Abstract
Background Late-life depression is associated with chronic illness, disability, and a poor prognosis. Primary care management may be in need of improvement.
Aim To compare the effects of an intervention programme that aims to improve the identification, diagnosis, and treatment of depression in patients aged ≥55 years with the effects of usual care.
Design of study Cluster randomised controlled trial.
Setting General practices in the Netherlands.
Method Trained GPs performed the intervention and their practice assistants conducted the screenings. Patients were screened with the 15-item Geriatric Depression Scale (GDS-15) and given a consultation with the GP who diagnosed depression with the mood module of the Primary Care Evaluation of Mental Disorders (PRIME-MD). Antidepressant treatment was proposed. Primary outcomes were measured with the Montgomery Åsberg Depression Rating Scale (MÅDRS). Trained independent research assistants performed independent evaluations in both arms.
Results Eighteen practices (23 GPs) were allocated to the intervention and 16 practices (20 GPs) to usual care. From June 2000 to September 2002, 3937 patients were screened; 579 patients had a positive score on the GDS-15, 178 had major depression, of whom 145 participated in the trial. MÅDRS scores for the intervention group dropped from 21.66 at baseline to 9.23 at 6 months, and the usual care group from 20.94 at baseline to 11.45 at 6 months. MÅDRS scores decreased during the year in both arms. For the intervention group, these scores increased between 6 and 12 months.
Conclusion The programme resulted in lower MÅDRS scores in the intervention group than in the usual care group, but only at the end of the intervention, at 6 months after baseline.
INTRODUCTION
Late-life depression is associated with chronic illness, disability and, frequently, a poor prognosis.1 In many countries, studies have been planned and undertaken to improve the identification and treatment of patients with major depression in primary care. Several interventions have been tested in randomised controlled trials, but most have limited effects.2–4 Screening may improve patient outcomes to some extent when it is coupled with system changes that help to ensure adequate treatment and follow-up,5 but different reviews come to different conclusions.6 Collaborative care — that is, structured care in which non-medical specialists (such case managers usually have a nursing background) have a greater role than is usually the case now in augmenting the care — appears to be the best candidate for improving the outcomes of depression in the short and longer term.3,7
As a result of comorbidity issues and competing demands in the surgery, countries with a strong primary care service, such as the UK and the Netherlands, have a need for systems that can be operated within the other conflicting demands of primary care, whereas other countries might base policy on a more specialised service.8 Many studies have looked at younger adults, but few have addressed older adults.
In one of the first randomised studies that was conducted in the US, a multifaceted intervention increased the identification and treatment of late-life depression, but there was no improvement in the severity of the depression or the disability.9 A second large US study reported positive results of a collaborative care intervention on many aspects of depression and wellbeing in older people, compared with usual care;10 however, these results were from a US managed-care setting, which makes it difficult to compare the results with a European primary care setting. A third US study supported the effectiveness of an intervention specifically aimed at reducing suicidal ideation, regardless of the severity of the depression, in older patients in primary care,11 to the extent that it reduced mortality.12 In the fourth, and perhaps most relevant, feasibility study in the UK, older patients with major depression in the intervention group were somewhat less likely to suffer from this condition at follow-up compared with those who had had usual care (odds ratio [OR] 0.32, 95% confidence interval [CI] = 0.11 to 0.93, P = 0.036).13
The aim of the study was to understand how a programme based on disease management concepts for older adults with major depression would compare with usual care in a setting in which usual care and access to it is well developed, as is the case in Dutch primary care. The main aim of the present study was to compare the effects of a disease management programme standardising the identification, diagnosis, and treatment of late-life depression in general practice with the effects of usual care.
METHOD
The background and design of this study have been published previously,14 as well as the economic results,15 but a short summary is presented here.
Setting and recruitment
GPs in the eastern region of West-Friesland in the Dutch province of North Holland were invited to participate in a randomised controlled trial comparing the effects of an intervention programme aiming to improve the identification, diagnosis, and treatment of depression in older people, with the effects of usual care. Cluster randomisation was done at practice level and performed by an independent statistician.
