Incidence of road injuries in Sri Lanka
[Download detailed results tables: Sri Lanka-WebTables]
Last Updated: April 13 2011
Important: Please note that the results presented here are preliminary. Additional adjustments will be necessary as the results are made consistent with the ongoing Global Burden of Disease (GBD-2005) project, for which the results presented here are an input. Final estimates of the GBD-2005 study will be released in late 2010.
About this report
Reliable statistics of road injuries are an essential input for describing the public health burden of injuries, evaluating the impact of safety policies, and benchmarking achievements. While injury surveillance systems are common in high income countries, most low and middle income countries are unlikely to have such capacity for several decades. In the interim, estimates should be derived by harmonizing injury statistics from the wide array of data sources that may be available in a country or region.
This report summarizes our findings for the incidence of deaths and non-fatal injuries from road crashes in Sri Lanka. It is one of a series of national road injury assessments that we are producing during the course of this project. The intended audience of these reports includes the global donor community, the international research community, and national health and transport policy makers. We are committed to keeping this project open-source and collaborative in nature. All readers are encouraged to provide feedback to help improve methods, incorporate other sources of information, and suggest more effective methods for communicating these results.
In 2003, road crashes resulted in 1,951 deaths in Sri Lanka representing an annual injury rate of 12.1 deaths per 100,000 people. In addition, over three hundred thousand people were victims of non-fatal injuries due to road crashes. The road injury death rate in Sri Lanka was over twice the death rate in countries with the best road safety performance (Sweden, UK, and Netherlands).
Injuries as a whole, including from unintentional and intentional causes, resulted in 13,591 deaths or 11.8% of all deaths in Sri Lanka in 2003. Road injuries are the third leading cause of injury deaths in Sri Lanka, after deaths from suicide and collective violence (Table 1). In 2003, road injuries account for nearly approximately 14% of all injury deaths in Sri Lanka.
How did we compute these estimates?
Our general methodology for estimating deaths and non-fatal injuries involves piecing together data from a wide array of sources that typically include death registers, hospital records, funeral records, health surveys, and police reports. This requires filling information gaps, mapping from varying case definitions, deriving population based incidence estimates from sources that may not track denominator populations, and appropriately reapportioning cases assigned to poorly specified causes. For a general description of the broad methodology, please visit the Methods-overview section of our website.
The following sections describe the specific data sources used, the estimation methods, and the key results for our estimates of road injuries in Sri Lanka. The analytical adjustments to the data introduce uncertainty in the estimates. Thus, wherever possible, we have outlined the effects of the adjustments on the estimates.
Overview of data sources
We estimated the incidence and distribution of road injury deaths in Sri Lanka using national death registration data. We estimated the incidence and distribution of non-fatal injuries using the results of the 2002/2003 World Health Surveys.
Estimates of road injury deaths
Our review of data sources for estimating national road injury deaths in Sri Lanka revealed two potential data systems: national death registration data and national police data. In this study, we have computed estimates of national road injury deaths from the former (death registration) and compared them with police reported data.
We obtained death registration data from the publicly available WHO Mortality Database. These data are tabulations of deaths recorded by national civil registration systems. Typically these systems record age, sex, and causes of death coded using principles of the International Statistical Classification of Diseases (ICD). We only analyzed data for the most recent five years available (1999-2003), which had causes of death coded to ICD-10 categories. Data from prior years was not analyzed because it is not reported at a level of detail that allows us to use our analytical estimation methods. The most recent cause of death data available for our analysis was for the year 2003. Death registration data collection for the years 2004 and 2005 was disrupted by the 2004 Tsunami in Sri Lanka. Data for the year 2006 was not available for analysis at this time.
Reclassification to GBD-2005 definitions: We reclassified age into 38 age-sex groups. The age definitions match those used by the GBD-2005 project and are available on the GBD-Injury expert group website. Click here to go directly to the age definitions.
We reclassified the ICD coded deaths to the definitions of road injuries (and other injuries) as recommended by the GBD-Injury expert group [Click here for full details]. These definitions map all ICD codes for external causes of injury to 48 fully-specified cause categories and 21 partially-specified and undetermined cause categories. The fully-specified cause categories include nine road-user categories:
Pedestrian (V01-V04, V06, V09),
Two-wheeler rider (V20-V29),
Three-wheeler occupant (V30-V39),
Car occupant (V40-V49),
Van occupant (V50-V59),
Truck occupant (V60-V69),
Bus occupant (V70-V79), and,
Other road injury (V80, V82, V83, V84, V85).
