Effect of Collaborative Care vs Usual Care on Depressive Symptoms in Older Adults With Subthreshold Depression: The CASPER Randomized Clinical Trial

Key Points

Question  Is collaborative care an effective method to reduce depressive symptoms in older people with mild depression?

Findings  In the CASPER randomized trial of 705 participants aged 65 years or older with subthreshold depression, those randomized to a collaborative care intervention had lower depression scores as measured by the Patient Health Questionnaire 9-item survey at 4-month follow-up compared with usual care.

Meaning  Among older adults with subthreshold depression, a collaborative care intervention reduced depressive symptoms at 4-month follow-up compared with usual care. The long-term efficacy of this intervention is unclear.

Importance  There is little evidence to guide management of depressive symptoms in older people.

Objective  To evaluate whether a collaborative care intervention can reduce depressive symptoms and prevent more severe depression in older people.

Design, Setting, and Participants  Randomized clinical trial conducted from May 24, 2011, to November 14, 2014, in 32 primary care centers in the United Kingdom among 705 participants aged 65 years or older with Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) subthreshold depression; participants were followed up for 12 months.

Interventions  Collaborative care (n=344) was coordinated by a case manager who assessed functional impairments relating to mood symptoms. Participants were offered behavioral activation and completed an average of 6 weekly sessions. The control group received usual primary care (n=361).

Main Outcomes and Measures  The primary outcome was self-reported depression severity at 4-month follow-up on the 9-item Patient Health Questionnaire (PHQ-9; score range, 0-27). Included among 10 prespecified secondary outcomes were the PHQ-9 score at 12-month follow-up and the proportion meeting criteria for depressive disorder (PHQ-9 score ≥10) at 4- and 12-month follow-up.

Results  The 705 participants were 58% female with a mean age of 77 (SD, 7.1) years. Four-month retention was 83%, with higher loss to follow-up in collaborative care (82/344 [24%]) vs usual care (37/361 [10%]). Collaborative care resulted in lower PHQ-9 scores vs usual care at 4-month follow-up (mean score with collaborative care, 5.36 vs with usual care, 6.67; mean difference, −1.31; 95% CI, −1.95 to −0.67; P < .001). Treatment differences remained at 12 months (mean PHQ-9 score with collaborative care, 5.93 vs with usual care, 7.25; mean difference, −1.33; 95% CI, −2.10 to −0.55). The proportions of participants meeting criteria for depression at 4-month follow-up were 17.2% (45/262) vs 23.5% (76/324), respectively (difference, −6.3% [95% CI, −12.8% to 0.2%]; relative risk, 0.83 [95% CI, 0.61-1.27]; P = .25) and at 12-month follow-up were 15.7% (37/235) vs 27.8% (79/284) (difference, −12.1% [95% CI, −19.1% to −5.1%]; relative risk, 0.65 [95% CI, 0.46-0.91]; P = .01).

Conclusions and Relevance  Among older adults with subthreshold depression, collaborative care compared with usual care resulted in a statistically significant difference in depressive symptoms at 4-month follow-up, of uncertain clinical importance. Although differences persisted through 12 months, findings are limited by attrition, and further research is needed to assess longer-term efficacy.

Trial Registration  isrctn.org Identifier: ISRCTN02202951

Depression is the second leading cause of disability worldwide.¹ One in 7 older people meet criteria for depression.² Effective therapeutic strategies are needed in older people with depressive symptoms who also have comorbid diseases and impaired quality of life.^2-4Quiz Ref ID There is limited research about older people with mild depressive disorders who have insufficient levels of depressive symptoms to meet diagnostic criteria (called subclinical, subthreshold, or subsyndromal depression)⁵ but also reduced quality of life and function.⁴ Subthreshold depression is a risk factor for more severe depressive illness.⁶ With increased interest in preventive approaches to depression,⁷ trials have focused on adults with subthreshold disorders.⁸ The focus of the current research was older people with low-level depressive symptoms.

