- Email: [email protected]
The prevalence of depressive disorders in the united kingdom
The Prevalence of Depressive Disorders in the United Kingdom Maurice M. Ohayon, Robert G. Priest, Christian Guilleminault, and Malijaı¨ Caulet Background: The prevalence of major psychiatric disorders in the general population is difficult to pinpoint owing to widely divergent estimates yielded by studies employing different criteria, methods, and instruments. Depressive disorders, which represent a sizable mental health care expense for the public purse, are no exception to the rule. Methods: The prevalence of depressive disorders was assessed in a representative sample (n 5 4972) of the U.K. general population in 1994. Interviews were performed over the telephone by lay interviewers using an expert system that tailored the questionnaire to each individual based on prior responses. Diagnoses and symptoms lists were based on the DSM-IV. Results: Five percent (95% confidence interval 5 4.4 – 5.6%) of the sample was diagnosed by the system with a depressive disorder at the time of the interview, with the rate slightly higher for women (5.9%) than men (4.2%). Unemployed, separated, divorced, and widowed individuals were found to be at higher risk for depression. Depressive subjects were seen almost exclusively by general practitioners (only 3.4% by psychiatrists). Only 12.5% of them consulted their physician seeking mental health treatment, and 15.9% reported being hospitalized in the past 12 months. Conclusions: The study indicates that mental health problems in the community are seriously underdetected by general practitioners, and that these professionals are highly reluctant to refer patients with depressive disorders to the appropriate specialist. Biol Psychiatry 1999;45:300 –307 © 1999 Society of Biological Psychiatry Key Words: Depressive disorder, Epidemiology
From the Centre de Recherche Philippe Pinel de Montre´al, Montre´al, Que´bec, Canada (MMO, MC); University of London, Academic Department of Psychiatry St. Mary’s Hospital, Paterson Centre, London, England (MMO, RGP); and Stanford University School of Medicine, Sleep Disorders Center, Stanford, California (CG). Address reprint requests to Maurice M. Ohayon, MD, DSc, Centre de Recherche Philippe Pinel de Montre´al, 10905, boulevard Henri-Bourassa Est, Montre´al, Que´bec, H1C 1H1, Canada. Received February 28, 1997; revised August 8, 1997; revised November 5, 1997; accepted November 7, 1997.
© 1999 Society of Biological Psychiatry
Introduction
T
he prevalence of major psychiatric disorders in the general population is difficult to pinpoint owing to widely divergent estimates yielded by studies employing different criteria, methods, and instruments. Depressive disorders, which represent a sizable mental health care expense for the public purse, are no exception to the rule. In the U.S., the 1981 Epidemiological Catchment Area (ECA) study, which employed the Diagnostic Interview Schedule (DIS) as an investigative tool, estimated the prevalence of major depression, as defined by the DSMIII, at 1.6% for men and 2.9% for women in the month prior to survey (Regier et al 1988). Utilizing the Composite International Diagnosis Interview modified by the University of Michigan (UM-CIDI), Blazer et al (1994) in the 1990 –92 National Comorbidity Survey (NCS) instead found the past-month prevalence of major depression, as per DSM-III-R criteria, to be 3.8% for men and 5.9% for women. Elsewhere in the world, summarizing the findings of a series of general population studies carried out in 10 countries with the DIS, the Cross-National Collaborative Group (Weissman et al 1996) noted annual rates of DSM-III major depression ranging from 0.8% in Taiwan to 5.8% in Christchurch, New Zealand. Where the U.K. is concerned, the 1978 –79 Camberwell Community Survey (Bebbington et al 1991) investigated depressive disorder, as defined by the ICD-9, using the PSE-CATEGO-ID. Here, the reported rate was 2.2% for men and 4.9% for women. A 1993 survey carried out in Great Britain by the Office of Population Censuses and Surveys (Meltzer et al 1995) queried 10,108 adults, instead, with the Clinical Interview Schedule–Revised (CIS-R) (Lewis et al 1992). Of these respondents, 14% scored 12 or higher on the CIS-R for the week prior to interview. The most common diagnosis, as per the World Health Organization International Classification of Diseases (World Health Organization 1993), was mixed anxiety and depressive disorder F41.2 at the rate of 7.7% (9.9% for women and 5.4% for men). Depressive disorder F32 was found in 2.1% of respondents (2.5% for women and 1.7% for men). In another study, Horwath et al (1992) 0006-3223/99/$19.00 PII S0006-3223(98)00011-0
Depressive Disorders in the U.K.
estimated that 24.0% of the general population presented with depressive symptoms, and that these individuals ran four times the risk of developing dysthymia and five times that of suffering a first episode of major depression within a year. The data reported in this article were collected in the course of an epidemiological survey conducted by telephone from June to October 1994, with the broader purpose of investigating sleep habits, sleep-related symptoms, and psychiatric and sleep disorders in a representative sample of the U.K. general population. Depression and associated risk factors were defined as per the DSMIV.
Methods and Materials Determination of the Sample The target population comprised noninstitutionalized residents 15 years of age or over (approximately 45,709,600 people). A representative sample was drawn using a stratified probabilistic approach based on the population distribution for the 11 areas of the U.K., as per 1991 census figures (Scotland, Wales, Northern Ireland, and the eight areas of England: North, Yorkshire and Humberside, East Midlands, West Midlands, East Anglia, South East, South West, and North West). The selection method designed by Kish (1965) served to randomly determine the household member to be interviewed, based on age, gender, and number of residents in the home. Excluded from the study were subjects who did not speak sufficient English or who suffered from a hearing or speech impairment or an illness that precluded being interviewed. Subjects who declined to participate or who gave up before completing half the interview were classified as refusals even though they might have met an exclusion criterion. Phone numbers were dialed at least 10 times at different times of the evening and on different days, including weekdays and weekends, before being replaced. Of the 6249 eligible subjects thus contacted, 4972 agreed to be interviewed, for a participation rate of 79.6%. These included 843 subjects (17.0% of total respondents) who initially refused to take part in the study but changed their minds after being called back. An exhaustive description of the sample selection can be found elsewhere (Ohayon et al 1996a).
Instrument Interviewers used the Sleep-Eval knowledge-based expert system (Ohayon 1994a, 1995a) to conduct the interviews. The Sleep-Eval system is specially designed to administer questionnaires and conduct epidemiological studies in the general population. This tool follows in the tradition of artificial intelligence technology. It includes a nonmonotonic, level-2 inference engine endowed with a causal reasoning mode that simulates the reasoning process of a psychiatrist. The causal reasoning mode enables the Sleep-Eval system to formulate a series of diagnostic hypotheses based on the responses provided by a subject. The
BIOL PSYCHIATRY 1999;45:300 –307
301
nonmonotonic, level-2 inference engine examines these hypotheses and confirms or rejects them through further questions and deductions. The system formulates initial diagnostic hypotheses on the basis of responses to a standard set of questions put to all subjects. Concurrent diagnoses are allowed in accordance with the DSM-IV (American Psychiatric Association 1994) and the International Classification of Sleep Disorders or ICSD (1990). The system terminates the interview once all diagnostic possibilities are exhausted. The system has been validated in various contexts and has been demonstrated to be reliable and valid (Ohayon 1994b, 1995b). In general practice, the system’s diagnoses were compared to those of four psychiatrists. An agreement between three of four psychiatrists was used as a gold standard. A consensus between the psychiatrists was achieved for 88 on 114 patients. This yielded a high kappa: .97 between the consensus diagnoses and the system’s diagnoses. Otherwise, the comparisons of the system’s diagnoses against those of individual psychiatrists (regardless of how psychiatrists agreed together) gave lower kappas, ranging from .44 to .78. Another study compared the system’s diagnoses against those of 10 psychiatrists (Philippe Pinel Institute) on 91 patients (St-Onge and Ohayon 1994). Most of these patients (about 60%) were diagnosed with a psychotic disorder. In this study, the kappa was .48 for specific diagnoses of psychotic disorders, mostly schizophrenia. Finally, the diagnoses obtained for 150 subjects of the general population when the system is used by lay interviewers (2) were compared against those obtained by two clinician psychologists. A kappa of .85 was obtained in the recognition of sleep problems and a kappa of .70 for insomnia disorders. A more exhaustive description of the Eval system and the methodology of the study can be found elsewhere (Ohayon et al in press).
Analyses Three series of questions were used to explore presence of depressive disorders at the time of interview. The first examined the feeling of being down or being depressed, the second covered feelings of hopelessness, and the third assessed loss of interest and lack of pleasure in activities formerly considered pleasant. Responses were analyzed against the following data categories: sociodemographics; current medication intake (indication, trade name, dosage, duration of intake, and prescriber); medical consultations and treatments in the past 12 months; and stressful events in the past 12 months. To compensate for any potential bias from such factors as an uneven response rate across demographic groups and the absence of telephones in some households (7% in the U.K.), a weighting procedure was applied to correct for disparities in the geographical, age, and gender distribution between the sample and the U.K. population. Results are based on weighted n values. Percentages for target variables are given with 95% confidence intervals (CIs) in parentheses (95% CI). Bivariate analyses were performed using the chi-square test with the SPSS software package, version 6.1 for Macintosh. Reported differences were significant at the .05 level or less.
