## Abstract

Background: Road traffic injuries (RTI) are a major cause of mortality and disability in the world. Only after significant losses have communities in developed nations taken necessary steps to prevent crashes and their consequences. Increase in road safety is related to increasing socio-economic development. We aim to study the trends in injury and death rates in a developing country, India, define sub-national variations, and analyse these trends in relation to economic and population growth. Methods: Public sector data from India were used to develop a standardized database on traffic injuries and indicator of economic development. The data were analysed using linear regression models to test the a priori hypothesis of a positive relationship between net domestic product (NDP), and injury and death rates from road crashes across states. Results: The absolute burden of RTI in India has been consistently rising over the past three decades. The reported rates are lower than those estimated by global health agencies and may reflect under-reporting. Population-based rates provide a better assessment of the public health burden of RTI than vehicle-based rates. There is an inverted U-shaped relationship between NDP and injury and death rates. Even with the limited data, Kuznets phenomenon is evident for within-country level comparisons. Conclusions: India and other developing countries could learn from the experience of highly motorized nations to avoid the expected rise in RTI and deaths with economic development, by currently investing in road safety and prevention measures.

## Introduction

Injuries are the third leading cause of mortality in the world, responsible for 5.8 million deaths annually.1 Road traffic injuries (RTI) are a global health problem and constitute a large majority of the deaths caused by all injuries. They are the ninth leading cause of Disability Adjusted Life Years (DALYs) lost worldwide and are projected to rise further by 2020.2 Low and middle income countries contribute 90% of the DALYs lost and 85% of the deaths from road crashes.