GPs were recruited during a meeting of the regional Continuous Medical Education (CME) section covering GPs in the region. Those who did not attend were sent a letter of invitation and were also contacted by telephone.
As the intervention was intended to be implemented by the GPs, randomisation took place at GP level (that is, in the practice building). Cluster randomisation was chosen to prevent the carrying over of results from the intervention group to the usual care group.
How this fits in
Primary care management of late-life depression needs improvement. In a practice randomised controlled trial, the effects of an intervention to improve identification, diagnosis, and treatment of depression in patients aged 55 years or over (18 practices), were compared with the effects of usual care (16 practices). Observed depressive symptoms scores halved during the year in both arms. The programme resulted in somewhat lower depressive symptom levels, but only after 6 months.
Patients
The inclusion criteria for patients were being aged ≥55 years and having a current major depressive episode. Any previous episode of depression that occurred more than 6 months ago was permitted. Exclusion criteria were psychosis, bipolar depression, severe social dysfunctioning, inability to communicate in the Dutch language, alcohol or drug misuse, cognitive impairment, and current use of antidepressants. Those patients who were ineligible were treated routinely by the GP or referred to specialist care.
Intervention group
The practice assistant carried out screening for depression in the waiting room. All patients aged ≥55 years were asked to complete the 15-item version of the Geriatric Depression Scale (GDS-15).16 The GDS-15 score ranged from 0–15, with higher scores indicating a greater likelihood of depression. Patients with a positive screening result (GDS score of ≥5) were asked to participate in a diagnostic interview administered by the GP. The GP used the mood module of the Primary Care Evaluation of Mental Disorders (PRIME-MD) to assess symptoms of major depression as classified in the Diagnostic and Statistical Manual of Mental Disorders (DSM®) IV.17 The GP diagnosed major depression clinically in patients with at least five depressive symptoms, including mood or activity. Trained research assistants administered the PRIME-MD interview to the control subjects at baseline and at follow-up. If there was a discrepancy between the two assessments and a patient had a positive PRIME-MD according to the interviewer and not according to the GP, the patient was referred back to the GP for diagnostic assessment.
Usual care group
To determine the effectiveness of the disease management programme it was necessary to have a control group that represented current actual practice, that is, a group comprising patients receiving usual care. In the usual care group the practice assistants presented the GDS-15 to all patients aged ≥55 years who visited the practice. The GPs and the practice assistants were unaware of the cut-off score, which was determined at the research centre. The researchers did not inform GPs about which of their patients were included in the study.
A research assistant administered the PRIME-MD diagnostic interview with control subjects who had a positive score on the GDS-15. A positive score indicated that a patient was eligible for participation in the study. Patients were asked not to inform their GPs that they were participating.
Intervention
Treatment offered by GPs was based on the guidelines for the treatment of depression issued by the Dutch College of General Practitioners (NHG).18 The treatment consisted of education and information, drug therapy, and supportive counselling. Education and information about depression consisted of information on its causes, treatment (in particular, drug therapy), prognoses, and self-help. The patients had to return once every 2 weeks during the first 2 months, and then once a month for a period of 4 months. The total treatment period lasted for 6 months.
