Abstract
Recent studies suggest there has been a secular increase in the prevalence of depression, self-harm and suicidal ideation in young people [1]. This evidence comes from several sources.
Population studies have used adult samples and indirect measures of prevalence of depression. These studies have not measured self-harm or suicidal ideation. The Epidemiological Catchment Area Study [2] and other studies from North America [3], the UK [4] and New Zealand [5], reported increasing rates of current and lifetime prevalence of major depression in those born later in the century. However, this is not a uniform finding. Other surveys using similar methodology did not show such an increase [2].
Retrospective population studies use methods that can lead to recall bias. For example, older participants may find early episodes of depression difficult to remember [6].
Retrospective cohort studies of relatives of adult patients with affective disorders also suggest an increase in rates of depression in later generations [7–9] and in siblings of young people with mood disorder [10].
Two studies attempted to measure secular changes in the prevalence of depression prospectively. A Swedish cohort showed an increase in mild depression between 1947 and 1972 but no change in severe depression [11]. In contrast, a recent examination of the cohort used in the Stirling County Study found that the prevalence of depression remained steady from 1952 to 1992 [12]. The former used a non-structured interview while the latter used a semi-structured interview. Both studies examined only adult subjects.
There are also studies in young people using questionnaires. For example, Achenbach and Howell [13] reported the change in prevalence of emotional and behavioural problems in community samples of 7− 16-year-olds between 1976 and 1989, using the parent-reported Child Behaviour Checklist (CBCL). They described a significant increase over time in scores on the anxious/depressed scale and other problem scales. However, there were no data from the young people themselves.
There is strong evidence that suicide rates have increased in many Western countries, especially in young men. However, obtaining accurate estimates of completed suicide is difficult. It is estimated that many deaths by suicide are not recorded as such, even in those countries where mortality by suicide is collected [14]. This notwithstanding, suicide rates for young men aged 15–24 trebled in Australia from 1964 to 1993 [15], [16]. A similar increase has also been found for young men in other countries [17], [18].
The prevalence of self-harm or suicidal ideation is more difficult to determine. Surveys of self-harm are often based on hospital presentations, when not all adolescents who have deliberately harmed themselves will visit a hospital, and even if they do, the incident may not be recorded as deliberate self-harm in the hospital file [14]. Studies of clinic populations have reported an increase in suicidal behaviours in young men [19]. However, changes in clinic populations may not be representative of the wider community, but may instead reflect changes in referral patterns or service use. For long-term prospective studies, it is also essential that a consistent measure be used. If a clinical diagnosis alone is used there is potential for bias, as criteria for diagnoses change.
The aim of this study was to examine whether there had been changes in prevalence of psychological disturbance over time in a population of adolescents referred for treatment. The hypothesis is that psychological disturbances in young people, in particular depression, deliberate self-harm and suicidal thoughts, increased later in the 20th century. Although this population is potentially biased in many ways, ascertaining whether changes had taken place may contribute to the understanding of this troubling phenomenon.
Method
Setting
The data was collected retrospectively from clinic records at a university-affiliated adolescent psychiatry service in Sydney, Australia (the Rivendell Unit). The Unit provides outpatient and inpatient services for young people living in New South Wales (population 6.6 million). Other studies in this clinic showed that 23% of the patients were from the top three socioeconomic levels, 60% from the three middle levels and 18% from the two lower levels [20]. The procedures of the clinic, referral patterns and services remained constant during the period examined. These and the process used to evaluate patients have been described in detail elsewhere [21].
Of the adolescents evaluated at the Unit from 1983 to 1998 inclusive, there was complete questionnaire data available for 4495 young people aged 12–17 years when assessed. They were born between 1966 and 1986, a 21-year period. The mean age at evaluation was 14.1 years (SD = 1.4 years) and 58% were male. It is not known how many of the patients evaluated had data missing. It is estimated that at least 10% did not complete the questionnaires. The characteristics of those with missing data are unknown.
Instruments
The subjects completed the Youth Self-Report (YSR) [22] and their parents the Child Behaviour Checklist (CBCL) [23]. Both the YSR and CBCL consist of eight narrow-band scales and two broad-band scales derived from 113 items rated 0–2. These include items on deliberate self-harm and suicidal thoughts. These are well known and widely used instruments in clinical practice and research.
