03. Develop operational definitions and a usable questionnaire
Lesson
LEARNING OBJECTIVES
Translate a construct into a reproducible measurement, choose appropriate response options and recall periods, and build a data dictionary before programming the form.
MAP OBJECTIVES TO VARIABLES
For each objective, identify the information actually required. Record the variable name, construct, exact question or instrument, units, response categories, timing, source, validation and missing-value rules. An available column is not automatically a valid measurement of the concept you need.
Distinguish nominal categories, ordered responses and quantities with units. A code such as 1 = source A and 2 = source B is a label, not a continuous score. Keep “unknown,” “not applicable” and “declined” distinguishable when the study needs that distinction; define how these will be represented in the analysis.
WRITE ONE CONCEPT PER QUESTION
Avoid double-barrelled items such as “Is your water safe and affordable?” A respondent may regard it as affordable but unsafe. Avoid wording such as “You always treat your water correctly, don't you?” that signals a preferred response.
Use a suitable recall period and neutral wording. Ask concrete events or practices where possible rather than relying on undefined terms like “often” or “healthy.” Response categories should fit the intended construct and accommodate relevant answers without forcing an inaccurate choice.
WORKED FICTIONAL ITEM
Purpose: record the main household drinking-water source during a defined period.
Draft: “During the past seven days, which source provided most of the drinking water used by this household?”
Specify local source categories, whether more than one source is relevant, how “most” will be explained, what happens for a tie or an unlisted source, and which definition the analysis will use. This is a draft requiring contextual and language testing, not a validated instrument.
For a symptom outcome, prespecify the symptom definition, respondent, recall period and skip logic. Do not label a self-report as a confirmed clinical diagnosis unless the measurement supports that description.
BUILD THE DICTIONARY AND FORM TOGETHER
Question identifiers must stay stable across documents, forms and exported data. Document units rather than combining metres and kilometres in the same unlabelled field. Define permitted dates and sensible consistency checks, allowing authorized review of genuine unusual values.
YOUR TASK
Create entries for five variables: outcome, exposure, one confounder, selection information and a process variable. Draft two neutral questionnaire items, then revise one leading or double-barrelled example.
Specify skip logic for an inapplicable question and demonstrate that it will not overwrite a meaningful answer with zero. Review which fields collect identifiers and whether each is necessary.
SELF-CHECK AND FEEDBACK
Does consistent wording alone establish validity? No; the measurement must capture the intended construct adequately. Should every skipped answer be coded as “no”? No; distinguish inapplicable or missing answers according to the design. Can category codes be averaged meaningfully? Not merely because they are numeric.
PORTFOLIO OUTPUT
A versioned questionnaire excerpt, five-variable codebook and objective-to-measurement map.
FURTHER READING
CDC Field Epidemiology Manual: Collecting Data.
https://www.cdc.gov/field-epi-manual/php/chapters/collecting-data.html
WHO STEPS Manual.
https://www.who.int/docs/default-source/ncds/ncd-surveillance/steps/part2.pdf
Translate a construct into a reproducible measurement, choose appropriate response options and recall periods, and build a data dictionary before programming the form.
MAP OBJECTIVES TO VARIABLES
For each objective, identify the information actually required. Record the variable name, construct, exact question or instrument, units, response categories, timing, source, validation and missing-value rules. An available column is not automatically a valid measurement of the concept you need.
Distinguish nominal categories, ordered responses and quantities with units. A code such as 1 = source A and 2 = source B is a label, not a continuous score. Keep “unknown,” “not applicable” and “declined” distinguishable when the study needs that distinction; define how these will be represented in the analysis.
WRITE ONE CONCEPT PER QUESTION
Avoid double-barrelled items such as “Is your water safe and affordable?” A respondent may regard it as affordable but unsafe. Avoid wording such as “You always treat your water correctly, don't you?” that signals a preferred response.
Use a suitable recall period and neutral wording. Ask concrete events or practices where possible rather than relying on undefined terms like “often” or “healthy.” Response categories should fit the intended construct and accommodate relevant answers without forcing an inaccurate choice.
WORKED FICTIONAL ITEM
Purpose: record the main household drinking-water source during a defined period.
Draft: “During the past seven days, which source provided most of the drinking water used by this household?”
Specify local source categories, whether more than one source is relevant, how “most” will be explained, what happens for a tie or an unlisted source, and which definition the analysis will use. This is a draft requiring contextual and language testing, not a validated instrument.
For a symptom outcome, prespecify the symptom definition, respondent, recall period and skip logic. Do not label a self-report as a confirmed clinical diagnosis unless the measurement supports that description.
BUILD THE DICTIONARY AND FORM TOGETHER
Question identifiers must stay stable across documents, forms and exported data. Document units rather than combining metres and kilometres in the same unlabelled field. Define permitted dates and sensible consistency checks, allowing authorized review of genuine unusual values.
YOUR TASK
Create entries for five variables: outcome, exposure, one confounder, selection information and a process variable. Draft two neutral questionnaire items, then revise one leading or double-barrelled example.
Specify skip logic for an inapplicable question and demonstrate that it will not overwrite a meaningful answer with zero. Review which fields collect identifiers and whether each is necessary.
SELF-CHECK AND FEEDBACK
Does consistent wording alone establish validity? No; the measurement must capture the intended construct adequately. Should every skipped answer be coded as “no”? No; distinguish inapplicable or missing answers according to the design. Can category codes be averaged meaningfully? Not merely because they are numeric.
PORTFOLIO OUTPUT
A versioned questionnaire excerpt, five-variable codebook and objective-to-measurement map.
FURTHER READING
CDC Field Epidemiology Manual: Collecting Data.
https://www.cdc.gov/field-epi-manual/php/chapters/collecting-data.html
WHO STEPS Manual.
https://www.who.int/docs/default-source/ncds/ncd-surveillance/steps/part2.pdf
My study notes
Stored only in this browser. Do not enter patient information.Saved in this browser. No account needed.