Cognitive Interviewing at the National Center for Health Statistics: Cognitive Interviewing at the National Center for Health Statistics Paul Beatty
Mario Callegaro
Karen Whitaker
Kristen Miller
Alfredo Calvillo
Cognitive Interviewing at the National Center for Health Statistics:: Cognitive Interviewing at the National Center for Health Statistics: Current Practice and New Advancements
NCHS Cognitive Methods Staff
Session Outline : Session Outline What is Cognitive Interviewing?
Cognitive Interviewing in Practice: Examples of identifying questionnaire problems
Cognitive Interview Samples
Quality Standards
Extending the Methodology
Questionnaire design problems: Questionnaire design problems Survey data appear precise and factual, but are actually complex estimates
Some possible threats to accuracy derived from the questionnaire:
Questions not understood as intended
Don’t adequately capture respondent experience
Pose an overly challenging response task
Problems may not be visible in the actual survey data
How can we find these before data collection?
“In the last year, have you been bothered by pain in the abdomen?”: “In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward
But suppose we ask:
What, to you, is your abdomen?
What, to you, is your abdomen?: What, to you, is your abdomen?
“In the last year, have you been bothered by pain in the abdomen?”: “In the last year, have you been bothered by pain in the abdomen?” Seems to be straightforward
But suppose we ask:
What, to you, is your abdomen?
What does it mean to be “bothered by pain” in the abdomen?
What period of time are you thinking about here specifically?
“In the last year, have you been bothered by pain in the abdomen?”: “In the last year, have you been bothered by pain in the abdomen?” Possible revisions:
Show shaded picture of abdomen
Drop “bothered”
Use “In the past 12 months”
Clear alternatives address these problems, with no apparent drawbacks
Shaded picture of the abdomen: Shaded picture of the abdomen
Old ideas about questionnaire design and the response process: Old ideas about questionnaire design and the response process Originally seen in terms of “stimulus/ response”– all we need to do is standardize the stimulus
Questionnaire design as art, not science
But more recently, productive collaborations between psychologists and survey methodologists have changed these ideas…
Cognitive stages involved in responding to survey questions: Cognitive stages involved in responding to survey questions Comprehension: Respondent interprets the question
Retrieval: Respondent searches memory for relevant information
Estimation/Judgment: Respondent evaluates and/or estimates response
Response: Respondent provides information in the format requested
(see references at the end of this presentation)
Probes in cognitive interviews : Probes in cognitive interviews Comprehension: What does 'dental sealant' mean to you?"
Frame of reference: What were you thinking about while answering? (Basis of response)
Encourage narrative: To learn what the short response to the question means
Recall: How did you figure your answer to that?
Confidence: How certain are you about that?
Paraphrase the question
General characteristics of cognitive interviews: General characteristics of cognitive interviews Conducted one-on-one
Interviewers are questionnaire design specialists-- investigators, not just data collectors
Questionnaire followed, but also used to generate discussion
Participants paid for their time and effort
QDRL laboratory: QDRL laboratory
Ceiling mounted camera: Ceiling mounted camera
Cognitive Interviewing in Practice: Cognitive Interviewing in Practice
Examples of respondent's tasks: For pedagogic reasons we discuss the respondent's four basic tasks as if they are sequentially ordered. However the respondent may in fact go back and forth among different tasks Examples of respondent's tasks
Examples of respondent's tasks: Examples of respondent's tasks We selected 4 examples trying to get each task as “pure” as possible, but you’ll see how all these tasks are strongly interrelated
Cognitive interview instruction: Cognitive interview instruction A cognitive interview is different from a standardized interview and from an everyday conversation
The researcher asks the questions in the standardized way but also asks the respondents to think aloud, highlight problems, express their opinion, make judgments on the questions…
Training the respondent for a cognitive interview: Training the respondent for a cognitive interview Each cognitive interview begin with a short training session where we illustrate the characteristic of the task. We also encourage the respondent to think out loud, to feel free to express his/her problems and difficulties in answering the questions.
Comprehension task: Comprehension task Do you use any assistive devices to help with mobility, communication, self care, accessing your workplace?
Yes
No Source: 1972/74 Social Security Administration Survey of Disabled and Non Disabled Adults
Comprehension Transcription:: Comprehension Transcription: That’s a mouthful of questions. Assistive device??
… Well, I guess we all could be classified. We use glasses. I guess glasses are assisted devices and so I guess almost everyone has to say yes to that and in my case I wear glasses and I also have hearing aids. That’s it. I don’t use a cane or anything…
Slide23: I think a lot of people may have trouble with that question though because it’s kind of stiff and formal. Like if you would say assistive devices such as … and give some examples, maybe a person might pick up on it a little bit…
…because right off , the first thing I think about is a walker or cane or something but you are telling about hearing, communication which includes hearing and speaking, seeing, all of those are communication devices. Anyhow.