302
BIOL PSYCHIATRY 1999;45:300 –307
Logistic regression was used to identify predictors of depressive disorders (Hosmer and Lemeshow 1989). Frequently, statistical packages assume that studied samples are designed on a simple random basis. When sample selection is based on more complex techniques, such as in our study, this leads to an underestimate of the variance, especially in categories where the n is small. Consequently our logistic regression was performed using the SUDAAN software, which allows an appropriate estimate of the standard errors from complex samples by means of a Taylor series linearization method.
Results Overall At time of interview, 25.5% (95% CI: 24.3–26.7%) of subjects responded positively to at least one of the three series of questions regarding depressive symptoms; however, only 250 subjects (weighted figure) or 5.0% (4.4 – 5.6%) of the sample received a depressive disorder diagnosis. This represents 20% of those who responded positively to one of the three series of screening questions.
Depressive Symptoms Regarding the first series of questions, 9.7% of the sample (or 35.3% of subjects with a depressive disorder) reported feeling down or depressed. Where the second and third series are concerned, hopelessness was reported by 7.7% of the sample (or 25.0% of subjects with a depressive disorder), and loss of interest or pleasure in activities formerly considered pleasant, by 15.5% (or 50.1% of subjects with a depressive disorder). Women were more likely than men to report feeling down or depressed [11.7% (10.5–12.9%) vs. 7.5% (6.4 – 8.6%)], whereas loss of interest or pleasure was mentioned in practically equal proportions by both sexes. In sum, although it was common for people in the general population to report at least one of the three classic symptoms of depression, only about one in five received a depressive disorder diagnosis.
Depressive Disorders As mentioned above, 5.0% of the sample received a depressive disorder diagnosis. These disorders were more frequent in women [5.9% (5– 6.8%)] than men [4.2% (3.4 –5%); p , .01], with a relative risk of 1.4 (1.1–1.8). Major depressive disorder, single episode (296.2x, DSMIV) was the most common diagnosis [3.3% (2.8 –3.8%)], followed by dysthymia disorder (300.4, DSM-IV) [1.5% (1.2–1.8%)]. Table 1 gives the prevalence of depressive disorders against sociodemographic variables. A higher proportion of separated, divorced, or widowed subjects than of married subjects were diagnosed with a depressive disor-
M.M. Ohayon et al
der. Similarly, higher rates of depressive disorders were noted among individuals living alone and among the unemployed receiving income support or social security benefits.
Medication, Alcohol, Caffeine, and Cigarettes Of the subjects diagnosed with a depressive disorder, 5.2% (n 5 13) reported using hypnotic drugs, 3.3% (n 5 8) used anxiolytics, 6.7% (n 5 21) took antidepressants, and 2.7% (n 5 7) consumed other types of psychotropic drugs. The most common drugs were temazepam (n 5 8), diazepam (n 5 7), nitrazepam (n 5 6), dothiepin (n 5 5), and fluoxetine (n 5 5). These were most often prescribed by a general practitioner (81.3%), with the remainder (18.7%) being obtained on a psychiatrist’s prescription. Subjects who presented with depressive symptoms but received no diagnosis also consumed psychotropics, but only half as often as those diagnosed with a depressive disorder. Hypnotic drugs were used by 2.0% (n 5 23) of subjects with depressive symptoms only, anxiolytics by 1.6% (n 5 19), antidepressants by 2.3% (n 5 27), and other psychotropic drugs by 1.5% (n 5 17). Psychotropic medications were prescribed by general practitioners in 78.2% of these cases, and by psychiatrists in 3.7%. Daily use of tobacco was more common among depressive disorder subjects than in the rest of the sample (33.8% vs. 22.2%; p , .0001). Depressive disorder subjects were also heavier alcohol drinkers (32.3% vs. 8.9% for $6 alcoholic drinks per day; p , .001). No differences were noted regarding daily intake of coffee or tea.
Medical Consultations, Hospitalizations, and Physical Illnesses The proportion of subjects who consulted a physician in the past 12-month period was higher for the group diagnosed with a depressive disorder (77.4% vs. 60.2%; p , .0001). The number of consultations, too, was higher among depressive disorder subjects; 23.4% of those who saw a doctor had six or more consultations compared with only 11.4% of no-diagnosis subjects (p , .0001). Nearly all the medical consultations were with a general practitioner (over 96% in both groups). Psychiatrists were consulted by only 3.4% of subjects diagnosed with a depressive disorder and 0.4% of the rest of the sample (p , .0001). Asked whether their physician had diagnosed in them a psychiatric illness, 16.1% of subjects diagnosed by the system with a depressive disorder responded affirmatively, compared with 3.8% of no-diagnosis subjects (p , .0001). Hospitalization in the 12-month period prior to interview was reported by 15.9% of subjects with a depressive
Depressive Disorders in the U.K.
BIOL PSYCHIATRY 1999;45:300 –307
303
Table 1. Prevalence of Current Depressive Disorders in the General Population of the U.K. According to Sociodemographic Characteristics
Age groups (years) 15–24 25–34 35– 44 45–54 55– 64 $65 Marital status Single Married Separated/divorced Widowed Household composition Lives alone Lives with spouse Lives with parents Lives with someone Employment Retired Unemployed Housewife Student Without work— other Blue collar/farmers Executives and white collar Craftsmen/tradesmen Liberal profession
n
Men [% (95% CI)]
Women [% (95% CI)]
Total [% (95% CI)]
859 935 855 711 631 980
4.8 (2.8 – 6.8) 5.6 (3.5–7.7) 3.8 (2.0 –5.6) 4.9 (2.7–7.1) 2.9 (1.0 – 4.8) 2.6 (1.0 – 4.2)
5.3 (3.2–7.4) 6.8 (4.5–9.1) 6.0 (3.8 – 8.2) 4.5 (2.3– 6.7) 6.1 (3.5– 8.7) 6.2 (4.2– 8.2)
5.0 (3.5– 6.5) 6.2 (4.7–7.7) 4.9 (3.5– 6.3) 4.7 (3.1– 6.3) 4.6 (3.0 – 6.2) 4.7 (3.4 – 6.0)
1488 2513 367 604
5.9 (4.2–7.6)a 2.8 (1.9 –3.7) 6.5 (2.5–10.5) 4.5 (1.4 –7.6)
6.1 (4.3–7.9) 4.5 (3.3–5.7) 8.8 (5.1–12.5) 7.9 (5.4 –10.4)
6.0 (4.8 –7.2) 3.7 (3.0 – 4.4) 7.9 (5.1–10.7)a 6.9 (4.9 – 8.9)a
1677 2321 598 376
5.5 (3.8 –7.2)a 2.8 (1.9 –3.7) 5.5 (3.0 – 8.0) 5.7 (2.5– 8.9)
7.0 (5.4 – 8.6) 4.6 (3.4 –5.8) 5.1 (2.5–7.7) 9.3 (5.0 –13.6)
6.4 (5.2–7.6)a 3.7 (2.9 – 4.5) 5.3 (3.5–7.1) 7.4 (4.8 –10.0)a
1119 186 305 317 270 920 1330 212 313
2.3 (0.9 –3.7) 8.8 (3.7–13.9)a — 4.3 (1.1–7.5) 8.5 (3.9 –13.1)a 3.4 (1.9 – 4.9) 4.0 (2.5–5.5) 5.4 (1.8 –9.0) 4.5 (1.0 – 8.0)
6.1 (4.3–7.9) 18.1 (8.7–27.5)a 4.9 (2.5–7.3) 4.8 (1.5– 8.1) 12.7 (7.0 –18.4)a 5.5 (3.2–7.8) 4.6 (3.0 – 6.2) 5.8 (0.0 –11.9) 4.1 (1.2–7.0)
4.5 (3.3–5.7) 11.9 (7.2–16.6)a 4.8 (2.4 –7.2) 4.5 (2.2– 6.8) 10.5 (6.8 –14.2)a 4.2 (2.9 –5.5) 4.3 (3.2–5.4) 5.5 (2.4 – 8.6) 4.3 (2.1– 6.5)
p , .05 compared with the lowest value in the same column.
a
disorder and 10.8% of those without (p , .05). Arthritis, heart disease, backache, and pain in limbs were more frequently observed in depressive disorders subjects than in the rest of the sample.
