The effect of a severe disaster on the mental health of adolescents: a controlled study

The effect of a severe disaster on the mental health of adolescents: a controlled study

Klik hier voor het onderzoek in PDF formaat, (noot webmaster)

Sijmen A Reijneveld, Mathilde R Crone, Frank C Verhulst, S Pauline Verloove-Vanhorick

---

TNO (Netherlands Organisation of Applied Scientific Research) Prevention and Health, Leiden, Netherlands (S A Reijneveld MD, M R Crone MSc, Prof S P Verloove-Vanhorick MD); Erasmus University Medical Centre Rotterdam-Sophia Children's Hospital, Rotterdam (Prof F C Verhulst MD); and Leiden University Medical Centre, Department of Paediatrics, Leiden (Prof S P Verloove-Vanhorick)

---

Correspondence to: Dr Sijmen A Reijneveld, TNO Prevention and Health, PO Box 2215, 2301 CE Leiden, Netherlands (e-mail:SA.Reijneveld@pg.tno.nl)

Summary

Background Disasters greatly affect the mental health of children and adolescents, but quantification of such effects is difficult. Using prospective predisaster and postdisaster data for affected and control populations, we aimed to assess the effects of a severe disaster on the mental health and substance use of adolescents.

Methods In January, 2001, a fire in a café in Volendam, Netherlands, wounded 250 adolescents and killed 14. In the 15 months before the disaster, all grade 2 students (aged 12-15 years) from a school in Volendam (of whom 31 were in the café during the fire), and from two other schools, had been selected as controls for a study. 124 Volendam students and 830 from the other two schools had provided data for substance use, and completed the youth self-report (YSR) questionnaire about behavioural and emotional problems. 5 months after the disaster, we obtained follow-up data from 91 (response rate 73·4%) Volendam adolescents and 643 (77·5%) controls from the other two schools. The primary outcome measures were changes in score in YSR categories of total problems, alcohol misuse, smoking, and substance use. We compared changes in scores between groups using logistic regression.

Findings Volendam adolescents had larger increases in clinical scores than controls for total problems (odds ratio 1·82, 95% CI 1·01-3·29, p=0·045) and excessive use of alcohol (4·57, 2·73-7·64, p<0·0001), but not for smoking or use of marijuana, MDMA (ecstasy), and sedatives. Increases in YSR scores were largest for being anxious or depressed (2·85, 1·23-6·61), incoherent thinking (2·16, 1·09-4·30), and aggressive behaviour (3·30, 1·30-8·36). Intention-to-treat analyses showed significantly larger for increases in rates of excessive drinking and YSR symptom subscales in Volendam adolescents than controls. Effects were mostly similar in victims and their classmates.

Interpretation Mental health interventions after disasters should address anxiety, depression, thought problems, aggression, and alcohol abuse of directly affected adoloscents and their peer group.

Introduction
Disasters have substantial effects on the mental health of children and adolescents including symptoms of depression and anxiety, substance abuse, and difficulties with concentration, memory, behaviour, school, and sleep.^1-5 Many of these effects seem to be on a continuum of stress responses, including post-traumatic stress disorder.^2,3,5,6

Evidence shows that mental health outcomes of disasters are determined by the characteristics of the disaster, the individual, and the environment. Effects are increased if a child perceives a strong life threat, is injured, or has relatives or close friends who are injured.^2-5 Older children tend to be more severely affected than younger children, especially with respect to self-reported emotional symptoms in girls,^2-5 and pre-existing psychopathological disorders.^2,3 Finally, a severe parental reaction, and long-term family and environmental disruption worsen mental health outcomes.^2-4

Because of the unexpected nature of disasters, most investigations include people who were exposed to the event, and sometimes also people not exposed who were selected after the disaster.^1,3,7 Such studies might have bias in their selection of exposed and unexposed groups, and use of retrospective information about health before the disaster.^3,8 These biases generally lead to an overestimation of effects.⁷ Prospective investigations that have data for mental health before disasters are rare. Although Bromet and Dew³ identifed four studies with such data, none of these had controls who were assessed at the same time as the affected group.^8-11

On Jan 1, 2001, at 0030 h, a fire started in a popular youth café on the dyke of Volendam, a town in the west of the Netherlands. The fire killed four adolescents immediately and wounded about 250.^12,13 Of the 203 victims who were admitted, mainly for burns, 40 had to stay for over a month. 21 were admitted outside the Netherlands because the capacity of the Dutch burn units was exceeded.^12,13 A total of 14 adolescents died.¹³ Almost all 300 adolescents who were in the fire lived in Volendam. The disaster had a great effect on the local community and the rest of the Netherlands.

At the time of the disaster, students of a school in Volendam and of two other schools in the Netherlands were controls in a study on the effects of a school health-promotion programme. The programme consisted of school-based interventions to prevent behavioural and emotional problems, smoking initiation, excessive use of alcohol, and use of psychoactive substances. The programme was planned to run from November, 1999, to April, 2001. Baseline measurements for the programme outcomes had been taken 15 months before the disaster. In the Volendam school (940 students in total), about 200 students were present at the café fire; of these, 140 were wounded. Thus, this investigation resembles an actual experiment in which adolescents would be assigned to groups--ie, a natural experiment.^14,15 We therefore analysed our observational data in accord with CONSORT criteria for the analysis of randomised controlled trials,¹⁶ to best obtain information from this study.