Outcome measures
Significant self-reported symptoms of anxiety/depression were considered present if the young person scored on or above the 98th percentile of the YSR anxious/depressed scale. Parent-reported anxiety/ depression was considered present if the CBCL had a score on or above the 98th percentile. The 98th percentile scores for all of the YSR and CBCL narrow-band scales were taken from the distribution of scores in a normative population [22], [23]. The other seven narrow-band scales were scored in the same way.
The prevalence of self-reported thoughts of self-harm and suicidal ideation was estimated from data in the YSR: Item 18 (‘I deliberately try to hurt or kill myself’) and Item 91 (‘I think about killing myself’). Similarly, the prevalence of parent-reported knowledge of self-harm and suicidal ideation was estimated from the CBCL Item 18 (‘Deliberately harms self’) and Item 91 (‘Talks about killing self’). In both the self and parent-reported items, suicidal thoughts or deliberate self-harm were considered to be present if there was a score of 1 (somewhat or sometimes true) or 2 (very true or often true).
A scale was excluded from the analysis if the answer to more than one of the items that contributed to the scale was missing. Up to 5% of each of the narrow-band scales were incomplete.
Statistical analysis
The data was analyzed through logistic regression (using the SAS 6.12. statistical package). The six outcome variables were: self-reported anxiety/depression; self-harm; thoughts of suicide; parent-reported anxiety/depression: parent-reported self-harm; and parent-reported talk of suicide.
Year of birth was divided into three 5-year cohorts and one 6-year cohort to calculate descriptive statistics. For logistic regression, year of birth was analyzed as a continuous variable. As in other population surveys [2] year of birth rather than age was used because the aim was to examine secular trends. Age was considered a potential confounder because depression and suicidal behaviour increase with age during adolescence [24–26].
The CBCL and YSR narrow-band scales were analyzed separately for young men and women, as the 98th percentile scores are different according to gender [22], [23] and the pattern of disorders also varies for young men and women. Sex was considered a confounder for deliberate self-harm and talking of suicide, because suicide attempts are more prevalent among young women [27], [28].
Results
Description of the study population
Table 1 describes the number of participants in each 5-year birth cohort (the first comprised 6 years), their age at the time of referral, gender and their total problem scores. These results can be summarized as follows.
Descriptive data of adolescents by year of birth
1 The mean age at the time of referral decreased with increasing year of birth (from 14.6 years in those born in 1966–1971, to 13.1 years in 1981–1986).
2 There was a significant change, without a clear linear trend, in parent-reported total problem scores with year of birth but there was no change in self-reported total problem scores.
3 The proportion of male and female adolescents varied with each cohort but there was an overall increase in the proportion of young men.
Agreement between parent and child reports was small to moderate and consistent with that reported in the literature. For example, correlation for anxiety/depression was r = 0.42 (p < 0.0001), while for selfharm and suicidal thoughts was r = 0.47 (p < 0.0001) and r = 0.34 (p < 0.0001), respectively.
Descriptive statistics for the key outcomes
These are shown in Table 2. In summary, there was no change in self-reported anxiety/depression for young men, while there was a significant change for young women, but without a clear linear trend. There was no change in self-reported self-harm or suicidal ideation. However, parent reports showed a gradual increase in self-harm from 20.7% in the first cohort to 27.7% in the final cohort. Similarly, parentreported talk of suicide increased from 33.5% in the first cohort to 42.0% in the final cohort.
Descriptive statistics for key outcomes in the analysis
Anxiety/depression
The results in Table 3 show that neither young people of either gender, nor their parents, reported changes in anxiety/depression over time when controlled for the effect of age. Age had a large effect on anxiety/depression. The odds of self-reported anxiety/depression increased by 37% per year of age for young women and 17% per year of age for young men. Odds of parent-reported anxiety/depression increased by 8% per year of age for female adolescents, but a similar increase was not observed in male adolescents.
Changes in the prevalence of anxiety/depression: self and parent report (controlled for age)
Self-harm
As shown in the upper half of Table 4 the reports by adolescents (YSR) did not show a significant increase in self-harm with year of birth, when controlled for age and sex. In contrast, odds of parentreported self-harm (CBCL) increased by 5% per year over the 21 years. Again, age had a large effect. Odds of self-reported self-harm showed a 19% increase per year of age, while the odds of parent-reported selfharm showed a 16% increase per year of age. Both young women and their parents showed twice the odds of young men or their parents to report self-harm.