Drug therapy consisted of the prescription of 20 mg of paroxetine once daily. This drug was chosen because, when the study started, it was the most commonly prescribed antidepressant in the Netherlands.19 During supportive counselling the patient and GP together identified one specific problem that bothered the patient. A practical problem was chosen, because it was not usual care to include more reflective forms of psychotherapy. It was planned that, in any encounter, specific attention should be paid to generic counselling. The intensity of the counselling was low. A generic type of activity scheduling (to enhance pleasurable activities) was recommended to the GPs that is probably as effective as problemsolving treatment.20 At the time of the trial (2002–2003), it was not yet possible to implement problem-solving treatment in Dutch primary care because there were no experienced trainers. The activity-scheduling concept was also used in the IMPACT study, but there it was combined with problem-solving treatment.10
Instruments used to measure outcomes
Primary outcome measures were the Montgomery Åsberg Depression Rating Scale (MÅDRS)21 and the PRIME-MD scores. The MÅDRS, which is a patient interview, measured the severity of symptoms and the effect of treatment over time. Sub-groups of MÅDRS scores (50% symptom reduction and a MÅDRS score of <10) were also assessed. Furthermore, recovery from depression, defined as the absence of a PRIMEMD diagnosis of major depression, was assessed four times: at baseline (after the diagnostic interview), and after 2, 6, and 12 months. These PRIME-MD measures were additional primary outcome measures. The total PRIME-MD score was also analysed. The Clinical Global Impression (CGI) scale was also used as an additional primary outcome;22 these data are reported in the online version of this article.
Secondary outcome measures were: the SF-36 Mental and Physical Composite Scales,23 subjective wellbeing, satisfaction, quality of life (visual analogue scale of the EuroQol), and activities of daily life. At baseline and after 12 months the scores for the Mini-Mental State Examination and one Diagnostic Interview Schedule item (are you depressed now?) were also recorded. The scores for the GDS-15 were assessed at all measurement points. A parallel article reports on the cost-effectiveness analysis.15
The measurements, which were repeated after 2, 6, and 12 months, were performed by research assistants who had attended one specific training session. The research assistants were blinded with regard to the treatment group to which the patients were allocated.
Statistical analysis
The calculation of the power was related to the primary outcome measures. Power analyses showed that, using continuous outcome measures, 70 patients were needed in each group to be able to detect a medium-sized treatment effect (Cohen's δ or ‘effect size’ of 0.50) in conventional analyses.24 The principal analysis of the primary outcome measures consisted of an analysis of variances based on intention to treat. Time-trend analyses were performed to determine whether there were differences in the results at any of the measurements points. The differences described apply to those patients who agreed to participate.
To account for the levelled structure of the data and the cluster randomisation procedure, multilevel analysis was performed on the MÅDRS primary outcomes and the additional PRIME-MD primary outcome variable.25 The levels were GP, patient, and measurement point. All models were specified in such a way that it was possible to determine whether at any measurement point there were differences in change from baseline between the treatment groups.
According to the study protocol, it was intended to base the analysis on matched pairs of GPs, however because of the great variety in GP characteristics and the cluster randomisation procedure, GPs were matched on the basis of propensity scores.26–30
Propensity scores were corrected for bias in treatment assignment using a large set of observed confounders and their interactions. In this case, it showed that the randomisation had not been completely successful, possibly due to the small number of practices that were randomised. In particular, some patient characteristics were unequally distributed over the two treatment groups. As such, all tests were corrected for propensity by including the propensity score as a covariate in the analysis of variance and multilevel models. As all tests were corrected for propensity, adjusted means for the primary and secondary outcomes are also reported. The baseline means are not adjusted. The statistical analyses were performed with SPSS (version 12).
RESULTS
Recruitment
The recruitment procedure yielded a total of 43 GPs (out of a total of 101) from 34 practices (out of a total of 70) who were willing to participate. There were 18 practices (including 23 GPs) allocated to the intervention group and 16 (including 20 GPs) to the usual care group (Figure 1). The inclusion of patients started in June 2000 and ended in September 2002. In total, 3937 patients were screened to achieve the required number of patients for the study. The total number included 145 patients with major depression.
Figure 1 Flowchart of the trial.
The uptake of the screening and diagnostic procedures was positive. Of the 95 patients with major depression in the intervention group, 70 agreed to participate in the study (74%), and of the 83 patients with major depression in the usual care group, 75 agreed to participate (90%). This may reflect the willingness of the patients and GPs to accept the diagnosis of depression and its subsequent treatment in the intervention group. In the control group no such acceptance was required, but the patients did have to agree to a diagnostic interview.
