The number of trauma occurrences on any given day may be affected by the weather conditions. Although previous studies have analysed the possible correlation between weather variables and trauma centre admissions, the results were conflicting.1^–3 For example, Bhattacharyya and Millham3 reported a positive correlation between trauma admissions and daily maximum temperature but a negative correlation between these admissions and rainfall. On the other hand, Atherton et al2 suggested that rainfall was not a determinant for trauma admissions. Most recently, however, Rising et al1 indicated that increased rainfall, as well as high temperature, was positively associated with trauma admissions.

The difference in the results of previous studies may be related to the use of admission data from each of the single trauma centres and also their analytical methods.1^–3 The admission data may not be able to capture the whole spectrum of trauma patients, because many of them are not severely injured and may be discharged without the need for admission. Trauma admissions in a single centre may not identify the population-level effect of weather on trauma occurrence. A large population-based database capturing a broader spectrum of trauma patients was needed. Furthermore, previous studies used correlation and/or linear regression models for analysing the association between weather conditions and trauma occurrence.1^–3 However, because weather variables are also autocorrelated over consecutive time, it is appropriate to adjust the autocorrelations among these time-series data, using a time-series analysis such as an autoregressive integrated moving average (ARIMA) model. Moreover, additional weather factors, such as humidity and barometric pressure, were not previously evaluated for their possible associations with trauma occurrence.

Therefore, in the current study, we aimed to evaluate the associations between trauma occurrence and major weather factors, including temperature, rainfall, humidity and barometric pressure. We used data on all ambulance transports throughout the Tokyo metropolitan area (one of the largest cities in the world) to examine a broader spectrum of trauma patients. In addition, we employed the ARIMA time-series model to analyse these variables as a result of a study design with time-series data. Furthermore, we also analysed the effects of holidays on the number of trauma patients, because one study indicated that holidays may affect the number of visits to hospital emergency departments.3

MATERIALS AND METHODS

Study population

We conducted the study in Tokyo, the capital prefecture of Japan, with a population of approximately 12 million in 2005, a land area of 2187 square kilometres and a temperate climate. As a prefectural institution, Tokyo Fire Department organises a one-tiered system covering the entire prefecture, with basic life-support ambulances based at 80 fire stations throughout the prefecture.

Given national insurance coverage for all individuals, the Japanese people are free to access an emergency department, either by walk-in arrivals or by emergency ambulance transports in a given local area. As ambulance services are free of charge, usage is encouraged. In addition, because insurance reimbursement requires trauma patients to submit the records of police and fire stations in the details of accidents, emergency ambulance services are used for almost all trauma patients when they are being transported to an emergency department.

Ambulances are mostly staffed by non-physician emergency medical technicians. Ambulance staff are required to collect digitalised data of clinical information on all patients transported to hospital emergency departments, including the follow-up of diagnostic information provided by emergency physicians. Previous ethical approval was obtained for this study from the Institutional Reserve Board of St Luke’s International Hospital.

Data collection

We reviewed data from the ambulance transport records of the Tokyo Fire Department for all patients transported to hospital emergency departments of Tokyo, Japan, during a one-year period from 1 January 2005 to 31 December 2005. Of these patients, we identified trauma patients including those from motor vehicle collisions, using follow-up diagnostic data. We collected data for the hourly occurrence of trauma based on the time of the ambulance call for each trauma patient and determined the total number of hourly occurrences.

We also collected hourly meteorological data for the Tokyo metropolitan area based on the official publications of the Japan Meteorological Agency during the same period. Hourly meteorological data included barometric pressure (hectopascal), ambient temperature (centigrade), relative humidity (percentage) and rainfall precipitation (mm). We also collected data on calendar variables (seasons and holidays). The four seasons were defined by the Japanese government meteorological classification: spring (March to May), summer (June to August), autumn (September to November) and winter (December to February). In this study, holidays included Saturday and Sunday, and national holidays were based on the government criteria of Japan.

Statistical analysis

The hourly occurrence of trauma including motor vehicle collisions was treated as a dependent variable. Independent variables included hourly data for the following meteorological factors: barometric pressure (hectopascal), ambient temperature (centigrade), relative humidity (percentage) and rainfall precipitation (mm). A factor for holidays was also included in the independent variables. Descriptive statistics were calculated for the demographics of trauma patients, as well as for patients from motor vehicle collisions and for meteorological factors.

In analysing time-series data, we used an ARIMA model, which has the ability to cope with the stochastic dependence of consecutive data and has become well established in the commercial and industrial fields.4^–6 Six main parameters were selected when fitting the ARIMA model: the order of autoregressive (p) and seasonal autoregressive (P), the order of integration (d) and seasonal integration (D) and the order of moving average (q) and seasonal moving average (Q). Consequently, the process is called ARIMA (p, d, q) (P, D, Q). The selection of ARIMA processes was conducted using Akaike’s information criterion, which measures how well the model fits the series.

A multivariable-adjusted ARIMA model was constructed for fitting the model with a generalised least squares regression by controlling autocorrelation and cross-correlations of time-series data. The final ARIMA model was constructed using a stepwise approach, including only significant variables as the covariates. The beta coefficient of significant variables represents the slope of the regression surface. In each significant variable, a plus code of beta coefficient indicates a positive correlation, whereas a minus code of beta coefficients indicates a negative correlation. All p values were two-sided and p<0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 15.0J with SPSS TRENDS system.

