 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
| |
|
||||||||||||||||
| |
|
||||||||||||||||
|
|
Performance of the Composite International
Diagnostic Interview Short Form for Major Depression in Community and
Clinical Samples
Abstract The CIDI Short Form is a brief survey instrument designed
to identify episodes of major depression. The instrument was developed
for inclusion in the US National Health Interview Survey, but has also
been used in the Canadian National Population Health Survey (NPHS). In
this study, data deriving from use of the CIDI Short Form in the NPHS
are compared to published data from the Mental Health Supplement of the
Ontario Health Survey, which utilized a fully validated structured interview:
the Composite International Diagnostic Interview (CIDI). In an additional
analysis, the sensitivity and specificity of the Short Form were evaluated
in relation to the full CIDI mood disorders section in a clinical sample
of 122 psychiatric in-patients. Relative to published data from the Ontario
Health Survey, application of the CIDI Short Form in the NPHS resulted
in an overestimation of major depression prevalence by approximately 50%.
In the clinical sample, the CIDI Short Form was highly sensitive (98.4%),
but not highly specific (72.7%). Active medical conditions, substance
use disorders and dysthymia were frequently observed among subjects with
false positive CIDI Short Form ratings. The CIDI Short Form appears to
overestimate the 12-month period prevalence of major depression when it
is applied in community samples. Since the Short Form does not make exclusions
for organically induced symptoms, it is probable that some subjects with
depressive symptoms secondary to physical illnesses and/or drug exposures
score above the instrument's threshold, perhaps leading to an elevation
in period prevalence rates. Introduction In the past 15-20 years, research in psychiatric epidemiology has shifted from an emphasis on psychiatric symptoms and general well-being to a focus on the distribution and determinants of mental disorders. This trend has been facilitated by the development of structured diagnostic interviews that can be conducted by lay interviewers. These instruments allow non-clinicians to collect data, greatly reducing data collection costs. Unfortunately, psychiatric structured diagnostic interviews tend to be lengthy, precluding their inclusion in large-scale surveys such as the US National Health Interview Survey and the Canadian National Population Health Survey (NPHS). Kessler and Mroczek (Survey Research Center, University of Michigan, unpublished observations) have recently developed a brief questionnaire that is designed to predict the occurrence of major depression. The instrument contains a subset of questions from the Composite International Diagnostic Interview (CIDI), which is a validated structured diagnostic instrument.1-3 Development of the short questionnaire began with exploratory regression analyses of CIDI survey data. This was followed by additional validation studies and modifications to question wording (Kessler and Mroczek, unpublished observations). The final result was an abbreviated instrument called the CIDI Short Form for Major Depression. This instrument can be administered, on average, in less than one minute. In recent editions of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IIIR and DSM-IV), major depression is defined as a period of at least two weeks characterized by at least five severe and persistent depressive symptoms. The symptoms must not represent a normal reaction to the death of a loved one and should not be due to an identifiable organic factor such as a physical illness or drug exposure.4,5 The questions contained in the CIDI Short Form inquire about the existence of individual depressive symptoms occurring during the same two-week period in the preceding 12 months. Therefore, inclusion of the CIDI Short Form in an epidemiologic survey leads to an estimate of the 12-month period prevalence of major depression in the population sampled. In the NPHS, an algorithm proposed by Kessler and Mroczek (unpublished observations) was used to score the CIDI Short Form. According to these researchers' validation data, there is a 90% probability of major depression (in the preceding 12 months) in subjects reporting five or more depressive symptoms in response to the CIDI Short Form questions. Hence, a cut-point of five may be used to categorize respondents as having, or not having, major depression. This procedure was used by Beaudet in an analysis of the NPHS major depression data.6 Performance of the CIDI Short Form in the National Population Health Survey The first objective of this study was to compare the 12-month period prevalence of major depression deriving from use of the CIDI Short Form in the NPHS to that of the Mental Health Supplement of the Ontario Health Survey. An estimate of the 12-month period prevalence of major depression using NPHS data has already been reported by Beaudet:6 5.6%. However, Beaudet's estimate applied to subjects aged 18 and over, whereas the Ontario survey reported data for subjects aged 15-64. Therefore, in order to facilitate this comparison, the 12-month period prevalence of major depression was recalculated from the NPHS data set for subjects aged 15-64. In the NPHS, a complex multistage stratified sampling procedure was employed. In the NPHS public use microdata file,7 this complex sampling was accounted for by the calculation of weights; hence, all estimates deriving from this data set are weighted estimates. In this study, confidence limits for these weighted estimates were