Logistic Regression Analysis The logistic regression model initially included 14 variables found to be significant in univariate analyses. The following variables subsequently proved insignificant and were consequently dropped: annual family income, household composition, educational level, and hospitalization in the past 12-month period. Table 2 gives the risk factors associated with the presence of a depressive disorder. Heavy alcohol consumption and heavy smoking were the strongest risk factors, with odds ratios of 3.4 and 3.8, respectively. The model also indicated that the following variables were significantly associated with the presence of a depressive disorder: unemployed; single, separated, divorced, or widowed; medical consultation in past 12 months; painful affliction; and bereavement in the past 12-month period.
Discussion Methodological Aspects TELEPHONE INTERVIEWS. A novel methodology was employed in this study, namely, telephone interviews performed with the help of an artificial intelligence tool (the expert system). This is an attractive alternative offering many advantages over face-to-face interviews. For one, telephone inquiries are estimated to be one fourth to one half as costly (Davis et al 1993; Weeks et al 1983). In addition, subjects can be selected based on a broader sampling framework (Fenig et al 1993), and the time needed to complete the study is shorter (12 weeks in our case). Previous studies have shown also that response rates are comparable to or better than those obtained with face-to-face interviews (Aneshensel et al 1982; Weeks et al 1983; Fenig et al 1993). Furthermore, the discomfort involved in responding to emotionally charged questions, as is the case in depression assessment, is mitigated by the partial anonymity granted by the telephone and, consequently, may reduce bias and increase the validity of responses (Fenig et al 1993).
304
M.M. Ohayon et al
BIOL PSYCHIATRY 1999;45:300 –307
Table 2. Factors Associated with Depressive Disorders: Multiple Logistic Regression Model Variable Woman Age groups (years) 15–24 25–34 35– 44 45–54 55– 64 Employment Unemployed Student Retired Marital status Single Separated/divorced Widowed Medical consultation Bereavement Heart disease Painful affliction Alcohol consumption 1–2 drinks/day 3–5 drinks/day $6 drinks/day Cigarette consumption 1–20 cigarettes/day 21–35 cigarettes/day .35 cigarettes/day
Odds ratio (95% CI)
p
1.3 (1.0 –1.8)
.07
1.0 (0.4 –2.3) 1.3 (0.7–2.8) 1.1 (0.5–2.2) 0.9 (0.4 –1.8) 0.8 (0.5–1.5)
.97 .44 .90 .66 .56
1.5 (1.1–2.1) 0.9 (0.4 –2.0) 0.6 (0.3–1.2)
.03 .85 .14
1.7 (1.1–2.8) 1.8 (1.1–2.8) 1.7 (1.2–2.4) 2.0 (1.4 –2.8) 2.7 (1.9 –3.9) 1.4 (0.7–2.7) 2.1 (1.4 –3.2)
.02 .01 .01 .001 .001 .34 .001
0.6 (0.3–1.2) 1.0 (0.5–2.1) 3.4 (1.4 – 8.4)
.16 .93 .01
1.4 (1.1–2.0) 1.9 (1.0 –3.4) 3.8 (1.3–11.4)
.06 .03 .02
Categories of reference: man; $65 years old; worker; married; no medical consultation; absence of bereavement; absence of heart disease; absence of painful affliction; no alcohol; no smoking.
Numerous studies have recently compared the reliability of psychiatric telephone interviews with face-to-face interviews and found good agreement between the two methods (Aneshensel et al 1982; Wells et al 1988; Watson et al 1992; Fenig et al 1993). For example, comparing a telephone-administered version of the depression section of the NIMH Diagnostic Interview Schedule (DIS) with its face-to-face version, Wells et al (1988) found sensitivity, specificity, and positive predictive values to be comparable. INSTRUMENT. There are also many advantages to using an expert system over traditional paper-and-pencil questionnaires. The time required to train interviewers is short (about 6 hours). Missing data are practically nonexistent, and no manual data entry is needed, which minimizes the risk of error during data transcription or encoding. In addition, the system’s reasoning ability shortens interviews for subjects without mental health problems by preempting irrelevant questions. Thus, mean duration of interview in this study was 33.93 6 18.02 min. It also ensures uniformity of administration as questions are
always asked in the same manner. Moreover, the knowledge base has the capacity to accommodate multiple classifications. For the purposes of this survey, the Eval system covered the diagnostic criteria of the DSM-IV and the ICSD. Finally, the structure of the knowledge-based system allows for multiple diagnoses whenever permitted by the classifications in question. A number of other successful attempts at using computers in epidemiological surveys have been reported (Clayer et al 1992). LIMITS. Although the participation rate for the survey was relatively low at 79.6%, it is comparable to those in other studies, namely, the NIMH survey (Regier et al 1988), which registered rates from 68% to 79%, and the NCS (Blazer et al 1994), where it reached 82.4%. In a attempt to minimize the refusal rate, subjects who initially declined to participate in the survey were called back within a 3-week interval. This procedure enabled us to boost the participation rate by 13.5%. There was a significantly higher proportion of initial refusers thus recuperated among the elderly: 22.5% among subjects aged 65 years or older, 15.8% among those aged 45– 64, and 14.8% among those aged 15– 44 (p , .0001). This is consistent with studies that analyzed the characteristics of refusers (Friedman and Wasserman 1978). Although reasons for refusing were not systematically collected, interviewers reported that fears and uncertainty regarding confidentiality were often voiced by respondents when called back. Also, the portion of U.K. households without a telephone is 7%, a rate somewhat higher than in other industrialized countries, such as France (6%) and Canada (2%). The rate is low enough, however, for it not to vitiate the validity of inferences (Groves and Kahn 1979; Thornberry and Masse 1988).
Outstanding Results The risk factors for depressive disorders include being female, unemployed, separated, divorced, or widowed. Such associations have been observed by others (Myer et al 1984; Tennant 1985; Cross-National Collaborative Group 1992; Meltzer et al 1995; Weissman et al 1996). Although we could not confirm that younger subjects were at higher risk for depressive disorders, a higher proportion of men with depressive disorders was observed in the younger age groups. The prevalence of depressive disorders (5%) was similar to that registered by Blazer et al (1994) in the NCS with the UM-CIDI, but twice as high as that observed by Regier et al (1988) in the ECA survey with the DIS. In the U.K., we found a higher rate of men with depressive disorders than reported in the Camberwell Community
Depressive Disorders in the U.K.
Survey (1991), where the PSE-CATEGO-ID was used (4.2% vs. 2.2%). Our rate, however, was lower compared with the figures obtained in 1993 by the U.K. Office of Population Censuses and Surveys with the CIS-R, where mixed anxiety and depressive disorder (F41.2, ICD-10 classification) was pegged at 7.7% (9.9% for women and 5.4% for men), and depressive disorders (F32, ICD-10) at 2.1% (2.5% for women and 1.7% for men). Several factors may account for these differences. First the studies differ in terms of sampling framework. For example, the ECA survey (Regier et al 1988) and the Camberwell Community Survey (Bebbington et al 1991) were both local surveys, whereas this study was performed with a representative sample of the total U.K. population. Second, these studies were conducted 10 years ago. In this connection, temporal trends for major depression were reported in 1992 by the Cross-National Collaborative Group (1992); overall, rates of major depression in the general population have tended to rise across generations, historical periods, and cultural boundaries. Finally, the diversity of assessment tools and classifications employed across studies to determine the prevalence of depressive disorders must be underscored. Our approach for diagnosing depressive disorders was similar to the structure of the UM-CIDI (Blazer et al 1994). In the present study, three series of questions were used to detect the possibility of depressive disorders instead of just the one, as is the case with the DIS (Regier et al 1988). The PSE-CATEGO-ID (Bebbington et al 1991) operates along different lines, using severity of depressive symptoms to find a correspondence with the ICD-9. In the study by the Office of Population Censuses and Surveys (Meltzer et al 1995), the diagnosis of mixed anxiety and depressive disorder was assigned whenever a score of 12 or higher was obtained on the Clinical Interview Schedule–Revised (Lewis et al 1992), which is based on the ICD-10 classification. It is important to note that, in this latter study, the criteria for specific anxiety and depressive disorders were not met. The diagnosis was, in this case, a catch-all category. Although a majority of subjects diagnosed by the system with a depressive disorder had consulted a physician in the past 12-month period (77.4%), only a few reported having been diagnosed with or treated for a depressive disorder (12.5% of the 250 subjects). This is consistent with other studies, which found that a small proportion of subjects with depressive disorders seek mental health treatment (Dew et al 1988; Olfson and Klerman 1992). Moreover, the situation is exacerbated by evidence that no diagnosis is made in about 50% of individuals who consult a physician (Golberg and Huxley 1980). In sum, depressive disorder appears to be a frequently
BIOL PSYCHIATRY 1999;45:300 –307
305
missed diagnosis and consequently remains untreated. This is a major health issue, considering the cost of untreated depression to both the individual and society (Rupp 1995; Rice and Miller 1995). Depressive symptoms, however, may be misleading general practitioners. The patient’s primary motive for consultation is rarely depressive mood (Schurman et al 1985) and more likely a symptom of depression, such as sleep disturbances or somatic complaints (e.g., headaches, backaches, fatigue and appetite change, or a physical disease). This was reflected in a community survey where subjects stated they felt uncomfortable telling their physician they thought they had an emotional or depressive problem and, consequently, were more likely to report only a related physical complaint or insomnia (Vize and Priest 1993). Over the past two decades, efforts have been made around the world to improve the ability of family physicians to detect mental disorders. In the U.S., the National Institute of Mental Health (NIMH) recommended that general practitioners receive greater training in recognizing mental disorders (Taube and Burns 1988). Studies assessing the impact of such training, however, have yielded mixed results. Jones et al (1988) assessed the impact of psychiatric training on family physicians and found that 5 years later they still missed 71% of depressive disorders. The Swedish experience shows that the beneficial effects of educational programs are time-related (Rutz et al 1992). They noted that, to ensure the long-term benefits of such programs, training should be repeated every 2 years. In a World Health Organization study carried out in 15 centers around the world (Golberg and Gater 1996), it was found that the detection of mental disorders by primary care physicians is related to the clinic’s style of service delivery, that is, in clinics where patients always see the same physician, recognition is higher. Detection is also higher when physicians work closely with psychiatrists and, needless to say, when the mental disorder is severe. Educational programs must be continuous to provide physicians with information on the constellation of symptoms that lead to a diagnosis of depressive disorder. This must be accompanied by efforts to elucidate the most appropriate treatment for these disorders, given that, as recent epidemiological studies have indicated, over 90% of psychotropic medications are prescribed by general practitioners (Ohayon et al 1996a). In the U.K., the Defeat Depression Campaign was launched in 1992 by the Royal College of Psychiatrists in association with the Royal College of General Practitioners. The purpose of the campaign was to encourage the population to seek help when depressed and to educate primary care physicians to better recognize and treat depression. At this point in time, the impact of the campaign has still to be assessed;
306
BIOL PSYCHIATRY 1999;45:300 –307
however, the present study, which was undertaken about 18 months after the campaign began, suggests that its goals and expectations have yet to be met.
