Abstract
Objective
To outline the utility of the Composite International Diagnostic Interview (CIDI) in the diagnosis of psychosis.
Method
Report current situation.
Results
The CIDI was designed as a fully structured interview to be used by lay interviewers. It generates false positive diagnoses in community surveys and false negative diagnoses in psychiatric settings. A new psychosis module has been developed to reduce these problems.
Conclusions
The diagnosis of psychosis by fully structured diagnostic interviews is difficult.
I am here as a committee member of the World Health Organization Composite International Diagnostic Interview (CIDI) Advisory Committee. The CIDI is the most widely used epidemiological screening tool and has been used in some 20 international surveys. It will be used over the next two years in ‘World Mental Health 2000’, a set of 10 different country surveys all of the order of 5000–15 000 respondents. By the end of the year 2000 it will have been administered to more than 400 000 adults in many different countries and languages. The focus of interest is in the more common disorders such as anxiety, depression and substance use disorders and people are not quite as interested in the low prevalence disorders like schizophrenia. In population surveys, it is very expensive to pick up enough people with psychosis to make understandable decisions about what the data is telling you. For this reason when we did our national mental health survey in Australia two years ago, Assen Jablensky headed what was called the Low Prevalence Disorders Survey [1] because it was realised that the national survey of 10 000 people would only return approximately 40 or 50 people who might have schizophrenia.
At another level there is a conceptual problem. The CIDI was designed to be administered by lay interviewers [2]. Professional interviewers ask all the questions listed on the paper but, as they have no clinical training to make judgements about what is psychosis or whether the response is reasonable or unreasonable, they are unable to judge the presence of psychosis. It may well be that something like the CIDI just cannot get near the diagnosis of psychosis. However, maybe it can, and that is what the present paper is about.
In the United States National Comorbidity Survey, the CIDI simply operationalised the diagnostic criteria in DSM-III-R. The questions that were asked in the National Comorbidity Survey were: Do you believe people were spying on you?; plotting against you?; reading your mind?; reading your thoughts?; Could you actually hear what another person was thinking?, etc. The University of Michigan (UM) CIDI does a little more than that, but not much. In the United States National Comorbidity Survey, 28% of Americans answered yes to one of these questions [3]. This demonstrated that asking people about psychotic phenomena is not very specific. Nevertheless, 1.3% met criteria for narrowly defined psychotic illness and 2.2% for what was later called nonaffective psychosis. The people with nonaffective psychosis were then telephoned and interviewed in depth. The detailed history of their illness and their symptoms was reviewed by Kendler, who has had a lot of experience in the community determination of psychiatric illness. Kendler et al. reported that about 50 of these people met strict criteria for nonaffective psychosis [4]. Hence, in the community, using questions that address the diagnostic criteria and that have face validity, the CIDI generates a number of false positives.
To see if this false positive problem was general, Cooper and colleagues examined a clinical sample using clinician check lists with two clinicians and demonstrated that there was a good relationship between those clinicians' check lists [5]. They were reliable and valid. The clinician check lists were again just an operationalisation of the diagnostic criteria (now DSM-IV). When the CIDI was compared with the check lists, the opposite effect was happening [5]. The CIDI was identifying positive symptoms and missing the negative symptoms and, therefore, in a clinical sample, the problem was false negatives, or underestimates of the prevalence of psychosis.
How do you solve the problem of an instrument that in one environment produces false positives, yet in another environment produces false negatives? The CIDI used to ask people whether they have a symptom described in the diagnostic criteria and also ask the interviewer (who was trained as an interviewer but not in mental health) whether they believed that this response was plausible. If the interviewer felt that the response was not plausible then that was scored as a positive symptom. The CIDI was therefore modified to be more focused on the plausibility of the response. For example, the questions are now phrased: Have you believed that people were spying on you?; How did you know these people were spying on you?; Who was spying on you?; Was it a lot of people, a few people or only one?; Was it someone you knew or a stranger?; What did they hope to find out?; and What were they trying to do? Therefore, instead of a simple scoring of the criteria if the interviewer decides, ‘yes, it is implausible’, a much more subtle scoring of the ‘yes’, ‘no’ and ‘uncertain’ codes is used before a symptom was approved as meeting the criteria for a diagnosis.
Section P of the original CIDI, which focused on all the negative symptoms of schizophrenia, was then examined. An example of one of the problems with the original CIDI was that it questioned lay interviewers about whether there were neologisms present, and cryptically defined this as ‘the use of made up or meaningless words’. This contributed to problems of reliability. It is not surprising that numerous false negatives were observed in a clinical sample where the negative symptoms in people who had the disorder for a long time are important. Therefore, we have written a new version of Section P in which each sign is now fully described so that lay interviewers can understand what they are being asked to rate. The modified version of the CIDI has defined everything in detail and the interviewers are trained to understand the meaning of these terms. There is now evidence that the interviewers are able to reliably make these judgements. We have compared a number of people in a clinical setting using the revised CIDI and the Schedule for Clinical Assessment in Neuropsychiatry. The revised CIDI is an improvement.
In the National Survey of Mental Health Wellbeing, there were actually a number of questions designed to act as psychosis screeners; some from the United Kingdom psychosis screening questionnaire and others developed by Jablensky [1]. Each of the main questions asks about a putative psychotic experience and then checks that it is not an ‘X-File’ phenomena. It questions whether the experience came about in a way that many people would find hard to believe; in other words, is it unusual or strange? For example, question 2 asks: ‘In the past 12 months have you had the feeling that people were too interested in you?’. Question 2a asks whether ‘things were arranged to have a special meaning or even that harm might come to you?’. Question 3: ‘Do you have any special powers that most lack?’. It then does the inverse and question 3a asks: ‘Do you belong to a group of people who also have these powers?’, to avoid falsely identifying, for example, a religious group. A further question (G4) is, of course, the one we all believe is the only real way to diagnose the disorder: ‘Has a doctor ever told you that you may have schizophrenia?’. We developed a scorer for this and the scores correlated with the CIDI. We now have to go back and rework our data against the original check lists which, after all, are a type of reliability standard. At the moment, in the survey, when you see people commenting about nonaffective psychosis, it means that people said ‘yes’ to G1A, G2A and said ‘no’ to G3A. Two or more of those get you into the category known as a possible nonaffective psychosis. The question about whether a doctor has ever told you that you may have schizophrenia did not turn out to be a useful addition to the other three questions.