After adjusting for confounding factors (i.e., daily mean temperature, RH and PM2.5), we confirmed that DTR was an independent risk factor on hospital admissions for exacerbations of CRD, represented by COPD, asthma and bronchiectasis. We also assessed the potential effect medication of season and population characteristics (i.e., sex and age). We found that the associations, as well as the dose-response curses, were diverse in the hot and cold seasons. Hospital admissions for COPD and asthma were associated with DTR. Adverse effects of DTR on bronchiectasis patients were only observed in female.
Most of the previous studies focused on the relationship of DTR and mortality, and respiratory and cardiovascular diseases were regarded as the main cause of mortality after short-term exposure of DTR [29]. Morbidity is another important outcome of exposure to DTR. The effects of DTR on respiratory-related emergency room visit and out-patients service was usually investigated in the previous study but few hospital admissions which reflects severe effects of DTR [30–33]. To fill this knowledge gap, we carried out a province-wide study to investigate the association of hospitalization for exacerbation of CRD, represented by COPD, asthma and bronchiectasis. Our study discovered a nonlinear DTR-CRD relationship in subtropical regions. High RRs of moderate and extreme high DTR deserved more attention. There are potential mechanisms linking DTR and hospitalization for exacerbation of CRD: 1) the host defense function of the respiratory system, nasal responses and airway mucociliary clearance could be influenced when the temperature of respiratory epithelium fluctuated [34, 35]; 2) increased DTR might enhance the transmission of virus and bacteria and resulted in the occurrence of exacerbations of respiratory diseases [29, 36].
In our study, the maximum lag effect of DTR was identified at lag0-6 (RR = 1.09 [95%CI 1.08 to 1.1]) (e-Table 1). Furthermore, the lag effect persisted even when the number of moving average days increased to 21 days (RR = 1. 09 [95%CI: 1.06 to 1.12]). A previous study also demonstrated that eight days moving average of DTR was associated with respiratory emergency room admissions [37]. Similar characteristics of the lag effect of DTR measures deserve great attention, especially in severe outcome variables like hospital admission. Identifying the significant effect period for the occurrence of disease is helpful for the prediction of the DTR-related adverse events [29].
Our study has focused on the whole range of DTR, including extreme DTR in both hot and cold seasons. In the hot season, the RRs of CRD hospitalization increased rapidly in the relatively low DTR (Fig. 2), as heat waves in summer could lead to adverse events, including mortality and morbidity [38]. However, insignificant effects of extremely low DTR were found in the cold season. Given the minimum temperature in the study period is 2.0 °C (Table 1), the adverse effect of cold spells was insignificant on hospitalization for CRD in Guangdong Province. Higher RRs in moderate and high DTR were observed in the cold season, demonstrating the importance of developing the preventive measures of adapting to large DTR in the cold season. For example, it is needed to provide home heating and timely clothing for large DTR in the cold season [39].
The associations of DTR with adverse outcomes of COPD patients have been confirmed in previous studies. A time-series analysis conducted in Shanghai city reported that the association between DTR and daily COPD mortality was significant [40]. The emergency room visit for exacerbations of COPD was associated with DTR in an ecological study in Taichung city, Taiwan [31]. However, a city-level time series analysis reported that the insignificant relationship between DTR and hospitalization for total COPD patients was found in Changchun, a northeastern city of China [21]. Using data from 21 cities and the method of meta-analysis, we demonstrated that the RR of hospital admissions for exacerbations of COPD was 1.11 (95%CI: 1. 08 to 1.12) with per IQR increase of DTR at lag0-6. Furthermore, a previous study in Changchun city observed the greatest estimates for males appeared at lag 7 days, which is in line with our maximum estimate at lag0-6. Several single-city studies have observed the association of DTR with adverse health outcomes, including emergency room visit and hospital admissions, of asthma patients [9, 41, 42]. Our study discovered that the maximum lag effect of DTR on asthma exacerbations was at lag0-6 and the changed slightly until lag0-14. Similar effects were found on emergency department admissions in Brisbane, Australia [41]. To our knowledge, the relationship between hospitalization for bronchiectasis and DTR was assessed firstly. No significant association between DTR and total bronchiectasis patients was found. However, we found that DTR was a risk factor for female bronchiectasis patients who has a higher rate of hospitalization than male patients [43]. Both exacerbations of COPD and asthma have been confirmed to be associated with DTR, but their associations have not been compared directly. The subgroup analysis on diverse CRD shows that COPD patients are most vulnerable to D TR (RR = 1.11[95%CI: 1.08 to 1.1]), and bronchiectasis patients are not sensitive to the temperature change within a day. These results enhanced the importance of reducing the adverse impacts of DTR on CRD patients, especially COPD patients.
To the best of our knowledge, this is the first multi-city study to examine the short-term effect of the DTR on daily hospitalization of CRD. However, there are several limitations of our study. Firstly, measures of DTR were mainly obtained from 21 fixed-site monitoring stations (the data of Chaozhou city and Foshan city was replaced by nearest monitoring site) rather individual exposure. Using city-wide meteorological measures could lead to exposure measurement errors that underestimate the adverse effect of the temperature variation within a day [44]. Secondly, the hospitalization data of each patient depend on the hospital address, not exactly the individual living region. We assumed people would go to the hospital near their living region in a critical situation, and the exposure measurement errors for inter-cities patients could not be solved. Thirdly, although we obtained data from fixed 227 hospitals, the existing CRD populations might have increased due to the aging process of society, which could influence our estimates.
Our findings may close the knowledge gap of the relationship between DTR and CRD and highlight the importance of preventive measures, such as providing home heating, suitable clothing for large DTR, staying indoor to avoid environment temperature variation.