Meyers BS, Sirey JA, Bruce M, Hamilton M, Raue P, Friedman SJ, Rickey C, Kakuma T, Carroll MK, Kiosses D, Alexopoulos G. Predictors of Early Recovery From Major Depression Among Persons Admitted to Community-Based ClinicsAn Observational Study. Arch Gen Psychiatry. 2002;59(8):729-735. doi:10.1001/archpsyc.59.8.729
Copyright 2002 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2002
Twenty years have elapsed since the National Institute of Mental Health Collaborative Depression Study reported on the early course and treatment of major depression within the mental health sector. Using similar methods, an observational study was conducted to assess relationships between initial depression severity, personality dysfunction and other baseline characteristics, subsequent treatment, and 3-month outcomes among persons admitted to public and voluntary sector outpatient clinics, including 1 academic program.
A 2-stage sampling technique was used to recruit subjects (N = 165) diagnosed by the Structured Clinical Interview for DSM-IV, Patient Version, as having a major depression episode. Sociodemographic and clinical characteristics were assessed at admission. Data on treatment and outcome were obtained at 3 months using structured instruments from the Longitudinal Interview Follow-up Evaluation. Logistic regression was used to assess hypothesized predictors of early recovery. Analyses were carried out in the total sample and after dichotomizing subjects by baseline depression severity.
Fifty (30.3%) of the 165 subjects met recovery criteria. Less than half of the subjects (45%) met criteria for adequate pharmacotherapy. Less severe depression, having received adequate antidepressant treatment, female sex, and being married independently predicted early recovery. In the more depressed subgroup, early recovery was associated with female sex. Among less severely depressed subjects, high personality dysfunction scores and being married were significant predictors.
Initial depression severity and receiving adequate pharmacotherapy predict early recovery in individuals with major depression seeking outpatient treatment. A minority of persons receive intensive antidepressant treatment. Less severe personality dysfunction and being married predicts early recovery among persons with less severe depression.
WE REPORT on 12-week recovery and its predictors among patients with major depression following admission to community-based mental health clinics and an academic outpatient program. The study builds on and extends earlier results from the naturalistic National Institute of Mental Health (NIMH) Collaborative Depression Study (CDS)1- 4 in the following 2 principal ways: (1) Recruitment includes admissions to nonacademic clinics. (2) The study is conducted following the widespread use of new classes of antidepressants. The relationship between specific predictor variables, including antidepressant treatment intensity, and recovery 12 weeks following admission is assessed.
Evidence from randomized controlled trials has demonstrated a 65% 6-week response rate in subjects who receive antidepressants compared with approximately35% in those who receive placebo.5,6 Results from efficacy studies stand in contrast to observational data generated by the CDS.1- 4 The CDS demonstrated that intensive somatic treatment prior to admission to the study's academic centers was infrequent.1 Furthermore, the intensity of pharmacotherapy following admission to participating academic centers was both low and unrelated to either the probability of recovery or the duration of depressive episodes.2 The CDS also examined other factors that influenced recovery rates under naturalistic conditions. Acute onset was found to predict early recovery, while protracted index episodes, underlying dysthymia, lower socioeconomic status, and a comorbid Axis I disorder were associated with chronicity.3
The NIMH Treatment of Depression Collaborative Research Program trial indicated that depression severity influences differential responsiveness to pharmacotherapy compared with psychotherapy or medication clinic placebo.7 This 16-week randomized trial demonstrated that 150 mg of imipramine hydrochloride treatment was superior to placebo and 2 manualized psychotherapies among subjects with Hamilton Depression Rating Scale (HDRS)8 scores of less than 20 only. There were no significant differences between placebo and either imipramine or the psychotherapies in the less severely depressed subgroup. Also, subjects with a coexisting personality disorder were less likely to achieve remission HDRS scores of 6 or less,9 a finding consistent with other studies demonstrating that comorbid personality dysfunction worsens the course of depression.10- 13
Data generated from subjects recruited at academic centers may have limited generalizability to patients treated in community settings.14 We report the effects of baseline patient characteristics and treatment intensity on the early course of major depression among outpatients admitted to 1 academic center and 5 community-based mental health clinics. We tested the hypothesis that depression severity and comorbid personality dysfunction would decrease the likelihood of early recovery, while intensive antidepressant therapy would increase the recovery rate. We also investigated the effects of sociodemographic factors, functioning, and type of service site. A secondary analysis was carried out using the HDRS cutoff of more than20 applied in the NIMH Treatment of Depression Collaborative Research Program trial to determine whether this level of severity influenced the effect of identified predictor variables.
