Abstract
Major depression is common in adolescents, causes significant impairment and is associated with long-term interpersonal problems and suicide [1, 2]. Although it can be effectively treated [3, 4], depression in adolescents often goes undiagnosed [5, 6]. Epidemiological studies show that depressed adolescents seldom seek help [6, 7] and have low rates of antidepressant treatment: 3% in a recent community survey [7]. This may be compounded by poor detection by health care providers [5, 8].
Use of a brief depression rating scale increases the detection of depressed patients in primary care [9]. Does the same occur in specialist adolescent mental health services? To our knowledge, this has not been explored. We describe changes in the rate of adolescents diagnosed with major depression in a psychiatric service following the routine use of an adolescent-completed depression rating scale (Center for Epidemiological Studies Depression Scale, CES-D) [10]. The hypothesis was that this intervention would result in an increase in the rate of diagnosis of major depression. This was examined in a retrospective, naturalistic manner by surveying the rates of diagnosis before and after the introduction of the CES-D. As Leginski et al. stated, ‘by rigorously exploiting the evaluation opportunities of natural experiments, we… advance the improvement of our services’ [11].
Method
Setting
This was a unit that provided ambulatory and inpatient services for children and adolescents with mental health problems. Staff consisted of experienced child and adolescent psychiatrists, psychologists, social workers, trainee psychiatrists and nurses. There was little staff turnover during the period of the study.
Study design
The study was designed in 1998. There was no plan for such an evaluation prior to then. All patients assessed from January 1993 until the introduction of the CES-D, in September 1995, and all those assessed after that point until March 1997, were included. The March 1997 cutoff was chosen because the CES-D scale was replaced then by the CESDC [12]. During the period covered by the study, the youth self-report (YSR) [13] and child behaviour checklist (CBCL) [14] were routinely mailed to patients by clerical staff prior to assessment. None of the adolescents were hospitalized at the time of assessment; 23% received inpatient or day-patient treatment subsequently.
Intervention
From September 1995, adolescents were asked to complete the CES-D [10] prior to the initial evaluation interview. The CES-D was mailed with the other questionnaires. About one quarter (n = 123, 23%) of the adolescents assessed after September 1995 did not complete the CES-D. The reasons for this varied (e.g. inadvertently the CES-D was not mailed, the adolescent refused). However, this was not due to clinicians selectively using the CES-D since they were not involved in the process of administering it.
The CES-D was chosen because it is a widely used, simple and easy to understand self-report scale. The CES-D contains 20 statements (e.g. ‘I felt depressed’, ‘I felt that everything I did was an effort’). Respondents are asked to rate each statement in relation to the previous week as: ‘rarely or none of the time’, ‘some or a little of the time’, ‘occasionally or a moderate amount of time’, or ‘most or all of the time’. Responses are scored from 0 to 3 (four items are scored in the reverse). The CES-D has similar psychometric properties in adolescents as in adults; the internal consistency is high (α = 0.8–0.9) [15].
Outcome measure
The proportion of patients given a clinical diagnosis of DSM-III-R [16] or, after 1994, DSM-IV [17] major depression.
Diagnostic assessment
Evaluation of all patients followed the same procedure. Prior to the interview, parents – or main caregiver – and adolescents completed the questionnaires. After reading the information provided by referring agent, school reports, completed questionnaires (CBCL, YSR) and scale scores (including the CES-D, when in use), two clinicians interviewed the adolescent and family. Usually, one was medical (child and adolescent psychiatrist or trainee psychiatrist) and the other non-medical (e.g. clinical psychologist) but both could occasionally be from either group. Both professionals conducted a family interview followed by separate interviews with parents and adolescent each by one clinician. This process took about 2 hours.
When the evaluation was completed, the two professionals made consensus DSM-III/IV diagnoses. One of the clinicians wrote a detailed assessment report. There were no specific rules about how consensus was to be reached and there are no data to examine this. Clinicians had been instructed not to use the CES-D or other questionnaires as screening or diagnostic instruments but as supplementary items of information. In particular, clinicians were told that the CES-D was ‘not a diagnostic test’.
An independent researcher (T.M.) extracted information in the medical record for the study. This included: questionnaire data, clinical diagnoses, age, gender, family status (two original parents – intact family, single parent, blended, other), whether a child psychiatrist or trainee psychiatrist (as opposed to psychologist or social worker) had been involved in the assessment and whether the patient had been subsequently admitted for inpatient or day-patient treatment.