In addition, there are two partially-specified sub-categories of road injuries:
Road injury - unspecified occupant, i.e. not pedestrian or bicyclist (V87-V88), and,
Road injury - unspecified road user (V89, Y85.0)
Table 2 shows the distribution of the 13,591 injury deaths reported in the death registration data for the year 2003. It should be noted that the 21 partially-specified categories have a hierarchical structure of specificity and many of these categories are not related with road injuries.
We assessed the quality of the death registration data based on the distribution of the number of deaths in the partially specified categories. As shown in Table 2, the quality of the death registration as measured by the fraction of death registration cases assigned to partially specified categories is reasonably good for estimating total road injury deaths in Sri Lanka. However, over three-fourth of cases assigned to road injuries do not have a road-user specified.
Reallocation of injury deaths coded to partially-specified causes:
The deaths classified to partially-specified and undetermined cause categories were redistributed over the fully-specified categories. All redistributions were done in proportion to the number of cases in the fully-specified causes within age-sex groups. The redistribution is done in 21 steps, one for each partially-specified category. This respects the information content in the hierarchical structure of the partially-specified categories. Thus, several of these steps do not affect the road injury estimates. For instance, the category of firearm: undetermined intent was redistributed over the categories for firearm: unintentional, firearm: self-inflicted, and firearm: inter-personal. This redistribution step has no effect on the road injury estimates. Only the redistribution of the following partially-specified categories has an effect on road injury estimates: Road injury - unspecified occupant, i.e. not pedestrian or bicyclist,
Road injury - unspecified road user,
Unspecified transport injury,
Unspecified unintentional injury,
Unspecified mechanism - undetermined intent,
Unspecified mechanism - unspecified intent,
The effect of these redistribution steps is evident in Figure 2, which shows the change in the estimated number of road deaths after each redistribution step. The death registration data has 1,781 deaths specified as road injuries to begin with. As expected from the distribution of deaths shown in Table 2, there is only a small increase in the estimates of road injury deaths during the redistribution of deaths coded to partially specified causes. This occurs when deaths coded to unintentional injuries- unspecified mechanism are reallocated. This results in an increase of about 9.5% in the road injury deaths.
Adjustments that have not applied yet: Two key adjustments that are likely to modify the road injury death counts have not been applied yet. First, we have not reallocated deaths coded to unspecified causes outside the ICD injury chapter. This reallocation has not been done yet because research into the causes of deaths coded to this category is currently ongoing. It should be noted that in Sri Lanka, almost a quarter (Table 2) of deaths are coded to this category, and the effect on road injury death counts could be substantial.
Second, we have made no adjustments to account for incomplete death registration because estimates of completeness of global death registration data are currently being developed. However, our preliminary comparison of total all-cause deaths reported in the death registration data analyzed by us with deaths reported by the UN Population Division suggests that death registration in Sri Lanka is near complete. It should be noted that both of these adjustments would increase the estimated death counts. Thus, the mortality results presented here likely underestimate the true number of road injury deaths.
Comparison of our estimates with other sources
Our estimate of road injury deaths in Sri Lanka are similar to those reported by official government statistics in the 2009 WHO Global Status Report on Road Safety and police reports (Figure 3). The official government statistics in the WHO report are based on police reports. Our analysis of road injury death rates in low- and middle income countries has usually found substantial under-reporting by police. However, this comparison with death registration data provides face-validty of the accuracy of road traffic death rates reported by police in Sri Lanka.
Road deaths by age and sex
Most road deaths in Sri Lanka occur among adult males (Figure 4a). Over 4/5th (83%) of all road injury deaths were men and over 4/5th (81%) of these men were adults older than 20 years. Death rates (Figure 4b) among men are five times those among women overall. Men have higher death rates for all age groups but the gender differential is smaller among the elderly. Both death counts and death rates rise dramatically in the transition in age from childhood to young adults (see age groups 10-14 years and 15-19 years). Death rates continue to rise with age and are highest among the elderly.