Prescription of antidepressants is not recommended as a first-line treatment for subthreshold depression because there is little evidence that they are effective. Psychological therapies may be more appropriate, but higher-intensity forms of therapy such as cognitive behavioral therapy, interpersonal therapy, or behavioral activation are generally reserved for people with more severe disorders.⁹ Collaborative care involves the provision of care by a trained case manager under the principles of chronic disease management.¹⁰ A meta-analysis reported that collaborative care is effective for people with depression meeting diagnostic thresholds,¹¹ but its ability to prevent depression in high-risk populations has not been examined. The objective of this study was to evaluate the effect of a collaborative care intervention in older people with subthreshold depression in a UK primary care setting.

The Collaborative Care for Screen Positive Elders (CASPER) trial was a pragmatic, multicenter, 2-group, parallel randomized clinical trial. Older adults with lower-severity depressive symptoms recruited in primary care were randomized to receive either usual care from their primary care physician or a collaborative care intervention in addition to their usual primary care.

This study was approved by the NHS Leeds East Ethics Committee on September 28, 2010 (10/H1306/61). Participants aged 65 years or older from 32 primary care practices in the North of England gave written informed consent between March 2011 and July 2013. Prior to the definitive trial, an internal pilot was conducted with 100 participants in which the age of participants was 75 years or older. The age cut point was reduced from 75 years to 65 years following advice from the trial steering committee to align the trial population with an age-specific demarcation in the United Kingdom wherein patients aged 65 years or older are treated by older persons’ mental health services.

Potential participants were identified by postal questionnaire and were eligible if they reported depressive symptoms on a standardized brief 2-item case-finding tool (the Whooley questions: [1] Over the past month have you been bothered by feeling down, depressed, or hopeless? [2] Over the past month have you been bothered by having little or no interest or pleasure in doing things?)¹² and were found to have subthreshold depression according to Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) (DSM-IV) criteria using the Mini International Neuropsychiatric Interview (MINI version 5.0),¹³ conducted by telephone by researchers trained by clinical coinvestigators. The participants’ primary care physicians excluded patients with known alcohol dependency, psychosis, recent suicidal risk, significant cognitive impairment, recent bereavement, or terminal illness on clinical grounds (based on their knowledge of the patient). People receiving psychological therapy were excluded. Participants receiving antidepressants remained eligible. Ethnicity was recorded by self-report to describe the diversity of participants.

Participants were allocated to collaborative care or usual care by a computer in a 1:1 ratio by simple randomization without blocking or stratification. Treatment allocation was concealed from study researchers at the point of recruitment using an automated computer data entry system, administered remotely by the York Trials Unit, which used a computer-generated code. None of the participants, primary care practices, or clinicians were blinded to treatment allocation. Researchers who assessed outcomes were blinded to treatment allocation.

Participants in the intervention group received a program of collaborative care designed specifically for older people with subthreshold depression and to accommodate long-term physical health problems (see Pasterfield et al¹⁴ for an extended description). Quiz Ref IDCollaborative care was delivered by a case manager with a background in mental health nursing or a graduate psychologist in 8 weekly sessions.¹⁵ The intervention consisted of telephone support and session-by-session symptom monitoring to track treatment response. The case managers were supervised and corresponded with the primary care physician or intervention psychiatrist where necessary. The first session was delivered face to face and subsequent sessions were delivered via telephone. A computer system was used to monitor care, and supervision of case managers was offered by S.G., D.B., J.D., D.E., D.F., and D.M. Participants were offered a structured program of behavioral activation.¹⁶ This brief psychological intervention addressed the behavioral deficits of depression such as avoidance of social interaction and the absence of rewarding activities.¹⁷ Participants already prescribed antidepressants were encouraged to continue medication, and primary care physicians were encouraged to initiate medication only in response to increasing depressive symptoms.⁹

Participants randomized to the control group received usual primary care. They received no additional care to the usual primary care management of subthreshold depression.

