¾ 20. Conclusions. The prevalence of depression in our primary care setting is high. The disorder was underdiagnosed in as many as 44.3% of the persons likely to have depressive disorder (especially women, widows and widowers, retired persons, persons who had experienced stressful life events, and frequent users of primary care services).
Introduction
The prevalence of depressive disorders in primary care (PC) and the physicians' ability to detect them are controversial topics of considerable interest.1 Recent years have seen the publication of many studies2-11 that have addressed the issue in terms of primary and psychiatric care. However, certain questions remain to be examined, in view of the methodological limitations of some published studies10,11.
The prevalence of depression estimated for the population of users of PC services should be distinguished from the prevalence in the general population, which is thought to be approximately 6% per year and 15% throughout a person's lifetime.12 The life-time prevalence for major depressive episode in industrialized countries has been estimated at 8%-12% in men and 20%-26% in women.1-3 In all industrialized countries the incidence of mood disorders is on the rise, and these disorders are appearing at younger ages.4 According to several epidemiological studies, the prevalence of major depression in PC is estimated at 5%-10%; subclinical depression may be present in a further 10%-30% of the population.5-7
Studies published to date on the prevalence of depression in PC have used two methods. Some have been based on screening tests such as the Beck Depression Inventory (BDI)4,5,8, the Center for Epidemiologic Studies Depression Scale (CES-D)6,9-11 and Zung's Self-Rating Depression Scale (SDS),13,14 followed by structured diagnostic interviews such as the Structured Clinical Interview for DSM-III-R (SCID),8-10,15 the Mini International Neuropsychiatric Interview (MINI) or the Schedules for Clinical Assessment in Neuropsychiatry (SCAN).5 Two-phase studies require the participation of clinical staff in the second (interview) phase. The second approach has been to use a single diagnostic interview such as the Composite International Diagnostic Interview (CIDI).16 The main problem with diagnostic interviews is that they require specialized staff and sufficient time to be administered correctly; these requirements make such instruments difficult to use in PC.8,11,14,17,18
A review of published studies suggests that PC physicians fail to detect depressive disorders, which remain undiagnosed in one-third to one-half of the cases.8,9,14,15,19,20 In a number of studies we reviewed, different reasons were given to explain the evidence: firstly, the fact that patients are often suffering from a combination of complaints when they consult their physician, and thus present with multiple somatic symptoms which can also be considered diagnostic criteria for depressive disorders.2 Secondly, the characteristics of depressed patients who visit their PC physician differ from those of patients who have been diagnosed and treated by a specialist. In the former, the disorder is less severe and has been present for longer,8,9,17 whereas the latter patients are considered incident cases: their clinical manifestations are more florid, they are more promptly referred to mental health services, and they more accurately fulfil the criteria for a psychiatric diagnosis. Finally, some studies have noted that PC physicians are more likely to detect depression if they are familiar with the patient's previous history of psychiatric disorder.9 If the PC physician's contact with the patient has been sporadic, the detection rate is lower.
The PC physician's ability to detect depressive disorder is of fundamental importance in preventing depression-related suicides, improving the patient's quality of life and reducing health care costs, as these patients are more frequent users of medical services and have significantly greater degrees of dysfunction than patients with chronic medical diseases.1,4 The present study was designed to determine the prevalence of depression in PC with a two-phase approach, to investigate the possibility of underdiagnosis of depressive disorders by PC physicians, and to suggest which factors might influence underdiagnosis.
Material and methods
Design and setting
This cross-sectional, descriptive study was done in the Gavà II Basic Health Area, an urban area located 10 km from Barcelona with a population of 19 000.
Subjects
A total of 400 persons between the ages of 18 and 65 years were chosen randomly from among those who consulted their PC physician at the health center. Eight PC physicians participated: 4 who saw patients in the morning shift, and 4 who saw patients in the afternoon shift. Persons with inadequate language abilities were excluded. Sample size was calculated for an expected prevalence of 20% with a 5% alpha error and 4% precision.
Forty participants (20 from the morning shift and 20 from the afternoon shift) were selected at random on each of 10 days. They were informed about the purposes of the study, and after the doctor had seen them they were asked to give their verbal consent to take part in the study. The data collection period lasted for 3 months, from November 1999 to January 2000.
