, unspecified (ICD-10 code F43.9) – which are used for those who are experiencing symptoms following a traumatic or severe stressor that do not meet the full diagnostic criteria for one of the other three disorders. We obtained diagnostic data from the DPCRR, which records data on patients who were admitted to a psychiatric in-patient hospital, received out-patient clinic psychiatric care, or received treatment at a psychiatric emergency department. Diagnostic data were also obtained from the DNPR, which contains patient data from out-patient clinics, somatic hospitals and emergency rooms. Registries include data since 1995, including treatment dates and up to 20 diagnoses per treatment entry for patients. Our prior validation study of the stress diagnoses in registries showed good validity for acute stress reaction, other reactions to severe stress and unspecified reactions to severe stress and showed high validity for PTSD and adjustment disorder.Reference Svensson, Lash, Resick, Hansen and Gradus13
Outcomes
The primary outcomes in the present study were incident diagnoses of depressive disorders (ICD-10 codes F32, F33 and F34.1, which include major depression and dysthymia), anxiety disorders (ICD-10 codes F40 and F41, which include phobic anxiety and generalised anxiety disorders) and behavioural disorders, including attention-deficit hyperactivity disorder (ICD-10 code F90), conduct disorders (ICD-10 code F91) and disorders of social functioning with onset specific to childhood and adolescence (ICD-10 code F94). Our study outcomes were similar to those examined in the literature and a recent longitudinal studyReference Conway, Raposa, Hammen and Brennan14 examining the impact of early life stressors on later psychopathology. However, substance use disorders were not included because of the low incidence rates in our sample in this developmental period.
Analyses
We conducted descriptive and stratified analyses to examine demographic characteristics and psychiatric disorders across categories of stress diagnoses at baseline (i.e. at the time of the initial stress diagnosis). We used Cox proportional hazards regression models to examine associations of severe stress or adjustment diagnoses with each outcome (i.e. depressive, anxiety and behavioural disorders). For each analysis of incident disorders only children without the outcome diagnoses (e.g. depressive disorders) at baseline were included in the analyses, and the model controlled for other disorders at baseline (i.e. anxiety and behavioural disorder diagnoses). Cumulative incidence curves were plotted to examine the occurrence of new-onset depressive, anxiety and behavioural disorders following the initial stress diagnosis (or the index date for the comparison cohort). Incidence curves were plotted for a 10-year period following an initial stress diagnosis owing to some small cell sizes in the subsequent follow-up period. We also examined the number of incident diagnoses following stress disorder diagnoses, and time to incident psychiatric disorder diagnoses.
This study was approved by the Institutional Review Board of Boston University, the Danish Health and Medicines Authority and the Danish Data Protection Agency (record no. 2012-41-0841).
Results
Table 1 displays descriptive data and rates of psychiatric disorders at the time of the initial diagnosis for the stress cohort and the comparison cohort. Sixty-three per cent of the stress diagnosis cohort members were girls and 73% of the children were between 12 and 15 years of age. The proportions of children with comorbid baseline psychopathology differed across the stress disorder groups. Depressive disorders were most prevalent among children diagnosed with an adjustment disorder (5.2%). Depressive (4.2%) and behavioural disorders (7.4%) were most common among children with PTSD. Additionally, 14% of children diagnosed with PTSD, 13% diagnosed with adjustment disorder and 10% diagnosed with reaction to severe stress unspecified had one or more psychiatric diagnoses at baseline, other than their stress-related diagnoses. Cumulative incidence curves (Fig. 1) show that children in the stress cohort had a higher incidence of depressive, anxiety and behavioural disorders following their severe stress or adjustment diagnoses than did members of the comparison cohort following their matched index date over the course of the study. On average children in the stress cohort were followed for 5.8 years (Table 1).
Fig. 1 Cumulative incidence curves for diagnoses of ICD-10 depressive, anxiety and behavioural disorders after receipt of a stress diagnosis in children 6–15 years of age in Denmark, 1995–2011.
Table 1 Characteristics of the ICD-10 stress disorders and comparison cohorts of children 6–15 years of age in Denmark, 1995–2011
PTSD, post-traumatic stress disorder; ORSS, other reactions to severe stress; RSSU, reactions to severe stress, unspecified.
