Transcript 1- Introduction to Health Survey Research and Total Survey Error

Health Surveys
January 2008
Diane Martin, MA, PhD
1
Why Surveys?
• Answer important questions
• Learn new skills, especially in
measurement
• Self admin. relatively inexpensive
• Complete fairly quickly
• Write one results paper and/or one
methods paper
2
Survey Research
• Large body of scientific evidence
• Art of survey design, practical
experience
• Tradeoff between methodological rigor
and cost
• Two major types of res. questions
– Describe (estimation)
– Explain associations (hyp. testing)
3
Advantages of Surveys
• Collect information only respondent can
answer
• Good for attitudes, beliefs, expectations,
personal behaviors, subjective
measures, individual experiences
• Information society (culturally defined)
4
Disadvantages of Surveys
• Respondent burden
• Nonresponse
• Negative influence of
gov’t/privacy/telemarketing
• Poor survey techniques,
everyone thinks they can do this
5
Survey - Last Resort
• Are existing data sufficient?
• Can secondary data be abstracted?
• Can observations be made?
6
Most Important Survey
Principles
• Minimize errors throughout process
• Encourage response, decrease
respondent burden
• Use tailored design
• Define measures carefully
• Pay attention to detail
7
Scientific Integrity
•
•
•
•
•
•
•
Responsible conduct of research
AAPOR code of ethics
Conflicts of interest
Elements of disclosure
Human subject issues
Interviewer falsification
Data mashing
8
Ethical Principles
• Respect for the person: informed
consent, rights respected
• Beneficence: subjects’ well-being, risks
vs benefits
• Justice: benefits & burdens distributed
fairly across people
9
UW IRB
• Exempt, ‘no risk’, anonymous, limited
publication
• Minimal risk, expedited review (maybe)
• Full review
10
Survey Content
• Don’t reinvent the wheel
• Keep it simple, short
• Move from a conceptual model with
major variables to operational
definitions of variables (measurement
matrix)
• Consider measurement issues
• Aday LA, Designing and Conducting
Surveys, 1996
11
Single Questions versus
Scale or Index
• Advantages of scale or index
– Increases sensitivity
– Decreases number of variables for
analysis
• Consider the scale validity, reliability,
measurement error
• Obtain help if constructing new scale12
Survey Mode
• Choose mode appropriate for
content and population
• Know mode advantages and
disadvantages
• Interviews: in person, telephone,
ACASI
• Self administered: in clinic, mail,
computer/internet
13
Total Survey Error
• Overall goal: decrease study error
• Know different types of errors
• Anticipate errors in all phases of
study!!
• Understand how errors affect results
• Take errors into account in analysis
and discussion
• Groves RM et al, Survey
Methodology, 2004
14
Tailored Survey Design
• Survey response is explained by Social
Exchange Theory
– Actions of individuals motivated by
expected return on their actions
• Create respondent trust, perceived rewards >
costs
• Use health marketing and communication
• Format for print or web Qs is crucial
• Pretest
• Dillman DA, Mail and Internet Surveys, 2007 15
Analysis
• Know unit of analysis
• Write hypotheses a priori, with clear
definitions of outcome variable and
independent (predictor) variables
• Write plan of analysis and make
mock tables
• Use Epi Info to analyze data
16
Disseminate Results
• Feedback results to respondents,
community, gov’t MOH
(presentations, one page summary)
• Write articles for newspapers
• Find a journal article that reports on a
study similar to yours – mimic it
17
Summary
• Survey - last resort
• Don’t reinvent the wheel
• Keep it short, simple
• Maximize social exchange
• Minimize errors and respondent
burden
18