The primary outcome was self-reported severity and symptoms of depression assessed by the 9-item Patient Health Questionnaire (PHQ-9)¹⁸ at 4 months. The PHQ-9 has a score range of 0 (least depressed) to 27 (most depressed). Secondary exploratory outcomes included PHQ-9 depression severity at 12 months and, at 4 and 12 months, dichotomized depression according to “depression diagnosis,” defined using an optimum cut point of a PHQ-9 score ≥10, which has been validated as a sensitive and specific criterion for DSM-IV major depressive disorder.¹⁸ We also studied a limited range of secondary exploratory outcomes of decrements and comorbidities associated with depression, including health-related quality of life measured by the 12-item Short Form survey (SF-12) mental component scale and physical component scale (score range, 0 [lowest level of health] to 100 [highest level of health])¹⁹; anxiety measured by the Generalized Anxiety Disorder Assessment (GAD-7; score range, 0 [no anxiety] to 21 [severe anxiety])²⁰ and self-reported prescribed mental health medication. Data were also collected on somatoform problems measured by the 15-item PHQ (PHQ-15; score range, 0-28 [higher scores indicate greater physical impairment; item on menstrual problems was excluded])²¹; and psychological resilience measured by the 2-item Connor-Davidson Resilience Scale (CD-RISC2), which has a score range of 0 to 8, with higher scores indicating greater psychological resilience,²² but these were not statistically evaluated. Questionnaires were administered by researchers blinded to treatment allocation. Resource use was ascertained from primary care records, and quality-adjusted life-years were measured using the EuroQol (EQ-5D-3L), with a score range of 0 (death) to 1 (perfect health),²³ although the cost-effectiveness analysis is not reported here. Death during follow-up was prespecified as an outcome and measured via linkage to mortality data from the UK Office for National Statistics. In our trial protocol, we indicated that the number of falls would be recorded, but we decided not to collect these data before the first participant was randomized.

To detect a small to medium standardized minimum effect size of 0.3 (based on a meta-analysis of previous trials of collaborative care,²⁴ corresponding to approximately 1.3 PHQ-9 score points) with 80% power and a 2-sided .05 significance level, 352 patients were required (176 in each group). Although this was an individually randomized trial, the sample size was inflated to account for potential clustering around case managers and potential loss to follow-up of 25%. The final sample size to be recruited was 658 patients, 329 in each group.

Patients were analyzed as part of the group to which they had been randomized (intention to treat) using a linear mixed model if they had valid primary outcome data at 4- or 12-month follow-up and baseline PHQ-9 and SF-12 physical component score data. The primary analysis model included as fixed effects time (4 or 12 months), treatment group, and time × treatment interaction, adjusting for PHQ-9 depression at randomization and physical/functional limitations (SF-12 physical component score) at baseline. The primary end point was the estimate of the intervention effect at 4 months.

To quantify the effect of the groupings by case managers, these were modeled separately in each treatment group. Additional variables associated with PHQ-9 scores at 4 months (age, sex, GAD-7, PHQ-15, mental health medication use, and history of depression based on MINI responses) were included as covariates in the primary analysis model. To investigate the effect of missing data on the treatment effect, any baseline variables associated with nonresponse at 4-month follow-up (ie, no valid PHQ-9 score) were identified and included as covariates in the primary analysis model. In light of the observed differential dropout, a multiple imputation model of the primary analysis was additionally included.

Analysis of secondary outcomes was exploratory, and no adjustments for multiple testing were applied. The estimate of the effect of the intervention on PHQ-9 scores at 12 months was extracted from the primary analysis model. For the dichotomous outcome of depression diagnosis at follow-up (PHQ-9 score ≥10), data were analyzed by logistic regression with Poisson regression models to calculate adjusted relative risks (RRs). For other exploratory continuous secondary outcomes (SF-12, GAD-7), statistical analyses were conducted using a similar model to the primary analysis. Other collected data (PHQ-15 and CD-RISC2) were summarized descriptively.

Analyses used 2-sided significance at the .05 level. All analyses were conducted in Stata version 13.1 (Stata Corp). The statistical analysis and reporting of the trial followed the Consolidated Standards of Reporting Trials guidelines.²⁵ The study followed a trial protocol (Supplement 1) and all analyses followed a prespecified statistical analysis plan (Supplement 2, with a description of any amendments to the protocol).

