Ciechanowski P, Wagner E, Schmaling K, Schwartz S, Williams B, Diehr P, Kulzer J, Gray S, Collier C, LoGerfo J. Community-Integrated Home-Based Depression Treatment in Older AdultsA Randomized Controlled Trial. JAMA. 2004;291(13):1569-1577. doi:10.1001/jama.291.13.1569
Author Affiliations: Departments of Psychiatry and Behavioral Sciences (Dr Ciechanowski), Health Services (Drs Ciechanowski, Williams, and LoGerfo and Mss Schwartz and Kulzer), Biostatistics (Dr Diehr), and Medicine (Dr LoGerfo), Health Promotion Research Center (Drs Williams and LoGerfo and Mss Schwartz and Kulzer), and School of Pharmacy (Dr Gray), University of Washington, Seattle; W. A. MacColl Institute for Healthcare Innovation, Center for Health Studies, Group Health Cooperative, Seattle (Dr Wagner); University of Texas at El Paso (Dr Schmaling); and Chronic Disease Prevention and Healthy Aging, Prevention Division—Public Health, Seattle and King County (Dr Collier).
Context Older adults with social isolation, medical comorbidity, and physical
impairment are more likely to be depressed but may be less able to seek appropriate
care for depression compared with older adults without these characteristics.
Objective To determine the effectiveness of a home-based program of detecting
and managing minor depression or dysthymia among older aldults.
Design and Setting Randomized controlled trial with recruitment through community senior
service agencies in metropolitan Seattle, Wash, from January 2000 to May 2003.
Patients One hundred thirty-eight patients aged 60 years or older with minor
depression (51.4%) or dysthymia (48.6%). Patients had a mean of 4.6 (SD, 2.1)
chronic medical conditions; 42% of the sample belonged to a racial/ethnic
minority, 72% lived alone, 58% had an annual income of less than $10 000,
and 69% received a form of home assistance.
Interventions Patients were randomly assigned to the Program to Encourage Active,
Rewarding Lives for Seniors (PEARLS) intervention (n = 72) or usual care (n
= 66). The PEARLS intervention consisted of problem-solving treatment, social
and physical activation, and potential recommendations to patients' physicians
regarding antidepressant medications.
Main Outcome Measures Assessments of depression and quality of life at 12 months compared
Results At 12 months, compared with the usual care group, patients receiving
the PEARLS intervention were more likely to have at least a 50% reduction
in depressive symptoms (43% vs 15%; odds ratio [OR], 5.21; 95% confidence
interval [CI], 2.01-13.49), to achieve complete remission from depression
(36% vs 12%; OR, 4.96; 95% CI, 1.79-13.72), and to have greater health-related
quality-of-life improvements in functional well-being (P = .001) and emotional well-being (P = .048).
Conclusions The PEARLS program, a community-integrated, home-based treatment for
depression, significantly reduced depressive symptoms and improved health
status in chronically medically ill older adults with minor depression and
Clinically significant depression affects 15% to 20% of elderly individuals
in the United States.1,2 Older
individuals are less likely than younger adults to have major depression3 but have comparable or higher rates of less severe
depressive disorders such as dysthymia and minor depression. Dysthymia is
a chronic depressive syndrome persisting for at least 2 years.4 According
to the Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition (DSM-IV) Appendix of Criteria
Sets and Axes Provided for Further Study, minor depression is depressed mood
and/or significant loss of interest, plus 1 to 3 other depressive symptoms
present nearly every day for at least 2 weeks, occurring in the absence of
dysthymia.4(pp719-721) Minor depression
and dysthymia both lead to significant disability.5,6 Accumulating
evidence from primary care settings suggests that both syndromes may respond
to medications and nonpharmacological therapies,7,8 though
less is known about treatment of these disorders, particularly minor depression,
in community settings.