Method
A sociodemographic questionnaire was administered to obtain information on the patient's contact details and stressful live events during the previous 6 months. We recorded the first 10 life events from the Paykel Scale of Life Events (death of a relative or friend, personal illness, accident involving self or significant person, pregnancy (own or significant person's), change in financial situation, change in job or home, divorce or separation, marriage, or job loss). Then the 21-item, self-administered version of the BDI screening test for depression was given (validated for the Spanish population by Conde, Esteban and Useros, 1976).3,21-27 The medical records of all 400 participants were reviewed to record the number and type of antidepressant or anxiolytic drugs currently used, presence of psychiatric symptoms recorded by the PC physician during the preceding 30 days, referral to a mental health center, previous history of psychiatric illness, presence of underlying somatic disease, date of first annotation in the medical record, and total number of visits to the health center since the first visit. Within 1 week of the initial interview, 40 randomly chosen subjects were interviewed with the MINI28 as a check for the validity of the BDI; 20 of these subjects scored ¾15 on the BDI, and 20 had scored >=16.
On the basis of the results with the MINI, the cutoff score for the BDI was recalculated, and its specificity and sensitivity were computed for the sample we tested. This optimized cutoff was used to identify true positives, true negatives, false negatives and false positives. The PC physician was informed of those patients who scored above the optimized cutoff value on the BDI.
Statistical analysis
The data were analyzed with the SPSSWIN package (v. 6.1). Prevalences and confidence intervals were calculated with the information from the cross-sectional survey and the data on sensitivity and specificity of the BDI, as calculated in the light of the MINI results. The method used was that developed by Tenenbein,29 which uses maximum likelihood estimates to calculate the confidence intervals. This method allowed us to group information from each of the two phases of the study. The Tenenbein method was also used to distinguish between prevalence stratified by sex. Because of the low number of cases per stratum when variables with several strata were analyzed, we used unadjusted prevalences for these comparisons. The confidence intervals for these prevalences were calculated with the exact method from the relevant tables, for n* P<10. In all other cases the data were transformed to a normal distribution to calculate confidence intervals.
Other statistical analyses included calculation of ROC curves, odds ratios (OR), chi-squared statistics, and comparison of the means (Student's t test and one-way analysis of variance).
Results
Table 1 shows the sensitivity, specificity and predictive values for different cutoff scores on the BDI calculated from the MINI scores as a reference. The best results were obtained with a cutoff score of 20/21. Persons who scored ¾20 were not considered cases; those who scored >=21 were. Unadjusted prevalence of depressive disorder in our PC setting, as calculated with this cutoff score, was 15.5% (BDI>=21). When the cutoff score was corrected to take into account the predictive values for the MINI as calculated with the Tenenbein method, the prevalence was 20.2% with a 95% confidence interval of 11-29%.
Table 2 shows the prevalence of depressive disorder according to different sociodemographic variables. Depressive disorders were found to be more frequent in women than in men (OR, 4,15; P<.01).
The ability of PC physicians to detect depressive disorders was analyzed by examining each of four categories: true positives (persons who scored above the BDI cutoff score and whose medical record contained an annotation by the PC physician of depressive disorder or a prescription for antidepressants or anxiolytics), true negatives (score below the BDI cutoff, no annotation of depressive disorder by the physician or prescription for antidepressants or anxiolytics), false negatives (score above the BDI cutoff, but medical record contained no annotation or prescription order), and false positives (scored below the BDI cutoff, but medical record contained annotation, prescription, or both). This analysis yielded 66.6% true negatives, 7% false negatives, 8.9% true positives and 17.5% false positives. False negative cases were more likely to be women (8.3%) than men (4.2%), as was also found for true positives (10.9% women vs. 4.2% men). Men were more likely than women to be true negatives (75.4% men vs 62.6% women) (P<.05). Among widows and widowers, 18.8% were found to be false negatives (P<.05); the corresponding figure for retired persons was 2.2% (P<0.05). There were no significant differences between strata classified by age or type of employment in the rates of any of the four categories for true or false positives or negatives.
The PC physician had made annotations of psychiatric symptoms for 12.3% of the patients; of these, 38.3% scored above the BDI cutoff. Of all patients who scored >=21 on the BDI, the PC physician had noted psychiatric symptoms in 30% (P<.0001) (Table 3). Of all subjects in the study, 15.6% had a previous history of psychiatric disorder; 23.6% were taking antidepressant or anxiolytic medication, and 4.7% had been referred to mental health services. Of the patients with no prior history of psychiatric disorder, the PC physician detected 18 of 323 cases, and of the 60 cases with a history of such disorders, 29 were detected (P<.0001). Of the subjects who were taking antidepressants or anxiolytics, 64.4% scored below the BDI cutoff; of those who scored above the cutoff, 46.7% were not taking such medication (P<.0001).