Results from the Cox proportional hazards regression models (Table 2) indicate that each of the stress disorders is associated with higher rates of depressive, anxiety and behavioural disorders. For example, the rate of depressive disorders among children diagnosed with an adjustment disorder was 6.8 (95% CI 6.0–7.7) relative to the control comparison group. All other stress disorders also had strong associations with each type of incident psychiatric disorder. We examined proportions of incident psychiatric comorbidity (supplementary Table 1 available at https://doi.org/10.1192/bjp.2019.247). Across all stress disorders, a greater proportion of children were diagnosed with incident comorbid psychiatric disorders, with rates of two or more incident diagnoses ranging between 3.0 and 5.3%, with an average of 4.4% for the full stress cohort, compared with 0.5% in the comparison cohort. With respect to time to incident psychiatric comorbidity for each type of disorder (supplementary Table 2), for the full stress cohort an incident depressive disorder was diagnosed 3.9 years after a baseline stress disorder, and incident anxiety and behavioural disorders were diagnosed after 4.5 and 2.5 years respectively. In the comparison cohort incident depressive disorders were diagnosed 5.9 years from baseline, and incident anxiety and behavioural disorders were diagnosed 6.3 years and 3.8 years from baseline respectively.
Table 2 Associations between baseline ICD-10 stress disorders and incident psychiatric comorbidity in children 6–15 years of age in Denmark, 1995–2011
RRadj, adjusted rate ratio; PTSD, post-traumatic stress disorder; n.p., results not presented because fewer than five incident psychiatric disorder cases were identified.
a. Full stress cohort models were adjusted for age, gender and baseline depressive, anxiety and behavioural disorders.
b. Gender-stratified models were adjusted for age and baseline depressive, anxiety and behavioural disorders.
We also examined effect modification by gender in the associations between stress disorders and incident psychiatric disorders (Table 2). Small cell sizes (<5) limited calculation of estimates of associations between stress disorders and incident psychiatric diagnoses in some cases. With respect to all incident diagnoses, the pattern across boys and girls was similar. For instance, for anxiety disorders, relative to comparison controls, boys with a baseline diagnosis of reaction to severe stress unspecified had a hazard ratio of 8.3 (95% CI 4.6–15) and girls had a hazard ratio of 6.0 (95% CI 4.4–8.0). The hazard ratio for the association between adjustment disorder and behavioural disorders for boys was 6.3 (95% CI 5.0–8.0) whereas for girls it was 10 (95% CI 7.8–13). Overall, there was no evidence of possible effect modification by gender on the multiplicative scale.
Discussion
To the best of our knowledge this is the first prospective longitudinal population-based study of school-aged children and adolescents examining rates of incident psychiatric disorders following diagnosis of a range of stress disorders. At baseline, our study sample had a higher proportion of girls diagnosed with a stress disorder than boys, and most children with stress disorder diagnoses were older (12–15 years of age). These findings are consistent with existing data indicating a higher prevalence of PTSD among women relative to men in adult samplesReference Gradus3 and in the limited available findings from another population-based study of US adolescents, which used the National Comorbidity Survey Replication Adolescent Supplement (NCS-A).Reference McLaughlin, Koenen, Hill, Petukhova, Sampson and Zaslavsky1 In the stress cohort a greater proportion of older children had stress disorder diagnoses than younger children. This may represent the developmental course of stress disorders such that symptoms may emerge or worsen in adolescence. This may also be consistent with other population-based data showing an increase in rates of potentially traumatic events with increasing age, although our study could not assess this directly.Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 At baseline, across the full stress cohort, behavioural disorders were most common, followed by depressive disorders.