Although higher rates of depression exist among medically ill,9- 12 socially
isolated,11,13 homebound,14- 16 or functionally impaired12,17 older adults, such characteristics
may also lead to inadequate recognition and treatment of depression. Even
among less functionally impaired older adults attending primary care, few
depressed patients receive guideline-level depression treatment18,19 such
as medications or psychotherapy.20 Untreated
depression in elderly persons negatively influences physical functioning,21 adaptation to medical illness,22 and
quality of life23 and can be associated with
increased morbidity and mortality,24,25 treatment
nonadherence,26 increased health care costs,18,26 and lower satisfaction with care.23,27
Because of the relatively high prevalence of less severe forms of depression
among elderly persons, we set out to add to the evidence base on treating
these disorders in community settings by conducting a randomized controlled
trial comparing a program for treating minor depression and dysthymia, the
Program to Encourage Active, Rewarding Lives for Seniors (PEARLS), with usual
care in medically ill, low-income, mostly homebound older adults. Since home-based
treatment can reduce depression among socially isolated, chronically ill seniors,16 we created a home-based intervention focusing on
problem-solving treatment (PST). We hypothesized that patients receiving the
PEARLS intervention would show greater improvements in depression and higher
quality of life compared with patients receiving usual care. We also examined
the effects of the intervention on changes in health care utilization.
This study was conducted in metropolitan Seattle, Wash, and approved
by the University of Washington review committee on human subjects. Between
January 2000 and May 2003, we sought to enroll adults aged 60 years or older
receiving services from senior service agencies or living in senior public
housing with DSM-IV minor depression or dysthymia
diagnostic criteria. Social workers at Aging and Disability Services and Senior
Services of Seattle/King County administered the 2-item Primary Care Evaluation
of Mental Disorders (PRIME-MD) depression screening tool28 during
routinely scheduled visits or telephone calls. We also recruited self-referred
individuals through letters mailed by collaborating agencies to their clients
or residents in affiliated public housing. All potential self-referred or
screen-positive participants were interviewed by trained research associates
using the Structured Clinical Interview for DSM-IV (SCID)29 as a second-level screen. Participants who were eligible
and provided written informed consent completed a baseline interview.
We estimated that the intervention and usual care groups would each
require 64 patients to have a 95% chance of detecting as significant (at a
2-sided, .05 α level) a mean difference of 0.30 in the depression outcome
measure, the Hopkins Symptom Checklist (HSCL) 20, derived from the revised
HSCL-90.30 We planned to enroll 71 patients
in each group, anticipating 10% participant attrition.
Agency social workers identified 1238 potential participants, 20% of
whom completed the SCID, and the majority of the 181 self-referred potential
participants (72%) also completed the SCID (Figure 1). Of 374 potential participants evaluated with the SCID,
224 (60%) had the following exclusion criteria: no depression, 29%; major
depression, 21%; bipolar disorder, 4%; psychosis, 2%; and substance abuse,
0.3%. We used 5-minute recall of 3 items and orientation to year, month, and
date from the Mini-Mental State Examination31 to
screen for cognitive impairment; patients scoring less than 3 of 6 points
(3% of potential participants) were ineligible to participate. Of 150 remaining
potential participants (40%), 12 refused participation and 138 consented to
participate and were enrolled in the study. For all participants who were
excluded because of a diagnosis of major depression, a letter from the study
psychiatrist (P.C.) was sent to the partipant's primary care physician indicating
the diagnosis and suggesting that he/she consider this probable diagnosis
in treatment of the patient.
Using a 50:50 randomization allocation ratio, investigators created
envelopes containing concealed assignment codes assigned sequentially to eligible
patients in blocks of 10 by a research associate. Investigators changed the
allocation ratio to 60:40 after 11 blocks to ensure that the desired sample
of intervention participants was recruited (≥71 participants). To account
for the possibility that clinical severity of recruited participants may have
increased over time, which would lead to greater severity in the treatment
group, we included a dummy variable in all analyses denoting whether patients
were randomized using the 50:50 or 60:40 allocation scheme.
Usual care participants received no additional services, but letters
sent to their regular physicians and social workers reported their depression
diagnosis with recommendations to continue usual care. All participants could
seek additional primary or specialty mental health care. Intervention participants
were scheduled to receive PST by PEARLS therapists (2 male and 1 female master's-level
social workers employed by community agency collaborators). Near the end of
recruitment, a female registered nurse trained in PST replaced 1 of the social
workers for 1 of 72 intervention participants. Therapists thoroughly reviewed
a PST treatment manual,32 completed 8 hours
of lectures, viewed videotapes, and participated in extensive role-playing.
