In this issue of the journal, Preziosi et al. [2004] report the first study to assess differences in the utilization of health care related to the presence of air-conditioning in office workplaces. Although the study was simple and cross-sectional, the data variables from questionnaires, and the findings subject to a variety of questions, the findings are striking enough to deserve clarification. The study used a large random national sample of French women assembled for another purpose (to study antioxidant nutrients and prevention of cancer and cardiovascular disease). Participants reported health services and health events in monthly questionnaires over 1 year, and in one questionnaire in the middle of that period also reported whether air-conditioning was in use at their workplace. Fifteen percent of participants reported air-conditioning at work. Analyses adjusting for age and smoking status of participants found increases in most outcomes assessed: use of specific kinds of physicians, sickness absence, and hospital stays. While the increases in odds ratios (OR) and 95% confidence intervals (CI) were statistically significant for only otorhinolaryngology [OR (95% CI) = 2.33 (1.35-4.04)] and sickness absence [1.70 (1.13-2.58)], other increases were notable--dermatology [1.6 (0.98-2.65)]; hospital stay [1.51 (0.92-2.45)], and pneumonology [2.10 (0.65-6.82)]. The least elevated outcomes were for general practice medicine [0.99 (0.65-1.48)] and global medical visits [1.18 (0.67-2.07)]. [Preziosi et al., 2004 ,(Table 2)] Odds ratios for relatively common health outcomes often lie farther from the null than the risk ratios most useful for quantifying the increase in risk. Risk ratios, or prevalence ratios (PRs, the equivalent measure of effect for cross-sectional data), have seldom been used because of the convenience and availability of logistic regression models that estimate odds ratios. With baseline prevalences ranging up to 85.7% in the data from Preziosi et al. [2004], PRs allow a more appropriate estimate of the increase in each outcome associated with the risk factor of air-conditioning. The increase in prevalence was roughly estimated as [100* (crude PR * adjusted OR/crude OR)-100] %. Based on the data in Table 2 of Preziosi et al. [ 2004], estimates for the increased prevalence associated with air-conditioned offices include increases of 120% in otorhinolaryngology visits, and 40% in sickness absence. If these associated increases represented valid causal relationships, it would indicate enormous costs for employers and for society associated with air-conditioning systems, from increased health care and for reduced workplace productivity from sickness absence, in addition to a large burden of disease on workers.