A Randomized Effectiveness Trial of Interpersonal Psychotherapy forDepressed Adolescents FREE

Adolescent depression is a highly prevalent disorder (1.6%-8.9% in ayear).¹ Untreated adolescent depression isassociated with substantial morbidity, including school dropout, teenage pregnancy,suicide, and substance abuse,^2- 3 aswell as considerable health expenditure.⁴ Adolescentswith mental health problems, including depression, are largely undertreated.Fewer than 3 of 10 adolescents with mental health problems in the United Statesreceive mental health services.^5- 7 Thereis a large gap between the use of and need for mental health services.⁸ In recent years, school-based health clinics haveemerged as an important treatment setting for adolescents with mental healthand general medical problems.^9- 12

Several efficacious treatments are available for depression, includingserotonin reuptake inhibitors,^13- 15 cognitivebehavior therapy, interpersonal psychotherapy, and various group therapies.^16- 22 Mostof these treatments have been assessed in university hospitals under controlledconditions. Even then, the general response rate for cognitive behavior therapy,the most widely tested psychotherapy, is approximately 60%, leaving room forimprovement.²³ It is not known whether evidence-basedpsychotherapy can be effectively delivered in community mental health settingsby professionals with little formal training to broadly representative patientsamples.

This report describes the results of an effectiveness study of interpersonalpsychotherapy for depressed adolescents (IPT-A) in school-based health clinics.The goal was to assess the feasibility, acceptability, and efficacy of deliveringIPT-A under prevailing resource constraints of urban public school–basedclinics.

The sample consisted of adolescents aged 12 to 18 years who were referredto mental health clinicians in 1 of 5 school-based health clinics (3 middleschools and 2 high schools). The sample had a mean age of 15.1 years (SD,1.9 years), and was 84% female and 71% Hispanic (Table 1). The schools are located in urban, impoverished areas ofNew York City, NY.

All adolescents referred for a mental health intake in the school-basedhealth clinic were eligible to undergo screening to participate in the study.To be eligible for enrollment, adolescents needed a Hamilton Depression RatingScale (HAMD)²⁴ score of 10 or higher and aChildren's Global Assessment Scale (C-GAS) score of 65 or lower²⁵ atinitial clinical intake and again at study baseline. In addition, eligibilityrequirements included parental consent and a DSM-IV diagnosisof major depression, dysthymia, adjustment disorder with depressed mood, ordepressive disorder not otherwise specified. Adolescents were not eligibleif they were actively suicidal or mentally retarded; had a life-threateningmedical illness or a current diagnosis of substance abuse disorder, psychosis,or schizophrenia; were currently in treatment for depression; or were takingantidepressant medication. English-speaking students were accepted at all5 schools. In 2 schools, monolingual Spanish-speaking students were acceptedas well. A child psychologist (K.P.D.) determined whether students met eligibilitycriteria using the Schedule for Affective Disorders and Schizophrenia forSchool-aged Children lifetime depression module–child report²⁶ and the baseline clinical interview.

If at any time an adolescent's symptoms or functioning appeared to beworsening (HAMD score, >25), they underwent evaluation by a child psychologistor psychiatrist masked to the treatment condition. If possible, adolescentswere offered an adjunctive treatment such as medication. Adolescents who receivedantidepressants during the course of the study could remain in the study.They were removed from the treatment protocol if it was clinically necessaryto refer them for a more restrictive level of care such as hospitalizationor partial hospitalization.

The study was approved by the institutional review boards for the 3hospitals that sponsored the 5 school-based mental health clinics and by theNew York City Board of Education. Informed consent was received from parentsor legal guardians, and assent was received from the adolescents.

Randomization occurred at clinician and student levels. With the useof a table of random numbers, mental health clinicians in the school clinicswere randomized to provide IPT-A or treatment as usual (TAU). Half of thetherapists at each of the 5 schools received IPT-A training. Adolescents wererandomized to IPT-A or TAU within schools. In 1 school, clinicians were routinelyrestricted to treating students on a particular floor of the school; therefore,randomization occurred at the level of the clinician, but not the student(n = 7).

Interpersonal psychotherapy is a time-limited psychotherapy that focuseson current problems. The procedures are specified in a manual.²⁷ Thegoals of IPT-A are to reduce depressive symptoms and improve interpersonalfunctioning by relating the symptoms to 1 or more of 4 problem areas (grief,role disputes, role transitions, and interpersonal deficits) and by developingstrategies for dealing with these problems.²⁸ Twoindependent studies have demonstrated the efficacy of IPT-A.^17- 18 TheIPT-A intervention was delivered as 12 sessions during a 12- to 16-week period.Therapists provided 8 consecutive 35-minute weekly sessions followed by 4sessions scheduled at any frequency during the ensuing 8 weeks.