Problem-solving treatment is a skills-enhancing behavioral depression treatment
based on the assumption that problems of daily life cause and maintain depressive
symptoms, and through systematically identifying and addressing these problems,
patients can achieve decreased depressive symptoms. Recent research has shown
it is not necessary for a patient's perception of problem severity to improve
for PST to work.33
Each PEARLS session included selecting from a list of 250 pleasant activities34 to engage in before the next session. The inverse
relationship between pleasant events and depression is an important aspect
of behavioral theories of depression.35
We modified PST sessions to provide greater emphasis on social and physical
activation (S.S. and J.K., PEARLS Research Team, unpublished data, January
2000). The goal of physical activation was to assist patients in developing
a regular physical activity program consistent with national recommendations36 for moderate activity of at least 30 minutes' duration
at least 5 d/wk. Therapists received information on baseline physical, social,
and recreational activities derived from the Community Healthy Activities
Model Program for Seniors (CHAMPS) instrument.37 Physical
activation began during the third or fourth PST session, allowing patients
to develop familiarity with problem-solving skills. The goal of social activation
was to increase patients' interactions outside the home by using a resource
list under the guidance of the therapist. Group activities encouraging peer
support were given highest priority.
Starting in September 1999, each therapist saw 3 to 6 training cases,
and, as with all participants, these participants gave permission for audiotaping
sessions. The PST trainer (K.S.) scored every training audiotape on 9 specific
and 1 global item from 0 (very poor) to 5 (very good). Feedback for improving
adherence to the PST method was provided. Therapists achieving global ratings
of at least 3 (satisfactory) on the majority of sessions with 3 training cases
were certified as PST proficient and permitted to see full study participants.38 A research coordinator sent the PST trainer every
fifth full study tape for adherence coding and feedback. The mean global rating
for 109 sessions was 3.15.
Patients were scheduled for eight 50-minute in-home sessions over 19
weeks, in weeks 1, 2, 3, 5, 7, 11, 15, and 19. Increasing time between later
sessions gave patients more opportunities to practice PST skills. The Patient
Health Questionnaire 9 (PHQ-9)28 was administered
at the beginning of sessions to track depression change. After 19 weeks, therapists
maintained monthly brief telephone contact with patients, during which they
would administer the PHQ-9 and assess patients' use of PST.
All intervention cases were reviewed weekly or biweekly by the study
psychiatrist (P.C.) during depression management team sessions, which all
therapists attended. Therapists carried 3 to 8 active cases and each case
required 5 to 10 minutes of discussion and supervision. For cases lacking
continued improvement after 4 to 5 weeks (most recent PHQ-9 scores >50% of
baseline and continuing or recurring dysphoria or anhedonia), the psychiatrist
contacted patients' primary care physicians to recommend initiating or adjusting
antidepressants and to assess potential medical and substance abuse etiologies
for depression. This is consistent with evidence-based chronic care models
using stepped-care targeted communication between specialists and primary
care physicians or specialists and patients for cases with unremitting symptoms.23,39
The psychiatrist also reviewed medical problems and medications for
all patients, occasionally clarifying details by contacting primary care physicians.
The psychiatrist infrequently called patients, after discussion at team sessions,
to clarify clinical issues (eg, suicidal ideation, diagnostic uncertainty).
The availability of such telephone contacts with physicians or patients optimized
patient safety and increased the comprehensiveness of the intervention.
Outcome assessments were conducted in person at baseline and at 6 and
12 months by trained interviewers not involved in the intervention and included
sociodemographic characteristics and a count of medical conditions. Depression
was assessed using the HSCL-20,25 which is
validated in medical patients and found to be highly reliable.40,41 With
a 0-to-4 severity range, an HSCL-20 cutoff score of 1.72 has previously been
associated with the highest positive predictive value for diagnosis of major
depression in adult primary care patients.42 Since
neuroticism predicts persistence of depression in primary care43 and
an increased risk for late-life depression associated with disability,44 7 NEO neuroticism scale45 items
were administered at baseline.
Health-related quality of life in functional, physical, social, and
emotional well-being domains was assessed at baseline and 12 months using
the Functional Assessment of Cancer Therapy Scale–General (FACT-G).46 The FACT-G is a generic core questionnaire of 27
items for health-related quality-of-life questionnaires targeted to management
of chronic illness and used and validated in cancer, other chronic conditions,
and the general population.47
Health care utilization was assessed in three 6-month periods from 6
months before to 12 months after baseline using the Cornell Services Index48 and was dichotomized by creating indicator-dependent
variables for each period based on having any emergency department visits,
any medical hospitalizations, or at least 5 outpatient visits (median number
of baseline outpatient visits). The proportion of patients with any form of
home assistance (eg, home aid, nursing, or meal delivery) was determined at
Patients were asked whether they used antidepressants during all 3 assessments.
