Author's response to reviews Title:Effect of ambient temperature on emergency department visits in Shanghai, China: a time series study Authors:
Jianming Xu,Wei-bing Wang
2014-01-01
Abstract:1. It is unclear why and how a moving average lag model was used to evaluate t he lag effects of temperature on EDVs. The reference group needs to be provided when the authors stated “The cold effect seemed to be more acute among individuals aged <45 years”. What is the meaning to say “there were consist e t effects on individuals aged ≥65 years”? Response: Thanks. At present, the models to assess the lag effects mainly include d the sing-day lag model and the multi-day lag model (moving average lag mode l). For example, we used the moving average lag model to evaluate the la g effects, given that single-day lag models may underestimate the cumulative effect o temperature on mortality or hospital admissions (references: Bell et al. Time-series studies of particulate matter. Annual Review of Public Health.2004,25,247-280.). Accordingly, we have revised the statement in the Abstract section. Methods: 1. More details are needed to explain how they selected a temperature breakpoint (commonly called the “threshold”). Why did they use a linear-threshold model t quantify the effect of ambient temperature as the temperature-EDV relation is likely to be non-linear? How was the lag structure determined? Response: Thanks. First, we plotted the exposure-response curve between ED visits and ambient temperature based on the GAM. While the relationship between mean temperature and ED visits followed a U-shape curve, we couldn’t get the relative risk (RR) of ED visits per 1°C change in temperature. As a result, we cut the curve into two parts based on the threshold and the each part was similar to linear. Then we used the linear-threshold model to quantify the effect of ambient temperature on ED visits. Lot s of studies have demonstrated the distribution of temperature on mortality or morbidity is nonlinear and usually has a U-, Vor J-shape, the key issue for the study of temperature-related mortality is to define the turning point, which determine s the magnitude of the estimated risk below or above the threshold. The threshold temperature is generally the optimum temperature corresponding to the lowest r isk of mortality of morbidity. The threshold temperature is often different for differ ent locations and populations because of local climate and population acclimatisation. The lag structure up to 3 days (lag3) was finally selected to a ssess the effect of ambient temperature on ED visits because the DMT effects appe ared to be stable after 3 lag days. 2. They utilized a time series model to assess the relationship betwe en EDVs and ambient temperature, while controlling for air pollutants (PM 10-2.5, SO2 and NO2), relative humidity, rainfall and wind speed. Why did rainfall and wind speed matter? Additionally, how was the multi-linearity issue between humidity and rainfall dealt with in the model? Response: Thanks. In our model, in order to observe the independent effects o ambient temperature on ED visits, we controlled for potential confounder s such as long-term trend, day of week, relative humidity, wind speed and air pollutant. Considering the multi-linearity issue between humidity and rainfal l, we didn’t include the rainfall in our model. 3. They stated: “the optimum temperature (OT), corresponding to the lowest point in the E-R curve, was about 12°C”. Was OT only visually observed or tested in he model? The authors claimed that “the risk of ED visits decreased at t emperatures below the OT ...”Should “decreased” be changed to “increased”? Response: Thanks. As the exposure-response curve shows, the OT was 12°C a s observed. Therefore we have revised the sentence “the risk of ED vis its increased at temperatures below the OT and then reversed the trend”. Results: 1. Why was mean temperature used as an indicator of ambient temperature? The reasons for using mean temperature should be provided in the Methods section. Why and how was lag 0-3 selected? The OT varied with age but how about gender? Response: Thanks. In some studies, daily maximum or minimum temperat ure, synthetic measure such as humidex or apparent temperature were used to examined the effects on EDVs. Daily measurements may have different i mpacts with average ambient temperature on EDVs. Sensitivity analyses from previous s tudie using multiple temperature metrics have found that the effect estima tes of all temperature metrics were similar. Moreover, considering daily mean temperat ur may be able to provide more easily interpreted results in a policy context and more familiar to the public, we use it for subsequent analyses. We have added the reasons in t he Methods section (please see page 8, paragraph 1). 2. Why and how was lag 0-3 selected? The OT varied with age but how about gender? Response: Thanks. We examined the cumulative effect of ambient tempe rature on ED visits from lag 0 to lag 3, given that previous epidemiologic studies in China found little evidence of significant associations with a lag over 3 day s (references: YZ Mo et al. Relationship between daily mean temperature and emergency departm nt visits for respiratory diseases: a time-series analysis. Journal of Peking University, 44(3):416-419). In addition, the lag structure up to 3 days (lag3) was final ly selected to assess the effect of ambient temperature on ED visits bec aus the DMT effects appeared to be stable after 3 lag days. Thus, we only calculate d the cumulative effect of lag 3 days. In our study, we repeated the same procedure to e xamin the associations stratified by gender and age (<45, 45-65, 65-75 and ≥75 years). By observing ED visit-ambient temperature curves in males and female s, we found that the OT was 12 °C. However, the OT varied with age. Discussion: 1. Most vulnerable groups (e.g., floating population, poor people and unemployed residents) may not involve in this study. Selection bias should be thoroughly discussed. Response: The study population was relatively older than the entire population in Shanghai (Page 10, paragraph 3). But we agree with you that floating population and other unemployed residents may have different impacts structures. We have extended the limitations thoroughly in the Discussion section. On behalf of my co-authors, I thank the reviewers and editors for their insightful comments and suggestions. We feel the incorporation of this fe edback has greatly improved this revised manuscript. Sincerely, Weibing Wang