A total of 37 134 patients from 38 primary care centers were invited by letter to participate between March 2011 and May 2013. Of 6693 patients who consented to be contacted and provided information about depressive symptoms, 4259 were excluded (largely on the basis of negative results on the 2-item depression screen), and 2434 patients were assessed for eligibility by the MINI diagnostic interview. Seven hundred five patients (29%) were identified to have subthreshold depression and were randomized into the trial (58% female; mean age, 77 [SD, 7.1] years); 344 were allocated to collaborative care and 361 to usual care. The remaining patients were classified either as fulfilling no criteria for depression (n = 1558 [64%]) or as meeting criteria for major depressive disorder (n = 171 [7%]) (Figure 1). Primary outcome data (PHQ-9 depression severity at 4 months) were available for 586 patients, representing a loss to follow-up of 16.9% (23.8% in the collaborative care group and 10.2% in the usual care group). At 12-month follow-up, 519 patients were retained, with a loss to follow-up of 26.4% at 12 months (31.7% in the collaborative care group and 21.3% in the usual care group).

The 2 groups were comparable at baseline (Table 1). The median depression severity score in both groups based on the PHQ-9 was 7, consistent with a low severity depression.¹⁸ Prescription rates of antidepressants were low at baseline (collaborative care, 10%; usual care, 14%).

Collaborative care was delivered by 18 case managers (mean case load of 19.1 randomized patients). Participants received on average 6 sessions (median, 7; range, 1-15) over 7 to 8 weeks, of which 2 were delivered face to face and 4 were delivered by telephone. The average session duration was half an hour.

At 4-month follow-up (primary outcome), there was a between-group difference of −1.31 PHQ-9 score points (95% CI, −1.95 to −0.67; P < .001) equivalent to a standard effect size of 0.3 in favor of collaborative care. At 12-month follow-up (exploratory outcome), a between-group difference remained (−1.33 PHQ-9 score points; 95% CI, −2.10 to −0.55; P = .001). Table 2 shows full results including sensitivity analyses.

In secondary exploratory analyses, the proportion of participants with a new depression diagnosis (PHQ-9 score ≥10) was lower in the collaborative care group at 12-month follow-up but not at 4-month follow-up (at 4 months: 17.2% vs 23.5%; difference, −6.3% [95% CI, −12.8% to 0.2%]; RR, 0.83 [95% CI, 0.61-1.27]; P = .25; at 12 months: 15.7% vs 27.8%; difference, −12.1%; 95% CI, −19.1% to −5.1%; RR, 0.65 [95% CI, 0.46-0.91]; P = .01) (Table 2).

In secondary exploratory analyses, the RR of being prescribed antidepressants was not different between the 2 groups at 4-month follow-up (RR, 0.73; 95% CI, 0.51-1.04; P = .08) or at 12-month follow-up (RR, 0.84; 95% CI, 0.60-1.19; P = .33).

In secondary exploratory analyses of the SF-12, the physical health of patients was better for collaborative care at 4- and 12-month follow-up (at 4 months: mean score difference, −2.83 [95% CI, −4.03 to −1.62]; standard effect-size d = 0.2; P < .001; at 12 months: mean score difference, −1.67 [95% CI, −3.06 to −0.27]; d = 0.1; P = .02) (Figure 2 and Table 2). The mental health of patients was better for collaborative care at 4 and 12 months in exploratory analyses (at 4 months: mean score difference, −1.88 [95% CI, −3.29 to −0.47]; d = 0.2; P = .009; at 12 months: mean score difference, −2.15 [95% CI, −3.70 to −0.59]; d = 0.2; P = .007) (Figure 2 and Table 2).