Sites were selected from Westchester County, New York, in consultation with the county's commissioner of community mental health (S.J.F.) to maximize socioeconomic and ethnic diversity. Westchester County has approximately 900 000 inhabitants residing in urban, semiurban, suburban, and semirural settings and includes Yonkers, the third largest city in New York State. Six clinics were selected to provide diversity of setting types and patient sociodemographic characteristics. Sites included the outpatient department of a teaching hospital(New York Presbyterian Hospital–Westchester Division), a second voluntary sector hospital-based clinic (Phelps Memorial Hospital Center, Sleepy Hollow), 3 county clinics, and a freestanding voluntary sector clinic. Two county clinics were selected from urban settings with 1 in downtown Yonkers. Clinic administrators assisted by coordinating the scheduling of new patient evaluations so that a research assistant would be present when the patient arrived.
Patients 18 years and older admitted consecutively to participating clinics during the 2 years from October 1, 1995, through December 31, 1997, were invited to participate in a 2-stage screening process described previously.15 Written informed consent was obtained prior to the administration of research instruments. Patients with scores of 16 or higher on the Center for Epidemiological Studies–Depression scale16 or meeting DSM-IV17 criteria for major depression on the Mini International Neuropsychiatric Interview18 were administered the Structured Clinical Interview for DSM-IV, Patient Version (SCID-P),19 to establish the presence of a current unipolar major depression. Cognitive functioning was assessed with the Mini-Mental State Examination.20 The SCID-P interviews were administered by college graduates or master's-level students who had completed a 1-week SCID-P training course. A clinical psychologist(P.R.) with 2-year research experience who had also completed the course reviewed each SCID-P.
Admitted patients who met criteria for major depression without having exclusion criteria were invited to participate in the longitudinal study. Exclusion criteria were a Mini-Mental State Examination score of less than24 or a history of alcohol or other substance abuse within the past month. Patients who had episodes of mania or psychosis, had another Axis I disorder other than comorbid dysthymia, and participated in a controlled treatment trial were excluded.
Of the 1180 outpatient admissions approached, 1106 (94%) consented to the initial screening. Of these, 689 (62%) screened positive for major depression and 455 of these (66%) consented to the SCID-P interview. Twenty-seven percent of those interviewed (n = 125) had an exclusion criterion and an additional120 (26%) failed to meet criteria for a major depression. The remaining 210 subjects (46%) received the full battery of baseline assessments and entered the follow-up phase.
Three-month data were obtained on 165 (79%) of the initial longitudinal sample, with the remaining 45 (21%) refusing to participate or otherwise lost to follow-up. Subjects without 3-month assessments did not differ significantly on baseline sociodemographic or clinical variables from those with 3-month data.
Baseline measures for subjects with major depression included the 17-item HDRS,8 the Global Assessment of Functioning Scale from DSM-IV,17 the Chronic Disease Score21 to assess medical comorbidity, the 36-Item Short-Form Health Survey,22 and the Multilevel Assessment Inventory23 to assess social support and independent functioning, and the Duke Social Support and Stress Scale24 to assess social support and independent functioning. The first 3 subscales of the 47-item version of the self-report Inventory of Interpersonal Problems25 were administered to screen for the presence of personality dysfunction. The Inventory of Interpersonal Problems is a self-report measure that inquires about personality attributes that are presumed to be enduring without providing a time frame. A cutoff of greater than 1.1 on these subscales has been validated against a semistructured interview to indicate the presence of a DSM-IV personality disorder among patients with major depression.25 Soon after the study began, the 5 general anxiety items from the Clinical Anxiety Scale26 were added to assess comorbid baseline anxiety as a predictor of early course. Data on recent use of medical and mental health services were collected using the Cornell Services Use Index27 (This scale is available from the authors on request). Subjects were provided nominal reimbursement for participation in each phase of the study.
As in the CDS, follow-up assessments used measures from the Longitudinal Interval Follow-up Evaluation (LIFE).28 Subjects were interviewed by telephone to determine the course of depression, treatment provided, and adverse effects experienced.