Patient characteristics
The characteristics of the patients who participated are presented in Table 1. The majority of the patients were female. In total, 83% of the patients had experienced a previous episode of depression. There were statistically significant differences in characteristics between the patients in the intervention group and the usual care group, which indicates that the randomisation procedure had not been completely successful. These differences were accounted for in the analyses by using a propensity score as a covariate (note that no covariables that are of obvious substantive interest as a confounder, for example, sex, are included in the propensity score).
Table 1 Patient characteristics.
Follow-up
After 2 months 141 patients were still in the trial, 126 were still in the trial after 6 months, and 123 patients completed the 1-year follow-up. A total of 22 patients dropped out for various reasons (Figure 1). There were no differences between the two study groups with regard to the proportion who dropped out. The 95% CI for the difference in follow-up of 9% between intervention group (56/70) and usual care group (67/75) was –3 to 22%.
Intervention
During the study period, 95% of patients in the intervention group and 94% of patients receiving usual care had at least one contact with their GP. Forty-six (66%) patients in the intervention group and eight (11%) patients receiving usual care received some form of specific mental health care (antidepressant medication or referral) during followup. Twenty-eight (40% patients in the intervention group received antidepressant treatment for at least 6 months, as recommended by Dutch depression guidelines.20
Primary outcomes
Table 2 presents the results of the analysis of variance.
Table 2 Results of multilevel analysis (mean ± standard error) on the primary outcome variables MÅDRS and PRIME-MD.a
MÅDRS
Baseline values for the MÅDRS were 21.66 ± 2.86 in the intervention group, but a steady decline in the scores for this scale occurred during treatment and follow-up, but scores went up slightly from 9.23 to 10.80 after 6 months. After 2 months the mean scores were 19.56 ± 3.32, at the end of the treatment (after 6 months) the mean scores were 9.23 ± 2.84, and at the end of the follow-up period the scores were 10.80 ± 2.85. In the group of patients receiving usual care the baseline values for the MÅDRS were 20.94 ± 2.48. In this group there was a steady decline in the scores during treatment and follow-up. After 2 months the mean scores were 19.58 ± 3.49, while at the end of the treatment, after 6 months, the mean scores were 11.45 ± 2.52, and at the end of the follow-up phase the scores had further declined to 10.09 ± 2.50. Trend analysis revealed statistically significant differences after 6 months between the MÅDRS scores in the intervention group and the group of patients receiving usual care (effect size δ = 0.28). At 12 months MÅDRS scores were lower for the usual care group than the intervention group.
To gauge the robustness of these findings, two additional analyses were applied to the MÅDRS scores: an analysis of patients with a score of <10 indicating recovery and an analysis of 50% symptom reduction (Appendix 1). Recovery was higher at 6 months in the intervention group than in the control group: 48% versus 27%; risk difference 21% (95% CI = 0.04 to 0.38). At follow-up, a statistically significant difference was found between the intervention group and the usual care group with regard to the 50% symptom reduction (2 months: 31% versus 16%; 6 months: 42% versus 26%).
PRIME-MD
Multilevel analysis showed that the PRIME-MD outcome (major depression) was not influenced by GP level, characteristics, treatment, or age. Baseline values for the number of depressive symptoms were 6.10 ± 0.80 in the intervention group (Table 2). Data on the PRIME-MD at 2 months could not be collected due to an administrative error. Dichotomised, after 6 months, 50% (n = 35) of the patients in the intervention group no longer had a diagnosis of major depression (PRIME-MD negative) versus 55% (n = 41) of the patients in the usual care group. This is a non-significant difference. After 1 year, 66% (n = 37) of the patients in the intervention group no longer had a diagnosis of major depression versus 64% (n = 43) in the usual care group. This is also a non-significant difference.