Asia is one of the regions where fatalities from road injuries have been rising steadily over the past decade.3 The magnitude of RTI is projected to increase by two and a half times in the next two decades in the South Asian region.4 India is the largest country in the South Asian region and has one of the highest rates of RTI per 1000 vehicles in the world.5 It has an annual Gross National Income per capita of US$480, with an overall health profile illustrated by an infant mortality rate of 67 per 1000 live births, child mortality rate of 93 per 1000 live births, and maternal mortality rate of 540 per 100 000 live births.6 A World Bank study has shown that the economic development of regions and nations is associated with an increase in the number of injuries and deaths from road traffic crashes.4 RTI, in international comparisons, like other environmental factors described by Kuznets curves,7 follow an inverted U-shaped pattern in relation to economic development.4 The goal of this study is to explore the relationship between RTI and fatalities and development at a national level in a developing country. The specific objectives are to analyse public sector data in India and study the trends in traffic injury and death rates over time, to identify the most appropriate indicator for assessing the public health impact of RTI, and to analyse the relationship between economic development and traffic-related mortality in India, in order to visualize whether it follows the predicted inverted U-shaped relationship. ## Methods Data sources for this study include the Ministry of Road Transportation and Highways, and the Census Bureau of India,8,9 both public sector sources, and state-specific economic data from India Infoline®.10 Data were extracted using Microsoft Excel® and used to generate RTI-specific injury and fatality rates. The population data from the Census Bureau were used without modifications. Vehicle counts were available at 5-year intervals from 1966 to 1996 and were reported for 12 month periods (1 April–31 March); while injury data were reported for each calendar year (January–December). Data from the years 1971–2001 were used for national level comparisons. If data were missing for a specific year, they were computed using the previous year's data. To remove annual variability in calculation of rates, data over 3 years (1999–2001) were added and averaged to generate annual rates for individual states and union territories (Delhi included as union territory). Union territories are regions under direct national government control and have no state legislature. These regions have higher economic development and literacy level as compared to the states by virtue of being directly financed by the central government and having a large population of government employed officials. Data from the recently created states (Jharkhand, Chattisgarh, and Uttaranchal) were analysed with the parent states for this study (Bihar, Madhya Pradesh, and Uttar Pradesh, respectively). The rates for both injuries and deaths were explored to examine the relationship between two denominators-rates per 100 000 people and per 1000 vehicles. Per capita net domestic product (NDP) has been used as the indicator of economic development for the financial year 2000–2001. This represents an adjusted Gross Domestic Product (GDP), defined as GDP minus depreciation.11 NDP has been converted to US$—43.50 Indian rupees for every US$. A directed acyclic graph (DAG) was used as a guiding pathway, reflecting the relation between primary variables of interest (effect of NDP on injury and death rates) and potential confounders (population and number of vehicles).12 This serves the purpose of defining the statistical analysis a priori, dependent upon the presumed pathway of causation and hence prevents over-adjusting for variables which are in the path of causation and not confounders. We identified population as a potential confounder and adjusted for the same, but number of vehicles was assumed to be an indirect determinant which is affected by NDP and population, and hence did not require adjustment after population was taken into account. The resulting data were entered into STATA® version 8.0 (College Station, TX) for further analysis. The variables for population and number of vehicles were log transformed to decrease variability, and simple and multiple linear regression models were used to analyse the data. Student's t-tests were used to analyse regional differences in mean values. To analyse the effect of NDP on injury and death rates, population-based rates were taken as outcome measures since they better represent the public health measure of RTI. Twenty-three states and two union territories were included in these analyses, for which NDP data were available. Non-parametric regression analysis and residual plots were analysed to assess non-linearity. The variability in the regression analysis was assessed using studentized residual and leverage plots. Outlying values were recognized as those with studentized residuals >2.5 units. A Spline model was fit to explain non-linearity, where applicable. Sensitivity analyses were performed using bootstrap regression technique, to account for non-normality in data, and re-analysis was done after removing outliers. We also performed a sensitivity analysis to explore the effect of adding number of vehicles (a covariate) to our final model. The P-value for statistical significance was taken as 0.05. ## Results There were 394 800 road crashes reported in India in 2001 that resulted in more than 382 000 injuries and 80 000 deaths on Indian roads. This amounted to an increase of more than 274 000 crashes (>200% increase), 65 000 deaths, and 310 000 injuries (>400%) from the year 1971 (figure 1). The numbers of road crashes and the injuries and deaths associated with them have been increasing by more than a constant proportion over the time period. There has been a 29-fold increase in the number of vehicles in the past three decades when compared with <2-fold (85%) increase in the overall national population. Figure 1 Road traffic crashes, morbidity, and mortality in India, 1971–2001 Figure 1 Road traffic crashes, morbidity, and mortality in India, 1971–2001 These events translate to rates of 1.45 deaths and 6.96 injuries per 1000 vehicles, or 7.90 deaths and more than 37 injuries per 100 000 people in 2001 (figure 2). This amounts to approximately one injury for every crash and one death for every five crashes with no adjustments. Over the past 30 years, the population-based rates of crashes, injuries, and deaths have been increasing, whereas vehicle-based rates appear to be decreasing. Figure 2 Rates of injury and death per 1000 vehicles and 100 000 people in India, 1971–2001 Figure 2 Rates of injury and death per 1000 vehicles and 100 000 people in India, 1971–2001 The state-wise (including union territories) distribution of injuries and death rates from road crashes over years 1999–2001 is highly variable between regions. The injury and death rates per 1000 vehicles show a marginal decrease over the 3-year period with no uniform trends among different states, while the population-based rates have remained almost constant. There is no correlation (R2 = 0.4 for injury rate, R2 = 0.001 for death rate) between the injury and death rates calculated with a population-based denominator and the vehicle denominator for different states and union territories. The distribution of rates, injury to death ratios, vehicles per person and per capita NDP varies between states and union territories (table 1). The states have statistically significantly higher vehicle-based rates than union territories, but the population-based rates have an inverse trend, although not significant. The mean number of vehicles per person in union territories is 4.3 times higher than the states. There was a statistically significant (P < 0.001) relationship between motorization level (number of vehicles per 100 people) and NDP, with the index increasing by 3.8 (95% CI = 2.93, 4.67) for every$100 increase in NDP.

