The questionnaire design imperative
A poorly designed questionnaire produces poor data. The most common problems:
- Too long, Completion rates drop dramatically beyond 15 to 20 minutes
- Ambiguous questions, "How severe are your symptoms?" means different things to different people
- Double-barreled questions, "Do you have pain and fatigue?" cannot be answered with a single yes/no
- Custom scales instead of validated instruments, Produces uninterpretable data
- No skip logic, Showing irrelevant questions wastes participant time and increases abandonment
Use validated instruments wherever possible
Always search for a validated instrument before writing a custom question.
A validated instrument has been:
- Developed with participant and clinician input
- Tested for reliability (consistent results across administrations)
- Tested for validity (actually measures what it claims to measure)
- Normed against a reference population
- Accepted by journals and regulatory agencies
Key validated instrument libraries
PROMIS (NIH): Physical function, fatigue, pain, sleep, anxiety, depression, social participation
→ healthmeasures.net
EQ-5D: Generic health related quality of life; widely used in health economics
→ euroqol.org
SF-36 / RAND-36: General health status
→ rand.org/health-care/surveys_tools/mos/36-item-short-form.html
NIH Toolbox: Cognitive, emotional, motor, and sensory function
→ nihtoolbox.org
Participant-specific instruments: Many diseases have gold-standard disease specific instruments. Your SAB should identify these.
Question types and best practices
Response scales
For symptom severity, use numeric rating scales (NRS) or Likert scales, not free text.
- NRS 0 to 10: "On a scale of 0 to 10, how severe is your pain today?", Simple, widely understood
- Likert (5-point): Never / Rarely / Sometimes / Often / Always, Good for frequency
- Visual analog scale (VAS): Continuous line from "None" to "Worst imaginable", More sensitive but harder to implement on paper
Date questions
For onset dates, provide a structured date picker with a "year only" or "approximate" option, many participants know the year of symptom onset but not the exact date. A "don't know / unsure" option is essential.
Branching logic
Use skip logic so participants only see relevant questions:
- "Have you ever had seizures?" → If NO, skip to next section
- "What medications are you currently taking?" → If NONE, skip medication detail section
Cognitive burden
- Use plain language (aim for 6th grade reading level)
- Avoid medical jargon; define terms when needed
- Group related questions in logical sections
- Provide progress indicators on long questionnaires
- Offer "save and return later" for multi-section forms
Questionnaire development process
- Draft based on scientific questions and SAB input
- Participant review, share draft with 3 to 5 participants for comprehension testing
- Cognitive interviewing, ask participants to "think aloud" as they answer questions
- Pilot test with 20 to 50 participants; measure completion rate and time
- Analyze pilot data, Are there items with very high "don't know" or skip rates? Do items perform as expected statistically?
- Revise based on pilot findings
- Final SAB approval
Frequency and burden management
- Enrollment questionnaire: Budget 20 to 30 minutes maximum for core elements
- Annual/biannual follow up: 10 to 15 minutes
- Brief check-ins: 5 minutes or less for high-frequency (quarterly or monthly) touchpoints
Shorter and more frequent is often better than long and infrequent, and yields better longitudinal data.