04. Design clinic: choose comparisons that answer the question
Lesson
LEARNING OBJECTIVES
Identify appropriate comparison participants, distinguish allocation concealment from blinding, and explain how design choices limit interpretation.
CASE A — A RARE OUTCOME
A fictional district has identified cases of an uncommon outcome. A case–control design might be efficient, but the comparison group must represent the population that produced the cases under the planned sampling scheme. Define a case, ascertain cases consistently and specify when each control was eligible.
Do not choose controls because they lack the exposure. Hospital controls may be convenient, but their reason for admission could itself relate to exposure. Matching can improve efficiency for selected factors but creates design and analysis requirements. Do not match on the exposure, automatically match on every variable or assume matching alone removes confounding.
CASE B — A DIAGNOSTIC TEST
A new screening method is intended for people attending routine services. Comparing obvious advanced cases with exceptionally healthy volunteers may exaggerate apparent accuracy relative to intended use. Plan recruitment across the relevant spectrum, consistent reference-standard verification, prespecified thresholds and handling of indeterminate results.
Readers need the purpose and timing of both tests, who interpreted them, what information they had and whether all eligible participants were included in analysis. A diagnostic table with perfect arithmetic can still come from a biased study.
CASE C — A PROGRAMME EVALUATION
A reminder programme could be assigned to individuals or whole clinics. Individual randomization can be complicated by contamination between participants. Cluster randomization may suit programme delivery, but the number of clusters, clustering and analysis must be planned appropriately.
Allocation concealment prevents recruiters predicting forthcoming assignment before enrolment. Blinding concerns knowledge after assignment and can affect care or assessment. When blinding is impractical, identify objective outcomes and standardized assessment strategies. Explain the primary contrast, such as assignment to the programme, and how deviations will be reported.
YOUR TASK
Choose two cases above. For each, write the recruitment rule, comparison definition, primary outcome, measurement timing, analysis implications and two major threats. Add a practical prevention measure for each threat.
Now review your own proposed study. Ask whether people in the compared groups could have entered the study by comparable rules and had comparable opportunities for outcome detection.
SELF-CHECK
1. Are unexposed controls necessarily appropriate? No; exposure-based selection can distort the OR.
2. Is a well-reported test threshold sufficient to establish applicability? No; recruitment, reference standard and intended-use population also matter.
3. Does randomizing ten clinics allow analysis as though all residents were independent? No; account for cluster allocation and dependence.
PORTFOLIO OUTPUT
Two design-clinic memos plus one revision to your own design. Record unresolved issues for methodological consultation rather than choosing a design solely because it is easiest to describe.
FURTHER READING
CDC Field Epidemiology Manual: Designing and Conducting Analytic Studies in the Field.
https://www.cdc.gov/field-epi-manual/php/chapters/design-conduct-analyze-field-studies.html
FDA: Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/statistical-guidance-reporting-results-studies-evaluating-diagnostic-tests-guidance-industry-and-fda
Identify appropriate comparison participants, distinguish allocation concealment from blinding, and explain how design choices limit interpretation.
CASE A — A RARE OUTCOME
A fictional district has identified cases of an uncommon outcome. A case–control design might be efficient, but the comparison group must represent the population that produced the cases under the planned sampling scheme. Define a case, ascertain cases consistently and specify when each control was eligible.
Do not choose controls because they lack the exposure. Hospital controls may be convenient, but their reason for admission could itself relate to exposure. Matching can improve efficiency for selected factors but creates design and analysis requirements. Do not match on the exposure, automatically match on every variable or assume matching alone removes confounding.
CASE B — A DIAGNOSTIC TEST
A new screening method is intended for people attending routine services. Comparing obvious advanced cases with exceptionally healthy volunteers may exaggerate apparent accuracy relative to intended use. Plan recruitment across the relevant spectrum, consistent reference-standard verification, prespecified thresholds and handling of indeterminate results.
Readers need the purpose and timing of both tests, who interpreted them, what information they had and whether all eligible participants were included in analysis. A diagnostic table with perfect arithmetic can still come from a biased study.
CASE C — A PROGRAMME EVALUATION
A reminder programme could be assigned to individuals or whole clinics. Individual randomization can be complicated by contamination between participants. Cluster randomization may suit programme delivery, but the number of clusters, clustering and analysis must be planned appropriately.
Allocation concealment prevents recruiters predicting forthcoming assignment before enrolment. Blinding concerns knowledge after assignment and can affect care or assessment. When blinding is impractical, identify objective outcomes and standardized assessment strategies. Explain the primary contrast, such as assignment to the programme, and how deviations will be reported.
YOUR TASK
Choose two cases above. For each, write the recruitment rule, comparison definition, primary outcome, measurement timing, analysis implications and two major threats. Add a practical prevention measure for each threat.
Now review your own proposed study. Ask whether people in the compared groups could have entered the study by comparable rules and had comparable opportunities for outcome detection.
SELF-CHECK
1. Are unexposed controls necessarily appropriate? No; exposure-based selection can distort the OR.
2. Is a well-reported test threshold sufficient to establish applicability? No; recruitment, reference standard and intended-use population also matter.
3. Does randomizing ten clinics allow analysis as though all residents were independent? No; account for cluster allocation and dependence.
PORTFOLIO OUTPUT
Two design-clinic memos plus one revision to your own design. Record unresolved issues for methodological consultation rather than choosing a design solely because it is easiest to describe.
FURTHER READING
CDC Field Epidemiology Manual: Designing and Conducting Analytic Studies in the Field.
https://www.cdc.gov/field-epi-manual/php/chapters/design-conduct-analyze-field-studies.html
FDA: Statistical Guidance on Reporting Results from Studies Evaluating Diagnostic Tests.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/statistical-guidance-reporting-results-studies-evaluating-diagnostic-tests-guidance-industry-and-fda
My study notes
Stored only in this browser. Do not enter patient information.Saved in this browser. No account needed.