Statistical analysis
Differences between groups were examined using the t-test or oneway ANOVA for continuous variables, and χ 2 test for categorical variables. Strength of association is expressed as odds ratio (OR).
The main analysis (SPPS for Windows, version 10, SPSS Inc, Chicago, Ill.) involved multiple logistic regression [18]. Diagnosis of major depression (present, absent) was the dependent variable. Two models were fit. The first model contained as predictors: (i) those variables significantly different in univariate analyses between depressed and non-depressed adolescents (these results are not presented but are available on request); and (ii) those variables that differed before and after CES-D use (presented in Table 1). With such potential confounders in place, the second logistic regression model fit added CES-D use as a further predictor, to evaluate the impact on diagnosis of the presence or absence of CES-D information, controlling for potential confounders. This strategy was performed twice, using, respectively, parent and child questionnaire data in the relevant scales. Separate analyses according to respondent are customary because no definite algorithms exist to combine data from various informants [19]. The scales were those described for the CBCL and YSR [13, 14].
Descriptive data according to CES-D use: counts or means and percent or (SD)
Results
There were 1310 patients included in the study. Forty-two per cent (n = 546) were female. Twenty-one per cent (n = 279) were assessed in 1993, 23% (n = 306) in 1994, 25% (n = 324) in 1995, 23% (n = 304) in 1996, and 7% (n = 97) in 1997. Data in Tables 1 and 2 show that characteristics of the patient population had changed little during the study period. Table 1 shows there were more females in the group that used the CES-D. This group was also less disturbed overall, had fewer externalising problems but more eating disorders.
Rates of diagnosis of major depression, depression scores, age and gender according to period and use of CES-D
Overall, 9% (n = 118: 46 males, 72 females) received a diagnosis of major depression; 13% (n = 72/546) of the females and 6% (n = 46/764) of the males. When the CES-D was not available the rate of major depression was 6% (n = 56/909; 4% of males, 9% of females) and 16% (n = 62/401; 10% of males, 22% of females) when it was used. That is, adolescents assessed when the CES-D was available were almost three times as likely (OR = 2.8, 95% confidence interval 95% [CI] = 1.9, 4.1) to receive a diagnosis of major depression scale; CBCL: child behaviour checklist; YSR: youth self report. depression as those assessed when the CES-D was not used; the increase was similar for boys and girls.
Because rates of diagnosis could have been influenced by changes in the patients referred (e.g. more females, more eating disorders), it was necessary to statistically control for these confounders, as described in the method. This was done through a multiple logistic regression, which confirmed the results of the univariate analysis (CES-D available = 1 or not = 0, entered in the second step: χ 2 improvement = 20.34, df = 1, p < 0.001, for parent reports; χ 2 improvement = 7.85, df = 1, p < 0.01, for youth reports). That is, when the effect of confounders were taken into account, adolescents were more likely to receive a clinician diagnosis of major depression when the CES-D was available and used (adjusted OR = 2.8, 95% CI = 1.8, 44.3 for parent reports; adjusted OR = 2.1, 95% CI = 1.3, 3.5 for youth reports).
Discussion
What are the possible reasons for our finding that routine use in a specialist adolescent psychiatric service of a self-report scale for depression prior to clinical assessment was associated with an increase in the proportion of adolescents diagnosed with major depression? There are, at least, five possibilities that could account for these findings and that we could address with our data.
1. It could be that adolescents diagnosed when the CES-D was and was not available differed. Data in Tables 1 and 2 show this not to be the case, with the exception of gender, total problems, externalising scores and eating disorders. However, these differences did not influence the change in rates of diagnosis of major depression to a significant extent, as demonstrated by the multivariate analysis.
2. The increased rate of major depression when the CES-D was present could suggest that as a result of the CES-D clinicians had lowered their threshold; that is, they diagnosed adolescents with fewer symptoms as depressed. If this were the case, it would be expected that scores on the anxious/depressed scale of the CBCL and YSR would be lower among those diagnosed with major depression when the CES-D was used. This was not so. Mean (SD) anxious/depressed score among those diagnosed with major depression was 14.4 (5.9) for adolescents who did not have the CES-D versus 15.7 (5.3) for those who had the CES-D (t = 1.23, df = 88, NS) (adolescent reports). The parallel results for parent reports were 18.5 (6.8) and 19.6 (6.9) (t = 0.73, df = 110, NS). Furthermore, adolescents diagnosed with major depression with and without the CES-D did not differ according to gender, age, comorbid diagnoses, whether they had an intact family, scores on the internalising, externalising and total problem scale of the CBCL and YSR and rates of subsequent admission to hospital. That is, the two groups diagnosed with major depression were very similar in all the characteristics we could examine with our data.
