Among the subjects who self-reported a physician diagnosis of depression in at least one of the questionnaires used in this cohort study, the percentage of confirmed diagnosis of depression was high. The percentage of subjects without real depression who had responded negatively to the questions regarding depression status in the questionnaires was even higher.
However, when several sample characteristics of the sample were analyzed, differences in the degree of confirmation regarding depression status were found. A more favourable response (better agreement between self-reported diagnosis and the gold standard) regarding the presence of depression was found for men, younger subjects, non-smokers, overweight and obese subjects and participants with a low level of physical activity during leisure time; but only physical activity was statistically associated with a correct positive classification of depression. Due to the high number of characteristics analysed, we can not exclude the possibility that some spurious associations could have occurred. However, higher prevalence of depression should lead to a higher proportion of confirmed depression. Depression has been reported to be inversely associated with physical activity in several epidemiological studies [5
]. Thus, a higher prevalence of depression or the presence of depressive symptoms (asthenia, tiredness) is expected among sedentary subjects.
In the same way, the proportion of correctly identifying a subject as free from depression was higher among non smokers. Smoking has also been associated to depression [7
The validity of the self-reported diagnosis of different diseases has been sufficiently studied in several populations in which the educational level of the participants was as high as it is in the SUN Project. The SUN Project uses similar methodologies to those of large cohort studies such as Nurses' Health Study or the Health Professionals Follow-up Study. In these cohort studies disease status is collected though questionnaires sent by mail every 2 years as we do. Predictive values obtained in their validation studies are generally high [13
]. The SUN Study has also assessed the validity of some questions included in the questionnaires regarding several diseases or conditions like hypertension [15
] or body mass index [16
] obtaining good results.
Most of the population studies which have analyzed depression have used depressive symptom scales to ascertain the disease: Diagnostic Inventory for Depression [17
], Center for Epidemiological Studies Depression Scale [18
], Hospital Anxiety and Depression Scale [19
] or Beck Depression Inventory [20
]. In the majority of the cases, the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (SCID-I) was used as gold standard to assess validity of the scales. However, an important limitation of the use of depressive scales is the arbitrary election of a cut-off point to define depression status. The confirmation of the disease for a specialist though a physician interview (categorical definition) is clearly the best option to assess a depressive disorder although, at this moment, it is controversial the use of a categorical classification to define a depressive status. Some authors defend the idea of a dimensional classification in which depression could be defined as a continuum ranged from absence of any symptom to the presence of symptoms with maximum intensity. On the other hand, under-diagnosis of depression occurs in the 44.3% of the patients coming to a primary care center [21
]. A recent study found a sensitivity of 40% and a specificity of 87% for a general practitioner diagnosis of a major depressive disorder as compared with that found through the SCID-I made by a psychiatrist [22
]. Therefore, it was necessary to carry out a validation study to rule out disease under-estimation in the questionnaires among our participants due to under-diagnosis. This situation should lead to a decrease of the proportion of confirmed non-depression.
The difficulty to classify depression correctly could have decreased the proportion of confirmed depression in the validation study. There are problems in differential diagnosis because depressive experiences vary from individual to individual. The co-occurrence of symptoms of anxiety and depression is very common. In a recent community study, the 56.3% of the sample with current major depressive disorder had also another mental disorder [23
]. On the other hand, some of the participants did not have a specific clinical report, only a verbal diagnosis from their physician. Thus, the depressive symptoms secondary to other primary psychiatric disorders or other pathology could be easily named depression.
We acknowledge that participation in the validation study was low (24.2% and 23.3% for depressed and non-depressed subjects, respectively) and therefore a selection bias cannot be excluded, conferring an artificially high validity to our results. To prevent this bias, we put special care to blind participants about the aim of the psychiatrist interview. Thus we tried to avoid that those participants who may be aware of incorrectly classifying themselves as depressed may be embarrassed of participating. Moreover, when we compared the socio-demographic characteristics of participants (n = 104) and non-participants (n = 333) we did not find any systematic difference regarding key variables (sex, age, smoking status, BMI, or physical activity) in their baseline assessment.
Finally, a small proportion of participants were not interviewed face to face by the specialists. They were interviewed by telephone, they sent their medical report by postal mail or they permitted the access to their reports at the University Clinic. Although a participant could lie or modify his/her responses when the interviewer is not present, the validity of depression assessment though the use of telephone has been demonstrated [24
]. Whereas the assessment of the disease could have been modified because of the use of a telephone interview, we do not expect any change regarding depression assessment using existing medical reports.