calculated using approximate methods described in the NPHS microdata file documentation. The 12-month period prevalence of major depression among subjects aged 15-64 in the NPHS was 6.3%, with 95% confidence limits of 5.7% and 6.8%. Offord et al. analyzed data from the Mental Health Supplement of the Ontario Health Survey.8 This survey used a probability sample from the Ontario population aged 15-64. The full CIDI was administered to each subject, and a 4.1% 12-month period prevalence of major depression was reported (the standard error associated with this estimate was 0.4, leading to approximate 95% confidence limits [CL] of 3.3% and 4.9%, a confidence interval that does not overlap with that derived from the NPHS data). One possible explanation for the difference between these two surveys is that the NPHS is based on a national sample, whereas the Ontario Health Survey was restricted to one Canadian province. However, when the NPHS data from the province of Ontario are considered alone, the 12-month period prevalence is similar to that of Canada as a whole: 6.5% (95% CL = 5.6%, 7.4%). A comparison of this prevalence with that reported by Offord et al.8 amounts, essentially, to a comparison of the CIDI Short Form with the full CIDI in independent samples. In relative terms, the CIDI Short Form overestimated the 12-month period prevalence of major depression by approximately 50%. A plausible explanation for the apparent overestimation of prevalence by the CIDI Short Form is a lack of specificity. False positive codings may occur for several reasons. First, depressive symptoms caused by excessive consumption of alcohol, ingestion of abused drugs, medication exposures or physical illnesses are not counted toward a diagnosis of major depression in the DSM-IIIR or the DSM-IV. Such organic exclusions are accounted for in the full CIDI, but not in the CIDI Short Form. Second, uncomplicated bereavement (normal grief) is not regarded as a mental disorder in the DSMs, even if severe and persistent depressive symptoms are present. Unlike the full CIDI, the CIDI Short Form incorporates no distinction between depressive disorders and bereavement. It is conceivable, of course, that the Short Form may be somewhat insensitive because it contains fewer questions focusing on specific depressive symptoms than the full CIDI. For example, the CIDI contains several questions about different manifestations of sleep disturbance; these include questions about initial, middle and terminal insomnia and also hypersomnia. Any of these disturbances, if sufficiently severe and persistent, would fulfil a DSM diagnostic criterion for sleep disturbance that counts toward a diagnosis of major depression. The Short Form contains only a single question referring to initial insomnia. Since depression is sometimes characterized by middle or terminal insomnia, or by hypersomnia, the CIDI Short Form will sometimes miss important manifestations of sleep disturbance. Since the Short Form appears to overestimate period prevalence, the false negative classifications (resulting from lack of sensitivity) are likely outweighed by false positive classifications (resulting from lack of specificity). This is not surprising for two reasons. First, since approximately 95% of the subjects in the sample do not have major depression, even small false positive misclassification rates will tend to increase the frequency observed. For example, in applying traditional models of misclassification bias9 under the assumption that the true prevalence is 4.1%, a 95% sensitivity (assuming perfect specificity) would bias the estimate downward to 3.9%, whereas a specificity of 95% (assuming perfect sensitivity) would bias the estimate upward by a much greater extent, to 8.9%. The CIDI Short Form in a Clinical Sample As outlined in the preceding section, a plausible factor influencing the performance of the CIDI Short Form in community samples may be a lack of specificity in subjects with depressive syndromes related to organic factors and bereavement. In order to evaluate this possibility, the mood disorders sections of the CIDI-Auto (an automated version of the full CIDI) and the CIDI Short Form were administered to a consenting series of 122 psychiatric in-patients. The subjects were consenting participants in a hospital-based case-control study of risk factors for major depression being conducted at the Calgary General Hospital. There were 58 males and 64 females in the sample, ranging in age from 18 to 75 years with a median age of 33 years. The 62 subjects with CIDI-Auto diagnoses of major depression (296.2x - single episode, 296.3x - recurrent, 296.5x - bipolar disorder, depressed) in the preceding 12 months were compared to the remaining 60 subjects on the basis of their performance on the CIDI Short Form. The Short Form was highly sensitive in this sample: 61 of 62 (98.4%) of the subjects with a full CIDI-confirmed major depressive episode scored five or more on the CIDI Short Form. This high degree of sensitivity of the instrument may reflect two aspects of the clinical sample: the severity of the disorders and their course in relation to the testing. Of the 62 cases where there was an episode of major depression in the preceding year, 52 (83.9%) of them were classified in the CIDI-Auto in the severe categories (296.23, 296.33 or 296.53). The single false negative subject on the CIDI Short Form, however, had a diagnosis of major depression, single episode, severe (296.23). Most of the disorders were active and had resulted in the current hospitalization; this may also have facilitated recall. In community samples, where disorders may have