Table 1

Summary statistics for RTI for states and union territories in India, 1999–2001 a

Annual rates

India

States (n = 25)

Union territories (n = 7)

P-value

Mean injury rate per 1000 vehicles (SD) 9.84 (9.44) 11.37 (10.03) 4.36 (3.61) 0.007
Mean injury rate per 100 000 people (SD) 56.2 (38.6) 50.4 (36.5) 77.1 (41.2) 0.106
Mean death rate per 1000 vehicles (SD) 1.92 (1.27) 2.20 (1.21) 0.83 (0.90) 0.009
Mean death rate per 100 000 people (SD) 11.3 (6.05) 9.98 (4.4) 15.8 (9.0) 0.145
Mean injury to death ratio (range) 4.8 (0–17.5) 4.87 (1.3–17.5) 4.7 (0–10.7) 0.909
Mean vehicles per 100 people (range) 11 (2–60) 6 (2–27) 26 (7–60) 0.035
Mean per capita NDP b (US$) (SD) 418.4 (2.13) 384 (184) 807 (122) 0.005 Annual rates India States (n = 25) Union territories (n = 7) P-value Mean injury rate per 1000 vehicles (SD) 9.84 (9.44) 11.37 (10.03) 4.36 (3.61) 0.007 Mean injury rate per 100 000 people (SD) 56.2 (38.6) 50.4 (36.5) 77.1 (41.2) 0.106 Mean death rate per 1000 vehicles (SD) 1.92 (1.27) 2.20 (1.21) 0.83 (0.90) 0.009 Mean death rate per 100 000 people (SD) 11.3 (6.05) 9.98 (4.4) 15.8 (9.0) 0.145 Mean injury to death ratio (range) 4.8 (0–17.5) 4.87 (1.3–17.5) 4.7 (0–10.7) 0.909 Mean vehicles per 100 people (range) 11 (2–60) 6 (2–27) 26 (7–60) 0.035 Mean per capita NDP b (US$) (SD) 418.4 (2.13) 384 (184) 807 (122) 0.005

a: Mean rates averaged over 3 years 1999–2001

b: Data only for 23 states and 2 union territories for the year 2000–2001

Injury and death rates are dissimilar among states and union territories and the trends generally do not conform to trends seen in the population growth rate or number of vehicles. The state of Kerala has the highest reported injury to death ratio in the country (17.42), as well as a higher number of injuries relative to other states with a similar population size and number of vehicles; while the state of Uttar Pradesh has a lower number of injuries than expected, based on the population and number of vehicles (figure 3A and B). The two states of Bihar and West Bengal both have lower numbers of deaths and injuries than similar regions.

Figure 3

Injury and death rates (A) per 100 000 people (B) per 1000 people for 32 states and union territories in India, 1999–2001

Figure 3

Injury and death rates (A) per 100 000 people (B) per 1000 people for 32 states and union territories in India, 1999–2001

The overall population-adjusted injury rate increased significantly by 11 per 100 000 people (95% CI = 3.7, 17.8) for every $100 increase in per capita NDP. The population-adjusted death rate also shows a similar significant increase of 2/100 000 people (95% CI = 1.4, 2.6) for every$100 increase in per capita NDP. There was no statistically significant relationship between total population and injury or death rates. Using the Spline model for NDP, the mean change in the injury and death rates for every $100 increase in NDP showed a significant rate increase up to a NDP of$400, a non-significant decrease when NDP ranged between $400 and$600, a significant (for injury rate) increase with NDP between $600 and$750 and a non-significant decrease when NDP rose to more than $750 (figure 4). The injury rate differed significantly between the four strata of NDP but the death rate for NDP less than$400 was not significantly different from the other strata (table 2).

Figure 4

(A) Death rate (1999–2001) by per capita NDP and (B) injury rate (1999–2001) by per capita NDP

Figure 4

(A) Death rate (1999–2001) by per capita NDP and (B) injury rate (1999–2001) by per capita NDP

Table 2

Change in 3 year (1999–2001) average injury and death rate per 100 000 people as a function of per capita NDP a for 25 states and union territories of India