This study was supported by the “Fond de la Recherche en Sante du Quebec” (FRSQ) and an educational grant from Synthelabo Group.
References American Psychiatric Association (1994): Diagnostic and Statistical Manual of Mental Disorders, 4th ed. Washington, DC: American Psychiatric Press. Aneshensel CS, Frerichs RR, Clark VA, Yokopenic PA (1982): Measuring depression in a community: A comparison of telephone and personal interviews. Public Opin Q 46:110 – 121. Bebbington PE, Hurry J, Tennant C (1991): The Camberwell Community Survey: A summary of results. Soc Psychiatry Psychiatr Epidemiol 26:195–201. Blazer DG, Kessler RC, McGonagle KA, Swartz MS (1994): The prevalence and distribution of major depression in a national community sample: The National Comorbidity Survey. Am J Psychiatry 151:979 –986. Clayer JR, McFarlane AC, Wright G (1992): Epidemiology by computer. Soc Psychiatry Psychiatr Epidemiol 27:258 –262. Cross-National Collaborative Group (1992): The changing rate of major depression: Cross-national comparisons. JAMA 268: 3098 –30105. Davis PB, Yee RL, Chetwynd J, McMillan N (1993): The New Zealand partner relations survey: Methodological results of a national telephone survey. AIDS 7:1509 –1516. Dew MA, Dunn LO, Bromet EJ, Schulberg HC (1988): Factors affecting help-seeking during depression in a community sample. J Affect Disord 14:223–224. Fenig S, Levav I, Kohn R, Yelin N (1993): Telephone vs face-to-face interviewing in a community psychiatric survey. Am J Public Health 83:896 – 898. Friedman M, Wasserman IM (1978): Characteristics of respondents and non-respondents in a telephone survey study of elderly consumers. Psychol Rep 42:714. Goldberg D, Gater R (1996): Implications of the World Health Organization study of mental illness in general health care for training primary care staff. Br J Gen Pract 46:483– 485. Golberg D, Huxley P (1980): Mental Illness in the Community. The Pathway to Psychiatric Care. London: Tavistock. Groves RM, Kahn RL (1979): Surveys by Telephone: A National Comparison with Personal Interviews. New York: Academic Press. Horwath E, Johnson J, Klerman G, Weissman MM (1992): Depressive symptoms as relative and attributable risk factors for first-onset major depression. Arch Gen Psychiatry 49: 817– 823. Hosmer DW, Lemeshow S (1989): Applied Logistic Regression. New York: Wiley. Kish L (1965): Survey Sampling. New York: Wiley. Lewis G, Pelosi AJ, Araya RC, Dunn G (1992): Measuring
M.M. Ohayon et al
psychiatric disorder in the community: A standardized assessment for use by lay interviewers. Psychol Med 22:465– 486. Meltzer H, Gill B, Petticrew M, Hinds K (1995): The prevalence of psychiatric morbidity among adults living in private households. OPCS Surveys of Psychiatric Morbidity in Great Britain: Report 1. London: HMSO. Myer JK, Weissman MM, Tischler GL, Holzer CE III, Leaf PJ, Orvaschel H, et al (1984): Six-month prevalence of psychiatric disorders in three communities: 1980 to 1982. Arch Gen Psychiatry 41:959 –967. Ohayon M (1994a): Use of an expert system (Eval) in mental health epidemiological surveys. In: Barahona P, Veloso M, Bryant J, editors. Proceedings of the 12th International Congress of the European Federation for Medical Informatics. Lisbon: Medical Informatics in Europe, pp 174 –179. Ohayon M (1994b): Validation of a knowledge based system (Adinfer) versus human experts. In: Barahona P, Veloso M, Bryant J, editors. Proceedings of the 12th International Congress of the European Federation for Medical Informatics. Lisbon: Medical Informatics in Europe, pp 90 –95. Ohayon M (1995a): Validation of expert systems: Examples and considerations. Medinfo 8:1071–1075. Ohayon M (1995b): Knowledge Based System Eval: Decisional Trees and Questionnaires. Quebec: Bibliothe`que Nationale du Que´bec. Ohayon MM, Caulet M (1996): Psychotropic medication and insomnia complaints in two epidemiological studies. Can J Psychiatry 41:457– 464. Ohayon MM, Caulet M, Priest RG, Guilleminault C (1996): Hypnagogic and hypnopompic hallucinations: Pathological phenomena? Br J Psychiatry 169:459 – 467. Ohayon MM, Guilleminault C, Paiva T, Priest RG, Rapoport DM, Sagales T, et al (1997): An international study on sleep disorders in the general population: Methodological aspects. Sleep 20:1086 –1092. Olfson M, Klerman GL (1992): Depressive symptoms and mental health service utilization in a community sample. Soc Psychiatry Psychiatr Epidemiol 27:161–167. Regier DA, Boyd JH, Burke JD Jr, Rae DS, Myers JK, Kramer M, et al (1988): One-month prevalence of mental disorders in the United States: Based on five epidemiologic catchment area sites. Arch Gen Psychiatry 45:977–986. Rice DP, Miller LS (1995): The economic burden of affective disorders. Br J Psychiatry 166(suppl 27):34 – 42. Rupp A (1995): The economic consequences of not treating depression. Br J Psychiatry 166(suppl 27):29 –33. Rutz W, von Knorring L, Walinder J (1992): Long-term effects of an educational program for general practitioners given by the Swedish Committee for the Prevention and Treatment of Depression. Acta Psychiatr Scand 85:83– 88. Schurman RA, Kramer PD, Mitchell JB (1985): The hidden mental health network. Treatment of mental illness by nonpsychiatrist physicians. Arch Gen Psychiatry 42:89 –94. St-Onge B, Ohayon M (1994): L’utilisation du syste`me Expertal dans un milieu de psychiatrie le´gale. Paris: Abre´ge´s du Congre`s de Psychiatrie et de Neurologie de Langue Franc¸aise, 112. Taube CA, Burns BJ (1988): Mental health services system research: The National Institute of Mental Health program. Health Serv Res 22:837– 855.
Depressive Disorders in the U.K.