Changes in the prevalence of parent and self-reported self-harm and suicidal ideation (controlled for age and sex)
Talks of killing self
These results are shown in the lower half of Table 4. There was no change with time in the YSR item ‘I think about killing myself’. There was a significant increase of 4% per year in the odds of parents endorsing the CBCL item ‘Talks about killing self’. These behaviours also increased with increasing age. There was a 17% increase in odds of self-reported talking about killing oneself per year of age and a 7% increase in parent reports. Male adolescents were less likely to talk about killing themselves (OR = 0.63) and parents were less likely to report suicidal ideation in their male children (OR = 0.47).
Discussion
The picture that emerges is complex and findings vary depending on the informant. Overall, however, there was no evidence of an increase in the prevalence of depressive symptoms in this referred group over 21 years based on either parent or adolescent reports. This was an unexpected finding but given the biases built into any clinic sample, it cannot be concluded that depression is increasing or not. In contrast, parent reports showed a large increase of 5% per year over the 21-year birth cohort for self-harm and 4% per year for suicidal ideation. However, young people themselves did not report similar changes in either behaviour. It is also of note that the young people themselves are asked if they are thinking of killing themselves whereas parents report their children talking of killing themselves. Unless the young people discuss such thoughts their parents will be unaware of them.
These results suggest that some of the previously reported increase in psychological disturbance in young people may be due to greater awareness of their symptoms by their parents. During the last 20 years there has been much media attention on youth suicide and depression in Australia, the teenage literature surprisingly often depicts suicide and psychiatric illness [29] and there have been several government-sponsored suicide prevention campaigns. This may explain some of the changes in parents' perceptions. A similar phenomenon was noted by Rey [21] who found that clinicians were 2.8-times more likely to diagnose major depression in young people when there was routine completion of a self-rating depression scale prior to assessment. This was probably because it raised the clinicians' awareness of the possibility of depression. It is also possible that the increased publicity about youth suicide has made young people more likely to talk about killing themselves, without necessarily increasing the number of attempts at self-harm or suicide.
It is of note that the mean age at assessment in this study decreased from 14.6 years in those born in 1966–1971 to 13.1 years in those born in 1982–1986. This is consistent with the hypothesis of enhanced parental awareness of symptoms, leading to referral at a younger age. Alternatively, it could also be due to symptoms starting earlier, to a decrease in parents' tolerance for disturbed behaviour, a reduction in families' support systems or better knowledge about or access to services, among other possible explanations. Based on anecdotal information, many clinicians in Sydney, Australia, believe that presentations to hospital of younger people with mental health problems have increased markedly in recent years. These findings appear to corroborate that observation but more studies will be required to confirm this phenomenon.
As expected, both age and sex were significant confounders. Also as expected, the numbers of young people who self-reported scores above the 98th percentile on the anxious/depressed narrow-band scale increased with age. The odds increased by 37% per year of age for young women and 17% per year of age for young men. However, the parent-reported anxious/depressed narrow-band scale showed only an 8% increase in odds per year of age for young women and no increase for young men. This is consistent with the well-known phenomenon that parents often do not detect depressive symptoms in their children, particularly male adolescents [30]. It also raises the possibility that young men are less likely to be referred for treatment of depression or anxiety, especially if the lack of awareness of their symptoms extends to other adults, such as teachers and mental health staff.
It would be useful to have data from other clinics and countries to see whether results are similar and to examine the underlying mechanisms.
Clinical implications
The parents in this study showed an increasing awareness of self-harm and suicidal ideation in their children, with increasing year of birth. There was no evidence for an increasing awareness of anxiety/depression. While this increasing awareness may be limited, it is still encouraging because it may lead to earlier referral for treatment. However, it is still clear that young people's self-reports are essential to obtain accurate information about their current mental state, especially for emotional problems. This becomes particularly important for older male adolescents, although parents of female adolescents are also less likely to have a full understanding of symptoms as their daughters increase in age. Services also need to note a trend for people of younger age to be referred for treatment.
Limitations
A major shortcoming of this study is that it is based on a clinic cohort, so the lack of changes found might not represent what has occurred in the general adolescent population. Clinic populations are subject to referral bias, where changes in referral patterns (themselves influenced by a multitude of factors) produce changes in the characteristics of the sample, independent of changes in the community. Using a clinic sample may produce results specific to the study population, which cannot be necessarily generalized to other clinics or the wider community. Also, rates fluctuate and it is not clear whether the same results would apply to a larger time span, for example 50 years. There was also an unknown number of patients (estimated at about 10%) who did not have questionnaire data. These probably were more disturbed and their inclusion may have changed the results. Therefore, these results will need to be confirmed by prospective studies or multiple-wave population-based data, although such studies are time-consuming and expensive.