At the end of the treatment (after 6 months) the mean level of depressive symptoms in the intervention group had declined to 2.80 ± 1.04, and at the end of the follow-up the scores were 3.23 ± 1.04. In the usual care group, the baseline number of depressive symptoms was 6.33 ± 1.01. Data collection on the PRIME-MD at 2 months also failed in this group. At the end of the treatment (after 6 months) the mean scores had declined to 3.99 ± 1.22, and at the end of the follow-up, the scores had further declined to 3.74 ± 1.21. Trend analysis revealed no statistically significant differences between the PRIME-MD scores in the intervention group and the usual care group at any measurement point.
No statistically significant differences were shown on the CGI.
Secondary outcomes
There were no statistically significant differences between the intervention group and the usual care group with regard to any of the secondary outcome measures (Appendix 2).
DISCUSSION
Summary of main findings
This study of the effectiveness of an intervention programme to improve the identification, diagnosis, and treatment of late-life depression in primary care shows that on the main outcome, the MÅDRS, at the end of the intervention period, a statistically significant difference was found in favour of the intervention group. It was remarkable that after 6 months and 1 year the patients in both groups had improved significantly on the MÅDRS primary outcome compared with at baseline. The same applies to the number of depressive symptoms (PRIME-MD): the majority of patients in both groups no longer had a diagnosis of major depression. At 12 months MÅDRS scores were lower for the usual care group than the intervention group. The mean decrease in MÅDRS scores was quite substantial, that is, a decrease of more than 50% in both groups.
Strengths and limitations of the study
According to the study protocol, the intention was to base the analysis on matched pairs of GPs. However, because of the cluster randomisation procedure and some variety in characteristics of GPs, it was decided to match according to propensity scores. These scores make it possible to correct for the influence of a large set of observed confounders and their interactions on the treatment assignment. Although there were no differences in depression severity level between the two groups at baseline, the clustered randomisation procedure had not been completely successful for some patient characteristics (such as sex).
It could be expected that this is a group with more than one health problem when they attend the GP, which will have an impact on the identification and treatment of the depression, if only because of competing demands on a GP's time.8 The required number of 70 per arm was not achieved for the entire follow-up of 1 year.
Comparison with existing literature
Several factors might explain the relatively modest effects of the intervention. The case-finding procedure applied led to the selection of patients with moderate major depression, as demonstrated by the MÅDRS scores (20–21) and by the relatively good prognosis of many patients. There are also are doubts about the effects of antidepressants in primary care.31 These arguments favour active follow-up and stepped care rather than an immediate prescription of antidepressants, that is, active management should only be commenced when patients do not recover.
In addition, the study asks whether the GPs who participated could have been more motivated to treat depression and whether they were perhaps more familiar with depression guidelines than other GPs in the Netherlands.18 The data does not support this assertion: a total of only 48% of subjects in the intervention arm received antidepressants for 6 months. Another Dutch study showed that more than 70% of older patients received antidepressants from their GP.32 The Dutch depression general practice guidelines are highly accessible and are integrated into the medical records systems, including the formularies, that most Dutch GPs use.21 Most Dutch GPs are, therefore, well aware of the guideline criteria.16
When it comes to recognition, the process of the diagnostic interview might have prompted patients to seek help where they might not have done otherwise. However, this was an intended effect of the intervention (‘a package’).
There are also differences between the organisation of primary care in Europe and in the US. The review of the effectiveness of disease management programmes for major depression in primary care contained only one trial that was performed in Europe.4 This education programme aimed at GPs did not improve the identification or recovery of patients from major depression. Two of three US trials that studied older subjects with major depression in primary care had positive results.9–11
Implications for future research and clinical practice
The results of this study support a stepped care or repeated assessment approach that is not too far from current usual care, but requires active follow-up by a nurse or perhaps via the internet. For an example of internet-based self-help and monitoring depression, see http://moodgym.anu.edu.au. Such projects require further large-scale testing within a practice setting.