Tennant C (1985): Female vulnerability to depression. Psychol Med 15:733–737. Thornberry OT Jr, Masse JT (1988): Trends in United States telephone coverage across time and subgroups. In: Groves RM, Biemer PP, Lyberg LE, Massey JT, Nicholls L, Waksberg J, editors. Telephone Survey Methodology. New York: Wiley, pp 25–50. Vize CM, Priest RG (1993): Defeat Depression Campaign: Attitudes towards depression. Psychiatr Bull 17:573–574. Watson CG, Anderson PED, Thomas D, Nyberg K (1992): Comparability of telephone and face-to-face diagnostic interview schedules. J Nerv Ment Dis 180:534 –535. Weeks MF, Kulka RA, Lessler JT, Whitmore RW (1983):
BIOL PSYCHIATRY 1999;45:300 –307
307
Personal versus telephone surveys for collecting household health data at the local level. Am J Public Health 73:1389 – 1394. Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S, Hwu HG, et al (1996): Cross-national epidemiology of major depression and bipolar disorder. JAMA 276:293–299. Wells KB, Burnam MA, Leake B, Robins LN (1988): Agreement between face-to-face and telephone-administered versions of the depression section of the NIMH Diagnostic Interview Schedule. J Psychiatr Res 22:207–220. World Health Organization (1993): ICD-10 Classifications of Mental and Behavioural Disorders: Diagnostic Criteria for Research. Geneva: World Health Organization.
From the Centre de Recherche Philippe Pinel de Montre´al, Montre´al, Que´bec, Canada (MMO, MC); University of London, Academic Department of Psychiatry St. Mary’s Hospital, Paterson Centre, London, England (MMO, RGP); and Stanford University School of Medicine, Sleep Disorders Center, Stanford, California (CG). Address reprint requests to Maurice M. Ohayon, MD, DSc, Centre de Recherche Philippe Pinel de Montre´al, 10905, boulevard Henri-Bourassa Est, Montre´al, Que´bec, H1C 1H1, Canada. Received February 28, 1997; revised August 8, 1997; revised November 5, 1997; accepted November 7, 1997.
© 1999 Society of Biological Psychiatry
Introduction
T
he prevalence of major psychiatric disorders in the general population is difficult to pinpoint owing to widely divergent estimates yielded by studies employing different criteria, methods, and instruments. Depressive disorders, which represent a sizable mental health care expense for the public purse, are no exception to the rule. In the U.S., the 1981 Epidemiological Catchment Area (ECA) study, which employed the Diagnostic Interview Schedule (DIS) as an investigative tool, estimated the prevalence of major depression, as defined by the DSMIII, at 1.6% for men and 2.9% for women in the month prior to survey (Regier et al 1988). Utilizing the Composite International Diagnosis Interview modified by the University of Michigan (UM-CIDI), Blazer et al (1994) in the 1990 –92 National Comorbidity Survey (NCS) instead found the past-month prevalence of major depression, as per DSM-III-R criteria, to be 3.8% for men and 5.9% for women. Elsewhere in the world, summarizing the findings of a series of general population studies carried out in 10 countries with the DIS, the Cross-National Collaborative Group (Weissman et al 1996) noted annual rates of DSM-III major depression ranging from 0.8% in Taiwan to 5.8% in Christchurch, New Zealand. Where the U.K. is concerned, the 1978 –79 Camberwell Community Survey (Bebbington et al 1991) investigated depressive disorder, as defined by the ICD-9, using the PSE-CATEGO-ID. Here, the reported rate was 2.2% for men and 4.9% for women. A 1993 survey carried out in Great Britain by the Office of Population Censuses and Surveys (Meltzer et al 1995) queried 10,108 adults, instead, with the Clinical Interview Schedule–Revised (CIS-R) (Lewis et al 1992). Of these respondents, 14% scored 12 or higher on the CIS-R for the week prior to interview. The most common diagnosis, as per the World Health Organization International Classification of Diseases (World Health Organization 1993), was mixed anxiety and depressive disorder F41.2 at the rate of 7.7% (9.9% for women and 5.4% for men). Depressive disorder F32 was found in 2.1% of respondents (2.5% for women and 1.7% for men). In another study, Horwath et al (1992) 0006-3223/99/$19.00 PII S0006-3223(98)00011-0
Depressive Disorders in the U.K.
estimated that 24.0% of the general population presented with depressive symptoms, and that these individuals ran four times the risk of developing dysthymia and five times that of suffering a first episode of major depression within a year. The data reported in this article were collected in the course of an epidemiological survey conducted by telephone from June to October 1994, with the broader purpose of investigating sleep habits, sleep-related symptoms, and psychiatric and sleep disorders in a representative sample of the U.K. general population. Depression and associated risk factors were defined as per the DSMIV.
Methods and Materials Determination of the Sample The target population comprised noninstitutionalized residents 15 years of age or over (approximately 45,709,600 people). A representative sample was drawn using a stratified probabilistic approach based on the population distribution for the 11 areas of the U.K., as per 1991 census figures (Scotland, Wales, Northern Ireland, and the eight areas of England: North, Yorkshire and Humberside, East Midlands, West Midlands, East Anglia, South East, South West, and North West). The selection method designed by Kish (1965) served to randomly determine the household member to be interviewed, based on age, gender, and number of residents in the home. Excluded from the study were subjects who did not speak sufficient English or who suffered from a hearing or speech impairment or an illness that precluded being interviewed. Subjects who declined to participate or who gave up before completing half the interview were classified as refusals even though they might have met an exclusion criterion. Phone numbers were dialed at least 10 times at different times of the evening and on different days, including weekdays and weekends, before being replaced. Of the 6249 eligible subjects thus contacted, 4972 agreed to be interviewed, for a participation rate of 79.6%. These included 843 subjects (17.0% of total respondents) who initially refused to take part in the study but changed their minds after being called back. An exhaustive description of the sample selection can be found elsewhere (Ohayon et al 1996a).
Instrument Interviewers used the Sleep-Eval knowledge-based expert system (Ohayon 1994a, 1995a) to conduct the interviews. The Sleep-Eval system is specially designed to administer questionnaires and conduct epidemiological studies in the general population. This tool follows in the tradition of artificial intelligence technology. It includes a nonmonotonic, level-2 inference engine endowed with a causal reasoning mode that simulates the reasoning process of a psychiatrist. The causal reasoning mode enables the Sleep-Eval system to formulate a series of diagnostic hypotheses based on the responses provided by a subject. The
BIOL PSYCHIATRY 1999;45:300 –307
301
nonmonotonic, level-2 inference engine examines these hypotheses and confirms or rejects them through further questions and deductions. The system formulates initial diagnostic hypotheses on the basis of responses to a standard set of questions put to all subjects. Concurrent diagnoses are allowed in accordance with the DSM-IV (American Psychiatric Association 1994) and the International Classification of Sleep Disorders or ICSD (1990). The system terminates the interview once all diagnostic possibilities are exhausted. The system has been validated in various contexts and has been demonstrated to be reliable and valid (Ohayon 1994b, 1995b). In general practice, the system’s diagnoses were compared to those of four psychiatrists. An agreement between three of four psychiatrists was used as a gold standard. A consensus between the psychiatrists was achieved for 88 on 114 patients. This yielded a high kappa: .97 between the consensus diagnoses and the system’s diagnoses. Otherwise, the comparisons of the system’s diagnoses against those of individual psychiatrists (regardless of how psychiatrists agreed together) gave lower kappas, ranging from .44 to .78. Another study compared the system’s diagnoses against those of 10 psychiatrists (Philippe Pinel Institute) on 91 patients (St-Onge and Ohayon 1994). Most of these patients (about 60%) were diagnosed with a psychotic disorder. In this study, the kappa was .48 for specific diagnoses of psychotic disorders, mostly schizophrenia. Finally, the diagnoses obtained for 150 subjects of the general population when the system is used by lay interviewers (2) were compared against those obtained by two clinician psychologists. A kappa of .85 was obtained in the recognition of sleep problems and a kappa of .70 for insomnia disorders. A more exhaustive description of the Eval system and the methodology of the study can be found elsewhere (Ohayon et al in press).
Analyses Three series of questions were used to explore presence of depressive disorders at the time of interview. The first examined the feeling of being down or being depressed, the second covered feelings of hopelessness, and the third assessed loss of interest and lack of pleasure in activities formerly considered pleasant. Responses were analyzed against the following data categories: sociodemographics; current medication intake (indication, trade name, dosage, duration of intake, and prescriber); medical consultations and treatments in the past 12 months; and stressful events in the past 12 months. To compensate for any potential bias from such factors as an uneven response rate across demographic groups and the absence of telephones in some households (7% in the U.K.), a weighting procedure was applied to correct for disparities in the geographical, age, and gender distribution between the sample and the U.K. population. Results are based on weighted n values. Percentages for target variables are given with 95% confidence intervals (CIs) in parentheses (95% CI). Bivariate analyses were performed using the chi-square test with the SPSS software package, version 6.1 for Macintosh. Reported differences were significant at the .05 level or less.
302
BIOL PSYCHIATRY 1999;45:300 –307
Logistic regression was used to identify predictors of depressive disorders (Hosmer and Lemeshow 1989). Frequently, statistical packages assume that studied samples are designed on a simple random basis. When sample selection is based on more complex techniques, such as in our study, this leads to an underestimate of the variance, especially in categories where the n is small. Consequently our logistic regression was performed using the SUDAAN software, which allows an appropriate estimate of the standard errors from complex samples by means of a Taylor series linearization method.
