We evaluated whether a disease management program for major depression in elderly primary care patients would be cost effective in comparison with usual care. Although antidepressant treatment rates were significantly higher in the intervention group, we found no statistically significant difference in total costs, effects, and cost-effectiveness between the 2 treatment groups.
Our findings are in line with 2 randomized trials that also found that disease management programs for elderly primary care patients with depression had no significant effect on depression severity.26,27
In contrast, Katon et al.28
found a significant effect, accompanied by a modest and insignificant increase in total health care costs. However, patients in that study had access to a depression care manager, which makes the intervention program much more intensive.
Several American observational studies reported much higher health care costs in depressed elderly patients than we do.4,6,28–30
These studies used computerized databases to estimate costs, whereas our cost estimates were based on patient reports of resource use. Also, our study population was on average about 10 years younger than the study populations in the abovementioned studies.
There may be several explanations for the nonsignificant results of this study. First, despite the fact that all patients were diagnosed as having major depression according to the PRIME-MD, most patients detected by our screening method were only mildly to moderately depressed according to their MADRS scores at T0.31,32
This may arise from the fact that screening in primary care typically leads to detection of mildly depressed patients.33
Because patients with milder forms of major depression are not likely to be included in trials evaluating the efficacy of antidepressants, it is unclear whether antidepressants are efficacious in these patients.34
Indeed, a recent review concluded that antidepressants have only moderate effects in older ambulatory patients with mild to moderate depression.35
Another partial explanation for the nonsignificant findings of this study may be the existence of a Hawthorne effect, although this seems inevitable in this type of research. The fact that GPs and patients knew they were participating in a study influences their behavior and perceptions and thereby may reduce any differences there might have been between the treatment groups. The fact that trained research assistants interviewed study patients at home may also have contributed to the Hawthorne effect. We tried to control for a Hawthorne effect by blinding usual care GPs to the screening results and the research assistants to the allocation status of the GPs. However, the Hawthorne effect may have been substantial, because, especially in the treatment of depression, attitudes, environment, time spent with patients, etc., will have an effect on the results. Related to the Hawthorne effect is that usual care patients frequently received GP care and physiotherapy for comorbid disorders. Treating comorbidity may improve outcomes of depression as well. This may also partly explain the relatively positive outcomes in usual care patients.
There are also 2 design aspects of our study that may have contributed to the nonsignificant findings of our study. First, we chose a pragmatic design, meaning that we tried to replicate everyday clinical practice as much as possible to enhance the generalizability of our findings. However, a disadvantage of a pragmatic design is that the contrast between the treatment groups may be diminished. Second, it was not possible to blind patients included in the usual care group. Although usual care patients were requested not to reveal to their GP that they were participating in this study, there is a risk that some patients informed their GP and subsequently received some kind of treatment for depression. However, as only a few usual care patients were prescribed antidepressants or referred to a mental health care provider, this problem seems to be small.
Initiation of mental health care treatment in both the intervention and the usual care group was associated with more severe depression at baseline. Thus, GPs use severity of depression as a criterion to initiate depression treatment, but this criterion alone seems insufficient to distinguish patients who will benefit from depression treatment from patients who will not benefit.
General practitioners of intervention patients had to agree with the depression diagnosis and had to be willing to prescribe antidepressants, which may have led to selection bias in the intervention group. In this case, it can be expected that intervention patients have more severe depression at baseline than usual care patients. Intervention patients were indeed somewhat more depressed than usual care patients at baseline, but the difference was small and not statistically significant. Therefore, we do not expect that this particular form of selection bias was very strong in our study.
The follow-up rate of 86% is very good for studies in elderly depressed populations. Patients who dropped out before the end of the study were older and more depressed than patients who completed all follow-up measurements. Moreover, although a sufficient number of patients was included in the study, the number of patients analyzed at 12 months was smaller than the required sample size to detect moderate clinical effects. However, the results of a missing value analysis using the Expectation Maximization algorithm36
did not differ from the complete case analysis (data not shown). Therefore, we do not expect that inclusion of patients who dropped out before the end of the study would have altered our conclusions.
Our study was underpowered to detect relevant differences in costs, which is reflected in wide confidence intervals for cost differences. This is a common problem in “piggy back” economic evaluations. Because the distribution of cost data is typically heavily skewed, very large numbers of study patients are needed to detect relevant cost differences.37
It is generally considered unethical to increase study sizes beyond the level needed to prove clinical effectiveness.
Another limitation to the economic evaluation presented in this article is the manner in which cost data were collected. In interviews, patients were asked about their health care utilization over the past 6 months. Subjects in general, and elderly subjects in particular may not be able to recall health care utilization reliably over such a long period. This may have introduced recall bias. Most likely, our estimates are an underestimate of the true utilization rates for frequently occurring resources such as visits to the GP. For more seldom occurring resources such as hospitalizations, we expect our estimates to be reasonably adequate.
A final limitation is that costs of production losses and informal care giving were not measured. As the differences in clinical outcomes and direct costs were small, we consider it unlikely that inclusion of lost productivity costs would have altered the results of our study. Because our study population consisted of elderly subjects, it is probable that many subjects received informal care. Future studies should attempt to measure these costs.
In conclusion, this disease management program for major depression in elderly primary care patients was not cost-effective in comparison with usual care. There were no significant differences in depressive symptoms, quality of life, and costs between the intervention and usual care group at 12 months. Therefore, based on these results, we recommend continuing usual care by GPs, which mainly consists of “watchful waiting.” In this situation, treatment for depression is initiated only when the GP diagnoses the patient as being depressed. We recommend that future research focuses on improvement of the detection of clinically important and treatable depression by GPs. Evidence is needed on indicators that help GPs in determining which patients may profit from depression treatment. Research is also needed on the (cost-) effectiveness of treatments for depression in elderly primary care patients, especially on the (cost-) effectiveness of treatments other than antidepressants, such as different forms of psychotherapy.