Persons who scored below the BDI cutoff consulted their PC physician a mean of 5.6 times per year (SD, 4.1), whereas patients with likely depressive disorder made a mean number of 8.9 visits per year (SD, 7.9) (P<.001). The usage rate by patients shown to be true negative cases was lower, with a mean of 5 visits per year (SD, 3.5), whereas this rate was higher in those found to be false negatives (10.7; SD, 2.1), false positives (7.6; SD, 5.2) or true positives (7.9; SD, 4.7) (P<.001).
The analysis of stressful life events showed that 55% of all participants had experienced at least one such event in the preceding 6 months. Of those with likely depressive disorder, 77% had experienced at least one stressful life event (OR, 3.2) in comparison to participants with no depressive disorder (P<.001) (95% CI, 66.4%-87.6%). Mean BDI score among persons who had experienced stressful life events was 12.7 (SD, 10.4), whereas mean score among those who had not had such events was 8.9 (SD, 8.2) (P<.0001). Unadjusted prevalence of depressive disorder in persons who had had a stressful life event in the preceding 6 months was 17.6% (95% CI, 12.6%-22.6%); the figure for persons who had not experienced such events was 7.8% (95% CI, 3.9%-11.7%). Participants who had experienced stressful life events scored higher on all items of the BDI (Table 4).
Discussion
The adjusted prevalence of depressive disorders in PC in the present study was 20.2%. Our findings showed that the PC physician correctly identified 55.7% of these patients.
The prevalence of depressive disorders we found is within the range of values other studies have reported. Klinkman et al10 reported a prevalence of 22% with the CES-D; Carmin and Klocek5 obtained a prevalence of 15.4% with a BDI cutoff score of 16/17, and Ferrer and Rodríguez,2 in a study in Galicia (Northwestern Spain), obtained a prevalence of 11.2% with a BDI cutoff of 15/16. It should be recalled that in the present study, we evaluated a wide range of depressive disorders, and included in our sample of participants some cases of dysthymic disorder, as seen in the results with the MINI (66.67% major depression, 26.67% dysthymia, 6.67% major depression and dysthymia). Other studies, in contrast, centered exclusively on major depression.7,17 The prevalence in our study is similar to that found by Klinkman et al10 in a sample of patients in Michigan (USA), and higher than the prevalence in the study by Ferrer and Rodríguez.2 The differences may be due to biological, environmental, cultural or age-related factors (for example, the sample in the study from Galicia included patients older than 15 years), or to the use of different cutoff scores.
The patient's sex is one of the main factors associated with depressive disorder.1 In women we found that the prevalence was fourfold as high as in men (26.8% vs 8.1%; OR, 4.15); in other words, more than one-fourth of the women and approximately one in twelve men who visited their PC physician for any reason were depressed. These figures are higher than those given by Gater et al for the prevalence of depressive disorder in the PC setting in 15 countries, which was 12.5% for women and 7.1% for men (OR, 1.8).14 The discrepancy should alert us to the need to improve the detection and treatment of cases, and thus decrease the utilization of health care services, improve the patient's social and familial situation and quality of life, and reduce the risk of self-injurious behavior.
Earlier studies suggested that the PC physician often misses depressive disorders, which go undiagnosed in between one-third and one-half of the cases.4,8-10,14,17 In the present study we considered detection by the PC physician to have been successful when the physician had made at least one annotation of depressive symptoms in the patient's medical record, or when the patient was taking antidepressants or anxiolytics. The rate of detection was 55.7%; this implies that 44.3% of the cases were missed; a figure similar to that in other studies.
Factors related with a lower rate of detection by PC physicians were patient's sex (women), marital status (widow or widower) and employment status (retired). These persons should therefore receive greater attention in PC situations to avoid a possible missed diagnosis, initiate treatment when necessary, improve the patient's quality of life and reduce health care costs.
We also noted a marked tendency for more women than men to be classified as false positives (annotation of depressive symptoms or prescription for antidepressants or anxiolytics in the medical record, BDI score below the 20/21 point cutoff) and true positives (annotation of depressive symptoms or prescription for antidepressants or anxiolytics in the medical record, BDI score above the 20/21 point cutoff). Men, on the other hand, tended more often to be classified as true negatives (no annotation of depressive symptoms or prescription for antidepressants or anxiolytics, BDI score below the 20/21 point cutoff). This pattern may be related with cultural factors, as women openly express their feelings more readily than men, and the physician might interpret certain negative emotions as symptoms of depression.
