The national burden of road traffic injuries in Thailand

In order to calculate YLL, we needed to determine age- and sex-specific RTI mortality rates in 2004. Approximately 60-65% of deaths registered in Thailand are coded to specific causes of death, while the remainder are assigned to ill-defined codes [16]. In this study, we used the total number of fatal RTIs estimated by the SPICE cause of death study [17]. The cause of death study was undertaken in a representative sample of 11,984 deaths from the 2005 vital registration data. For each death record, the cause of death was ascertained from two sources. The first source was a household interview with a close relative of the deceased using a standardized verbal autopsy (VA) tool. For 9,819 (82%) of the 11,984 deaths, a VA interview was completed, and a cause of death was assigned. The second source of causes of death was from a medical records review for those who died at a health facility where they had been treated prior to death. Of the 11,984 deaths, the interviews ascertained that 4,644 deaths took place in a hospital, of which 3,316 cases (71.2%) had adequate information in their medical records to be assigned a "certified cause of death" by medical experts. The redistribution of causes of death more than doubled the number for hospital deaths and almost doubled the number of home deaths assigned to RTIs. In both cases, the specificity of the vital registration system was high, but sensitivity was low. More detail on the estimates of mortality for Thailand in 2004-05 is available from previous publications [12, 17‐20].

An uncertainty analysis was done using Monte Carlo simulation with the Ersatz program (http://​www.​epigear.​com, Brisbane, Australia). The VA cause attribution was validated with the subsample of deaths for which both a medical record and a VA interview were available (2,558 of 3,316) [12, 17]. In this study, the corrected cause profiles were applied to total deaths using proportional mortality by age and sex. The sampling uncertainty of the redistributed causes of death was quantified using Monte Carlo simulation. Multinomial distributions were assumed for VA, medical record, and validation samples, and the distribution of causes of deaths was obtained by sampling simultaneously from the multinomial distributions. The misclassification matrix that governs the mapping from ill-defined to specific cause was assumed to have a Dirichlet distribution. However, other uncertainty, such as that around cause attribution in the medical records substudy, is not included.

YLL was calculated by multiplying the mortality estimates by the GBD standard life expectancy for a death at each age [21]. Lognormal distributions were assigned around the number of fatal RTIs.

To quantify the total number of nonfatal RTI victims, we used three national injury datasets. First, hospital data for 2004 were provided by the National Health Security Office (NHSO), which covers all hospitals under government authority. Second, the Bureau of Epidemiology, Ministry of Public Health, provided Injury Surveillance data (IS) from 1999 to 2004. Both datasets contain International Statistical Classification of Disease and Related Health Problems (ICD-10) codes for injury admissions (S00-T98 and V01-V98). Third, the Health and Welfare Survey data (HWS-2005), a survey of 19,914 households and 67,815 people, were obtained from the National Statistics Office (NSO). For our calculations, we used hospital data rather than IS because of its greater coverage. However, as the hospital admission database does not fully cover all hospitals in Thailand (largely missing private hospitals), we used the HWS-2005 survey to set an envelope of the total number of admissions based on the reported frequency of hospital admissions from all causes in the past year by type of hospital (community, provincial, tertiary or university, private, and other). Using inflation factors based on the HWS-2005 (Table 1), the results from the hospital admission database were extrapolated to the whole country in 2004. This resulted in estimates of i) the number of RTIs by age and sex; ii) causes and nature of injuries; and iii) the ratios between the numbers of deaths, admitted RTI victims, and nonadmitted RTI victims in IS and hospital data.

Estimating the number of YLDs required information on the duration of long-term disability, which equates with the life expectancy of the people with these disabilities. We distinguished two types of long-term disabilities: those that reduce life expectancy, and those that do not. The duration of long-term disabilities that we assumed do not carry a risk of death from complications, such as amputated limbs, was obtained directly from Thai life tables [17]. The duration of other long-term disabilities associated with an increased risk of death due to complications (including skull fracture, spinal cord injury, femur fracture, intracranial injury, burns, and injured nerves [27]) was calculated by multiplying mortality risks in the Thai life tables by relative risks of mortality as used in previous burden of disease studies [11, 22]. These assumptions are supported more by expert judgment than by empirical evidence, but the impact on the overall results is likely to be small, given that YLD makes up only a small part of the overall RTI-related burden. Furthermore, variations in the relative risk of dying will only have a modest impact on YLD, partly due to the fact that health gains and losses in future years are subject to discounting. The duration assumptions for short-term disabilities were adopted from the GBD study (Table 2) [23].

In 2004, road traffic crashes resulted in 567,000 victims in Thailand (Table 3). There were 24,800 (95% uncertainty interval (UI): 22,400-27,200) RTI deaths, equating to a crude RTI death rate of 40 per 100,000 people (95% UI: 36-44), 66 per 100,000 people in males (95% UI: 59-73) and 14 per 100,000 people in females (95% UI: 12-16). The age-adjusted death rate due to RTIs (using WHO's World Standard Population 2000-2025 [28]) was 39 per 100,000 people (64 and 14 for male and female, respectively). There were 126,000 hospital admissions and 417,000 emergency department visits for RTI.

Thailand lost 673,000 DALYs (95% UI: 546,000-881,000) due to RTIs in 2004 (Table 4). The majority (88%) of DALYs lost were due to premature mortality. The RTI burden mostly affected men (82%) and was higher in men than in women at all ages. Of the 77,800 DALYs due to disability, 95% were from the long-term consequences. Young adults were disproportionately affected by RTIs; 69% of DALY loss occurred in victims aged 15 to 44.

