Alcohol-related problems and depression are highly prevalent in general practice [1]. More than half of all persons with alcohol-related problems obtain all of their care from their general practitioner. Hence, primary care has a considerable potential as a site for prevention of and interventions for alcoholism [2]. Screening and interventions for alcoholism in primary care has been demonstrated to reduce drinking [3].

However, depression and alcoholism often are under-diagnosed and untreated in family practice. For example, family physicians in Australia failed to identify 65% of CAGE-defined alcohol abuse [CAGE: an acronym of key words found in the following screening questions for alcoholism: "Have you felt that you need to c ut down on your drinking?" "Have people a nnoyed you by criticizing your drinking?" "Have you ever felt bad or g uilty about your drinking?" "Have you ever had a morning e ye-opener?" [4]]. Physicians in the United States (US) detected only 49% of the current alcohol problems diagnosed using the Diagnostic Interview Schedule (DIS) [5]. An international multi-centre study conducted by the World Health Organization (WHO) revealed that physicians recognize just 42% of all cases of Major Depression diagnosed by means of the Composite International Diagnostic Interview (CIDI) [6]. These studies examined the percentage of simple agreement between physicians' diagnosis and diagnostic tools, or assessed characteristics that influence the detection of mental illnesses. However, few studies have evaluated patient complaints and doctors' diagnoses as potential causes of diagnostic errors pertaining to mental health.

Consequently, the aims of this survey were: 1) to evaluate the accuracy of a family physician's diagnosis of depression and alcoholism using a validated screening instrument; and 2) to assess patient complaints and physician diagnoses that result in alcoholism and depression being overlooked in family practice.

In this survey of a single family physician's practice, the kappa statistic, which measured the degree of agreement between the family physician and the J-MINI for major depression, was 0.43. Compared to the diagnostic accuracy by psychiatrists reported in J-MINI validation study by Otsubo and colleague [8], the family doctor in this survey had an acceptable level of diagnostic accuracy for major depression. Otsubo and colleagues reported that the kappa statistic estimating agreement between expert psychiatrists' diagnoses using diagnostic criteria and the J-MINI is 0.36 for major depression and 0.26 for alcoholism [8]. Spitzer and Fleiss used the kappa statistic to measure inter-observer agreement between psychiatrists before the era of research diagnostic criteria, and estimated a kappa of 0.41 for affective disorder (including neurotic or manic-depression) and 0.71 for alcoholism [10]. Even using the Structured Clinical Interview for DSM-III-R (SCID), kappa statistics estimated between mental health professionals were 0.42 for current major depression, and 0.76 for alcoholism, when they were used in general population [11]. For the Primary Care Evaluation of Mental Disorders (PRIME-MD), kappa statistics were 0.61 for major depression and 0.71 for alcoholism between primary care physicians and mental health professionals [12].

Because it corrects for chance agreement, the kappa statistic definitely is useful for calculating inter-observer concordance. A limitation, however, is that it depends upon disease prevalence or base rate. Consequently, estimates tend to be low when prevalence is low (especially below 10%), even if a high level agreement of agreement is observed [13–15]. Among the 112 patients screened in this survey, the rates of major depression and alcoholism were 4.5% and 7.1%, respectively. Both major depression and alcoholism have a reported prevalence of 3.7% among new adult patients in Japanese family practice, when the J-MINI is used [16]. Using the Diagnostic Interview Schedule (DIS), the prevalence among first-visit adult patients seen in general medicine outpatient clinics in Japan is reported to be 4.7% for major depression and 9.4% for alcoholism [17]. Among outpatients in hospital-based general practice in the US, the prevalence of major depression ranged from 7 to 19%, and of alcoholism from 3 to 7% [12]. The prevalence has been shown to vary dependent upon the diagnostic tools and the settings in which surveys are conducted.

The doctor in this survey, despite exhibiting an acceptable level of diagnostic accuracy for major depression, relative to experts, nonetheless missed all cases of alcoholism. Interestingly, almost every patient for whom a MINI-diagnosed mental health disorder was missed by the doctor had some acute illness, like a common cold or asthma. Because the reason for their visit was specific to physical symptoms, mental health problems might have been overlooked, except for the structured screening instrument. In typical medical practice, physicians often are not very successful diagnosing and treating alcohol-related problems or mental illnesses, likely due not only to their clinical skills, but also the patients' expectations and the constraints of the health care system [2]. In fact, even the MINI, which is the shortest structured psychiatric interview available, may be impractical in a busy clinical practice. Mental health screening may be required in primary care settings.

Our study does have several limitations. First, our results were obtained from a single clinical practice and the sample size was small. These two limitations restrict both the generalizability of our results and the confidence we can place upon them. Second, the physician who served as the attending in this study has extensive experience detecting depression in a primary care setting, and was a member of the research team, which may have led to a higher detection rate of depression than in most clinical practices. Third, convenience, rather than random, sampling was used for a variety of practical reasons, including the lack of research funding and our reluctance to over-burden staff. Further research is warranted using validated Japanese version self-report diagnostic tools, and a larger and more representative sample.

In this survey, even though almost half of major depressive episodes were identified, a rate that is acceptable relative to previously-reported experts, all cases of alcoholism were missed. Because a detailed psychiatric interview might be impractical in a busy clinical practice, a mental health screening instrument might be beneficial in primary care settings, especially for alcohol-related problems

The Institutional Review Board of the Japanese Association of Family Medicine approved this research.