Most studies use a mix of validated instruments and questions written for the study. The validated ones are fixed; the rest are where studies go wrong. A step-by-step guide to writing questions that measure what you think they measure.
Free sandbox · No credit card · 21 CFR Part 11 aligned
Daily check-in
Question 2 of 5
In the last 24 hours, how often did you feel short of breath while walking?
The short version
Step by step
Write down exactly what you want to measure, in whom, and how the answer will be used in analysis.
Search for validated instruments for that concept and population; check licensing and translations.
Draft study-specific items with clear wording, recall periods and response options.
Cognitive interviews with a handful of target participants, then a pilot on the real device.
Writing items
Questionnaire problems are usually wording problems. Participants answer the question they think you asked, and if it can be read two ways, some will read it the other way. The fixes are simple and well established.
"How often did you feel tired and irritable?" asks two things. A participant who was tired but not irritable has no correct answer. Split double-barrelled questions.
"In the last 7 days" or "today" belongs in every item, not only in the instructions. Match the recall period to how often the questionnaire is given: a two-week recall collected weekly overlaps itself.
Offer as many positive as negative options, cover every possible answer (including "not applicable" where relevant), and use the same scale direction throughout. Frequency (never to always), intensity (none to very severe) and agreement scales measure different things; do not mix them within one item.
"How much has the treatment improved your sleep?" assumes improvement. "How would you rate your sleep in the last 7 days?" does not.
Write for a reading age around 11 to 12 where possible, avoid medical jargon and abbreviations, and use official translations for validated instruments. For your own items, translate and back-translate.
| Analyte | Result | Unit | Ref. range |
|---|---|---|---|
| Haemoglobin | 13.8 | g/dL | 13.0 to 17.0 |
| White blood cells | 6.2 | 10^9/L | 4.0 to 10.0 |
| Platelets | 412 | 10^9/L | 150 to 400 |
Response formats
| Format | Good for | Watch out for |
|---|---|---|
| Numeric rating scale (0 to 10) | Intensity of symptoms such as pain | Anchors must be defined at both ends. |
| Verbal rating scale | Frequency or severity in words | Keep categories balanced and ordered. |
| Visual analogue scale | Fine-grained continuous ratings | No preset marker; some participants find it harder. |
| Yes / no | Events that clearly happened or not | Loses information about degree. |
| Multiple choice | Several things that can all apply | Needs "none of the above". |
| Number or time | Counts, durations, clock times | Units and ranges defined. |
| Free text | Exploration, other symptoms | Not analysable at scale; use sparingly. |
Build it
Capture offers ten question types (short text, number, linear scale, slider, single choice, multiple choice, dropdown, date, time and long text) with skip logic and required fields. Visual analogue scales have no preset value, and the phone layout wraps long items rather than truncating them. Preview the participant view on your own phone before the study starts.
How much pain do you feel right now?
How much pain do you feel right now?
NoneMildModerateSevereFree sandbox with every question type.
Commonly five to seven labelled options for agreement or frequency scales, or a 0 to 10 numeric scale for intensity. Keep direction and labels consistent across items.
Include one when not knowing is a genuine, meaningful answer. Otherwise it can become an easy way out, so decide item by item.
Use a validated instrument when one exists for your concept and population. Write your own only for what no instrument covers.
The time window a question asks about, such as the last 24 hours or the last 7 days. Put it in each question and match it to how often the questionnaire is given.
Asking a few people like your participants to think aloud while answering, to check they understand each question as intended.
As short as possible for the endpoints. Burden drives missing data, especially for repeated questionnaires.
Yes. The free sandbox includes every question type and a phone preview.
Build, preview and test your questionnaire. Free sandbox.