The crude rate of DALY loss was 10.8 per 1,000 people per year and 17.8 and 3.9 for males and females, respectively (Table 5). Standardized to the World Standard Population [28], these rates were 10.6, 17.2, and 4.1 per 1,000, respectively. The highest DALY rate was 21 per 1,000 at ages 15 to 29.

The use of Thai data on the average DW and the proportion of long-term disability for 2004 yielded almost double the amount of YLD in comparison to the estimate based on GBD assumptions. This difference in YLD was statistically significant but is small in comparison to the uncertainty estimated around YLL. Like mortality, YLD from long-term disability peaks at ages 15 to 29.

We used the best available RTI data for estimating the burden of RTIs in Thailand in an effort to avoid the limitations that often affect studies in low- and middle-income countries, such as underenumeration, limited reporting systems, and large uncertainties in the number of cases and deaths from specific causes of injury [5]. First, estimates of the number of fatal RTIs were obtained from the recent cause of death study [18, 19]. This study reduced the number of deaths attributed to ill-defined codes and those missed through underregistration, major limitations of fatal RTI data in Thailand and other developing countries [5, 16, 29]. The number of fatal RTIs from cause of death studies was 1.8 times the report by the Royal Thai police in 2004. Second, we used recent local data to estimate the proportions and severity of long-term disabilities. Third, we included nonfatal, nonadmitted RTI cases. Though it increased the overall burden of RTIs only slightly, their inclusion allows the estimation of treatment costs that can be avoided by reducing RTI rates and thus facilitates economic evaluation of interventions.

Although we used the best available data, these data were not without flaws. In the absence of a nationally representative dataset, we had to estimate the number of nonfatal RTI victims by combining elements of different datasets. Comparison of different data sources (the IS and the hospital data) shows that the evidence on the magnitude of the RTI problem is inconsistent. This is likely due to differences in the reporting system, coverage, and quality of data in each dataset [11, 30]. For example, hospital data in 2004 showed a decline in the total number of nonfatal RTIs since 1999. In 2004, 126,000 RTI victims were admitted to a hospital, down from 162,000 in 1999. In contrast, the Royal Thai police reported an increase in RTIs to about 14,000 fatal and 94,000 nonfatal cases in 2004, compared to 12,000 and 48,000, respectively, in 1999 [31], but this is likely due to a decrease in underreporting [32, 33] and still well below our estimates.

The total number of deaths due to RTIs in this study is similar to the study in 1999 (24,413 cases), but 1.75 (UI: 1.45-2.01) times as high as figures reported in documents from the Ministry of Public Health [11, 31]. These reports rely on mortality statistics in Thailand, which provide incorrect counts of cause of death, with large proportions attributed to "unknown causes" [16].

Premature mortality contributed the major part (88%) of DALY loss due to RTIs. This is a considerably higher proportion than estimated in Australia (72-73%), a Swiss canton (70%), Iran (62%), and Serbia (57%) [22, 34‐36]. An important reason for this higher proportion of fatal RTIs is the popularity of motorcycles, which are affordable for low-income families and young drivers [31, 37]. In 2002, 68% of all vehicles registered in Thailand were motorcycles. In 1999, about 35% of all vehicle crashes involved motorcycles, and this increased to 45% in 2005. As a result, 65% (in 1999) and 82% (in 2005) of victims of fatal RTIs were motorcyclists [25, 37]. The high fatality rates are partly because only 32% of motorcyclists wore helmets during the daytime and 9% at night, and only 30% of helmets comply with the Thai industrial standard [37, 38]. An alternative explanation is that our study underestimates the nonfatal burden. However, the agreement between the hospital data and the data from the IS system suggests that our estimates are fairly accurate and that the high contribution of mortality to the RTI burden may be a specific Thai phenomenon.

Males, particularly those aged 15-29, are an extreme high-risk group as they contributed 47% (319,000 DALYs) to the total RTI burden in Thailand. This is similar to previous reports in Thailand estimating the male burden at three to five times the female burden, and similar to the situation in European countries (Austria, Denmark, Netherlands, Norway, England, and Wales) and Iran [7, 30, 39]. Thirteen percent (83,000 DALYs) of the burden of RTIs occurred in children less than 15 years old.

Results from a Thai follow-up study of long-term disability from RTIs led to a significantly higher estimate of YLD compared to an estimate based on conventional GBD assumptions. We used an average DW regardless of the nature of injuries. This approach is simpler and less error-prone than calculations that use estimates of the proportions of RTI victims with specific injury diagnoses and the related risks and severity of permanent injury. This empirically derived weight of 0.57 was higher than any of the GBD disability weights except for spinal cord lesion[13]. This higher average DW was the main cause of the difference in the long-term YLD as the average proportion of long-term disability was similar to that used in GBD studies [40].

The ratio of deaths (24,000) to admissions (126,000) and nonadmissions (417,000) of 1:5:17 in this study is similar to the ratio found in Bangladesh (1:8:28) [41]. It differs from the World Health Organization's ratio of 1:35:70 and raises questions about the use of global average data for specific countries [1].

Our study reveals an apparent stagnation in the absolute size of the RTI burden between 1999 and 2004, despite a population increase of 3.9% over that period and increases in the number of vehicles from 6.3 million in 1999 to 16.6 million in 2002 and 26 million in 2007 [2, 31, 37]. This suggests that policy measures that have been implemented or intensified since 1999 may have prevented an increase in the burden of RTI. Such measures include general and sobriety checkpoints, raising public awareness of helmet and seatbelt use through mass media campaigns, speed limit enforcement, vehicle inspections, and traffic engineering, such as better road design [2, 31, 37]. It also suggests that greater efforts to reduce the burden of RTI in Thailand are urgently needed.