NDP range (US$) Unadjusted Adjusted for population Sensitivity analysis Boot strap regressionb Adjusting for number of vehicles also After removing outliersc Injury rate <400 29.8 (10.2, 49.5) 30.7 (9.8, 51.6) 30.7 (10.2, 57.6) 28.7 (5.5, 51.8) 23.9 (7.8, 39.9) 400–600 −14.6 (−40.4, 14.9) −15.9 (−47.4, 15.5) −15.9 (−45, 130.9) −20.2 (−57.3, 16.9) −12.3 (−36, 11.3) 600–750 67.9 (10.2, 125.7) 71.4 (7.9, 135) 71.4 (0, 144) 63.3 (−10.7, 137.3) 71.6 (−24.1, 119) >750 −20.9 (−55.0, 13.5) −21.3 (−56.8, 14.1) −21.3 (−86.8, 21.5) −20.3 (−56.9, 16.3) −21.2 (−47.8, 14.4) Death rate <400 2.9 (1.2, 4.7) 3 (1.1, 4.9) 3 (1.1, 4.7) 2.1 (0.4, 3.8) 3.7 (2, 5.4) 400–600 1.8 (−0.9, 4.4) 1.7 (−1.2, 4.5) 1.7 (−1.2, 5.9) −0.2 (−2.9, 2.6) 0.7 (−1.9, 3.2) 600–750 4.2 (−0.9, 9.4) 4.4 (−1.3, 10.1) 4.4 (0, 8.7) 0.8 (−4.7, 6.4) 5.6 (−0.6, 10.6) >750 −0.7 (−5.7, 5.3) −0.8 (−3.9, 2.4) −0.8 (−2.7, 0.5) −0.3 (−3.0, 2.5) −0.9 (−3.6, 1.9) NDP range (US$)

Sensitivity analysis

Boot strap regressionb

Adjusting for number of vehicles also

After removing outliersc

Injury rate
<400 29.8 (10.2, 49.5) 30.7 (9.8, 51.6) 30.7 (10.2, 57.6) 28.7 (5.5, 51.8) 23.9 (7.8, 39.9)
400–600 −14.6 (−40.4, 14.9) −15.9 (−47.4, 15.5) −15.9 (−45, 130.9) −20.2 (−57.3, 16.9) −12.3 (−36, 11.3)
600–750 67.9 (10.2, 125.7) 71.4 (7.9, 135) 71.4 (0, 144) 63.3 (−10.7, 137.3) 71.6 (−24.1, 119)
>750 −20.9 (−55.0, 13.5) −21.3 (−56.8, 14.1) −21.3 (−86.8, 21.5) −20.3 (−56.9, 16.3) −21.2 (−47.8, 14.4)
Death rate
<400 2.9 (1.2, 4.7) 3 (1.1, 4.9) 3 (1.1, 4.7) 2.1 (0.4, 3.8) 3.7 (2, 5.4)
400–600 1.8 (−0.9, 4.4) 1.7 (−1.2, 4.5) 1.7 (−1.2, 5.9) −0.2 (−2.9, 2.6) 0.7 (−1.9, 3.2)
600–750 4.2 (−0.9, 9.4) 4.4 (−1.3, 10.1) 4.4 (0, 8.7) 0.8 (−4.7, 6.4) 5.6 (−0.6, 10.6)
>750 −0.7 (−5.7, 5.3) −0.8 (−3.9, 2.4) −0.8 (−2.7, 0.5) −0.3 (−3.0, 2.5) −0.9 (−3.6, 1.9)

95% Confidence intervals are shown in parenthesis

a: Per 100 US$increase b: P-value for swilk test for residuals = 0.02 for injury rates and 0.31 for death rates c: State of Kerala (rstudent = 4.02) for injury rates and West Bengal (rstudent = 2.8) for death rates Using the sensitivity analysis techniques described above, we found that population did not significantly affect the variable coefficients. The increase in death rate with NDP$600–750 became statistically significant using bootstrap regression. The addition of the number of vehicles covariate to the model modified the coefficients by a large number in most cases (table 2). The co-linearity between the population and vehicle variables was very high (VIF for vehicles = 26.6 and population = 29.5).