Six- and 12-month depression outcomes included mean HSCL-20 depression scores;
treatment response, defined as at least a 50% decrease in HSCL-20 score from
baseline; and complete depression remission, defined as HSCL-20 score less
than 0.5.49 Costs of providing PEARLS were
based on direct personnel salaries and included the costs of time to conduct
in-person contacts and associated travel time, all therapist and psychiatric
telephone contacts, depression management team sessions, and maintenance of
PST quality by the PST trainer.
Two-tailed t tests and χ2 tests
were used to compare the intervention and usual care groups at baseline. To
evaluate group differences of HSCL-20 and FACT-G domain scores at assessment
points, we used mixed-effects regression analyses.50,51 To
evaluate treatment group differences in depression treatment response, depression
remission, or health care utilization at all assessment points, we used mixed-effects
ordinal regression analyses.52 These procedures
permit inclusion of patients with missing data, ensuring an intention-to-treat
analysis as randomized. Age, sex, NEO neuroticism scale score, dysthymia,
and randomization allocation scheme were fixed covariates for each analysis.
In all models, the intercept was always assumed to be random, and the time
effect (baseline, 6 months, and 12 months or baseline and 12 months) was fixed.
Marginal maximum likelihood methods generated maximum likelihood estimates
for group, time, and group × time interaction effects. Models were evaluated
using a z statistic (maximum likelihood estimate
divided by standard error). In the case of a nonsignificant group ×
time interaction, the models were refit without an interaction.
For regression analyses evaluating depression treatment response and
depression remission, 6 cases were missing outcome data at both 6 and 12 months,
so the method of last observation carried forward was first used,53 ensuring an intention-to-treat analysis. As a sensitivity
analysis, the mixed-effects ordinal regression analysis was run without the
6 cases as a completer analysis, and results were identical.
When significant interaction or group effects were detected in various
models, the following were calculated for each follow-up assessment point:
for HSCL-20 score differences, multiple regression analyses were used to estimate
the adjusted group mean change; and for depression treatment response, depression
remission, or health care utilization, logistic regression was used to obtain
adjusted odds ratios. For the FACT-G subscales, adjusted mean change scores
over the year were computed for each treatment group.
MIXOR, version 2.0, and MIXREG, version 1.2 (Don Hedeker and Robert
D. Gibbons, University of Illinois at Chicago) as well as SPSS, version 11.0
(SPSS Inc, Chicago, Ill) were used for analyses. P<.05
was considered statistically significant.
The sample (Table 1) consisted
mostly of women (79%), and had a mean age of 73 years (SD, 8.5 years). Few
patients (11%) were married or living with a partner, and 72% lived alone.
Forty-two percent belonged to a racial/ethnic minority (African American,
36%; Asian American, 4%; Hispanic, 1%; and American Indian, 1%), 58% had an
annual income of less than $10 000, 42% had an education level below
or equivalent to high school, and 69% received home assistance.
About half of patients (48.6%) had dysthymia and half (51.4%) had minor
depression, with a sample mean HSCL-20 score of 1.3 (SD, 0.5), which is consistent
with subthreshold depression and lower than the score (1.72) previously associated
with major depression in primary care.42 Patients
reported having a mean of 4.6 (SD, 2.1) of 10 chronic medical conditions (median,
5.0; range, 0-10; 4 had no chronic medical conditions). Since the intervention
group had significantly more dysthymia than the usual care group (61% vs 35%; χ21= 9.5; P = .002) and
less neuroticism (NEO neuroticism scale range, 1-5; mean scores, 2.9 [SD,
0.7] vs 3.1 [SD, 0.7]; t136 = 2.1; P = .04), we controlled for these differences in subsequent
analyses. There were no significant baseline differences in cognitive screening
or in health care utilization in the prior 6 months in the domains of mental
health, medical outpatient and emergency department use, medical hospitalizations,
or use of home assistance.
Intervention patients received, during the 19-week active phase, a mean
of 6.6 visits (SD, 2.5; median, 8.0; range, 0-8; 3 had no visits) and, during
the 33-week follow-up phase, a mean of 3.5 telephone contacts (SD, 2.7; median,
5.0 contacts; range, 0-7; 23 patients could not be or asked not to be contacted).