Significant exploratory between-group differences in anxiety symptoms (GAD-7) were observed in favor of collaborative care at 4-month follow-up (mean score difference, −1.08; 95% CI, −1.64 to −0.52; d = 0.3; P < .001) and at 12-month follow-up (mean score difference, −1.01; 95% CI, −1.61 to −0.42; d = 0.2; P = .001) (Figure 2 and Table 2).

Mean scores for physical health problems (PHQ-15) at 4 months were, for collaborative care, 7.4 (SD, 3.99), and for usual care, 9.1 (SD, 4.28). At 12 months, PHQ-15 mean scores were, for collaborative care, 8.1 (SD, 4.03), and for usual care, 9.2 (SD, 4.53). No statistical comparison was made (Figure 2).

Resilience (CD-RISC2 ) mean scores at 4 months were , for collaborative care, 6.2 (SD, 1.71), and for usual care, 5.7 (SD, 1.71); at 12 months, CD-RISC2 mean scores were, for collaborative care, 6.1 (SD, 1.71), and for usual care, 5.7 (SD, 1.77). No statistical comparison was made.

A total of 23 participants died during 12-month follow-up, 5 patients in the collaborative care group (1.5% of randomized patients) and 18 patients in the usual care group (5.0% of randomized patients), which was statistically significant (χ²₁ = 6.97; P = .008). All causes of death and their potential relatedness to the trial treatment were assessed independently and presented to the data monitoring and ethics committee in line with procedures for serious adverse events. Approximately 81% of deaths were categorized as unrelated to treatment and 18% as unlikely to be related to treatment. The data monitoring and ethics committee assessed that the recorded causes of death could not be reasonably attributed to either the intervention or control treatment. The exploratory observed group difference in mortality was therefore treated as a chance result.

Quiz Ref IDThe main finding from this randomized trial is that a collaborative care intervention reduced the PHQ-9 score at 4-month follow-up compared with usual care. In secondary exploratory analyses, the PHQ-9 score was also lower at 12-month follow-up in the collaborative care group, but high attrition rates reduce confidence in this result. For populations with case-level depression, a successful treatment outcome has been defined as 5 points on the PHQ-9.²⁶ This magnitude of benefit was not observed in either group of the trial when comparing scores before and after treatment, although this result would be anticipated given the lower baseline PHQ-9 scores in populations with subthreshold depression. The between-group difference was 1.31 PHQ-9 points, which is a small to medium effect size according to Cohen^9,27 and consistent with the Cochrane meta-analysis of collaborative care¹¹ but is not large when judged against a clinical difference of 5 points advocated in more severe disorders.²⁸ In additional secondary exploratory analyses, collaborative care prevented the onset of depression diagnosis (as ascertained by the PHQ-9) by 12.1% at 12 months but was nonsignificant at 4 months in an exploratory analysis.

Treatment of older people with subthreshold depression has been insufficiently studied. The results of the trial are consistent with other research regarding collaborative care for depression in older people.^11,16,29 However, few studies to date have examined the effectiveness of collaborative care in older people and explored the ability to prevent lower-severity depression symptoms from progressing to the point of case level depression. Quiz Ref IDBehavioral activation is a relatively simple type of treatment that could be taught to and administered by a wide range of health care professionals.³⁰

We noted a reduction in mortality for people who received collaborative care, but independent case-by-case review of deaths was not thought to be linked to the intervention. This was an unexpected finding and one that deserves further study in future trials. It is possible that people randomized to the control group were more unwell and that this might have influenced the primary outcome. However, this seems unlikely because the numbers who died were small in relation to the size of the trial and there was evidence of balance between groups at baseline on measures of symptoms severity and quality of life. In addition, all secondary outcomes are also exploratory.

Quiz Ref IDThere were several limitations to the study. First, ascertainment of depression diagnosis was exploratory and did not use a standardized diagnostic interview.³¹ Second, there was differential retention and attrition between the 2 groups. Participants who wished to discontinue the collaborative care intervention fully withdrew from the trial at the same time, including follow-up. It is possible that participants who withdrew may have presented a different outcome profile to those who continued, which may have biased the treatment effect. Third, there was no follow-up after 12 months.