A noteworthy finding was the high frequency of depressive disorder in our sample: persons who scored above the BDI cutoff visited their PC physician 58.9% more often than did those who scored below the cutoff. The greater use of health care services by patients with depressive disorder has been described in earlier studies.1,14
Of the patients on antidepressant medication, 64.4% scored below the BDI cutoff point. This might reflect good compliance with pharmacological treatment leading to improvements in symptoms and lower scores on the BDI, or it might reflect unnecessary use of antidepressant or anxiolytic medication in some patients. Because of the cross-sectional design of the present study, we cannot say which of these two hypotheses is the more likely. The same explanation applies for the 28.3% of the participants who had a previous history of psychiatric symptoms but who did not score above the BDI cutoff: at the time of the interview for this study they might have had no clinically active manifestations.
Any evaluation of our findings should take into account the fact that the data were obtained in two phases. Except for prevalence calculations, all persons who scored above the BDI cutoff were considered cases for the purposes of our analyses. Although the BDI is not a diagnostic instrument, the data we used to validate the BDI scores against the scores obtained on the MINI--a diagnostic interview used as the reference instrument--are in good agreement for positive predictive value, negative predictive value, sensitivity and specificity. We used a BDI cutoff score calculated specifically for our population, rather than using reference values as other studies have done.4,5,7 In the subsample of 40 patients who completed the MINI, we found 12.5% false negatives (BDI score below the cutoff but diagnosed as depressed with the MINI) and 5% false positives (BDI score above the 20/21 cutoff, but no diagnosis of depression with the MINI). Of the patients for whom the MIDI indicated a diagnosis of depression, 46.2% had some annotation of previous psychiatric problems in their medical record. Of those for whom the MIDI was not diagnostic of depression, 20% were taking antidepressants or anxiolytics, whereas of those who were diagnosed with the MINI as having depression, 39.4% were not on antidepressant medication.
In our analysis of the PC physician's detection of depression we considered annotation in the medical record of psychiatric symptoms or prescription of antidepressants or anxiolytics, as evidence of such detection. This might have led to errors of interpretation, as there might have been cases of depression that were correctly identified by the PC physician but not entered in the medical record for some reason, eg, short duration of the visit, different reason for consulting, or a chronic history of depressive symptoms noted in the chart more than 1 month before the visit included in the present study. Thus we were dependent on the quality of the medical record, and some patients may have been receiving treatment with antidepressants or anxiolytics although this was not noted in the chart. In their study, Klinkman et al10,11 asked physicians about the presence of depressive disorder and its severity, and concluded that physicians distinguished between persons with and without depression on the basis of their prior knowledge of the patient's psychiatric history and familiarity with the patient. In other words, the more often they had seen the patient, the more likely they were to correctly detect depression.
Another limitation of the present study is the fact that we used a prevalent sample, ie, a sample that included persons with depressive disorder, who consulted their physician more frequently. Our sample did not contain patients whose depression had already been correctly detected and diagnosed, and who therefore no longer consulted their PC physician. Nor did our sample include patients who no longer used PC services because they had already been referred to mental health services.
Our study confirms the high prevalence of depressive disorders in PC services in our setting; PC physician should therefore be on the alert for these disorders. The patients in whom depression is most often overlooked, and who therefore deserve particular attention, are women, widows and widowers, retired persons, persons who have experienced stressful life events in the preceding months, and those who consult their PC physician more frequently than average. Once the disorder has been diagnosed, treatment should be started promptly to lower costs and reduce the risk of suicide, and increase the patient's quality of life.
Acknowledgments
We thank Dr. Neus Parellada, health technician at the Distrito de Atención Primaria Baix Llobregat Litoral, and Drs. Emilia Caramés, Ángel Espín, Inma García and María Soler, family physicians at the Área Básica de Salud Gavà II, for their invaluable help with this study.
This project was carried out with the support of the Agencia de Evaluación de Tecnología Médica (15/47/98).
Correspondence: Dr. Josep Maria Haro. Centro de Salud Mental Gavà, Sarriá, 13-15, 08850 Gavà (Barcelona), España. E-mail: 27652jha@comb.es