The TAU condition was whatever psychological treatment the adolescentswould have received in the school-based clinic if the study had not been inplace. The psychotherapy varied but closely resembled supportive counseling.Most adolescents in the TAU group received individual psychotherapy. Eightadolescents received 1 to 3 additional family/parent sessions, and 5 participatedin group therapy.

Participating school clinicians included 11 social workers and 2 doctoral-levelclinical psychologists. Before participating, the clinicians had little formalspecialized training in screening and identifying depression in adolescentsand no training in evidence-based psychotherapy. Training in the screeningbattery was provided to all clinicians. A subset of 6 social workers and 1psychologist were trained in IPT-A. The training included reading the IPT-Amanual, 2 half days of didactic training, and weekly supervision throughoutthe project by 2 of us (L.M. or K.P.D.). Treatment adherence and competencewere measured by checklists completed by therapists and supervisors. Overallcompetence in IPT-A was rated as satisfactory (mean, 3.3; SD, 0.87) on a scaleof 1 (poor) to 5 (excellent). One of the 7 IPT-A therapists who treated asignificant number of cases was a low outlier on the competency scale. Themean competence score without him was 3.6 (SD, 0.55). A future report willexamine therapist adherence and competence in detail.

The primary domains assessed included (1) depression symptoms as measuredby the clinician-rated HAMD and self-reported Beck Depression Inventory (BDI)²⁹; (2) global functioning as measured by the clinician-ratedClinical Global Impressions scale (CGI)³⁰ andC-GAS; and (3) social functioning as measured by the Social Adjustment Scale–Self-Report(SAS-SR).³¹ Higher scores on the HAMD and BDIindicate a greater number of symptoms; on the CGI and C-GAS, better functioning;and on the SAS-SR, worse functioning. Scores on the HAMD can range from 0to 74; on the BDI, from 0 to 39; on the C-GAS, from 0 to 100; and on the SAS-SR,from 1 to 5.

All patient assessments were performed in both treatment conditionsby a psychologist or a social worker masked to the patient's treatment conditionand were not shared with the treating clinicians. The assessments were conductedat baseline and weeks 4, 8, 12, and 16, or at early termination from the protocol.At week 12, the mask was accidentally broken in 2 cases for one evaluator.After week 12, a second evaluator completed the assessments of these cases.

Three independent evaluators were trained in the assessments beforethe onset of the study and participated in a reliability study to establishcomparability between their ratings. School-based clinicians also participatedin the reliability study on the screening instruments (HAMD and C-GAS). Reliabilityfor the Schedule for Affective Disorders and Schizophrenia for School-agedChildren was obtained by the 3 independent evaluators rating 7 audiotapedinterviews (κ = 0.76). Reliability of the HAMD was obtained from 21clinicians (including school clinicians and independent evaluators) rating20 audiotaped HAMD interviews (intraclass correlation coefficient, 0.84).To obtain reliability for the C-GAS, the 21 clinicians rated 20 case vignettes(intraclass correlation coefficient, 0.83).

The comparability of the patients in the IPT-A and TAU groups was examinedfor demographic characteristics, including sex, ethnicity, living in a single-parenthome, receiving public assistance, and diagnoses, using the χ² test.Mean age and paternal and maternal education were analyzed using an unpaired t test. Between-group differences for major outcome measures(ie, BDI, HAMD, C-GAS, and CGI) at baseline were calculated using unpaired t tests.

Analyses were conducted for the intent-to-treat sample (n = 63). Overallefficacy of treatment was assessed by conducting an analysis of covariancecontrolling for pretreatment scores on all major outcome measures by treatmentcondition at termination. Clinical recovery was defined as a HAMD score of6 or less and a BDI score of 9 or greater. Data were analyzed using a χ² test for the percentage of patients who reached these cutoff scoresat the point of termination. The α level was .05 (2 sided).

Post hoc analyses examined differential treatment effects by severityof depressive symptoms and age. Those who were in the highest HAMD quartilescore (≥22) and in the lower half of a median split of the C-GAS were consideredto be the most severely ill. Age was stratified as 12 to 14 years and 15 to18 years. A separate stratified analysis by each of these 3 baseline measureswas conducted to evaluate the treatment effect.

