Situation: Over 3 weeks, 46 residents from many barangays of a municipality developed acute bloody diarrhea. They had bought food from many different sources, so no list of all exposed persons exists. The public health nurse helps the municipal epidemiology and surveillance unit compare the cases with well residents (controls) on what they ate before the illness. Older residents were more likely both to eat the flavored ice and to develop bloody diarrhea. For a similar study the team is planning in the next outbreak, which combination of measures should it include in its plan so that age does not distort the odds ratio? 1. Match each control to a case of similar age 2. Double the number of controls in every age group 3. Analyze the results separately for each age group 4. Enroll as controls only residents who ate the ice

Age is linked to both the exposure and the outcome, so it is a potential confounder. Confounding is controlled at the design stage by matching (or restriction) and at the analysis stage by stratification or multivariable analysis. Adding controls improves precision but does not remove confounding, and choosing controls by exposure status would make the comparison of exposure meaningless.

In this outbreak investigation, the team is using a case-control design because there is no defined cohort of all exposed persons. The goal is to compare the odds of exposure (eating the flavored ice) between cases (residents with bloody diarrhea) and controls (well residents). However, age is associated with both the exposure and the outcome: older residents were more likely to eat the flavored ice and were also more likely to develop bloody diarrhea. This makes age a confounder. If age is not addressed, the crude odds ratio will mix the effect of the ice with the effect of age, producing a distorted estimate.

Confounding can be controlled at two stages: during study design (by matching or restriction) and during data analysis (by stratification or multivariable adjustment). The question asks which measures should be included in the plan for the next outbreak so that age does not distort the odds ratio.

Measure	Does it control confounding by age?	Reason
1. Match each control to a case of similar age	Yes	Individual matching forces the case and control to have the same or nearly the same age, so age can no longer differ systematically between the two groups. This removes age as a source of bias at the design stage.
2. Double the number of controls in every age group	No	Increasing the number of controls improves statistical precision (narrower confidence intervals), but it does not change the distribution of age between cases and controls. The imbalance in age remains, so confounding is not removed.
3. Analyze the results separately for each age group	Yes	Stratified analysis calculates the odds ratio within each age stratum and then combines them (e.g., Mantel-Haenszel). Because age is held constant within each stratum, its confounding effect is controlled at the analysis stage.
4. Enroll as controls only residents who ate the ice	No	Selecting controls based on exposure status would make cases and controls identical with respect to the exposure. This destroys the ability to compare exposure frequency, which is the entire purpose of a case-control study. It does not address age confounding.

In the cited review of case-control studies on postoperative infections, 64.3% of the 42 studies that used individual matching selected age as a matching criterion ^[1]. This reflects how commonly age is recognized as a potential confounder in clinical and epidemiologic research. Matching on age is a design-stage strategy that prevents age from distorting the exposure-outcome association before any data are collected.

Key point! Matching and stratification are complementary, not competing, approaches. A well-designed study may match on age and still stratify by age in the analysis to account for any residual imbalance from imperfect matching. Watch out! Adding more controls or selecting controls by exposure status does not address confounding. Only measures that make age comparable between cases and controls—or that hold age constant during analysis—can remove the distortion.

Therefore, the correct combination is matching controls to cases by age and analyzing results separately by age group, which corresponds to measures 1 and 3.