BACKGROUND: The relationships between cold temperatures and cardio-respiratory mortality in the elderly are well documented. We wished to determine whether similar relationships exist with consultations in the primary care setting and to assess the lag time at which the effects were observed. METHODS: Generalized additive models were used to regress time-series of daily numbers of general practitioner (GP) consultations by the elderly against temperature, after control for possible confounders and adjustment for overdispersion and serial correlation. Consultation data were available from between 38 452 and 42 772 registered patients aged >or=65 years from 45-47 London practices contributing to the General Practice Research Database between January 1992 and September 1995. RESULTS: There was little relationship between consultations for respiratory disease and mean temperature on the same day as the day of consultation. However, a strong association was apparent with temperature levels up to 15 days previously, with an increase in consultations being observed particularly as temperatures drop below 5 degrees C. Every 1 degrees C decrease in mean temperatures below 5 degrees C was associated with a 10.5% (95% CI: 7.6%, 13.4%) increase in all respiratory consultations. No relationship was observed between cold temperatures and GP consultations for cardiovascular disease. CONCLUSIONS: Our study suggests a delayed effect of a drop in temperature on consultations for respiratory disease in the primary care setting. Information such as this could be used to help prepare practices to anticipate increases in respiratory consultation rates associated with low temperatures.