The World Mental Health Composite International Diagnostic Interview
The World Mental Health Survey Initiative
Health and Work Performance Questionaire
NCS-R Screener notes to all users:
The SC section is perhaps the most distinctive feature of the interview schedules used in the NCS family of surveys compared to previous DIS and CIDI surveys. As detailed elsewhere (Kessler et al., 1999; 2000; 1998)[1], [2], [3], pilot work for the NCS made it clear to us that quite a few respondents under-report mental disorders in DIS and CIDI surveys (and presumably also in other diagnostic surveys that use interviews with a stem-branch structure). One reason for this is that they want the interview to end quickly and they realize after a few sections that there is a stem-branch structure in which saying "no" to the first question will result in the interview being dramatically shortened. As the papers cited above discuss, there is a rather substantial survey methodology literature on this sort of phenomenon. The standard way to deal with it is to administer a full set of stem questions before any probes are administered. And that's exactly what we did when we designed the interview schedule for the NCS. Specifically, we consolidated all the diagnostic stem questions for all the CIDI diagnoses included in the NCS into one part of the first section of the interview and asked these questions as a set. Positive responses were then used as the foundation for probing throughout the remaining diagnostic sections of the interview.

This consolidation of diagnostic stem questions also allowed us to deal with two other reasons for under-diagnosis in DIS and CIDI surveys: that some respondents are not motivated to engage in the active memory search needed to give complete and accurate answers to lifetime diagnostic stem questions, and that other respondents, even when motivated, are unable to give accurate responses because of incomplete memory. As detailed in the above-cited papers, the SC section used respondent motivation techniques developed by Charlie Cannell and his colleagues to deal with the motivation problem. In addition, we embedded several feedback, decomposition, and pacing strategies developed by Cannell into the SC section to help deal with the problem of incomplete memory.

The logic of the NCS-R SC section is very similar to the SC section in the baseline NCS, but we also added a few new things into the section to meet the somewhat different needs of the new survey. Here's a question by question annotation of the overarching rationale for the various components of the section.

SC1-5. We wanted to start with some basic descriptive questions that would constitute a warm-up for the respondent. As we had an interest in residential mobility that we wanted to explore in NCS-R, we decided to include these as the first warm-up question. Age and sex and marital status were also recorded in order to have these responses available for skip instructions in later sections. Height and weight were asked next to continue with the theme of asking basic descriptive questions that convent to R that we're interested in health.

SC6. Notice that some numbers are missing throughout the instrument, while others have an odd numbering scheme (e.g., CC50.15). These are due to the fact that the instrument went through a number of revisions after it was programmed as a CAPI interview. Numbering changes in CAPI is currently very difficult, with the result that we never renumbered but rather deleted numbers when questions were dropped and added new fractional numbers when new questions were added.

SC7. This is used as a stem question for the TB section.

SC8.1-9. These are standard questions asked in a great many other health surveys that we asked for benchmarking purposes as well as to add to the context-setting effect we wanted to establish in the early questions.

SC9b-g. These are questions originally asked in the 1957 Americans View Their Mental Health Survey and again in a 1976 replication of that survey. We have a number of questions in NCS-R, including the entire WU section, that were selected so as to provide a third data point of trend data in comparison to these two earlier surveys.

SC9h. This question is designed to assess a dimension of poor health that is seldom assessed in dimensional measures of health-related quality of life (QOL) as an intervening variable in planned analyses linking information about chronic conditions (independent variables) to health utilities (dependent variables) through dimensional measures of perceived health and health-related QOL (intervening variables). SC8.1-9b are other measures that were designed to be used as intervening variables in this same way along with the questions in the 30 Day Functioning (FD) section.

SC9.01-04. These questions were added as secondary outcomes based on consultation with our collaborator Niko Pronk.