Regardless of the type of stress disorder diagnosis, the rates for all three types of incident comorbid disorders – anxiety, depressive and behavioural disorders – were elevated compared with the comparison cohort, which includes children who may have experienced trauma or severe stress but had no stress or adjustment disorder diagnoses. Thus, severe stress disorders in childhood may confer a transdiagnostic diathesis for a range of mental disorders rather than disorder-specific risk. These findings are also consistent with those from multiple adult samples, including adults with stress diagnoses in the larger Danish stress cohort.Reference Gradus, Antonsen, Svensson, Lash, Resick and Hansen4,Reference Walsh, McLaughlin, Hamilton and Keyes15 Aetiological research suggests that regulatory mechanisms (e.g. stress- and immune-related mechanisms) are affected by traumatic and stressful events and play a role in disease risk for multiple psychiatric disorders.Reference Danese and Lewis16 This is consistent with the above-mentioned population-based cross-sectional study of US adolescents using NCS-A data examining trauma and other severe stressors, which found an increased risk for a range of psychiatric comorbidities in adolescents,Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 and with our findings that children in the stress cohort were more likely to be diagnosed with a greater number of incident psychiatric diagnoses.
To the extent that stress disorders represent a range in ICD-10 disorder severity, depending on the number and types of symptoms (e.g. children who do not meet criteria for PTSD are diagnosed with another, less severe, stress disorder such as an adjustment disorder), our findings also suggest that being diagnosed with a more or less severe stress disorder is not differentially associated with risk for future diagnoses. For instance, in our sample, children across all stress diagnoses had comparable risk for all incident psychopathology. One possible explanation for this finding is that the structure of PTSD may differ across developmental stages and the criteria may be more representative of adult manifestationsReference Cao, Wang, Cao, Zhang and Elhai17,Reference Ayer, Danielson, Amstadter, Ruggiero, Saunders and Kilpatrick18 and hence, diagnostic specificity across stress diagnoses may not be comparable to that in adult samples. For instance, studies with children have found that those exposed to traumaReference D'Andrea, Ford, Stolbach, Spinazzola and van der Kolk19 or severe stressorsReference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 exhibit a broad range of disorders. However, adult studies have also reported similar findings, where individuals diagnosed with PTSD and those with subthreshold PTSD symptoms (i.e., those who suffer from post-traumatic stress symptoms but do not meet diagnostic criteria for PTSD) present with comparable functional impairment and risk for incident psychiatric comorbidityReference Gradus, Antonsen, Svensson, Lash, Resick and Hansen4,Reference McLaughlin, Koenen, Friedman, Ruscio, Karam and Shahly20,Reference Marshall, Olfson, Hellman, Blanco, Guardino and Struening21 and also other deleterious outcomes, such as risk for re-traumatisation, suicidality and all-cause mortality.Reference Gradus, Antonsen, Svensson, Lash, Resick and Hansen4 An alternative explanation stems from evidence suggesting that symptom severity rather than the total number of symptoms or specific symptom constellation required for an ICD-10 diagnosis is associated with individual functional impairment.Reference Cohen and Scheeringa7 In addition to symptom severity, the impact of trauma or severe stressors on a range of psychological or biological outcomes (for instance the multiple domains of functioning proposed within the National Institute of Mental Health's Research Domain Criteria framework) maybe associated with future psychopathology. Finally, individual variations in the latency of symptom development may also affect the course of psychopathology. In our study, findings related to time to incident psychopathology indicate that psychiatric comorbidity was evident earlier in the stress cohort than in the comparison cohort. Future research needs to examine whether differences in the time to incident psychopathology diagnoses represent a natural disorder course (i.e. manifestation of pathophysiology following trauma or severe stressor exposure), or an artefact of how diagnoses are made (i.e., symptoms of incident disorders may be recognised sooner as part of a post-trauma or severe stress presentation than in the absence of a traumatic or severe stressor-related disorder).
With respect to gender-related findings, our study did not identify a pattern of differences, which is consistent with findings from the NCS-A data, indicating no differences in psychiatric outcomes between adolescent boys and girls who had experienced a range of childhood adversities.Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2 It is possible that gender-related differences noted in adult samples become more clearly established later in the life course.
Strengths and limitations
It is notable that the unpredictable nature of most traumatic events and severe stressors makes data collection for stress studies uniquely challenging, and extensive primary data collection is often prohibitive in terms of the cost and time involved. Thus, a majority of studies tend to be retrospective and to focus on an index trauma or severe stressor and/or PTSD, rather than a range of post-traumatic diagnoses or reactions. Thus, the use of prospectively collected longitudinal data from a large unselected sample that draws from the full country population is a methodological strength of this study.