No patients were seen in person by the psychiatrist. However, the psychiatrist
made 52 telephone contacts during the intervention, 37 with patients' physicians
to discuss the following: antidepressant recommendations (19), laboratory
tests (10), cognition (7), other medications (4), alcohol abuse rehabilitation
(2), sleep apnea (1), falls (1), and pain complaints (1). Fifteen antidepressant-related
calls occurred during the active phase and 4 during the follow-up phase and
involved care of 14 patients. Eleven calls (8 in the active and 3 in the follow-up
phases) were recommendations to start an antidepressant, and 8 (7 in the active
and 1 in the follow-up phases) were recommendations to adjust medications.
Nine psychiatrist calls were made to patients to assess study participation
(7), suicidal thoughts (1), and confusion (1). One call to a patient's case
manager addressed the patient's living situation, and 1 call to Adult Protective
Services initiated an assessment of suspected elder abuse.
At baseline, 36% of patients were taking antidepressants with no significant
group differences at baseline and at 6 and 12 months. Seven patients (9.7%)
in the intervention group compared with 4 (6.1%) in the usual care group started
an antidepressant, while 5 (6.7%) in each group stopped an antidepressant
during the study. With a definition of psychosocial interventions as "visits
with a counselor, therapist, psychotherapist, or mental health provider,"
4 usual care patients reported visits during the first 6 months, 3 reported
such visits in both 6-month periods, and 3 reported visits in the last 6-month
For HSCL-20 depression score, there was a significant group ×
time interaction (z = −2.40; P = .02) due to significant group differences in HSCL-20 at 6 months
(t = −4.85; P<.001)
and 12 months (t = −2.20; P = .03) (Table 2). For
depression treatment response and depression remission, there were no significant
group × time interactions (z = −1.91; P = .06 and z = −0.90; P = .37, respectively), so the models were refit, resulting
in significant group effects (z = 4.19; P<.001 and z = 3.00; P = .003, respectively) and nonsignificant time effects (z = −0.45; P = .65 and z = −0.82; P = .41, respectively). Table 2 shows that these significant group
effects are due to the odds of a 50% depression treatment response or of complete
remission being significantly higher for the intervention group at 6 and 12
There were significant group × time interactions for the FACT-G
functional well-being (z = 3.22; P = .001) and emotional well-being (z = 1.97; P = .048) domains at 12 months. For functional well-being,
these results were due to a significant mean change in the intervention group's
scores (0.52; 95% confidence interval [CI], 0.29-0.74) but not in that of
the usual care group (0.09; 95% CI, −0.14 to 0.33) over 12 months. For
emotional well-being, these results were due to a significant mean change
in the intervention group's scores (0.33; 95% CI, 0.14-0.52) but not in that
of usual care group (0.11; 95% CI, −0.09 to 0.31) over 12 months. For
FACT-G social and physical well-being, there were no significant group ×
time interactions or time or group effects.
For emergency department visits and medical outpatient visits there
were no significant group × time interactions (z =
−0.25; P = .80 and z =
−1.29; P = .20, respectively), so models were
refit for main effects; there were still no group effects (z = −1.09; P = .28 and z = −0.18; P = .86, respectively) or
time effects (z = −1.32; P = .19 and z = −1.19; P = .23, respectively). For hospitalizations, there was no group ×
time interaction (z = −0.70; P = .48); with refit models there was no time effect (z = 0.57; P = .57), but there was a trend-level
group effect (z = −1.82; P = .07) (Table 3).
Mean costs per patient of providing the PEARLS intervention were $422
for PST sessions, $28 for follow-up telephone calls, $12 for psychiatric telephone
calls, $87 for psychotherapy quality assurance, and $81 for depression management
team sessions. Total mean cost per patient was $630.
The PEARLS intervention resulted in significantly lower severity and
greater remission of depression compared with usual care at 6 and 12 months.
Six-month outcomes are comparable with those of a multisite trial of elderly
primary care patients with major depression or dysthymia that included PST,
in which more than 65% of intervention patients took antidepressants by 6
months23 compared with less than 40% in PEARLS. A nonsignificant
increase in depression in the intervention group by 12 months may have resulted
from a significant decrease in therapist-guided problem solving and physical
and social activation in these relatively disabled and socially isolated individuals,
though the influence of decreased nonspecific supportive contact with therapists
in the follow-up period cannot be ruled out as contributing. Future studies
may benefit from extra PST sessions, focusing on activation, in patients relapsing
or not remitting during follow-up. On the other hand, a previous study of
younger primary care patients with major depression has shown 6 sessions of
PST over 11 weeks to be effective54; it is
possible that patients responding more quickly to the PEARLS intervention
(in the active phase) may be able to switch to monthly telephone check-ins
earlier in treatment.