Retrieval task: Retrieval task How old were you when this (high blood pressure) was first diagnosed?
Source: US/Canada Joint Health Survey
Retrieval Transcription: Retrieval Transcription Huh, shoot I was in my forties but I don’t remember exactly when …
Because you said that it was in 1996
Yeah, I remember this so good because I moved back here in 95 … I know I was in my forties
(interviewer speech is written in blue)
Judgment task: Judgment task Some people who have health conditions, impairments, or disabilities get help from other people in order to get around, lift or carry things, communicate, keep track of things, or remember things.
As a result of your compression fracture, or your hearing or your shoulder, do you require help from other people?
Yes
No Source: Disability Statistics Institute
Judgment Transcription: Judgment Transcription I hate to say require, but of course I did require it when that compression fracture happened first but now I’m back doing everything myself and I do have the neighbors come in and I have my lawn mowed instead mowing it and I have my gardening done instead of doing it and so forth… so I guess I required help…
Response task: Response task Does a physical condition or mental health problem reduce the amount or the kind of activity you can do at home?
Yes, sometimes
Yes, often
No Source: Canadian Cooperative Survey
Response Transcription: Response Transcription Hmm there’s something I can’t do so hmmm and there’s things that I can but with difficulty or with aid hmm so actually I don’t have an answer to that I guess I would be if I have one always I guess we fall in the always
You really would like an always category
Yes
Part 3: Cognitive Interview Samples: Part 3: Cognitive Interview Samples
Cognitive Interviewing Samples: Cognitive Interviewing Samples Generally small and focused (sometimes 12 interviews). Reasons:
Intensive labor effort
Rich data, time consuming to analyze
But with such samples, how can we infer to the population?
Need to select people with the characteristics of greatest interest
Some sample considerations : Some sample considerations Keep the overall goal in mind: maximize chances of discovering problems, rather than inference
Glaring problems can be discovered quickly, with more subtle problems captured later
The value of “one significant case”
How many interviews should be conducted?: How many interviews should be conducted? More is better; but there are diminishing returns
Useful to get:
Participants with average (or below) knowledge and experience
Some demographic variety
Coverage of questionnaire rather than population
Possible sources of participants: Possible sources of participants Newspaper advertisements
Flyers
Special interest organizations, trade organizations, non-profit groups
Word-of-mouth
Database of previously used participants
Source pros and cons: Source pros and cons Newspaper advertisements
Able to reach a large number of people
Tend to get a high-volume of calls over a short period of time
Advertising rates can be expensive
Have to screen for and weed-out “professional” research participants
People who read the newspaper tend to be better educated
Source pros and cons cont.: Source pros and cons cont. Flyers
Able to target people in specific geographic locations
Tend to get a low-volume of calls trickling in over a period of weeks
Inexpensive
Often need permission to post flyer which can tap into resources/time
Source pros and cons cont.: Source pros and cons cont. Special interest organizations, trade organizations non-profit groups, etc.
Able to target very specific people
Tend to get a low-volume of calls trickling in over a period of weeks
May need their IRB approval before recruiting for participants
Participants may be biased
Source pros and cons cont.: Source pros and cons cont. Word-of-mouth
Tend to get a low-volume of calls trickling in over a period of weeks
Database of previously used participants
Good for when you need a few people with very specific characteristics
Shelf-life of a participant ranges from 6-12 months
What goes in the advertisement/flyer: What goes in the advertisement/flyer Catchy header
Description of who you are looking for with what specific characteristics/conditions
What the study is about
Time involved (1 hour; 1 ½ hours, etc)
Amount of payment/token of appreciation
Contact name & phone number
Affiliation
The telephone screener: The telephone screener Comprised of a series of questions designed to elicit both demographic information, as well as information specific to the study
Needs to be concise, yet long enough to determine eligibility
Videotaping interviews: Great for illustrating question problems to the sponsor. “A picture is worth a thousand words.”
Valuable analysis tool (supplement interviewer notes, behavior coding, timing of sections, etc.)