The LIFE's Psychiatric Status Rating (PSR) uses a 6-point scale to quantify the severity of depressive symptoms relative to full major depression criteria. Application of the PSR uses the LIFE method of dividing the 3-month time frame by significant landmarks (eg, holidays, birthdays, and life events) and then determining the subject's level of symptoms in relation to the specified dates. Recovery was defined as being asymptomatic (PSR ratings of 1) or having 1 or more symptoms of no more than a mild degree (PSR of 2) over the preceding2 weeks. The LIFE's 8-week PSR duration of recovery criterion was shortened to 2 weeks for the purpose of identifying cases of recovery occurring within the first 12 weeks. An investigator (B.S.M.) with experience using these measures trained and supervised the research assistants.
Antidepressant intensity was quantified initially on a 5-point scale adapted from the LIFE's Composite Antidepressant Scale, with ratings added for newer classes of antidepressants (scale available from us on request). In the revised scale, a score of 3 was assigned for 125 mg or more of desipramine hydrochloride, 50 mg or more of nortriptyline hydrochloride, 150 mg or more of venlafaxine hydrochloride, 50 mg or more of sertraline hydrochloride, 20 mg or more of fluoxetine hydrochloride, 20 mg or more of citalopram hydrobromide, or 20 mg or more of paroxetine hydrochloride.
The LIFE's Composite Antidepressant Scale score was collapsed to assess intensity of antidepressant treatment. Adequate treatment was defined as a weekly mean LIFE Composite Antidepressant Scale score of 3 or more for a minimum of 4 consecutive weeks receiving 1 or more antidepressants and inadequate treatment defined as having received antidepressants below this dose or duration. Subjects who did not receive any antidepressant treatment comprised a third group.
Delineation of dose and duration of antidepressant use applied significant anchor dates to subdivide the 3-month follow-up period.28 Patient report was used to determine the dose and duration of antidepressant treatment and compliance with medication. The use of psychotherapy was assessed using self-report data elicited through the Cornell Services Use Index.
Data analyses were carried out to construct a parsimonious and clinically informative model describing the likelihood of recovery at 12 weeks based on the hypothesized predictors. Bivariate analyses were also carried out for potential associations between early recovery and both baseline sociodemographic and clinical variables. The bivariate analyses used χ2 tests with continuity adjustment for 2-sample comparisons or the Fisher exact test for cells with expected frequencies of fewer than 5 for categorical variables and t tests for continuous data. Data are reported as mean (SD). Two-tailed P values of less than .05 were considered statistically significant in the bivariate analyses.
To reduce the number of factors eligible to be used for construction of the predictive model, we added only those variables with nominal 2-sided P values of .10 or less from the bivariate analyses into the hierarchical logistic regression after entering the hypothesized predictors, followed by age and sex. Adjusted odds ratios (ORs) and 95% confidence intervals(CIs) were calculated for each predictive variable. All values are reported as mean (SD).
The 165 subjects with 3-month follow-up data had a mean age of 44.6(17.2) years and 13.4 (2.8) years of education. Sixty-four percent of the sample was female (n = 106) and 30% were members of racial minority groups(n = 66). One hundred eight subjects (66%) received an antidepressant.
Fifty subjects (30.3%) met criteria for recovery at 3 months. An additional52 subjects (31.5%) had not recovered but no longer met criteria for a major depression at the 3-month assessment. There were not significant associations between sociodemographic or site characteristics and early recovery (Table 1). Recovery seemed to be associated with the clinical variables of lower baseline HDRS scores (t163 = 3.38, P = .001), lower anxiety scores (t132 = 2.45, P = .02), and having interpersonal problem scores below the cutoff for personality dysfunction (χ21 = 4.81, P = .03). The 21 subjects with concurrent dysthymia constituted 13% of the sample. Although the presence of dysthymia was associated with a lower3-month recovery rate (8% vs 15%), this difference did not reach statistical significance (χ21 = 0.90, P= .34). Early recovery was associated with better baseline functioning as indicated by higher scores on the Global Assessment of Functioning Scale (t163 = −2.06, P= .04) and the 36-Item Short-Form Health Survey subscales of physical functioning(t129.7 = −3.37, P = .001) and pain perception (t160= 2.62, P = .01).
One hundred eight (65%) of the 165 subjects received antidepressants;74 (69%) received a new class of antidepressant. The proportions of subjects treated with an antidepressant from a new class and those treated with a tricyclic antidepressant that met criteria for adequacy were highly comparable (68% vs 70%). Overall, only 74 (45%) of the 165 subjects met criteria for an adequate dose for 4 weeks. Of these 74 subjects, 28 (38%) recovered, compared with16 (28%) of 57 subjects who did not receive any antidepressants and only 6(18%) of 34 who received antidepressants at inadequate doses (χ22 = 4.7, P = .09) (Table 2). Self-reported compliance with pharmacotherapy (χ21 = 3.80, P = .004) was associated with recovery among subjects for whom antidepressants were recommended. There were no differences between recovered and nonrecovered subjects in whether psychotherapy had been recommended and number of therapy sessions or remaining in treatment over the 12 weeks.