Two statistical approaches were used to examine changes over time inthe outcome measures. A repeated-measures analysis of variance (ANOVA) wasconducted on each of the outcome measures of interest³² with2 treatment groups; the number of time categories varied, with the outcomemeasure ranging from 4 (weeks 0, 4, 8, and 12) for HAMD to 2 for C-GAS (weeks0 and 12). For the 6 cases in which attrition occurred, the end point wascarried forward. Although this approach is easily interpretable, it assumesthat variances and covariances of observations across time points are homogeneous(assumption of compound symmetry). Violation of this assumption will frequentlyresult in type I errors greater than the nominal values, therefore a randomcoefficients regression analysis³³ was performedto complement the repeated-measures ANOVA. This method accommodates attritionof subjects by using available data without carrying over end points and isnot restricted to the assumption of compound symmetry. The primary parametersthat were estimated by the random coefficients regression are overall slopesfor each treatment supplemented by tests of differences among slopes.

Of the 509 adolescents who underwent screening from April 1, 1999, throughJanuary 31, 2002, 183 (35.9%) were initially eligible for baseline assessment,but 23 (12.6%) of the 183 were determined to be ineligible by the baselineassessment. Thus, 160 (31.4%) of those undergoing screening were eligiblefor the study. Of the 160 eligible, 96 (60.0%) refused to participate (Figure 1). Refusal to participate was notrelated to adolescent sex. The most common reasons for refusal were relatedto the requirements intrinsic to a research protocol, including not wantingto be randomized (n = 25 [13.7%]) and not wanting a parent contacted (n =26 [14.2%]). Other reasons for exclusion each represented less than 5% ofthe sample. No adolescents were excluded at screening or baseline becauseof current medication treatment.

The treatment groups did not significantly differ in background demographiccharacteristics (Table 1). Otherdiagnoses regarded as possible or probable by results of the clinical interviewat baseline were 20 patients (32%) with anxiety disorders, 5 (8%) with oppositionaldefiant disorder, 10 (16%) with substance use, and 4 (6%) with attention-deficit/hyperactivitydisorder.

Of the 64 participants initially enrolled in the study, 57 completedthe full protocol. Reasons for attrition are displayed in Figure 1. In the IPT-A condition, 1 adolescent was referred to theemergency department for suicidality and hospitalized for 1 week. In the TAUcondition, 1 adolescent was referred to the emergency department, hospitalizedovernight, and withdrawn from the study. Four adolescents were referred formedication treatment during the protocol, including 2 in IPT-A and 2 in TAU.One of the 2 adolescents in IPT-A was prescribed medication during her weeklonghospitalization, but discontinued its use after 1 to 2 weeks. The second IPT-Apatient received medication for attention-deficit/hyperactivity disorder,but discontinued its use after 2 weeks. One TAU patient was referred for butnever received medication, and the second received the medication during anovernight hospitalization but was nonadherent after several days.

At baseline, there were no significant group differences in depressionsymptoms on the HAMD or BDI. Mean baseline score was 18.6 on the 24-item HAMD,whereas the mean BDI score was 21.4. On the HAMD, adolescents who receivedIPT-A compared with those who received TAU reported significantly greaterdecreases in depressive symptoms (week 12, 8.7 vs 12.8; P = .04) (Table 2). Onthe BDI, adolescents who received IPT-A reported fewer depressive symptomsthan those who received TAU (8.4 vs 12.3; P = .14)at week 12 (Table 3).

At termination, 17 (50%) of the adolescents in IPT-A and 10 (34%) ofthose in TAU met the HAMD recovery criterion (score, ≤6). On the BDI, 25(74%) of the adolescents in IPT-A compared with 15 (52%) of adolescents inTAU met the recovery criterion (score, ≤9) (χ²₁ =3.9; P = .048). Effect sizes for the difference intreatment outcomes are shown in Table 2 and Table 3.

The repeated-measures ANOVA showed a statistically significant interactionbetween treatment and time (F_3,169 = 3.79; P = .01). Post hoc pairwise contrasts of the treatment means at individualtime points revealed significant differences between mean scores that emergedat week 8, with IPT-A reducing the mean score by a difference of 4.1 pointsgreater than TAU (t = −3.18; P = .003) (Figure 2). TheBDI analysis similarly showed a statistically significant interaction betweentreatment and time (F_3,169 = 2.88; P =.04) and significant differences between mean scores also at week 8, whenIPT-A reduced the mean score by a difference of 5.42 compared with that ofTAU (t = −3.34; P =.001).

