A study finds that children who eat breakfast have higher school scores. Breakfast may help, but another factor could influence both: household routine, food security, sleep, school attendance or caregiver time. A factor linked to both the supposed cause and the outcome is called a confounder.
Think of a hidden third path
If A and B appear together, ask whether C could be connected to both. Confounding does not mean the finding is fake. It means the simple story “A caused B” may be incomplete.
For example, children who spend more time outdoors may sleep better. Daylight and movement could matter, but family schedules, housing, neighbourhood safety and device routines may also affect both outdoor time and sleep.
Use the three-arrow check
- Write the exposure or behaviour the headline highlights.
- Write the outcome researchers measured.
- List factors that might influence both.
Draw arrows from each possible confounder to both boxes. Then check whether researchers measured those factors and how they handled them.
What “adjusted for” means
Researchers may use statistical methods to compare people who are more alike on measured characteristics. An adjusted result can be more informative than an unadjusted one, but adjustment is not magic. It cannot fully correct a variable that was measured poorly, and it cannot include a factor the researchers never collected.
The Office for National Statistics distinguishes sampling uncertainty from non-sampling problems such as inaccurate answers, nonresponse and analysis errors. MedlinePlus advises checking who was studied, the design, duration, funding and whether the result applies to you.
Do not adjust away the pathway
Sometimes a third variable is not a confounder but part of how an effect might happen. If an activity improves sleep partly because it increases movement, adjusting for movement could remove part of the relationship researchers want to understand. That is why the paper’s reasoning matters, not just the length of its variable list.
Questions parents can ask
- What factors differed between the groups before the outcome?
- Which were measured and which may be missing?
- Were the measures reliable and collected before the result?
- Did the authors discuss residual confounding?
- Does another design point in the same direction?
Pair this with the guide to cause versus association and the guide to interpreting one study.
Choose a proportionate response
A confounded study can still generate a useful question. It should not, by itself, justify blame, a costly purchase or a health decision. For a low-risk idea, you might test a small reversible change. For medical or safety choices, compare authoritative guidance and consult a qualified professional.
A fair summary sounds like: “These things were linked in this group. Researchers accounted for some differences, but other explanations may remain.” That sentence is less dramatic than a headline and usually more honest.
Sources and further reading
A quick test for confounding
When a headline links one family habit with one child outcome, draw three boxes: the habit, the outcome, and “other things that may affect both.” Fill the third box before deciding what the result means. For example, families who read more often may also differ in available time, access to books, adult education, language, or the child’s existing interest. A careful study tries to measure and account for plausible alternatives, but adjustment is never magic: researchers can only adjust for variables they measured well.
Then look for the authors’ own limitations. Useful phrases include “residual confounding,” “unmeasured confounding,” and “observational design.” These are not admissions that the study is useless. They tell you how cautiously to interpret it. A consistent pattern across different groups and methods is more reassuring than one dramatic result.
What to do at home
Use the finding as a question, not a command: “Would this be reasonable for our family to try?” Choose changes that are low-risk, affordable, and compatible with your child. Observe what happens without assuming your household experiment proves cause. For health, development, or safety decisions, bring the full study—not only the headline—to a qualified professional.