Participants may not wish to be taped and decline to participate in the study
Videotaping interviews
QDRL control room: QDRL control room
If you…they will come: If you…they will come Design a good ad/flyer
Get them “bought-in” during the initial screening call
Make reminder calls a few days ahead of the scheduled appointment
Greet participants
Pay them well
Part 4: Quality Standards for Cognitive Interviewing: Part 4: Quality Standards for Cognitive Interviewing
Probing and “think alouds” : Probing and “think alouds” Our examples focus on probing, but another option is to encourage think-alouds (“tell me what you’re thinking while answering”)
Pros: interviewer more free to listen and less potential for bias
Cons: Unnatural; may not reflect actual cognitive processes, and may actually interfere with them
Concurrent vs. retrospective probes: Concurrent vs. retrospective probes Concurrent probes: between questions
Pros: Question is very fresh on the mind
Cons: Potential bias; switching of tasks can be distracting
Retrospective probes: probe only at end of questionnaire
Pros: Avoids bias and task-switching
Cons: Long gap btw question and probe
Mode of cognitive interview : Mode of cognitive interview Cognitive interviews designed to be intensive and face-to-face; most survey interviews used to be that way
Now many surveys carried out by phone– should cog interviews also be?
Pros: Realism of cognitive task
Cons: Awkward interview situation
Compromise: lab-based telephone interview with remote monitoring
TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES: TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES 10. Not asking the question exactly as it is written.
9. Forgetting to get the participant to answer the question.
8. Not following up on an indication that the participant has responded incorrectly.
7. Not checking to see if the answer is indeed correct.
TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES: TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES 6. Cutting the participant off and encouraging them to say as little as possible.
5. Forgetting to pursue how participants are interpreting key words.
4. Waiting long periods of time before writing up notes.
TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES: TEN BIGGEST COGNITIVE INTERVIEWING MISTAKES 3. Asking hypothetical questions.
2. Asking probes that suggest there is a right answer. For example, “You only go to licensed tattoo parlors, right?”
1. Telling the participant that it is really they who have misunderstood the question.
Cognitive Interviewing in Rural Mississippi : Cognitive Interviewing in Rural Mississippi Southern, rural county in Mississippi
21 Participants
Poorer than typical lab participants
Less education than typical lab participants
All had telephones and televisions
Why conduct cognitive interviews with the poor and less educated?: Why conduct cognitive interviews with the poor and less educated? 1. Practical: To improve estimates. Survey questions can be improved to be more inclusive.
2. Theoretical: To improve understanding of the question-response process.
Joint Canada United States Survey of Health: Joint Canada United States Survey of Health Jointly conducted by the National Center for Health Statistics and Statistics Canada
General Health Questions, including subjective health, access to care, chronic conditions, cancer screening, smoking and limitation questions
What we learned about the question-response process:: What we learned about the question-response process:
Cannot expect a respondent knows how “to be a survey respondent”: Cannot expect a respondent knows how “to be a survey respondent” That an impromptu response is required
That their response must be categorizeable
That the formality of the question-response process will be grasped
Cannot expect respondents to make sense of vague or elusive words: Cannot expect respondents to make sense of vague or elusive words Incorrectly inferring the meaning of an abstract word
Unable to respond to scaled items: “mild,” “moderate,” “severe,” “extreme”
Completely misunderstand the entire question
Cannot expect participants to make mathematical calculations: Cannot expect participants to make mathematical calculations “How old were you when your high blood pressure was first diagnosed?”
Age--- no problem
Year---problem
Years ago---problem
Cannot expect respondents to answer within another system of knowledge: Cannot expect respondents to answer within another system of knowledge Chronic conditions:
“Do you have chronic bronchitis?”
“Do you have asthma?”
“Do you have coronary heart disease?”
“Do you have angina?”
“Do you have congestive heart failure?”
Why conduct cognitive interviews with the poor and less educated?: Why conduct cognitive interviews with the poor and less educated? 1. Practical: To improve estimates. Survey questions can be improved to be more inclusive.
2. Theoretical: To improve understanding of the question-response process.
Cross Cultural/Cross Linguistic Testing: Cross Cultural/Cross Linguistic Testing Growing Needs
Hispanic Groups
Translation & interviewing problems
Census 1990 data: Census 1990 data Language Spoken at Home and Ability to Speak English for Persons 5 Years and Over Ranked by Number Who Speak English Less Than "Very Well“
United States 1990 Ranked by Total Number of Speakers
Slide63: Census 2000 data
National Adult Literacy Survey --- 1992 : National Adult Literacy Survey --- 1992 literacy tasks:..
reading a bus schedule
using an automatic teller machine
understanding a judge’s instructions to a jury
Five proficiency levels
Level 1 being least proficient
Level 5 being most proficient.
http://nces.ed.gov/naal/
What are the findings of the study? : What are the findings of the study? 21% to 23% at the lowest level= 40 to 44 million people
Level 1 perform tasks involving “brief, uncomplicated text,” such as totaling the entry on a bank deposit slip or locating information in a short news article, but many do so with difficulty.
Slide66: 25 to 28% = 50 million American adults, functioning at Level 2.
Interestingly, many respondents at Levels 1 and 2 did not consider themselves “at risk” because of their literacy skills.
