Presentation is loading. Please wait.

Presentation is loading. Please wait.

Using Behavior Coding to Evaluate the Effectiveness of Dependent Interviewing Joanne Pascale QUEST Conference Ottawa, Canada April 26, 2007.

Similar presentations


Presentation on theme: "Using Behavior Coding to Evaluate the Effectiveness of Dependent Interviewing Joanne Pascale QUEST Conference Ottawa, Canada April 26, 2007."— Presentation transcript:

1 Using Behavior Coding to Evaluate the Effectiveness of Dependent Interviewing Joanne Pascale QUEST Conference Ottawa, Canada April 26, 2007

2 Outline Dependent interviewing Research questions addressed with behavior coding Methods Findings Summary and Recommendations

3 Dependent Interviewing (DI) Used in longitudinal studies Carries reports from one wave into the next interview wave E.g.: ‘Last time you said you were receiving benefit X. Are you still receiving benefit X?’

4 Purpose of DI Potential for smoother, smarter, more efficient interview Literature demonstrates: –Reduced burden –Reduced item non-response –Reduced seam bias –Respondents want and expect it

5 Research Questions How is DI being implemented in the field: Do interviews read questions as worded? Do respondents affirm or dispute the fed-forward data from prior wave? Does this vary by ‘style’ of DI?

6 Methods: The ELSA Study English Longitudinal Study on Aging (ELSA) Panel study of 50+ begun in 2002 Follow-ups every 2 years CAPI face-to-face

7 Methods: The ELSA Pilot Vehicle for DI test: pilot of Wave 3 January, 2006 4-week field period Recorded onto laptops using CARI 17 interviewers 104 pilot interviews

8 Methods: DI Topics and Styles 3 different topic areas: –Demographics –Health conditions Eye (e.g. glaucoma) Cardio-vascular (e.g. high blood pressure) Chronic (e.g. arthritis) –Vehicle ownership 5 different styles of DI

9 Demographics Item Wording 1. Does John still live here? 2. Can I just check, is John’s date of birth January 1, 1960? 3. Our records show that when we last interviewed you, you had a child called Billy, whose date of birth is July 1, 2005. Are these details correct?

10 Health Item Wording 4a. Our records show that when we last interviewed you in [DATE], you said that you had [EYE CONDITION]. 4b. Do you still have [EYE CONDITION]? [same routine for CVD and chronic conditions]

11 Vehicle Item Wording 5. At the time of last interview, you or someone in your household owned [VEHICLE]. Is that vehicle still owned by you or someone in your household?

12 Behavior Coding Listened to several pilot tapes to develop code frame First-level exchange sufficient Developed Interviewer, Respondent and Outcome codes

13 Interviewer Codes Read as worded/slight change Major change FF statement became a question FF question became a statement Other major change Omitted Inaudible Interviewer/Other

14 Respondent Codes Adequate Affirmed FF item Disputed FF item Inadequate Answer/Elaboration Clarification Inaudible Respondent/Other

15 Demographics Findings Interviewer BehaviorRespondent Behavior Exact/slight change Q read as statement Q omitted Adequate Affirmed Adequate Disputed Does NAME still live here? 403318811 Can I check, is NAMES’s DOB? 57371910 Our records show child’s name, DOB. Details correct? 7980895

16 Health Findings Interviewer BehaviorR Behavior Exact/slt change S read as Q Q read as S OmittedAdequate Affirmed Adequate Disputed Last time you reported EYE cond 6238--06210 Do you still? 63--16 -- Last time you reported CVS cond 6320--0875 Do you still? 76--313-- Last time you reported CHRONIC cond 4134--5854 Do you still? 61--1418--

17 Vehicle Findings Interviewer BehaviorRespondent Behavior Exact/slight change Q read as S OmittedAdequate Affirmed Adequate Disputed Last time you owned …vehicle. Still own? 8284746

18 Summary: Interviewers Qs read as worded: 40-82% Non-standard reading varied: –Demographics & vehicle: questions turned into statements or omitted –Health: statements of what was reported last time were turned into questions; question “is that still” omitted

19 Summary: Respondents Adequate answer: 72-94% Disputed fed-forward data: 0-10% –Confirmed prior wave report but said no longer the case –Denied prior wave report –Disagreed with details of prior wave report

20 Recommendations Avoid simple statements of prior wave data For topics unlikely to change over time (e.g.: DOB) avoid re-asking; verify accuracy of first report Style of DI needs to be tailored to item Prior wave data could be challenged; need to allow for corrections


Download ppt "Using Behavior Coding to Evaluate the Effectiveness of Dependent Interviewing Joanne Pascale QUEST Conference Ottawa, Canada April 26, 2007."

Similar presentations


Ads by Google