In the random regression analysis to linearize change in HAMD scores,the time scale was transformed by using the natural logarithm of week + 1.Thus, slope estimates approximate change in HAMD score per unit of changein logarithm of week after baseline. Significantly different slopes over timewere found between the 2 treatment groups (treatment × time interaction,F_1,114 = 8.85; P = .004). The IPT-A grouphad a significantly faster rate of improvement relative to the TAU group duringthe 12 weeks. Slope estimates (per natural logarithm of week in treatment)for the 2 treatments ±SE were −4.55 ± 0.54 for IPT-A and−2.29 ± 0.53 for TAU. The rate of improvement for the IPT-A groupon the HAMD was 1.99 U per time more rapid than in the TAU group. The averagedifference between treatments was 4.10 points at week 8 and 5.21 points atweek 12.

Adolescents receiving IPT-A compared with those receiving TAU reportedsignificantly greater improvement in overall functioning on the C-GAS (F_2,60 = 5.03; P = .04) at week 12 (Table 2). A repeated-measures analysisto determine changing patterns with time was not performed because there wasonly a single assessment made after baseline, at week 12.

At week 12, adolescents who received IPT-A as compared with those whoreceived TAU were rated as significantly less ill (F_2,60 = 4.89; P = .03) (Table 2).At week 12, adolescents who received IPT-A compared with those who receivedTAU had significantly greater improvement in their symptoms (F_1,60 =5.3; P = .03). The effect size for CGI severity ofillness was 0.48 (95% confidence interval [CI], 0.15-0.81) and for improvementwas 0.59 (95% CI, 0.24-0.94).

At week 12, adolescents who received IPT-A compared with those who receivedTAU reported significantly greater improvement in dating (F_2,60 =4.7; P = .03) and overall social functioning (F_2,60 = 7.0; P = .01) and a trend for greaterimprovement in family functioning (F_2,60 = 2.8; P = .10) on the SAS-SR. The repeated-measures ANOVA showed that patternsin variation of mean overall social functioning during the 12 weeks of treatmentwere significantly different between the 2 conditions (treatment × timeinteraction, F_3,167 = 4.40; P = .003)(Figure 3). Post hoc pairwise comparisonsof treatment means at individual time points showed that significant differencesemerged at week 8. The IPT-A condition reduced the mean score by a differenceof 0.24 points compared with that of the TAU condition (t = −2.24; P = .03).

The random regression analysis showed that rates of improvement overtime in the 2 treatment groups were significantly different (treatment ×time interaction, F_1,114 = 9.80; P = .002).Although overall social functioning scores on average decreased linearly duringthe 12 weeks for IPT-A, these scores in TAU on average remained relativelyunchanged. Slope estimates for the 2 treatments ±SE were −0.06± 0.01 for the IPT-A and −0.02 ± 0.01 for TAU, which translatesinto an average difference between treatments of 0.27 points at week 8 and0.45 points at week 12.

Adolescents treated with IPT-A received a mean of 10.5 therapy sessionsin 16 weeks compared with a mean of 7.9 therapy sessions for those who receivedTAU (P = .01). Part of the beneficial effects ofIPT-A could be due to a greater dosage of therapy. Conventional methods foradjusting for covariates (number of sessions) were not appropriate for thisstudy, because only 3 IPT-A patients received fewer than 8 sessions comparedwith 12 TAU patients. The 3 IPT-A patients dropped out of treatment, whereasonly 2 of the 12 TAU patients with fewer than 8 sessions dropped out. Thissuggests that dropout status and number of sessions are confounded in IPT-A.To explore this volume further, we compared the subset of TAU patients whohad at least 8 sessions (n = 17) with the IPT-A patients with at least 8 sessions(n = 30). There was a significant reduction in HAMD scores among the IPT-Apatients compared with the TAU patients in this subgroup (week 12, t = 2.55; P = .01). Of the 29 TAU patients,17 had at least 8 sessions, whereas 12 had fewer than 8 sessions. We exploredthe possibility that among the TAU patients, those who had more than 8 sessionsmay be more severely ill than those with fewer sessions, a finding that wouldresult in the previous comparison being biased in favor of IPT-A. These 2subgroups of TAU were compared on their baseline, week 12, and week 16 HAMDscores. These analyses showed that there was no significant association betweennumber of sessions and baseline HAMD scores (18.47 vs 18.17; P = .87) for TAU patients. In addition, no statistically significantdifferences were found between the number of sessions and HAMD scores at weeks12 and 16 in these 2 TAU groups. In fact, those with at least 8 sessions showedsomewhat greater improvement than those with fewer than 8 sessions.