resolved months prior to evaluation, the Short Form may be less sensitive. Of the 60 subjects without a major depressive episode in the previous year, 27 had bipolar disorders coded as manic, mixed or unspecified at the time of interview. Many of these subjects had scores of five or greater on the CIDI Short Form. However, these subjects could not be interpreted as false positives because the CIDI-Auto could not confirm that there had been no episode of major depression in the previous year (the interview codes their disorder and the nature of the most recent episode). Of the remaining 33 subjects, there were nine false positives, so that the false positive rate in this sample was 27.3% (specificity = 72.7%). Four of these nine subjects had one or more active medical conditions, three were diagnosed with substance-use disorders. These are both possible reasons for false positive codings, as described above. In addition, four of the subjects had dysthymia. Dysthymia is a chronic form of depressive disorder, typically characterized by fewer and less severe depressive symptoms than major depression. The distinction between these different categories of depressive disorder may require a more extensive evaluation of symptom frequency and severity than can be accomplished using the CIDI Short Form. Summary A comparison between the NPHS findings and those of the Ontario Health Survey suggests an overestimation of period prevalence by the CIDI Short Form. A tendency to overestimate prevalence may be expected since the Short Form's questions focus exclusively on depressive symptoms. The instrument does not exclude depressive symptoms that are related to physical illnesses or drug exposures, nor does it differentiate those circumstances where severe depressive symptoms are not pathological, such as bereavement. The frequency of active physical illnesses and substance use disorders in this study among false positive clinical subjects without major depression provides some confirmation of this theoretical possibility. The brevity of the Short Form makes its inclusion in large-scale general population surveys feasible. Since large-scale national surveys such as the NPHS use probability sampling and incorporate numerous, diverse measures of health status determinants, the inclusion of the CIDI Short Form in the NPHS and other studies may offer opportunities for the advancement of mood disorders epidemiology. However, data presented here indicate that the CIDI Short Form is vulnerable to errors in diagnostic classification. In analytical epidemiologic research, this may translate into systematic error in the estimation of population parameters such as relative risk. The direction of bias may result in a dilution of observed effect (situations where the misclassification rates do not depend on exposure to the risk factors in question: non-differential misclassification bias) or an over- or underestimation of observed effect (in situations where the misclassification rates differ among the different risk factor exposure groups: differential misclassification bias).10 Until additional validation data for the CIDI Short Form are available, estimates deriving from its use must be considered vulnerable to misclassification bias. In studies where the full application of a structured diagnostic interview is feasible, this should be regarded as the preferred method of identifying cases of major depression. In situations where the use of a structured diagnostic interview is not feasible, the use of the CIDI Short Form may be an alternative. However, investigators choosing to use the CIDI Short Form should strongly consider application of a study design that can account for the possibility of misclassification bias. This generally entails the inclusion of an internal or (less preferable) external validation study into the data collection procedures so that error rates associated with the imperfect measure can be estimated and accounted for in the analysis.11 References 1. World Health Organization. The Composite International Diagnostic Interview, Version 1.1. Researcher's manual. Geneva: WHO, 1994. 2. Wittchen HU. Reliability and validity studies of the WHO-Composite International Diagnostic Interview (CIDI): a critical review. J Psychiat Res 1994;28:57-84. 3. Robins LN, Wing J, Wittchen HU, Helzer JE, Babor TF, Burke J, et al. The Composite International Diagnostic Interview. An epidemiological instrument suitable for use in conjunction with different diagnostic systems and in different cultures. Arch Gen Psychiatry 1988;45:1069-77. 4. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington (DC): American Psychiatric Association, 1994. 5. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 3rd ed. Revised. Washington (DC): American Psychiatric Association, 1987. 6. Beaudet MP. Depression. Health Reports 1996;7(4):11-24. (Statistics Canada Cat 82-003.) 7. Statistics Canada (Health Statistics Division). National Population Health Survey, 1994-95. Public use microdata files. Ottawa, 1997. 8. Offord DR, Boyle MH, Campbell D, Goering P, Lin E, Wong M, et al. One year prevalence of psychiatric disorder in Ontarians 15 to 64 years of age. Can J Psychiatry 1996;41:559-63. 9. Rogan WJ, Gladen B. Estimating prevalence from the results of a screening test. Am J Epidemiol 1978;107:71-6. 10. Kleinbaum DG, Kupper LL, Morgenstern H. Chapter 12: "Information bias." In: Epidemiologic research. New York: Van Nostrand Reinhold, 1982:220-41. 11. Greenland S. Variance estimation for epidemiologic effect estimates under misclassification. Stat Med 1989;8:1031-40. Author Reference
|
 |
|||
|
Last Updated: 2002-10-29 |  |
|