Among older adults with subthreshold depression, collaborative care compared with usual care resulted in a statistically significant difference in depressive symptoms at 4-month follow-up, of uncertain clinical importance. Although differences persisted through 12 months, findings are limited by attrition, and further research is needed to assess longer-term efficacy.

Corresponding Author: Simon Gilbody, PhD, University of York, Health Sciences, Heslington, York, Yorkshire YO10 5DD, England (simon.gilbody@york.ac.uk).

Author Contributions: Dr Gilbody and Ms Keding had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Gilbody, Adamson, Atherton, Bailey, Ekers, Foster, Hamilton, Hewitt, Holmes, Mitchell, Pervin, Richards, Spilsbury, Woodhouse, McMillan.

Acquisition, analysis, or interpretation of data: Gilbody, Lewis, Adamson, Birtwistle, Bosanquet, Clare, Delgadillo, Ekers, Gabe, Gascoyne, Haley, Hamilton, Hargate, Hewitt, Keding, Lilley-Kelly, Meer, Overend, Pasterfield, Spilsbury, Traviss-Turner, Trépel, Ziegler, McMillan.

Drafting of the manuscript: Gilbody, Atherton, Bailey, Ekers, Overend, Pervin, Trépel, Woodhouse.

Critical revision of the manuscript for important intellectual content: Gilbody, Lewis, Adamson, Bailey, Birtwistle, Bosanquet, Clare, Delgadillo, Ekers, Foster, Gabe, Gascoyne, Haley, Hamilton, Hargate, Hewitt, Holmes, Keding, Lilley-Kelly, Meer, Mitchell, Pasterfield, Richards, Spilsbury, Traviss-Turner, Trépel, Ziegler, McMillan.

Statistical analysis: Gabe, Hewitt, Keding, Trépel.

Obtained funding: Gilbody, Adamson, Ekers, Hewitt, Holmes, Spilsbury, McMillan.

Administrative, technical, or material support: Gilbody, Lewis, Atherton, Bailey, Birtwistle, Bosanquet, Clare, Delgadillo, Ekers, Foster, Haley, Hamilton, Lilley-Kelly, Mitchell, Overend, Pasterfield, Pervin, Traviss-Turner, Trépel, Woodhouse.

Supervision: Gilbody, Lewis, Adamson, Bailey, Bosanquet, Delgadillo, Ekers, Hamilton, Mitchell, Richards, Spilsbury, Traviss-Turner, Trépel, Ziegler, McMillan.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Group Information:Trial steering committee members: Mike Beckett, York MIND, York (now at the Retreat, York); David Geddes, NHS North Yorkshire & York and Clifton Medical Practice, York (now at the NHS Commissioning Board, Leeds); Alison Layton (chair), NEYNL CLRN and Research and Development, Harrogate District Hospital (now at Yorkshire and Humber CRN); Waquas Waheed, Lancashire Care NHS Foundation Trust, Preston (now at National Primary Care Research and Development Centre, University of Manchester, Manchester); members of the CASPER Trial Management Group. Data monitoring and ethics committee: David Kessler (chair), University of Bristol, Bristol; Judy Leibowitz, Camden PCT, London (now at Camden and Islington NHS Foundation Trust); Stephen Walters, Medical Statistics and Clinical Trials, Health Services Research, ScHARR, University of Sheffield.

Funding/Support: This project was funded by the UK National Institute of Health Research Health Technology Assessment Programme (project number 08/19/04).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Disclaimer: The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the HTA, NIHR, NHS, or Department of Health.

Additional Contributions: We acknowledge the contribution of Sarah Nutbrown, MSc (University of York), who developed site-specific procedures for the trial and David Torgerson, PhD (York Trials Unit, University of York), who advised on trial design. Ms Nutbrown was employed by the University of York using funds allocated to the CASPER trial and Dr Torgerson’s salary was partially supported by the CASPER grant award. We thank the users of mental health services and caregivers who were part of the advisory group established at the end of the pilot phase; their insights and understanding helped improve the relevance and readability of the study documentation.