Write questions

A badly written question produces bad data every week, for every patient. These twelve rules protect the quality of what you collect.

​

The twelve rules

1. One idea per question. “How were your sleep and your energy?” has no interpretable answer. If there’s an “and” in the middle, it’s probably two questions. 2. An explicit recall window. Always write “in the last 7 days” or “over the past week”. Without it, each person answers about a different period, and some answer about yesterday. In a study of 515 participants published in the Journal of Urology in 2019, Flynn and colleagues compared daily diaries with 7-day and 30-day recall: every correlation came in above 0.60. 3. Observable behavior, not moral judgment.

• Before: “Were you disciplined with your diet?”
• After: “On how many of the last 7 days did you manage to follow the plan at your main meals?”

4. No judging words. Out of the vocabulary: slipped, failed, cheated, sinned, fell off the wagon, did badly. These words trigger shame, and shame produces underreporting in exactly the data you need most. 5. Avoid the agree/disagree format. The Pew Research Center describes acquiescence bias: some people tend to agree with any statement. Prefer item-specific scales (“how would you rate”, “how often”). 6. Reduce social desirability on sensitive topics. Respondents underreport alcohol and drug use and overstate socially valued behaviors. Normalize it in the wording: “Most weeks have something unexpected. On how many days this past week did you…”. 7. The question has to work when the answer is zero. “How many more times did you work out than last week?” assumes they worked out. “How many training sessions did you do in the last 7 days?” works with zero. 8. Mutually exclusive and exhaustive options. If a real case fits none of the options, the patient picks anything and the data dies right there. Include “not applicable” where it fits. 9. The patient’s language, not the chart’s. No “fluid intake”, “dietary plan adherence”, “gastrointestinal symptomatology”. Water, sticking to what we agreed, bowels and stomach. 10. Twenty words maximum in the wording. If it doesn’t fit, the question is doing more than one thing. 11. Numbers only where memory holds up. “How many days” and “how many times” work. “How many grams of protein” doesn’t, unless the patient uses an app and you know it. 12. Nothing goes in without a decision behind it. Before including any question, answer this: what changes in my approach, or in the patient’s perception, depending on the answer? If the answer is nothing, cut it.

​

Before and after

​

Conditional explanation

LiveClin lets you require a justification only when the patient picks a specific option. It’s the feature that resolves the conflict between wanting context and not wanting to tire the patient out with free text: you only ask for the why once the what has raised a flag. Example:

In the last 7 days, how did following the plan at your main meals go?

• I managed it on almost every day (great)
• I managed it on about half the days (neutral)
• I managed it on only a few days (poor, with a required explanation: “What got in the way most this week?”)
• I basically couldn’t (terrible, with a required explanation: “What happened this week?”)

Someone who’s doing well answers in eight seconds. Someone who isn’t hands you the context you would otherwise have to hunt for in the appointment. See how to set this up in the Create questions tutorial.

​

Next steps

References for this page

• Flynn, K. E. et al. (2019). Can 7 or 30-day recall questions capture self-reported lower urinary tract symptoms accurately? Journal of Urology.
• Pew Research Center, Writing Survey Questions: acquiescence bias, social desirability, double-barreled questions.