Results Overall At time of interview, 25.5% (95% CI: 24.3–26.7%) of subjects responded positively to at least one of the three series of questions regarding depressive symptoms; however, only 250 subjects (weighted figure) or 5.0% (4.4 – 5.6%) of the sample received a depressive disorder diagnosis. This represents 20% of those who responded positively to one of the three series of screening questions.
Depressive Symptoms Regarding the first series of questions, 9.7% of the sample (or 35.3% of subjects with a depressive disorder) reported feeling down or depressed. Where the second and third series are concerned, hopelessness was reported by 7.7% of the sample (or 25.0% of subjects with a depressive disorder), and loss of interest or pleasure in activities formerly considered pleasant, by 15.5% (or 50.1% of subjects with a depressive disorder). Women were more likely than men to report feeling down or depressed [11.7% (10.5–12.9%) vs. 7.5% (6.4 – 8.6%)], whereas loss of interest or pleasure was mentioned in practically equal proportions by both sexes. In sum, although it was common for people in the general population to report at least one of the three classic symptoms of depression, only about one in five received a depressive disorder diagnosis.
Depressive Disorders As mentioned above, 5.0% of the sample received a depressive disorder diagnosis. These disorders were more frequent in women [5.9% (5– 6.8%)] than men [4.2% (3.4 –5%); p , .01], with a relative risk of 1.4 (1.1–1.8). Major depressive disorder, single episode (296.2x, DSMIV) was the most common diagnosis [3.3% (2.8 –3.8%)], followed by dysthymia disorder (300.4, DSM-IV) [1.5% (1.2–1.8%)]. Table 1 gives the prevalence of depressive disorders against sociodemographic variables. A higher proportion of separated, divorced, or widowed subjects than of married subjects were diagnosed with a depressive disor-
M.M. Ohayon et al
der. Similarly, higher rates of depressive disorders were noted among individuals living alone and among the unemployed receiving income support or social security benefits.
Medication, Alcohol, Caffeine, and Cigarettes Of the subjects diagnosed with a depressive disorder, 5.2% (n 5 13) reported using hypnotic drugs, 3.3% (n 5 8) used anxiolytics, 6.7% (n 5 21) took antidepressants, and 2.7% (n 5 7) consumed other types of psychotropic drugs. The most common drugs were temazepam (n 5 8), diazepam (n 5 7), nitrazepam (n 5 6), dothiepin (n 5 5), and fluoxetine (n 5 5). These were most often prescribed by a general practitioner (81.3%), with the remainder (18.7%) being obtained on a psychiatrist’s prescription. Subjects who presented with depressive symptoms but received no diagnosis also consumed psychotropics, but only half as often as those diagnosed with a depressive disorder. Hypnotic drugs were used by 2.0% (n 5 23) of subjects with depressive symptoms only, anxiolytics by 1.6% (n 5 19), antidepressants by 2.3% (n 5 27), and other psychotropic drugs by 1.5% (n 5 17). Psychotropic medications were prescribed by general practitioners in 78.2% of these cases, and by psychiatrists in 3.7%. Daily use of tobacco was more common among depressive disorder subjects than in the rest of the sample (33.8% vs. 22.2%; p , .0001). Depressive disorder subjects were also heavier alcohol drinkers (32.3% vs. 8.9% for $6 alcoholic drinks per day; p , .001). No differences were noted regarding daily intake of coffee or tea.
Medical Consultations, Hospitalizations, and Physical Illnesses The proportion of subjects who consulted a physician in the past 12-month period was higher for the group diagnosed with a depressive disorder (77.4% vs. 60.2%; p , .0001). The number of consultations, too, was higher among depressive disorder subjects; 23.4% of those who saw a doctor had six or more consultations compared with only 11.4% of no-diagnosis subjects (p , .0001). Nearly all the medical consultations were with a general practitioner (over 96% in both groups). Psychiatrists were consulted by only 3.4% of subjects diagnosed with a depressive disorder and 0.4% of the rest of the sample (p , .0001). Asked whether their physician had diagnosed in them a psychiatric illness, 16.1% of subjects diagnosed by the system with a depressive disorder responded affirmatively, compared with 3.8% of no-diagnosis subjects (p , .0001). Hospitalization in the 12-month period prior to interview was reported by 15.9% of subjects with a depressive
Depressive Disorders in the U.K.
BIOL PSYCHIATRY 1999;45:300 –307
303
Table 1. Prevalence of Current Depressive Disorders in the General Population of the U.K. According to Sociodemographic Characteristics
Age groups (years) 15–24 25–34 35– 44 45–54 55– 64 $65 Marital status Single Married Separated/divorced Widowed Household composition Lives alone Lives with spouse Lives with parents Lives with someone Employment Retired Unemployed Housewife Student Without work— other Blue collar/farmers Executives and white collar Craftsmen/tradesmen Liberal profession
n
Men [% (95% CI)]
Women [% (95% CI)]
Total [% (95% CI)]
859 935 855 711 631 980
4.8 (2.8 – 6.8) 5.6 (3.5–7.7) 3.8 (2.0 –5.6) 4.9 (2.7–7.1) 2.9 (1.0 – 4.8) 2.6 (1.0 – 4.2)
5.3 (3.2–7.4) 6.8 (4.5–9.1) 6.0 (3.8 – 8.2) 4.5 (2.3– 6.7) 6.1 (3.5– 8.7) 6.2 (4.2– 8.2)
5.0 (3.5– 6.5) 6.2 (4.7–7.7) 4.9 (3.5– 6.3) 4.7 (3.1– 6.3) 4.6 (3.0 – 6.2) 4.7 (3.4 – 6.0)
1488 2513 367 604
5.9 (4.2–7.6)a 2.8 (1.9 –3.7) 6.5 (2.5–10.5) 4.5 (1.4 –7.6)
6.1 (4.3–7.9) 4.5 (3.3–5.7) 8.8 (5.1–12.5) 7.9 (5.4 –10.4)
6.0 (4.8 –7.2) 3.7 (3.0 – 4.4) 7.9 (5.1–10.7)a 6.9 (4.9 – 8.9)a
1677 2321 598 376
5.5 (3.8 –7.2)a 2.8 (1.9 –3.7) 5.5 (3.0 – 8.0) 5.7 (2.5– 8.9)
7.0 (5.4 – 8.6) 4.6 (3.4 –5.8) 5.1 (2.5–7.7) 9.3 (5.0 –13.6)
6.4 (5.2–7.6)a 3.7 (2.9 – 4.5) 5.3 (3.5–7.1) 7.4 (4.8 –10.0)a
1119 186 305 317 270 920 1330 212 313
2.3 (0.9 –3.7) 8.8 (3.7–13.9)a — 4.3 (1.1–7.5) 8.5 (3.9 –13.1)a 3.4 (1.9 – 4.9) 4.0 (2.5–5.5) 5.4 (1.8 –9.0) 4.5 (1.0 – 8.0)
6.1 (4.3–7.9) 18.1 (8.7–27.5)a 4.9 (2.5–7.3) 4.8 (1.5– 8.1) 12.7 (7.0 –18.4)a 5.5 (3.2–7.8) 4.6 (3.0 – 6.2) 5.8 (0.0 –11.9) 4.1 (1.2–7.0)
4.5 (3.3–5.7) 11.9 (7.2–16.6)a 4.8 (2.4 –7.2) 4.5 (2.2– 6.8) 10.5 (6.8 –14.2)a 4.2 (2.9 –5.5) 4.3 (3.2–5.4) 5.5 (2.4 – 8.6) 4.3 (2.1– 6.5)
p , .05 compared with the lowest value in the same column.
a
disorder and 10.8% of those without (p , .05). Arthritis, heart disease, backache, and pain in limbs were more frequently observed in depressive disorders subjects than in the rest of the sample.
Logistic Regression Analysis The logistic regression model initially included 14 variables found to be significant in univariate analyses. The following variables subsequently proved insignificant and were consequently dropped: annual family income, household composition, educational level, and hospitalization in the past 12-month period. Table 2 gives the risk factors associated with the presence of a depressive disorder. Heavy alcohol consumption and heavy smoking were the strongest risk factors, with odds ratios of 3.4 and 3.8, respectively. The model also indicated that the following variables were significantly associated with the presence of a depressive disorder: unemployed; single, separated, divorced, or widowed; medical consultation in past 12 months; painful affliction; and bereavement in the past 12-month period.