A majority of those at Level 1 and almost all those at Level 2 described themselves as being able to read English “well” or “very well
Typical cognitive problem: Typical cognitive problem Original: EN LOS ULTIMOS 3 MESES, ¿asistió usted a una escuela o universidad?
Alt. Form: EN LOS ULTIMOS 3 MESES, ¿estaba tomando clases en una escuela o universidad?
At any time IN THE PAST 3 MONTHS, have you attended a regular school or college?
Both our respondents as well as our interviewers had a lot of problem with this syntax because they interpreted the word"asistió" in terms of "assisting" in the cognitive vein of being a teachers aide.
Cross cultural referencing: Cross cultural referencing Spanish : ¿Cuál es él titulo o nivel de escuela mas alto que usted ha terminado?
English: What is the highest degree or level of school you have completed?
Answer: Llegue a la cuarta!
I got to the fourth!
Translation conflicts: Translation conflicts Race, Hispanic Origin
Education / reading ability
Citizenship; acquiring citizenship; legal status
Social Security – country specific
Foster children
Regional glossary needed:: Regional glossary needed: “you don’t know beans!”
Habichuela, frijol, poroto
Mexico vs others: Cigars, cigarettes?
Pavo versus guajolote;
Rentar versus alquiler,
Third world colliding with 1st world and its technology. : Third world colliding with 1st world and its technology. ACASI with semi-literate groups
The structured interview as an inquisition
Interviewers often seen as agents of the government.
As social workers or do-gooders.
Interviewers often community leaders; help get access
Are also as limited as the respondents in education and linguistic skills.
Slide72: Respondents fearful
Respondents assume we already “know” the answers; tell us “the truth”?
Interviewers often receive minimal training
Often do not have bilingual supervisors to assess quality control.
Interviewers will often not know why a problem exists but only that it does.
Is this an interview? Respondent “Response Bias”
Parallel problem in English and in Spanish: Parallel problem in English and in Spanish step #1: the question is originally asked very simply
step #2: a review suggests “tweaking”—this often means adding clarification clauses in the sentence
step #3: we use formal phrasing so as to have correct “English” (or Spanish)
result: we come off as stilted or even incomprehensible to the respondent.
KISS keep it simple (and understandable)
Some final comments: Some final comments Questionnaire design is a slow process, done in stages (iterative expert reviews, cognitive tests, field pretests)– allow lots of time
Don’t ignore potential questionnaire errors just because they are less visible
Solving problems up front is far preferable to discovering you’ve got problems with a questionnaire that has been used for years
Slide75: QDRL
Working papers link
www.cdc.gov/nchs/products/pubs/workpap/workpap.htm
References: books on Cognitive Aspect of Survey Methodology (CASM): References: books on Cognitive Aspect of Survey Methodology (CASM) Schwarz, N., Sudman, S. (Eds.) (1995). Answering Questions: Methodology for Determining Cognitive and Communicative Processes in Survey Research. San Francisco: Jossey-Bass.
Sirken, M. G. et al (Eds.). (1999). Cognition and Survey Research. New York: Wiley
Stone, A. et al. (Eds.) (2000) The Science of Self-Report. Implications for Research Practice. Mahwah, NJ: Lawrence Erlbaum Associates.
Sudman, S. Bradburn, N., Schwarz, N. (Eds.). (1996). Thinking About Answers: The Application of Cognitive Processes to Survey Methodology (Jossey-Bass Social and Behavioral Sciences Series). San Francisco: Jossey-Bass.
Tanur, J. (Ed.) (1992). Questions About Questions. Inquires into the Cognitive Bases of Surveys. New York: Sage.
Tourangeau, R., Rips, L.J., Rasinski, K.A. (2000). The Psychology of Survey Response. Cambridge: Cambridge University Press.
Books on the interview process: Books on the interview process Fowler, F.J., Mangione, T.W. (1990) Standardized Survey Interviewing. Minimizing Interviewer-Related Error. (Applied Social Research Methods Series, Vol. 18). New York: Sage.
Maynard, D. W. et al (Eds.) (2002) Standardization and Tacit Knowledge. Interaction and Practice in the Survey Interview. New York: Wiley.
Books on Qualitative Research: Books on Qualitative Research Glaser, B. G., Strauss, A.I. (1967) The Discovery of Grounded Theory. Strategies for Qualitative Research. Chicago: Aldine.
Lincoln, Y.S., Guba, E. G. (1985) Naturalistic Inquiry. New York: Sage.
Newman, I. and Benz, C.R. (1998) Qualitative – Quantitative Research Methodology. Exploring the Interactive Continuum. Carbondale, IL: Southern Illinois University Press