Methods

Participants In the affected school, 124 students aged 12-15 years (mean 13·6) had completed the baseline questionnaire. Of these, 31 were at the fire, 17 of whom were wounded. The other two schools were unaffected by disasters and both were located in the middle of the Netherlands, about 80 km from the affected school. 830 students aged 12-15 years (mean 13·8) had completed the baseline questionnaire. These students are controls in this study. At baseline, all adolescents were informed about the nature of the original study, and that a second measurement would be taken.

The original study protocol had been approved by the medical ethics committee of TNO Prevention and Health, Leiden, and included consent from the parents' council of every school. We informed the committee about the fire. After discussing the situation with the headmaster of the affected school, we continued the study with the inclusion of additional questions about satisfaction with school-based care after the disaster and whether students' needs had been met. Results were reported to the school management immediately after follow-up, to assist in further post-disaster school planning.

Procedures In September-October, 1999, students had completed a baseline questionnaire consisting of the youth self-report questionnaire (YSR),^17,18 questions about smoking,¹⁹ and use of alcohol, marijuana, sedatives and hypnotics, and MDMA (ecstasy). YSR scores, smoking, and use of alcohol were the primary endpoints of the original study, and remained so in the postdisaster analyses. The YSR was used to assess the adolescent's report of their behavioural and emotional problems in the preceding 6 months. Its reliability and validity have been replicated for the Dutch translation from English.^17,18 We also obtained data for sex, age, level of education (lower or higher--ie, junior secondary vocational education or general secondary education US higher), ethnic background (of child and both parents--at least two born in the Netherlands, or other), religion (Christian, other, or none), and parental employment (one or both parents working more than 3 days per week, or other).

As planned, all students completed the same questionnaire in May, 2001. To ensure comparability of information, we took no measures to increase the response of Volendam adolescents. We added some questions to the end of the Volendam students' questionnaires regarding their exposure to the fire, and school-based postdisaster care. Data for fire exposure of those lost to follow-up were provided by the school management. Responses were anonymous and could only be linked by school code, sex, date of birth, and initials of the adolescent.

For the analyses, we dichotomised all outcome measures, using cut-offs as planned in the original study. Of the YSR, we used only the items relating to mental health problems and calculated scores for nine symptom subscales, two broad groups of symptoms designated internalising and externalising (each comprising several symptom subscales of emotional and behavioural problems, respectively), and a total problems score. Adolescents were allocated to a normal or clinical range, with the 98th percentile of the Dutch normative sample as cut-off for symptom subscales, and the 90th percentile for other scales.¹⁸ Smoking was defined as at least one cigarette per month, excessive use of alcohol as at least five drinks on one occasion in the preceding 2 weeks, and use of other psychoactive substances as use at least once ever.

Statistical analysis Because of the nature of the study, we did not calculate power beforehand. We compared the adolescents of the affected school who provided follow-up data with those in control schools to look for changes between baseline and follow-up in all categories, using logistic regression. Next, we assessed whether the size of the changes was affected by sex or exposure to the disaster.^2-5 We repeated all analyses, adjusting for background characteristics that were significantly different between affected and control schools (age, educational level, ethnic origin, religion, and parental employment). We also did an intention-to-treat analysis of data from all individuals, including those lost to follow-up. We assumed no change in those lost to follow-up since baseline. All p values are two-sided and have not been adjusted for multiple comparisons. We used ² tests for all proportions except for when the expected counts were less than 5, in which case we used Fisher's exact test. For continuous data we used t tests, and F tests in ANOVA.

Role of funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, writing of the report, or the decision to submit this report for publication.

Results
We received complete data for the baseline measurement from 954 adolescents, of whom 124 attended the affected school. Of these, 734 (77%) provided us with at least some follow-up data, with slightly fewer from the affected school (91 [73·4%]) than from the others (643 [77·5%], p=0·31). The figure shows participant flow and follow-up. Table 1 shows baseline characteristics of all adolescents. Demographic characteristics but not baseline values of outcome measures differed significantly between the affected school and control schools. Demographic differences were mostly due to real differences between the schools; eg, the affected school and one of the control schools were denominational--Roman Catholic and Protestant, respectively. We also analysed whether affected and control schools differed in loss to follow-up for characteristics that showed significant overall differences in loss to follow-up. We did this by assessing the interaction between school (affected or control) and follow-up status (analysed or lost to follow-up). None of these interactions were significant (range of p values: 0·08-0·95). Differences were thus not due to selective loss to follow-up, since loss to follow-up by characteristic did not differ between the affected school and control schools. The exposure to the fire was, however, much higher in the group lost to follow-up than in the analysed group. In the analysed group (n=91), 14 adolescents (15%) were present at the fire, of whom five (5%) were wounded; in the lost to follow-up group (n=33), these figures were 17 (52%) and 12 (36%), respectively (both p<0·0001), one of whom died.

---

---

The proportion of adolescents with behavioural and emotional problems, as measured by the YSR, increased in all three schools (table 2). Increases were larger in the affected school than in the control schools for the total problems scale (p=0·045), and the anxious or depressed (p=0·014), thought problems (ie, incoherent thinking) (p=0·028), and aggressive behaviour (p=0·012) scales. These differences increased after adjustment for background characteristics in which the affected and control schools differed significantly, except for the differences in total problems score. Intention-to-treat analysis yielded lower effect estimates, with two remaining significant.

---

---

The proportion of adolescents using various substances increased (table 3). All increases were larger in the affected school than in the control schools, but differences were only significant for excessive drinking (p<0·0001). This result was also seen in the intention-to-treat analyses. A Bonferroni adjustment for comparisons of primary endpoints²⁰--YSR total problems score, smoking, and excessive use of alcohol--resulted in only alcohol use being significant.

---