SC9.1-10.14. This long series of questions was developed in collaboration with Tom Hale from the Department of Labor (DOL) as a set of test questions to assess Disability as defined by the Americans with Disabilities Act (ADA). DOL was asked by Congress to develop a measure of ADA Disabilites that can be administered in the Current Population survey to track rates of Disability in the U.S. The NCS-R questions were used as a test database for this purpose. A preliminary classification of respondents with ADA Disabilities was made based on these responses. Clinical re-interviews with these respondents plus a random sub-sample of presumably not disabled respondents are planned to be carried out to generate a gold standard designation. Standard psychometric procedures will then be used to boil down the NCS disability questions to select the subset that most accurately predict the clinician gold standard classifications.

SC19. This is known as a commitment probe. See the papers cited in the first paragraph of this memo for a discussion of the rationale for including this type of question. It should be noted that the very small number of respondents who say no to this question are terminated from the remainder of the interview. Our experience in a number of surveys has been that roughly 0.5% of respondents are terminated at this point in the interview. Our feeling about this, in a nutshell, is that it's a waste of effort to collect additional data from people who tell us that they are not going to take the interview seriously.

SC20-35. These are the key questions in the SC section: the diagnostic stem questions for many remaining sections of the interview. Note that interviewers are instructed to read the questions SLOWLY. This instruction is reinforced in training. "Slow" reading is a pace of approximately two words per second. "Normal" interviewer reading pace according to the General Interviewer Training (GIT) procedures on which the design of the NCS family of interviews is based, is approximately three words per second.

The notion of GIT will come up on a number of occasions in the notes to this interview. Every survey has, either implicitly or explicitly, GIT standards that interviewers are expected to follow. When a survey is carried out without explicit GIT rules in place, the implicit rule is that interviewers can do whatever they want to do with respect to such things as question pace, eye contact with respondents, methods for handling digressive respondents, methods for handling don't know responses, the use of directive and nondirective probes, etc. Many of the professional survey firms in the United States use the GIT rules developed in the 1950s by Charlie Cannell in the Survey Research Center (SRC) at the University of Michigan. These are the GIT rules used in the NCS family of surveys. New SRC interviewers go through a 40-hour training program in these GIT rules before they receive study-specific training for their first survey and they receive periodic GIT refreshed courses. The SRC GIT training manual can be purchased from the publications division at SRC.

Although I won't comment on the specifics of the stem questions in the SC20-35 series, I want to point out several general features of those questions. First, they often consist of question sets that are alternative ways of asking the same question. Respondents are generally skipped over the other questions in the set as soon as they answer any one positively. We can see this in the very first pair of questions, which deal with panic attacks. Question SC20 is the standard panic stem question. We came to discover in pilot work, though, that some people with panic say no to the first question because they do not think of their attacks as attacks of "fear", but rather focus on the physical feelings as the core of the attacks. These people think of the attacks as attacks of suddenly experiencing various uncomfortable physical sensations. SC20a is a "second chance" sort of question, asked only of people who do not endorse SC20, aimed at picking up these people who conceptualize their panic attacks in physical terms. SC26-26b is another similar set of alternative wordings that deals with the diagnostic stem for generalized anxiety disorder.

Second, readers should be aware that great care went into developing the precise wording for these stem questions. We began with the stem wording in the DIS and earlier versions of the CIDI and then consulted both with clinical experts and with patients to determine whether they felt that the stem questions captured the essence of the experience shared by people with these disorders. It is important to recognize that diagnostic stem questions are, in effect, mini portraits designed to paint a quick sketch for respondents in order to prime a memory schema about past experiences. We need to know the sorts of categories used to store these experiences in memory in order to paint the portraits that will tap into the memory schema. A number of changes were made in conventional CIDI stem questions based on this way of thinking. For example, focusing again on SC20, the standard CIDI question asks respondents about sudden fears that occur "in situations where most people would not be afraid". This phrase is a useful one from a formal point of view in defining a panic attack. However, our pilot work showed that the inclusion of this phrase throws off quite a few people who have panic because they think of the "situation" as the occurrence of the attack with all its associated psycho-physiological symptoms. I woke up in the middle of the night with my heart pounding and sweating and thinking that I was going to die. That's the sort of "situation" they have stored in memory. This is not the sort of situation "where most people would not be afraid" even though it is formally true that lying in bed at night is not the kind of "situation" in which most people are afraid. Based on the realization that this phrase was interfering with accurate memory search, we deleted it from the panic stem question.