Discussion Methodological Aspects TELEPHONE INTERVIEWS. A novel methodology was employed in this study, namely, telephone interviews performed with the help of an artificial intelligence tool (the expert system). This is an attractive alternative offering many advantages over face-to-face interviews. For one, telephone inquiries are estimated to be one fourth to one half as costly (Davis et al 1993; Weeks et al 1983). In addition, subjects can be selected based on a broader sampling framework (Fenig et al 1993), and the time needed to complete the study is shorter (12 weeks in our case). Previous studies have shown also that response rates are comparable to or better than those obtained with face-to-face interviews (Aneshensel et al 1982; Weeks et al 1983; Fenig et al 1993). Furthermore, the discomfort involved in responding to emotionally charged questions, as is the case in depression assessment, is mitigated by the partial anonymity granted by the telephone and, consequently, may reduce bias and increase the validity of responses (Fenig et al 1993).
304
M.M. Ohayon et al
BIOL PSYCHIATRY 1999;45:300 –307
Table 2. Factors Associated with Depressive Disorders: Multiple Logistic Regression Model Variable Woman Age groups (years) 15–24 25–34 35– 44 45–54 55– 64 Employment Unemployed Student Retired Marital status Single Separated/divorced Widowed Medical consultation Bereavement Heart disease Painful affliction Alcohol consumption 1–2 drinks/day 3–5 drinks/day $6 drinks/day Cigarette consumption 1–20 cigarettes/day 21–35 cigarettes/day .35 cigarettes/day
Odds ratio (95% CI)
p
1.3 (1.0 –1.8)
.07
1.0 (0.4 –2.3) 1.3 (0.7–2.8) 1.1 (0.5–2.2) 0.9 (0.4 –1.8) 0.8 (0.5–1.5)
.97 .44 .90 .66 .56
1.5 (1.1–2.1) 0.9 (0.4 –2.0) 0.6 (0.3–1.2)
.03 .85 .14
1.7 (1.1–2.8) 1.8 (1.1–2.8) 1.7 (1.2–2.4) 2.0 (1.4 –2.8) 2.7 (1.9 –3.9) 1.4 (0.7–2.7) 2.1 (1.4 –3.2)
.02 .01 .01 .001 .001 .34 .001
0.6 (0.3–1.2) 1.0 (0.5–2.1) 3.4 (1.4 – 8.4)
.16 .93 .01
1.4 (1.1–2.0) 1.9 (1.0 –3.4) 3.8 (1.3–11.4)
.06 .03 .02
Categories of reference: man; $65 years old; worker; married; no medical consultation; absence of bereavement; absence of heart disease; absence of painful affliction; no alcohol; no smoking.
Numerous studies have recently compared the reliability of psychiatric telephone interviews with face-to-face interviews and found good agreement between the two methods (Aneshensel et al 1982; Wells et al 1988; Watson et al 1992; Fenig et al 1993). For example, comparing a telephone-administered version of the depression section of the NIMH Diagnostic Interview Schedule (DIS) with its face-to-face version, Wells et al (1988) found sensitivity, specificity, and positive predictive values to be comparable. INSTRUMENT. There are also many advantages to using an expert system over traditional paper-and-pencil questionnaires. The time required to train interviewers is short (about 6 hours). Missing data are practically nonexistent, and no manual data entry is needed, which minimizes the risk of error during data transcription or encoding. In addition, the system’s reasoning ability shortens interviews for subjects without mental health problems by preempting irrelevant questions. Thus, mean duration of interview in this study was 33.93 6 18.02 min. It also ensures uniformity of administration as questions are
always asked in the same manner. Moreover, the knowledge base has the capacity to accommodate multiple classifications. For the purposes of this survey, the Eval system covered the diagnostic criteria of the DSM-IV and the ICSD. Finally, the structure of the knowledge-based system allows for multiple diagnoses whenever permitted by the classifications in question. A number of other successful attempts at using computers in epidemiological surveys have been reported (Clayer et al 1992). LIMITS. Although the participation rate for the survey was relatively low at 79.6%, it is comparable to those in other studies, namely, the NIMH survey (Regier et al 1988), which registered rates from 68% to 79%, and the NCS (Blazer et al 1994), where it reached 82.4%. In a attempt to minimize the refusal rate, subjects who initially declined to participate in the survey were called back within a 3-week interval. This procedure enabled us to boost the participation rate by 13.5%. There was a significantly higher proportion of initial refusers thus recuperated among the elderly: 22.5% among subjects aged 65 years or older, 15.8% among those aged 45– 64, and 14.8% among those aged 15– 44 (p , .0001). This is consistent with studies that analyzed the characteristics of refusers (Friedman and Wasserman 1978). Although reasons for refusing were not systematically collected, interviewers reported that fears and uncertainty regarding confidentiality were often voiced by respondents when called back. Also, the portion of U.K. households without a telephone is 7%, a rate somewhat higher than in other industrialized countries, such as France (6%) and Canada (2%). The rate is low enough, however, for it not to vitiate the validity of inferences (Groves and Kahn 1979; Thornberry and Masse 1988).
Outstanding Results The risk factors for depressive disorders include being female, unemployed, separated, divorced, or widowed. Such associations have been observed by others (Myer et al 1984; Tennant 1985; Cross-National Collaborative Group 1992; Meltzer et al 1995; Weissman et al 1996). Although we could not confirm that younger subjects were at higher risk for depressive disorders, a higher proportion of men with depressive disorders was observed in the younger age groups. The prevalence of depressive disorders (5%) was similar to that registered by Blazer et al (1994) in the NCS with the UM-CIDI, but twice as high as that observed by Regier et al (1988) in the ECA survey with the DIS. In the U.K., we found a higher rate of men with depressive disorders than reported in the Camberwell Community
Depressive Disorders in the U.K.
Survey (1991), where the PSE-CATEGO-ID was used (4.2% vs. 2.2%). Our rate, however, was lower compared with the figures obtained in 1993 by the U.K. Office of Population Censuses and Surveys with the CIS-R, where mixed anxiety and depressive disorder (F41.2, ICD-10 classification) was pegged at 7.7% (9.9% for women and 5.4% for men), and depressive disorders (F32, ICD-10) at 2.1% (2.5% for women and 1.7% for men). Several factors may account for these differences. First the studies differ in terms of sampling framework. For example, the ECA survey (Regier et al 1988) and the Camberwell Community Survey (Bebbington et al 1991) were both local surveys, whereas this study was performed with a representative sample of the total U.K. population. Second, these studies were conducted 10 years ago. In this connection, temporal trends for major depression were reported in 1992 by the Cross-National Collaborative Group (1992); overall, rates of major depression in the general population have tended to rise across generations, historical periods, and cultural boundaries. Finally, the diversity of assessment tools and classifications employed across studies to determine the prevalence of depressive disorders must be underscored. Our approach for diagnosing depressive disorders was similar to the structure of the UM-CIDI (Blazer et al 1994). In the present study, three series of questions were used to detect the possibility of depressive disorders instead of just the one, as is the case with the DIS (Regier et al 1988). The PSE-CATEGO-ID (Bebbington et al 1991) operates along different lines, using severity of depressive symptoms to find a correspondence with the ICD-9. In the study by the Office of Population Censuses and Surveys (Meltzer et al 1995), the diagnosis of mixed anxiety and depressive disorder was assigned whenever a score of 12 or higher was obtained on the Clinical Interview Schedule–Revised (Lewis et al 1992), which is based on the ICD-10 classification. It is important to note that, in this latter study, the criteria for specific anxiety and depressive disorders were not met. The diagnosis was, in this case, a catch-all category. Although a majority of subjects diagnosed by the system with a depressive disorder had consulted a physician in the past 12-month period (77.4%), only a few reported having been diagnosed with or treated for a depressive disorder (12.5% of the 250 subjects). This is consistent with other studies, which found that a small proportion of subjects with depressive disorders seek mental health treatment (Dew et al 1988; Olfson and Klerman 1992). Moreover, the situation is exacerbated by evidence that no diagnosis is made in about 50% of individuals who consult a physician (Golberg and Huxley 1980). In sum, depressive disorder appears to be a frequently
BIOL PSYCHIATRY 1999;45:300 –307
305
missed diagnosis and consequently remains untreated. This is a major health issue, considering the cost of untreated depression to both the individual and society (Rupp 1995; Rice and Miller 1995). Depressive symptoms, however, may be misleading general practitioners. The patient’s primary motive for consultation is rarely depressive mood (Schurman et al 1985) and more likely a symptom of depression, such as sleep disturbances or somatic complaints (e.g., headaches, backaches, fatigue and appetite change, or a physical disease). This was reflected in a community survey where subjects stated they felt uncomfortable telling their physician they thought they had an emotional or depressive problem and, consequently, were more likely to report only a related physical complaint or insomnia (Vize and Priest 1993). Over the past two decades, efforts have been made around the world to improve the ability of family physicians to detect mental disorders. In the U.S., the National Institute of Mental Health (NIMH) recommended that general practitioners receive greater training in recognizing mental disorders (Taube and Burns 1988). Studies assessing the impact of such training, however, have yielded mixed results. Jones et al (1988) assessed the impact of psychiatric training on family physicians and found that 5 years later they still missed 71% of depressive disorders. The Swedish experience shows that the beneficial effects of educational programs are time-related (Rutz et al 1992). They noted that, to ensure the long-term benefits of such programs, training should be repeated every 2 years. In a World Health Organization study carried out in 15 centers around the world (Golberg and Gater 1996), it was found that the detection of mental disorders by primary care physicians is related to the clinic’s style of service delivery, that is, in clinics where patients always see the same physician, recognition is higher. Detection is also higher when physicians work closely with psychiatrists and, needless to say, when the mental disorder is severe. Educational programs must be continuous to provide physicians with information on the constellation of symptoms that lead to a diagnosis of depressive disorder. This must be accompanied by efforts to elucidate the most appropriate treatment for these disorders, given that, as recent epidemiological studies have indicated, over 90% of psychotropic medications are prescribed by general practitioners (Ohayon et al 1996a). In the U.K., the Defeat Depression Campaign was launched in 1992 by the Royal College of Psychiatrists in association with the Royal College of General Practitioners. The purpose of the campaign was to encourage the population to seek help when depressed and to educate primary care physicians to better recognize and treat depression. At this point in time, the impact of the campaign has still to be assessed;
306
BIOL PSYCHIATRY 1999;45:300 –307
however, the present study, which was undertaken about 18 months after the campaign began, suggests that its goals and expectations have yet to be met.