Differential treatment effects were found on the HAMD and C-GAS by ageat study entry (12-14 vs 15-18 years). A significant correlation was foundbetween age and severity of illness measured on the CGI (r = 0.33; P = .008) and age and global functioningmeasured on the C-GAS (r = 0.36; P = .004). Specifically in the older group, IPT-A compared with TAUwas more effective as demonstrated by a significantly lower HAMD score atweek 12, with an average difference of 7.6 points (P =.006), and a significantly higher C-GAS score at week 12, with an averagedifference of 6.6 points (P = .02), controlling forthe baseline scores. The difference between the 2 treatments was minimal andnonsignificant in the younger group. The interaction between age and treatment(IPT-A vs TAU) in the adolescents 15 years and older was marginally significantto predict HAMD scores (P = .055) and statisticallysignificant to predict C-GAS score (P = .03) at week12.

When baseline severity of illness was measured by the HAMD (using ≥22as the top quartile of severity), IPT-A patients compared with TAU patientswithin the high-severity subgroup had significantly lower HAMD scores (differenceof −9.3; P = .04) and higher C-GAS scores (differenceof 13.6; P = .04) at week 12. The group differencewas statistically nonsignificant in the low-severity HAMD subgroup. Baselineseverity of impairment was measured by C-GAS using a median split rather thanquartiles because of the limited variance in the C-GAS scores. Patients treatedwith IPT-A compared with those treated with TAU within the high-impairmentC-GAS subgroup had significantly lower HAMD scores (difference, 7.8; P = .02) and higher C-GAS scores (−13.8; P = .006) at week 12. The difference between the 2 groups was statisticallynonsignificant in the low-impairment subgroup.

At week 16, a telephone follow-up interview was conducted with all patients(n = 62) after completion of any remaining sessions between weeks 12 and 16.No attempt was made to contact 1 patient who did not attend any therapy sessions,and this patient was not included in the analysis. Assessments administeredby telephone by the independent evaluators included the HAMD and C-GAS. Thestudy findings were maintained at week 16. On the HAMD, the IPT-A patientscompared with the TAU patients reported significantly greater decreases indepressive symptoms (6.9 vs 10.6; P = .04), an effectsize of 0.51 (95% CI, 0.003-1.02), and a trend toward greater improvementin overall functioning on the C-GAS (F_2,59 = 3.72; P = .06).

Much has been written about the gap between clinical practice and clinicalresearch. As Weisz and colleagues³⁴ have stated,the crucial question is whether community clinicians can actually benefitfrom the findings of outcome research, and an important task is to study theapplication of research therapy to community settings. The present study helpsto bridge the gap by successfully applying modified research therapy proceduresfor the training of school-based clinicians to deliver an evidence-based therapy.Specifically, the results demonstrate the effectiveness of IPT-A comparedwith TAU for the treatment of adolescent depression in school-based healthclinics in impoverished urban communities in New York City.

Adolescents who receive IPT-A have significant reductions in depressionsymptoms and impairment by clinician report and self-report and significantimprovement in overall and specific domains of social functioning. Depressedadolescents treated with IPT-A rather than TAU improved faster and were significantlybetter after 8 consecutive weekly sessions of IPT-A, and this is not merelya dosage effect. The largest treatment effects occurred in the older and/ormore severely depressed adolescents, consistent with findings in the adultIPT literature.³⁵ This finding suggests thatmilder depression in younger adolescents can be more easily treated with supportivepsychotherapy, whereas more severe depression is more effectively treatedwith a structured treatment specifically targeted for adolescent depression.The current findings extend treatment effects observed in carefully controlledclinical trials with depressed adolescents^17- 18 totreatment in school-based health clinics, and are an important first stepin the study of the transportability of treatments from the laboratory tothe clinic.