In addition to this strength, our study has important limitations that should be kept in mind when evaluating our results. First, this sample includes only children who had a stress diagnosis diagnosed by a mental health professional. It is possible that children diagnosed by a general practitioner, who may present with a less severe presentation, were under-sampled in the stress cohort. Second, given the healthcare and social support systems for Danish individuals (i.e. tax-supported universal healthcare that enables all members to receive healthcare at no cost) relative to other countries with available population-based studies of children (e.g. NCS-A in the USA) and the lack of racial/ethnic variability in our sample, our study findings need to be replicated in other samples and countries. Third, despite the large sample size, the frequencies of certain stress disorders (e.g. PTSD) in our study were low compared with other population-based studies.Reference McLaughlin, Koenen, Hill, Petukhova, Sampson and Zaslavsky1 Accordingly, some of the cell sizes for gender-related stratified analyses were too small to reliably examine associations. These associations need to be examined further in future research. Within the current sample, PTSD is rare and thus the strength of bias from exposure misclassification in observed associations will be driven by specificity. A previous validation study of stress diagnoses in the DPCRR found that no one in the comparison cohort had PTSD,Reference Svensson, Lash, Resick, Hansen and Gradus13 indicating that specificity is 100%. Non-differential misclassification of PTSD with perfect specificity is expected to have limited impact on validity. Fourth, this study does not include data on traumatic events or severe stressors preceding, or subsequent to, stress disorder diagnosis. Causal pathways for incident psychiatric disorders may occur through stress disorders or through unmeasured traumatic events, severe stressors or other mechanisms. Future studies accounting for traumatic events or severe stressors and psychiatric diagnoses are necessary.
Implications
Results from this study suggest a need to examine a broad range of post-traumatic or severe stress reactions using a transdiagnostic multidimensional approach, akin to the Research Domain Criteria framework proposed by the National Institute of Mental Health. One option is to use person-centred approaches (e.g. latent class or latent profile analyses) that identify homogeneous subsets of individuals on the basis of their post-trauma or severe stressor diagnoses or symptoms, other relevant indicators (e.g. gender, symptom severity, functional impairment, latency of symptom development) and characteristics of the trauma or severe stressor history itself (e.g. chronicity of stressors or revictimisation,Reference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2,Reference Adams, Moreland, Cohen, Lee, Hanson and Danielson22–Reference Pratchett and Yehuda24 timing of trauma or severe stressor exposureReference Pratchett and Yehuda24 and type of trauma or severe stressorReference McLaughlin, Green, Gruber, Sampson, Zaslavsky and Kessler2) within a larger heterogeneous sample. For instance, one study using this approach examined psychopathology outcomes in a sample of 815 adults at age 20 following exposure to ‘early social stress’ (perinatal through age 5) and found support for two broad latent dimensions of internalising and externalising psychopathology, rather than disorder-specific vulnerability.Reference Conway, Raposa, Hammen and Brennan14 With respect to treatment implications, our study findings lend support to the increasing attention on developing transdiagnostic, rather than disorder-specific, assessment and treatment models.Reference D'Andrea, Ford, Stolbach, Spinazzola and van der Kolk19,Reference Gutner, Galovski, Bovin and Schnurr25 Aetiologically relevant operationalisation of traumatic or severe stress reactions via multidimensional clinical outcomes within a longitudinal framework are necessary first steps to diagnostic clarity, which is central to prevention and intervention efforts.
This study shows that rates of incident depressive, anxiety and behavioural disorders in school-aged children are comparable across a range of stress disorders. Future research within a longitudinal framework, particularly in diverse samples, is needed to examine transdiagnostic approaches in the course of post-trauma psychopathology.
Funding
J. L. G., T. L. L. and H. T. S. were funded by the National Institute of Mental Health, grant numbers R01MH109507, R01MH110453 and R21MH094551.
Author contributions
J.L.G., H.T.S., T.L.L. and D.K.F. contributed to the acquisition of the data. A.B., T.J., K.C.K. and J.L.G. made substantial contributions to the conception or design of the work. A.B., D.K.F., K.C.K., T.J. and J.L.G. were involved in the analysis. All authors were involved in interpretation of data for the work, either drafting the work or revising it critically for important intellectual content, provided final approval of the version to be published, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2019.247.
eLetters
No eLetters have been published for this article.