Questionnaire

A questionnaire is a structured set of questions designed to collect standardized data from respondents in survey research. Questionnaires can include closed ended items (with predefined response options such as multiple choice or Likert scales), open ended items (allowing free text responses), or a combination.^[1] They are used across disciplines to measure attitudes, behaviors, knowledge, health status, and a wide range of other constructs.

Questionnaires are among the most commonly used data collection instruments in social science, health research, and education. When well designed, they produce reliable, valid, and comparable data across large samples at relatively low cost. The design of a questionnaire directly affects data quality: poorly worded questions, biased response options, inappropriate scales, and confusing layouts can introduce measurement error, reduce response rates, and compromise the interpretability of results.^[1] Researchers investing in questionnaire development and validation contribute tools that may be used across hundreds of subsequent studies, magnifying the impact of good or poor design decisions.

Questionnaire development typically follows a structured process. The researcher begins by defining the constructs to be measured, informed by the research question and existing literature. Items are drafted to capture each construct, with attention to clarity, specificity, and the avoidance of double barreled questions (which ask about two things at once), leading questions (which suggest a preferred answer), and ambiguous language.^[1][2]

Response formats are selected based on the measurement goals. Likert scales (typically five or seven points) measure the intensity of agreement, frequency, or satisfaction.^[2] Visual analogue scales allow respondents to mark a continuous scale. Dichotomous items (yes or no) are efficient but sacrifice precision. For factual questions (age, education level), categorical or numerical response options are appropriate.

The questionnaire undergoes pilot testing with a small sample to identify problems with comprehension, flow, and response patterns. Cognitive interviewing, in which respondents think aloud while answering, or answer follow up probes about how they understood each question and arrived at their answer, can reveal misinterpretations that would not be detected through quantitative pilot testing alone.^[3] After revision, the questionnaire is administered to a larger sample for psychometric evaluation, including assessment of reliability (internal consistency via Cronbach's alpha, test retest reliability) and validity (construct validity, criterion validity); content validity is judged earlier, when experts review the draft items against the definition of the construct.^[2]

Question order affects responses. Sensitive questions are generally placed later in the questionnaire, after rapport has been established. The overall length affects completion rates, with shorter questionnaires generally producing higher response rates. Mode of administration (online, paper, telephone, in person) also influences response patterns, and researchers should consider mode effects when interpreting results or comparing across studies.^[1]

Use questionnaires when the goal is to collect standardized self report data from a sample of respondents. They are appropriate for surveys, epidemiological studies, clinical trials (for patient reported outcomes), and any study requiring structured measurement of subjective or factual information. Use validated questionnaires whenever possible, as developing and validating a new instrument is time consuming and reduces comparability with prior research. If no validated instrument exists for the construct of interest, invest in rigorous development and validation before using the new instrument in a study.

A public health team develops a 25 item questionnaire to assess COVID 19 vaccine hesitancy among healthcare workers. Items are drafted based on a literature review and expert consultation. Cognitive interviews with 12 nurses reveal that three items are consistently misunderstood, and these are revised. A field test with 150 respondents yields a Cronbach's alpha of 0.88 for the full scale and a two week test retest intraclass correlation coefficient (ICC) of 0.82. Exploratory factor analysis identifies three subscales: safety concerns, efficacy doubts, and institutional distrust, so internal consistency is also reported for each subscale, because a high alpha across all 25 items does not show that they measure one construct. The final questionnaire is administered online to 4,500 healthcare workers, achieving a 62% response rate (2,790 completed questionnaires from 4,500 invited).