Although other research efforts have tried to disseminate treatmentson a larger scale³⁶ and study their effectivenessin comparison with community care,³⁷ this studyis one of the first to assess the effectiveness of an evidence-based treatmentfor adolescent depression as delivered by school-based clinicians. Concernsabout implementation have been expressed regarding necessary adaptations intraining and supervision, although little has been written about actual implementation.The results of this study are significant, despite the small sample size,in that they support the ability to implement evidence-based practices outsideof university settings. Although organizational factors such as patient volume,staffing patterns, referral sources, and resource availability influenceddelivery of the intervention and necessitated protocol modifications, it wasstill possible to implement IPT-A in a clinically effective manner. Modificationsincluded abbreviated therapist training, monitored treatment adherence withoutaudiotapes or videotapes, shortened session duration, flexibility in schedulingowing to school calendar constraints, acceptance of heterogeneous diagnosticprofiles, and parental notification of treatment when otherwise unnecessaryin the school clinics.³⁸

The study involved training on-site clinicians rather than bringingin expert clinicians. A frequently asked question is what is needed to changeclinician practice. Training was abbreviated compared with efficacy studiesthat have extensive therapist training.³⁹ Aftera brief didactic session, clinicians received 1 hour of supervision per 4to 5 cases per week without the use of videotapes or audiotapes. It may bepossible to streamline this training as graduate programs move toward trainingin evidence-based treatments and clinicians come to the job with greater knowledgeof the treatments.

Although medication is seen as a frontline treatment for adolescentdepression,⁴⁰ it is not easily accessed inthe school clinics, and minority families are generally reluctant to acceptantidepressants.⁴¹ More recently, concernshave developed regarding the effects of antidepressant medications for childrenand adolescents.⁴² With little physician time,clinics are unable to offer medication and must refer adolescents to nearbyhospital centers. The 4 students who were prescribed antidepressant medicationshowed poor adherence with their regimens. An important future study wouldbe to assess the effectiveness of medication alone and combined with psychotherapyfor the treatment of more severely ill students.

This study has several limitations. The predominantly female sampledoes not permit the power to test for sex differences in treatment outcome.The size of the study sample also precludes analyses of treatment effectivenessin the middle vs high school students and other school or environmental effectsthat may facilitate or impede successful implementation of the intervention.At one school, therapists, but not students, were randomly assigned; however,this school contributed only 7 cases. Delivery of TAU may have been alteredby our observation of its practice, and this likely made TAU more effectivethan normal. Attempts to monitor the number of visits attended in the TAUcondition led to increased outreach and contact with the adolescents. Finally,the current study represents a relatively small-scale demonstration. An importantchallenge ahead is to expand this dissemination effort to a larger numberand wider range of school-based clinics.

Although there was little attrition of students randomized to treatment,there was a significant rate of refusal to participate (60.0%) before baselineassessment. The most common reasons for refusal were related to research requirements,particularly parental notification and consent rather than a refusal to receivepsychotherapy. At present, adolescents receive treatment in the clinics withoutparental knowledge. Many wish to retain this privacy and therefore declineto participate. Outside the study, we would expect better participation intreatment; however, such research requirements are likely to continue to hinderthe conduct of research in this setting. We have no reason to believe thatIPT-A is less effective with adolescents who declined participation for reasonsrelated to technical research demands, although we are unable to evaluatethe effectiveness of IPT-A with those who refused study participation.

The present study demonstrates a successful implementation of a briefpsychosocial intervention (IPT-A) delivered by community-based cliniciansin urban school-based health clinics serving minority students. The IPT-Aintervention is a viable and effective model for organizing treatment deliveryin this population and setting. The intervention can be learned and deliveredby social workers and psychologists who work in these clinics. We found consistentreduction of depression symptoms and improved social and general functioningwith IPT-A. This setting provided a rigorous test of the effectiveness ofIPT-A. Similar or more robust intervention effects would likely be achievedin school settings with more clinical resources. Given the critical role thatschool clinics play in the provision of care for adolescents,⁴³ thesefindings are of public health importance for improving the delivery of mentalhealth care to underserved, depressed adolescents.

Corresponding author and reprints: Laura Mufson, PhD, New York StatePsychiatric Institute, 1051 Riverside Dr, Unit 24, New York, NY 10032 (e-mail: MufsonL@childpsych.columbia.edu).

Submitted for publication May 23, 2003; final revision received December24, 2003; accepted January 20, 2004.

This study was funded by grant 6 HS5 SM52671-02-1 from the SubstanceAbuse and Mental Health Administration, Rockville, Md, and grant 5 P30 MH60570-03from the Psychotherapy Core of the National Institute of Mental Health ChildPsychiatry Intervention Research Center, Rockville.

We thank Hector Bird, MD, and Jami F. Young, PhD, for their thoughtfulcontributions to the project and readings of earlier drafts; Cora de Leon,MSW, for her hard work; the patients and the school-based health clinics fortheir participation in this project; and Marilyn Camacho for her tirelessefforts on behalf of the project and manuscript preparation.