There was a significant difference between the academic and nonacademic sites in the distribution of antidepressant treatment intensity classified as adequate, inadequate, or none (χ22 = 8.92, P = .02). A greater proportion of subjects treated at the academic site met criteria for adequate antidepressant treatment (53% vs 36%, χ21 = 4.41, P = .04). Nevertheless, type of service site was not associated with differences in early recovery(Table 1).
Hierarchical logistic regression for the 163 subjects with complete data sets was conducted. Hypothesized predictors of baseline depression severity, meeting criteria for personality dysfunction, treatment adequacy, age, and sex were entered in the model first. More severe depression (OR = 0.89; 95% CI, 0.81-0.97; P = .01) was associated with a decreased likelihood of early recovery (Table 3). Classifying antidepressant intensity into adequate, inadequate, or none demonstrated that subjects who had adequate treatment were more than 3 times more likely to recover than subjects who received inadequate treatment (OR = 3.2; 95% CI = 1.1-9.5; P = .04). Being married (OR = 2.4;95% CI = 1.1-5.3; P = .03) was the only other baseline characteristic or treatment variable that remained significant in the logistic regression. There were no significant interactions between regression terms in the model. Because most of the 42 geriatric subjects (91%) were treated at the academic site that emphasizes antidepressant therapy, we repeated the model using dummy variables for site and for age 60 years and older. The resulting model was also significantly associated with early recovery. We then compared the 2 models and found that the addition of older age and site variables did not change the model significantly (Wald χ26 = 7.4, P = .29). The more parsimonious model is reported.
Based on the absence of a drug-placebo difference in patients with HDRS scores of less than 21 in the NIMH Treatment of Depression Collaborative Research Program trial,14 we repeated the logistic regressions after subdividing subjects into those with HDRS scores of more than 20 (n= 72) and those with scores of 20 or less (n = 101). Of predictors in the original model, only female sex remained a significant predictor of early recovery in the subgroup of more severely depressed subjects (OR = 1.1; 95% CI = 1.2-106.5; P = .04). Despite an increased OR, adequate antidepressant treatment was not a significant predictor in the smaller subgroup of more severely depressed subjects (OR = 4.7; 95% CI = 0.76-29.2; P = .10). In the subgroup of less depressed subjects, subjects who met criteria for personality dysfunction had a decreased likelihood of early recovery (OR = 0.32; 95% CI = 0.13-0.79; P= .01) while those who were married were more likely to recover (OR = 3.6;95% CI = 1.4-9.3; P = .01).
Only 30% of 165 outpatient admissions with major depression achieved the recovery criterion of 2 weeks with no more than minor symptoms over the subsequent 12 weeks. Despite evidence that psychiatrists are 1.6 times more likely to write antidepressant prescriptions than a decade ago,29 only 65% of the 165 outpatients received an antidepressant during the 3 months following a clinic admission and only 44% met criteria for 4 weeks at an adequate dose. Of subjects who had antidepressants prescribed, most (69%) did meet criteria for adequacy.
In contrast to CDS findings,3,4 intensive pharmacotherapy was associated with early recovery. Classifying subjects as having received adequate antidepressant treatment, inadequate intensity, and no antidepressants demonstrated a significant independent effect on recovery rates. These are the first data of which we are aware demonstrating that meeting a criterion for adequately intense antidepressant therapy predicts early course recovery under naturalistic conditions. Follow-up observational data from the CDS did demonstrate that remaining at the level of somatic treatment associated with recovery reduced the rate of early recurrence.30
The naturalistic design precludes determining relationships between dose and treatment response. Although the effect of adequate antidepressant treatment decreased to a trend level in the more severely depressed subgroup of 61 subjects, the smaller sample size limited power to demonstrate a significant effect. Alternatively, patients who received higher doses would include both subjects who had their doses raised because of nonresponse and those who recovered as a result of having received higher doses. The possible overrepresentation of treatment-resistance among subjects who received adequate pharmacotherapy would decrease the likelihood of finding a treatment effect. The same selectivity process would act among less severely depressed subjects.