Such age and sex patterns in deaths and death rates are consistent with those seen in other countries. For the most part, the gender differentials in death rate are a result of higher exposure to road traffic among men in combination with higher risk-taking behavior. Similarly, the age pattern of death rates partly reflects patterns of exposure and partly case-fatality rates. While exposure to road traffic declines with age among older populations, the bio-mechanical tolerance to injury (i.e. the likelihood of death in the event of a crash) also declines, resulting in increasing death rates with age.
Road deaths by type of road-user
ICD-coded death registration data typically allows disaggregation of road traffic deaths into road-user categories. However, in the Sri Lankan death registration data, only a small fraction of (24%, Table 2) road traffic deaths have an identified road-user type. Thus using death registration data for estimating such disaggregation is likely to lead to biased results. However, our analysis of death registration data has shown that police data do not under-report road traffic deaths in Sri Lanka. Thus, Figure 5 presents the road-user distribution of road injury deaths rates based on police records. Most road deaths (36%) in Sri Lanka occur among pedestrians (Figure 5). Vehicle occupants comprise an additional third of road injury deaths.
Estimates of non-fatal road injuries
The incidence of non-fatal road injuries can be estimated from various sources, including police reports, hospital administrative records, and population surveys. Among these, police reports are widely known to underreport road injury cases in low income countries as well as in high income countries. Hospital records have the advantage of providing detailed medical descriptions of injuries making classification of injuries by severity possible. However, estimation of population rates is difficult without investing substantial efforts in identifying the hospital catchment population. Thus, population based health and injury surveys are the most reliable sources of information for incidence of road injuries, especially in information-poor settings.
We estimated the incidence of non-fatal road injuries in Sri Lanka using the 2002-2003 World Health Surveys (WHS), a nationally represented household survey that included questions on road injury involvement. These surveys, which were conducted by the World Health Organization, provide a unique opportunity for cross-country comparisons of non-fatal injuries because they asked the same set of questions in 53 countries, most of which are low- or middle- income countries. The surveys included the following set of road injury related questions in their household module:
In the past 12 months, have you been involved in a road traffic accident where you suffered from bodily injury?
When (in the last 12 months) did the accident happen?
Within the last 30 days/ 1-2 months ago/ 3-5 months ago/6-12 months ago/ Do not know
Did you receive any medical care or treatment for your injuries?
Where did you first receive care?
On-site, ambulance/Hospital/Outpatient facility/Private physician/ Traditional healer/ Other
Was it government operated or private?
Government operated/ Private
How soon after the traffic accident occurred did you first receive care?
In one hour or less/ In over one hour but within 24 hours/More than 24 hours later
In Sri Lanka, 6698 interviews were conducted as part of the WHS, with a 99% response rate. We imputed the response for 5.6% of records that had a missing response to the questions related with accidents. We replaced the missing values for this question with the predicted value from a logistic regression model fit to the cases with a non-missing response. The model used the response to this question as the dependent variable and the following independent variables: gender, age groups, place of residence, permanent income quintile, country, marital status, education, occupation, self-rated health, visual acuity (seeing and recognizing a person from across road), and alcohol consumption as predictor dummy variables. We used survey (svy) commands in Stata 10 for the analysis of the WHS.
Non-fatal injuries by age, sex, and residence
Our analysis of the 2002/2003 WHS suggests that over three hundred thousand people in Sri Lanka are involved in non-fatal road crashes annually. This corresponds to 1.6% [95% CI:1.1 -2.1%] of the population, which is lower (2.04%) for the combined population of the 50 WHS countries analyzed by us.
Although road injury incidence is significantly higher among residents of urban areas than residents of rural areas (Figure 6). Men have significantly higher rates of non-fatal road injury incidence than women (Figure 7), with a point estimate that is almost four times higher. Road injury incidence is higher among young adults (18-44 years) than the oldest group but the difference is not statistically significant. It should be noted that while the non-fatal injury rate among those older than 65 years is much lower than among young adults, the death rate among the elderly (Figure 4b) is the highest, pointing to high case fatality rates among the elderly.
These country reports were produced as part of a project funded by the World Bank Global Road Safety Facility. The results presented here are based on secondary data analysis of data collected by various national and international agencies. We are grateful to Mrs. Susantha Ranadheera and her team at the Vital Statistics Unit of the Registrar General’s Department of Sri Lanka for providing the detailed breakdown of cause of death data for Sri Lanka. We are also thankful to Dr. Eeshara Kottegoda and Prof. Amal Kumarage for providing access to relevant data for the analysis.