This study was supported by the “Fond de la Recherche en Sante du Quebec” (FRSQ) and an educational grant from Synthelabo Group.
References American Psychiatric Association (1994): Diagnostic and Statistical Manual of Mental Disorders, 4th ed. Washington, DC: American Psychiatric Press. Aneshensel CS, Frerichs RR, Clark VA, Yokopenic PA (1982): Measuring depression in a community: A comparison of telephone and personal interviews. Public Opin Q 46:110 – 121. Bebbington PE, Hurry J, Tennant C (1991): The Camberwell Community Survey: A summary of results. Soc Psychiatry Psychiatr Epidemiol 26:195–201. Blazer DG, Kessler RC, McGonagle KA, Swartz MS (1994): The prevalence and distribution of major depression in a national community sample: The National Comorbidity Survey. Am J Psychiatry 151:979 –986. Clayer JR, McFarlane AC, Wright G (1992): Epidemiology by computer. Soc Psychiatry Psychiatr Epidemiol 27:258 –262. Cross-National Collaborative Group (1992): The changing rate of major depression: Cross-national comparisons. JAMA 268: 3098 –30105. Davis PB, Yee RL, Chetwynd J, McMillan N (1993): The New Zealand partner relations survey: Methodological results of a national telephone survey. AIDS 7:1509 –1516. Dew MA, Dunn LO, Bromet EJ, Schulberg HC (1988): Factors affecting help-seeking during depression in a community sample. J Affect Disord 14:223–224. Fenig S, Levav I, Kohn R, Yelin N (1993): Telephone vs face-to-face interviewing in a community psychiatric survey. Am J Public Health 83:896 – 898. Friedman M, Wasserman IM (1978): Characteristics of respondents and non-respondents in a telephone survey study of elderly consumers. Psychol Rep 42:714. Goldberg D, Gater R (1996): Implications of the World Health Organization study of mental illness in general health care for training primary care staff. Br J Gen Pract 46:483– 485. Golberg D, Huxley P (1980): Mental Illness in the Community. The Pathway to Psychiatric Care. London: Tavistock. Groves RM, Kahn RL (1979): Surveys by Telephone: A National Comparison with Personal Interviews. New York: Academic Press. Horwath E, Johnson J, Klerman G, Weissman MM (1992): Depressive symptoms as relative and attributable risk factors for first-onset major depression. Arch Gen Psychiatry 49: 817– 823. Hosmer DW, Lemeshow S (1989): Applied Logistic Regression. New York: Wiley. Kish L (1965): Survey Sampling. New York: Wiley. Lewis G, Pelosi AJ, Araya RC, Dunn G (1992): Measuring
M.M. Ohayon et al
psychiatric disorder in the community: A standardized assessment for use by lay interviewers. Psychol Med 22:465– 486. Meltzer H, Gill B, Petticrew M, Hinds K (1995): The prevalence of psychiatric morbidity among adults living in private households. OPCS Surveys of Psychiatric Morbidity in Great Britain: Report 1. London: HMSO. Myer JK, Weissman MM, Tischler GL, Holzer CE III, Leaf PJ, Orvaschel H, et al (1984): Six-month prevalence of psychiatric disorders in three communities: 1980 to 1982. Arch Gen Psychiatry 41:959 –967. Ohayon M (1994a): Use of an expert system (Eval) in mental health epidemiological surveys. In: Barahona P, Veloso M, Bryant J, editors. Proceedings of the 12th International Congress of the European Federation for Medical Informatics. Lisbon: Medical Informatics in Europe, pp 174 –179. Ohayon M (1994b): Validation of a knowledge based system (Adinfer) versus human experts. In: Barahona P, Veloso M, Bryant J, editors. Proceedings of the 12th International Congress of the European Federation for Medical Informatics. Lisbon: Medical Informatics in Europe, pp 90 –95. Ohayon M (1995a): Validation of expert systems: Examples and considerations. Medinfo 8:1071–1075. Ohayon M (1995b): Knowledge Based System Eval: Decisional Trees and Questionnaires. Quebec: Bibliothe`que Nationale du Que´bec. Ohayon MM, Caulet M (1996): Psychotropic medication and insomnia complaints in two epidemiological studies. Can J Psychiatry 41:457– 464. Ohayon MM, Caulet M, Priest RG, Guilleminault C (1996): Hypnagogic and hypnopompic hallucinations: Pathological phenomena? Br J Psychiatry 169:459 – 467. Ohayon MM, Guilleminault C, Paiva T, Priest RG, Rapoport DM, Sagales T, et al (1997): An international study on sleep disorders in the general population: Methodological aspects. Sleep 20:1086 –1092. Olfson M, Klerman GL (1992): Depressive symptoms and mental health service utilization in a community sample. Soc Psychiatry Psychiatr Epidemiol 27:161–167. Regier DA, Boyd JH, Burke JD Jr, Rae DS, Myers JK, Kramer M, et al (1988): One-month prevalence of mental disorders in the United States: Based on five epidemiologic catchment area sites. Arch Gen Psychiatry 45:977–986. Rice DP, Miller LS (1995): The economic burden of affective disorders. Br J Psychiatry 166(suppl 27):34 – 42. Rupp A (1995): The economic consequences of not treating depression. Br J Psychiatry 166(suppl 27):29 –33. Rutz W, von Knorring L, Walinder J (1992): Long-term effects of an educational program for general practitioners given by the Swedish Committee for the Prevention and Treatment of Depression. Acta Psychiatr Scand 85:83– 88. Schurman RA, Kramer PD, Mitchell JB (1985): The hidden mental health network. Treatment of mental illness by nonpsychiatrist physicians. Arch Gen Psychiatry 42:89 –94. St-Onge B, Ohayon M (1994): L’utilisation du syste`me Expertal dans un milieu de psychiatrie le´gale. Paris: Abre´ge´s du Congre`s de Psychiatrie et de Neurologie de Langue Franc¸aise, 112. Taube CA, Burns BJ (1988): Mental health services system research: The National Institute of Mental Health program. Health Serv Res 22:837– 855.
Depressive Disorders in the U.K.
Tennant C (1985): Female vulnerability to depression. Psychol Med 15:733–737. Thornberry OT Jr, Masse JT (1988): Trends in United States telephone coverage across time and subgroups. In: Groves RM, Biemer PP, Lyberg LE, Massey JT, Nicholls L, Waksberg J, editors. Telephone Survey Methodology. New York: Wiley, pp 25–50. Vize CM, Priest RG (1993): Defeat Depression Campaign: Attitudes towards depression. Psychiatr Bull 17:573–574. Watson CG, Anderson PED, Thomas D, Nyberg K (1992): Comparability of telephone and face-to-face diagnostic interview schedules. J Nerv Ment Dis 180:534 –535. Weeks MF, Kulka RA, Lessler JT, Whitmore RW (1983):
BIOL PSYCHIATRY 1999;45:300 –307
307
Personal versus telephone surveys for collecting household health data at the local level. Am J Public Health 73:1389 – 1394. Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S, Hwu HG, et al (1996): Cross-national epidemiology of major depression and bipolar disorder. JAMA 276:293–299. Wells KB, Burnam MA, Leake B, Robins LN (1988): Agreement between face-to-face and telephone-administered versions of the depression section of the NIMH Diagnostic Interview Schedule. J Psychiatr Res 22:207–220. World Health Organization (1993): ICD-10 Classifications of Mental and Behavioural Disorders: Diagnostic Criteria for Research. Geneva: World Health Organization.