3. Clinicians’ awareness of major depression and diagnostic practices may have changed during the study period. If this were the case, a gradual increase in the rates of diagnosis over time might have been expected. To indirectly estimate this, adolescents were divided into four groups according to when they had been evaluated. These data are presented in Table 2. Rates of major depression spiked when the CES-D was available but there was no change otherwise. Rates were surprisingly similar across the periods when the CES-D was not used (about 6%). In particular, during the era after introduction of the CES-D, the 123 adolescents who happened to not have CES-D information yielded a rate of diagnosis very similar to those from the earlier period when the CES-D was not in use.
4. It may be that the CES-D is a better instrument that gave clinicians information about depression not available otherwise. Examination of sensitivities and specificities of the YSR anxious/depressed scale and CES-D showed they were comparable (area under the receiver operating characteristic curve was 0.85 and 0.89, respectively). Further, correlation between CES-D and YSR anxious/depressed scale scores (r = 0.74) approached the upper limit defined by their respective reliabilities. This suggests that clinicians had at their disposal similar questionnaire data about depressive symptoms throughout the study period and that lack of information was not an important reason for the change in diagnostic practice (whether or not such information was actually used).
5. It may have been that clinicians were less sensitive to the possibility of major depression in adolescents with higher levels of disturbance when the CES-D was not available, perhaps as their attention was focused on other problems (e.g. conduct disorder) when adolescents were very disturbed. In such a case, availability of the CES-D may have reminded them of the possibility of depression. To examine this explanation, the total problems scales of the CBCL and YSR were used as a measure of the overall level of disturbance. Rates of major depression were compared between adolescents in the top quartile and the rest when the CES-D was used versus not used. For parent report, those in the top quartile were not more likely to be given a diagnosis of major depression (5%; 13/242) than the rest (7%; 39/603) when the CES-D was not available. When the CES-D was used, those in the top quartile were more likely to be given a diagnosis of major depression: 25% (20/79) and 13% (39/301), respectively (χ 2 = 7.29, df = 1, p < 0.01) (data missing for 85 cases, 6%). In fact, when the CES-D was used, those in the top quartile in the total problems scale were six times as likely (95% CI = 2.8, 12.7) to be diagnosed with major depression as those who did not have the CES-D. Similar results were obtained with adolescent self-report data.
In summary, explanations 1 to 4 are not supported by the data, while there is support for explanation 5.
Interpretation of these results ought to take into account several limitations. Firstly, the study had a retrospective, ecological design. However, because uniform questionnaire data were available throughout, it allowed statistical control for those differences in the characteristics of the patients on which this study had data. Conversely, the study has the advantage that clinicians’ diagnostic behaviour could not have been influenced by the knowledge that such an evaluation was taking place. A second limitation was that clinical diagnosis of major depression was used as the main outcome measure. There are no data about the reliability or validity of these diagnoses. They were made by consensus between two clinicians but we have no systematic information on how consensus was reached; nor can we ascertain whether clinicians’ under- or over-diagnosed major depression. Nevertheless, there were no differences between adolescents diagnosed with major depression with and without the CES-D, before and after the introduction of this scale, in the level of depressive symptoms reported (by patient or parent) or other characteristics. A third shortcoming was that DSM-III R and DSM-IV criteria were used. Both are very similar, the difference being that DSM-IV specifically requires impairment (criterion C) [17], p.327]. That is, diagnosis became more restrictive with DSM-IV; this would have resulted in a lower rate of major depression in the patients evaluated later in the study period. This was not so. A fourth limitation was that high scores on the anxiety/depression scale of the CBCL and YSR are not synonymous with major depression. However, there is evidence supporting the validity of these measures [20] and their convergence with major depression in adolescents [21]. Finally, results from this unit may not generalize to other services.
Conclusion
A depression rating scale may alert clinicians to the possibility of major depression and sensitise parents and adolescents.
If our results are confirmed, this is a cost-effective procedure, particularly with more disturbed adolescents.
Footnotes
Acknowledgements
We thank Paul Simpson for his help.