Background/aim Recent epidemiological studies investigating the modifying effect of air temperature in ozone-mortality associations lack consensus as how to adjust for nonlinear temperature effect in addition to including an interaction term.

Methods We evaluated the influence of temperature confounding control on temperature-stratified ozone-mortality risks in a time-series setting in eight European cities and 86 U.S. cities, respectively. To investigate potential residual confounding, we additionally incorporated next day’s ozone in models with differing temperature control.

Results Using only a categorical variable for temperature or only controlling nonlinear effect of low temperatures yielded highly significant ozone effects at high temperatures, but also significant residual confounding in both regions. Adjustment for a nonlinear effect of temperature, especially high temperatures, substantially reduced ozone effects at high temperatures and residual confounding. For example, when using a distributed lag nonliner temperature term, risk estimates at low, medium, and high temperatures were 0.23% (95% CI: −0.09% to 0.55%), 0.23% (95% CI: −0.06% to 0.53%), and 0.36% (95% CI: 0.04% to 0.68%) in European cities, and 0.11% (95% CI: −0.31% to 0.54%), 0.17% (95% CI: −0.07% to 0.41%), and 0.59% (95% CI: 0.32% to 0.85%) in U.S. cities.

Conclusion Inadequate control for confounding by air temperature leads to residual confounding and an overestimation of the temperature modifying effect in studies of ozone-related mortality.