Comparison with existing literature
We did not find a statistically significant difference of MDD underrecognition rates between study groups. This negative result contrasts somewhat with the efficacy of screening guidelines reported in controlled clinical trials [
7,
8,
15]. The failure to detect differences in the secondary outcome of depression undertreatment is, nevertheless, consistent with prior reports that point to the ineffectiveness of screening alone to resolve this issue in PC [
14,
16].
With regard to underrecognition of depression, a study done by Caballero et al. [
26] carried out in PC setting in Spain, that used a similar design to our study, reported that 54% of PC patients with MDD in Spain remain unrecognized. However, in our study we found that only 41% of control group patients remained unrecognized. This may have contributed to reduce between-group differences in this study.
Adherence to the screening among the PCPs in the intervention group was suboptimal: only about half of them used the screening in their practice frequently or very frequently during the study. In contrast, 90% of PCPs considered that its acceptance was good or very good. Our adherence data are better than the results reported by Harrinson et al. [
17], however, worse than Kirkaldy et al. [
18]. A study conducted by Harrison et al. [
17] in primary care centers through the United States showed that, despite the recommendations raised by the USPSTF, only the 2.3% of the physicians performed the screening for depression. Kirkaldy et al. evaluated the screening program implemented in a medical center for one month and found that the 97% of PCPs performed the screening [
18]. Our results on acceptance and adherence suggest that many PCPs were willing to implement the screening recommendations but were unable to do that frequently. PCPs in our study setting have a very short consultation time with each patient, they have no support staff for depression care and are not audited and paid for screening [
19]. Besides, PCPs have other conflicting clinical priorities [
19,
28]. These factors may prevent physicians from establishing new systems, even if acceptable and feasible [
29]. The suboptimal adherence found in our study is in accordance with a previous study where PCPs had a positive view of the NICE guideline for depression, but its impact was compromised by resource and practitioner barriers to implementation [
30]. The degree of implementation of new interventions into routine clinical practice is markedly influenced by contextual factors in the setting where they have to be applied [
31].In order to improve the quality care for depression, organizational barriers to the implementation of a depression guideline, in the complex realities of PC, is a significant factor to take into account. In that sense, it should be mentioned that collaborative care in primary care has shown effectiveness in the management of depression [
32,
33]. However, this is a multi-component approach which requires the involvement of additional staff like nurses. Currently, it is not a usual practice for the management of depression in Spain [
33]. The aim of this study was to test the implementation of a specific recommendation regarding depression screening in the current everyday reality of the primary care setting in Spain.
The present study was based on the 2002 USPSTF recommendations [
7,
13]. These recommendations are in accordance with EU guidelines that are known and used in Spain [
5], and with Canadian guidelines [
5,
6]. However, we chose the USPSTF guidelines because they were more detailed and had more information regarding their implementation. The version issued in 2002 recommended screening only in those cases where the patient could have adequate follow-up and treatment in case depression was recognized. The 2009 recommendations are more restrictive and specify that the screening should not be applied unless there is support from additional staff to provide depression care [
8]. PCPs in our study lack additional staff, and as previously discussed, this could in part explain the suboptimal implementation of the recommendations by some PCPs randomized to the intervention group, and therefore the study results.
Consistent with early research, somatic symptoms were a frequent reason for consultation and were associated with underdiagnosis of depression [
34]. This finding supports the additional vigilance in patients with somatic symptoms. Also, PCPs should be aware that, in patients with medical illnesses, an interview focussed on the affective and cognitive symptoms of depression could be helpful and easier to apply [
35].
Finally, psychiatric comorbidities and inability to work were associated with recognition. Maybe these conditions can sound an alarm to the PCP, so he/she is more driven to look for other diagnosis.
Strengths and limitations of the study
Because of the pragmatic, cluster-randomized design, this study allowed to assess the effectiveness rather than the efficacy (under controlled conditions) of depression screening. This may be construed as a strength because the evidence obtained under everyday practice conditions serves as a basis for decisions about healthcare policies [
36]. To evaluate the recognition of the depression episode, the participating PCPs reviewed the patients’ medical records themselves, so this may have introduced a collection bias, so a possible deflation of underrecognition rates cannot be ruled out. However, this procedure was done in both study groups. The use of an independent evaluator would have been optimal. However, this option was ruled out after the evaluation of the pros and cons, taking into account the difficulties of its implementation in the 69 participating centres. There was not any specific time-frame for the revision of the medical record. This adds complexity to the collection of information and could be considered a limitation. However, it allows the capture of current episodes of long duration. Feasibility, acceptance and adherence to the screening of depression were reported by PCPs themselves which adds a recall bias and a social desirability bias into the results, so an overestimation cannot be ruled out. It should be mentioned, that for a small percentage of patients, the study visit was the first contact with the participating PCP, so this may have contributed to the reported underrecognition rates of depression. However, the percentage was balanced between both groups and the inclusion of these patients reflects more precisely the clinical reality of PC. The study was powered to find a difference between groups of 15% or greater; smaller differences, that could be considered clinically relevant by some clinicians, were therefore undetectable. This aspect should be taken into account for future research, especially in pragmatic trials were differences may be smaller than in controlled trials. The focus on high-risk patients, based on USPSTF 2002 recommendations, would have contributed to undermine the chances of showing an effect of the intervention, since the recognition of depression in this group could be higher than among all adults. The intervention was provider-dependent, that is, the PCPs had to do the screening. Support from nursing staff might have had a greater chance of yielding positive results. It could be suggested that screening may hardly be effective in depression because of its episodic nature. Nevertheless, we do not consider this as an explanation for our results, because most patients had been suffering from depression for several months and had consulted their PCP several times.