The finding of decreased early recovery in men, including among the more severely depressed subjects, adds to a conflicting literature using different methods.31 Selection factors may contribute to these findings in that men who seek treatment in community outpatient clinics may do so because of a more persistent form of depression.
The finding of increased early recovery among married persons overall and particularly among those with mild depression has not been reported previously. In an epidemiological sample, persons who were either married or single had lower rates of persistent depression than individuals who had had a change in marital status due to separation, divorce, or widowhood.31 In contrast, marital status did not predict time to recovery in the clinical sample of the CDS.4 It may be that the largely in patient composition of the CDS subjects (75%), including 15% who met criteria for psychotic depression, washed out a possible effect of marital status, particularly among persons with less severe depression.
Depression severity affected the association between screening positive for personality dysfunction and early recovery, with the significant effect observed only in the subgroup of mildly depressed subjects (P = .01). In the absence data on personality dysfunction data at 3 months, we are unable to assess whether this association on a self-report measure was a function of being depressed. Other factors including negative life events32 and the presence of anxiety disorder symptoms33 have been demonstrated to delay recovery from major depression under naturalistic conditions. Although bivariate analyses suggested that subjects who recovered early had less anxiety at admission(P = .02), severity of anxiety did not independently predict recovery in a model using the 134 subjects with baseline anxiety scores. Similarly, subjects who did not recover had lower baseline scores in multiple functional domains, but baseline functioning did not have an effect in the logistic regression that controlled for other factors. Thus, the lower baseline36-Item Short-Form Health Survey physical functional and pain perception scores in nonrecovered subjects may have been accounted for by a factor such as depression severity in the regression.
This study was limited by the small sample size relative to the number of factors that may influence early recovery. By testing for the effect of hypothesized predictors before entering other factors identified by the bivariate associations, the analysis limited the likelihood of a type I statistical error. Also, the numbers of subjects in specific predictor categories (eg, ethnic minority or having comorbid dysthymia) may have been too small to demonstrate a significant effect. Furthermore, factors such as physical functioning that were marginally associated with early recovery in the regression (data not shown) require further study to more definitively investigate their predictive effects.
Because of the numerous patient, provider, and setting factors that might influence the administration of psychotherapy and differences in forms of this treatment, we are unable to determine whether psychotherapy administered alone or in combination with pharmacotherapy contributed to early recovery. Similarly, we are unable to assess the effects of different forms of insurance benefits on the treatments provided or early course.
Inventory of Interpersonal Problems and HDRS assessments were not repeated at 3 months. Therefore, we are unable to determine the extent to which personality dysfunction was a feature of the depressed state or whether identified predictors of recovery also predicted net decreases in depression severity.
Finally, interpretation of these data must be considered in light of the definition of recovery used. The CDS focused primarily on the longer-term course of depression and defined recovery as 8 consecutive weeks of a PSR of 1 or 2 rather than the 2-week criterion used in this study. The low recovery rate identified using the less rigorous 2 consecutive weeks of no or minimal symptoms is noteworthy.
The finding that under one third of individuals seeking treatment for major depression within the mental health sector recover within 3 months speaks to the need for longer and more intensive treatment. The frequency of recovery is markedly below rates reported in efficacy trials.5,6 The higher recovery rates reported from efficacy trials than naturalistic studies may result from screening out patients who are unlikely to respond from controlled trials. Although the use of systematically applied inclusion criteria suggests that the patient sample was similar clinically to subjects recruited into controlled studies, subjects participating in antidepressant trials are more likely to have a forced titration up to adequate dosages.
These data suggest that identifiable sociodemographic, clinical, and treatment factors can influence the early course of depression and are consistent with the need to improve our management of depression in community-based mental health settings. Initiatives designed to reverse the documented undertreatment of major depression in primary care34 will be achieved more easily after undertreatment within the mental health sector is corrected.14,35
Submitted for publication May 3, 2000; final revision received September26, 2001; accepted October 1, 2001.
This study was supported by grant MH53816 from the National Institute of Mental Health, Bethesda, Md.
We thank Mark Olfson, MD, for his valuable consultations during this project. We also thank Tara DiDomenico, Claire Mackay, Brooke Myers, Lauren Picone, and Mark Russakoff, MD, for their help.
Reprints: Barnett S. Meyers, MD, Department of Psychiatry, New York–Presbyterian Hospital, Westchester Division, 21 Bloomingdale Rd, White Plains, NY 10605(e-mail: firstname.lastname@example.org).