RESULTS

We identified a total of 643 849 patients who were transported by ambulance during the one-year study period in Tokyo. Of these, there were 226 339 trauma patients (35% of all patients). They included 94 916 patients from motor vehicle collisions (42% of all trauma patients). Trauma was ranked as top of the most frequent reasons for ambulance calls. The mean age was 46 years for all trauma patients and 37 years for patients from motor vehicle collisions. Male gender represented 59% of all trauma and motor vehicle collisions. The proportions of patients requiring subsequent admission to hospital wards were 22.5% for trauma patients and 11.5% for motor vehicle collisions.

Regarding the mean hourly numbers of all trauma and motor vehicle collisions, the highest occurrence of both all trauma and motor vehicle collisions was identified as between 17:00 and 18:00 hours, whereas the lowest occurrence of all trauma and motor vehicle collisions was noted between 04:00 and 05:00 hours. The dual peaks were found in the number of motor vehicle collisions both in the morning and evening rush hours and the nadir was identified at midnight. Regarding the hourly meteorological data, the mean ambient temperature was highest at 14:00 hours and lowest at 05:00 hours. Figure 1 shows the hourly mean number of trauma patients and the mean ambient temperature, illustrating the correlation of these two variables.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 No of trauma patients and the significance of ambient temperature.

In terms of the mean monthly numbers of all trauma and motor vehicle collisions, the greatest mean daily number of both all trauma and motor vehicle collisions was identified in December. The lowest mean daily number of all trauma transports was noted in May, whereas that of motor vehicle collisions transports was noted in January. In terms of the mean monthly values of meteorological data, the mean monthly rainfall precipitation was greatest in July. There were many more dates of holidays in December and August 2005 in the Japanese calendar.

The ARIMA (0, 0, 2) (1, 0, 0) model was fitted for significant meteorological covariates associated with the occurrence of all trauma (table 1). The significant factors included higher temperature (p<0.001), increased rainfall (p<0.05) and holidays (p<0.001). Barometric pressure and humidity were not associated with the occurrence of transporting all trauma patients. Based on the beta coefficients (non-standardised coefficients), the mean number of all trauma transports would be increased at 0.098 by the increase of each degree in temperature, increased at 0.133 by the increase of each millimetre in rainfall precipitation and increased at 0.991 on holidays.

When the ARIMA (1, 0, 3) (1, 0, 1) model was fitted for significant meteorological covariates associated with the occurrence of motor vehicle collisions (table 2), the significant factors also included higher temperature (p<0.05), increased rainfall (p<0.01) and holidays (p<0.01). Barometric pressure and humidity were also not associated with the occurrence of motor vehicle collisions. Based on the beta coefficients (non-standardised coefficients), the mean number of transports for motor vehicle collisions would be increased at 0.025 by the increase of each degree in temperature, increased at 0.121 by the increase of each millimetre in rainfall precipitation and increased at 0.261 on holidays.

DISCUSSION

In our time-series analysis based on a large population database, high temperature, rainfall and holidays were significantly associated with the occurrence of trauma and these factors were also significantly associated with motor vehicle collisions. Barometric pressure and humidity were not associated with the occurrence of trauma. There are several implications of the results of our study. Public agencies may be able to use the results to announce safety precautions regarding trauma and motor vehicle collisions to the general public. Ambient temperature may be used as additional information for determining speed limit criteria for automobiles.1 Furthermore, for local governments, the number of available emergency departments to accept trauma patients can be adjusted according to these meteorological factors.3

The strength of our study is that we analysed relationships among meteorological factors based on large, population-based and hourly data including 226 339 trauma occurrences and 8759 hourly data for meteorological factors. Our analysis also included patients with less severe injuries and those not admitted to hospital wards. In addition, we also performed a separate analysis for motor vehicle collisions. Furthermore, we used a statistical method to adjust autocorrelations and cross-correlations of time-series data.

Several studies have been conducted to evaluate the relationship between weather factors and the occurrence of trauma that were based on the daily admission volume from single trauma centres.1^–3 As patients who were not admitted to trauma centres because of less severe injury were not included for the analyses of these studies, their results may not reflect the potential relationships of trauma occurrence with meteorological factors. Nevertheless, the results of our study are in line with some of their findings, including higher temperature, increased rainfall and holidays as significant factors for trauma occurrence.

The possible mechanisms linking this association between higher temperature and the occurrence of trauma may include increased outdoor activities on days with higher temperatures. The increased outdoor activities are likely to put people at risk of traumatic injuries and motor vehicle collisions. Wearing thin clothes on days with higher temperatures may also contribute to the higher risk of these injuries.

Regarding the association between increased rainfall and the occurrence of trauma, people may be at increased risk of accidents when driving on slippery roads during rainy days.2 Using umbrellas and rain itself may contribute to a higher risk of traffic injuries by the narrowing of visual fields. Because people are also likely to use automobiles and be more hurried on rainy days, heavy traffic along with high speeds increases the risk of traumatic injury.

In terms of increased trauma on holidays, Holleman et al7 and Atherton et al2 provided evidence for a calendar pattern affecting the number of hospital visits. Bhattacharyya and Millham3 also suggested increased trauma admissions on holidays and holiday seasons including July and August. On holidays, people are likely to go outside into unfamiliar places, which may be a risk for accidents, and people may also become tired returning back home from remote areas during holiday travel.

The limitation of our study is that our results were based on a one-year study period and we could not analyse seasonal circadian patterns.8 Future studies to follow population-based data over several years are needed to analyse seasonal circadian patterns of trauma, including motor vehicle collisions.

In conclusion, high temperatures, rainfall and holidays are related to an increased occurrence of ambulance transports of trauma patients, including those from motor vehicle collisions. Based on continuous monitoring information on these conditions, public agencies may be able to provide safety precautions for the general public. Motor vehicle drivers, motor cyclists, cyclists and pedestrians are also able to use this weather information to take more care in preventing traffic accidents.