Relatively few patients initiated antidepressants and no net increase
in antidepressant use occurred between groups. Adequacy of dosing may have
been better and duration of use longer in intervention patients, but we report
only categorical use of antidepressants over 6-month periods. Ensuring timely
antidepressant refills and providing greater patient and physician education
about dealing with adverse effects and stigma related to antidepressants may
further enhance adequacy in this effectiveness intervention.
Despite the significant effect of the PEARLS intervention, only a third
of patients in the intervention group and 12% in the usual care group experienced
remission. This is similar to other late-life depression studies using antidepressants
and psychotherapy23 and may result from a reduced capacity of such
patients to respond fully to depression treatment. Psychosocial conditions
that increase the risk for depression (eg, low income, disability, social
isolation) are not likely to have changed substantially during the intervention.
Our program resulted in significant improvements in functional well-being
(eg, acceptance of illness, enjoyment of recreational activities) and emotional
well-being (eg, increased satisfaction in coping with chronic conditions)
at 12 months. Despite our focus on physical and social activation, quality
of life in physical and social well-being domains did not differ significantly
between groups, possibly because of significant physical disability and social
isolation in this population.
Recent studies suggest that depression is associated with higher health
care costs and utilization among the elderly. Unützer et al18 have
shown that depression is associated with higher ambulatory medical costs in
older adults over 4 years. A recent study of 9000 elderly patients screened
for a depression trial by Katon et al55 demonstrated
47% to 51% higher total ambulatory and inpatient medical costs in depressed
compared with nondepressed patients after adjustment for chronic medical illness.
We found trend-level group differences, with the intervention group less likely
than the usual care group to report having any medical hospitalizations during
the study period. Depressed patients report more physical symptoms,22 and untreated depression may result in poorer medical outcomes in patients
with cardiovascular disease,56,57 diabetes,58,59 and other chronic illnesses. Such
factors may have contributed to a greater likelihood of hospitalizations in
patients in whom depression was less effectively treated, and this should
be explored further in studies with larger sample sizes.
While our therapists worked for the participating community agencies,
the intervention, including team meetings, was provided as a parallel health
service to usual case management. We believe PEARLS can be disseminated within
community agencies already providing care for isolated, low-income older adults
by adding depression management to established case management. Most social
service agencies have psychiatric and mental health consultants who, with
minimal training, could provide appropriate oversight and quality assurance
of case managers also trained to provide the PEARLS intervention, and deliver
targeted brief telephone communication with physicians and, possibly, with
patients with more severe or complex depression (eg, significant medical or
psychiatric comorbidity or suicidal ideation).
There were several limitations to this study. The sample size was moderate
and limited to 1 urban geographical area. Similar multisite studies in larger,
more representative samples would improve generalizability. Also, in this
study we did not have access to automated health care records, relying instead
on self-reported medical comorbidity and health care utilization, which is
susceptible to social desirability and recall biases. However, use of retrospective
reports of health care utilization has been found to yield valid results,60- 63 and
recall bias may have been reduced by our assessment of categorical use of
any services vs none in the domains of emergency department and hospital utilization.
Future studies using Medicare utilization data or automated pharmacy data
will improve assessment of health care utilization and adequacy of medications
and allow for cost-effectiveness analyses of the PEARLS intervention. Future
studies will also be invaluable in determining the relative contribution of
the intervention's components—PST, behavioral/social activation, pleasant
activity scheduling, antidepressant use, and supportive contact with therapists—in
successfully decreasing depression severity. Finally, despite random assignment,
we had unequal baseline proportions of dysthymia and minor depression, with
intervention participants having a greater proportion of dysthymia at baseline
compared with usual care participants. We controlled for this difference in
baseline dysthymia prevalence in all subsequent analyses.
This is one of the first studies to show that by partnering with community
agencies, it is possible to target and effectively treat depressed, frail,
elderly adults using primarily nonpharmacological treatments such as psychotherapy.
Dissemination of PEARLS within existing community social service programs
has the potential to significantly improve the well-being and function of
depressed older adults served by these programs.