Third, we varied the length of the stem questions to deal with the fact that the core situations that define mental disorders vary substantially from one disorder to the next. SC24 is the most complex stem question. This question deals with mania. Note that there are a number of declarative sentences followed by a question. It is important, in writing complex questions of this sort, to use declarative sentences at the onset so as not to have the respondent answer prematurely. This same approach is used throughout the instrument, with especially complex questions broken into two or more sentences that begin with a declarative sentence and then follow with a question.

Another type of complex stem question is found in the SC27 series, where we ask about specific phobias. Given the nature of this disorder, it made sense to break down the question into a series of sub-questions about different specific fears. The SC27.2-27.4 series is a rare example of the SC section probing beyond the diagnostic stem questions. This was done because specific fears are so common in the general population that it was necessary to screen for persistence (SC27.2), severity (SC27.4), and avoidance (SC27.3) in order to reduce the number of people who screened into the specific phobia section.

Fourth, we made active use of visual cues in administering the screening questions. Each respondent had a Respondent Booklet (RB) that they held in their hands during the interview. Interviewers refered respondents to specific pages of this booklet at times when the visual cues were used. SC30, for example, begins with the parenthetical interviewer instruction "(RB, PG 5)", which instructs the interviewer to say "Please turn to page 5 in your booklet" before reading question SC30. This shorthand is found throughout the interview, as the RB is used in many different sections of the interview schedule.

SC36. This checkpoint routes respondents to appropriate sections depending on their endorsement of diagnostic stem questions in the SC section. You will see a similar checkpoint at the end of most diagnostic sections in the interview schedule, the exception being a few sections that are administered to all respondents because it was impossible to develop stem questions (e.g., the sections on Obsessive Compulsive Disorder (OCD) and nonaffective psychosis, each of which requires the respondent to respond to a large number of symptom questions).

In the paper and pencil interview (PAPI) version of this interview schedule, which is used in most countries participating in the World Health Organizatio's World Mental Health (WMH) surveys, a "Reference Card" is used to organize responses to these diagnostic stem questions. Instead of interviewers having to refer back to the SC section throughout the interview, they can refer to the reference card to check on which stem questions were endorsed or not. It is important to bear in mind that the NCS-R interview, although scripted in the print copy as a PAPI instrument, was administered using CAPI. This accounts for the fact that some of the interviewer checkpoints are much more complex than those that an interviewer could realistically be expected to manage in a real-life interview without the aid of a computer to guide the skip logic.

[1] Kessler, R. C., D. K. Mroczek, and R. F. Belli. (1999). Retrospective Adult Assessment of Childhood Psychopathology. In D. Shaffer, C. P. Lucas and J. E. Richters (Eds.), Diagnostic Assessment in Child and Adolescent Psychopathology, (pp. 256-84). New York, NY: Guilford Press.

[2] Kessler, R.C., Wittchen, H.-U., Abelson, J., & Zhao, S. (2000). Methodological issues in assessing psychiatric disorder with self-reports. In A.A. Stone, J.S. Turrkan, C.A. Bachrach, J.B. Jobe, H.S. Kurtzman, & V.S. Cain (Eds.), The Science of Self-Report: Implications for Research and Practice (pp. 229-255). Mahwah, NJ: Lawrence Erlbaum Associates.

[3] Kessler, R.C., Wittchen, H-U., Abelson, J.M., McGonagle, K.A., Schwarz, N., Kendler, K.S., Knäuper, B., & Zhao, S. (1998). Methodological studies of the Composite International Diagnostic Interview (CIDI) in the US National Comorbidity Survey. International Journal of Methods in Psychiatric Research, 7(1), 33-55.