ADHD glossary · Family perspectives
Confounding: ask about a shared cause behind a comparison
In its guidance on non-randomized intervention studies, Cochrane describes confounding as arising when shared causes influence both the intervention chosen and the outcome being examined. [1]
General education and practical perspectives. Editorial standards.
For an everyday reading task, ask whether a proposed explanation could affect both sides of the relationship. Be specific about those two connections. Merely naming something else present in the situation does not finish that reasoning.
The taller display stands
Imagine a fictional report comparing stalls at a neighborhood exhibition. Stalls with taller display stands record more visitors. A promotional post credits the stands with attracting people.
At home, Rafi asks how exhibitors chose their stands. His sister notices a possible shared explanation: experienced exhibitors might be more likely to bring tall stands and also have established mailing lists that bring visitors. The family writes this as a possibility to check, not a discovery about the fictional exhibitors.
They separate the two proposed connections: experience could influence the stand selected; experience could also influence attendance through an existing audience. If either connection lacks support in the actual report, that stays visible. They do not declare the stand useless, accuse exhibitors of deception, or calculate an adjusted result from guesses.
The family can choose a display stand that fits their materials while remaining uncertain about the attendance claim. This story is invented and contains no real exhibition data.
Put the alternative explanation into words
For a claim you are reading, try completing these optional sentences:
The comparison is between ___. The outcome is ___. A possible shared cause is ___. It might influence the choice because ___. It might influence the outcome because ___. The source does / does not address this possibility at ___.
If you cannot explain both connections, keep investigating or set the idea aside. You do not have to produce a list of every imaginable factor. A short, relevant question is more useful for this task than a page of unsupported possibilities.
What should you ask the source?
Look for how the groups came to differ, what information was collected before the outcome, and which alternative explanations the authors considered. If the report says it “adjusted for” something, ask what that something was and what limitations the authors retained. Do not translate the phrase into “every other explanation has been removed.”
This page does not teach statistical adjustment. It supplies a way to name a question you might bring to a fuller methods explanation.
Keep related questions distinct
Correlation focuses on the relationship reported. Causation focuses on whether one thing produced a difference in another. Confounding focuses on a particular complication involving a shared cause. Observational study focuses on what investigators observed rather than changed. Use the links according to the question you have, not as a ladder for ranking relatives' opinions.
A teen can ask an adult about the comparison without having to disprove it. A parent can say that a possible explanation is still only possible. A partner need not become the household statistician. Keep the discussion about evidence and wording; do not label a person “a confounder” or turn the exercise into surveillance of family behavior.
Source
[1] Sterne JAC, Hernán MA, McAleenan A, Reeves BC, Higgins JP. Chapter 25: Assessing risk of bias in a non-randomized study. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Cochrane; 2024. Chapter updated October 2019; accessed September 5, 2026. Section 25.2.1 supports [1]. No risk-of-bias instrument or clinical example is reproduced.
Related guides
Sources and what they support
These sources support the specific statements identified below. Our practical exercises are suggestions to consider, not evidence of a treatment effect.
-
Chapter 25: Assessing risk of bias in a non-randomized study — Cochrane
Source type: institutional.
Accessed .
Source check, September 5, 2026: Scoped source check. Passage: Section 25.2.1 Confounding, opening paragraph, retrieved line 77; scope and limits lines 79–85; authors/citation lines 50 and 60. Limits: Definition in non-randomized intervention-study context. No ROBINS-I questions, risk grades, tables, clinical examples, statistical adjustment procedure, or version-status recommendation reproduced.
Supports: In its guidance on non-randomized intervention studies, Cochrane describes confounding as arising when shared causes influence both the intervention chosen and the outcome being examined.
Correction check: Inspected publisher title, chapter citation/update line, and relevant definition section; searched the retrieved chapter for correction and retraction markers. No notice affecting this definition found. Matches for correction in Chapters 1 and 10 concern methods, not a correction notice. No exhaustive publisher-history or bibliographic clearance performed.
Updates and corrections
Published .
Found a factual error or something unclear? Use the contact route on our Trust page. Read our research method for the evidence approach.