## Discussion

The public health importance for countries to assess and evaluate their data on the magnitude and impact of RTI cannot be overstated. The data presented here have been gathered by the Government of India through various state governments, which in turn report incidents in police records over the specified period. However, police records are neither nationally standardized nor corrected for under-reporting. Generally, police reporting levels are lower in developing countries relative to high income countries.13

Over the past three decades, commensurate with the development of the nation, the burden of injuries and deaths due to road traffic crashes has been steadily increasing in India. Even 20 years ago in rural India, traffic injuries were the cause of about 20% of medico-legal deaths.14 While our study reveals a 4-fold ratio of injuries to deaths, it is often estimated that the actual number of injuries is about 20 times that of deaths.2,15 This discrepancy is attributed to greater under-reporting of less severe injuries relative to deaths at state and national levels.15,16

There has been a steady decline in vehicle-based death and injury rates in India, but a growing trend in the population-based rates of injuries and deaths. This seemingly divergent trend can be explained since vehicle-based rates are decreasing due to a disproportionate influx of vehicles on Indian roads, while population-based rates are increasing because of an absolute increase in the number of crashes relative to the rate of population growth. The observed death rate per 100 000 people is lower than the WHO estimated rates for low–middle income countries (20.2 per 100 000 people) and high income countries (12.6 per 100 000),2 which might be another indication of the potential under-reporting in police data, although the WHO is not country specific. The erratic decrease in the reported rates every 5 years (till 1995) for vehicles and every 10 years for population-based rates are due to the updating of information at those intervals.

The lack of correlation between the injury and death rates calculated per 100 000 people and per 1000 vehicles, two commonly used indicators for estimating or monitoring the burden of injuries, highlights the importance of using indicators appropriately. The vehicle denominator is sensitive to the number of registered vehicles and is limited by the fact that it does not take into account non-motorized transport. The actual public health impact of road crashes is better represented by population-based rates and can be used to estimate the severity of crashes and compare the impact of RTI with different conditions.2

The discrepancy in both injury and death rates between union territories and states in India can be attributed to higher vehicular penetration and lower population density in the union territories. It can also be ascribed to the greater economic development in these regions, as indicated by the difference in mean per capita NDP. Some states, e.g. Bihar, report significantly lower rates than the national average. These regions could have fewer crashes or, more likely, they could have a higher level of under-reporting. Under-reporting being the cause becomes more obvious when the differences are compared across indicators. Some states, like Uttar Pradesh seem to selectively under-report injuries relative to deaths, as is evident by the much lower injury rate as compared to death rate for the state in context of regions with similar population and vehicles. The high injury to fatality ratio noted in Kerala substantiates this fact. Kerala is India's most literate state with the lowest population growth, and probably has better level of government functioning, data collection, and reporting, and hence closely matches the estimated national injury to death ratios. Deaths in India are registered under a vital registration system, which makes reporting mandatory, whereas no such surveillance system exists for injuries; therefore, deaths are more likely to be reported in government data relative to injuries.

The relationship between population-based injury and death rate and the per capita NDP of the states and union territories reflects the association of development with RTI. Higher economic development, through a variety of factors such as a greater number of vehicles, results in a higher number of crashes. The rise in rates is then followed by a plateau and decline, most often related to increased investment in road safety measures, regulations, and public transport. This trend is similar to the Kuznets phenomenon, originally defined for other environmental factors, showing that rising economic influence in a society is associated with increase in environmental health hazards, until a certain time (high economic growth) after which there is a downward trend in the hazard with further increase in affluence, producing a U-shaped curve.7 The relation between motorization level and economic development observed in our analyses is similar to the global analysis and is a major factor in the rising part of the inverted U-shaped Kuznets curve.4 This analysis recognizes the confounding effect of number of vehicles which was expected from our prior DAG since the number of vehicles is highly correlated with the economic development of a community.

The use of public sector data may be considered a limitation of this study. This data only enumerates reported incidents; police records are particularly associated with under-reporting, and vital registration data with misreporting. The types of traffic crashes reported in public data are not known and were assumed to include injuries or deaths from all mechanisms. However, we believe that the analysis of such data is critical both for defining a set of ‘minimalist’ estimates and for documenting gaps in information. All the states and union territories did not have data on NDP and this limited our power of analysis. NDP also may not be the best surrogate indicator for development but data limitations restricted use of other possible indicators. This paper does not analyse specific contributory factors to the described relationship between economic development and RTI, e.g. types of vehicles or traffic mix on roads. Also, the indicators injury and mortality indicators used in this study may not be useful in international comparisons, where the denominator of distance traveled is commonly used to good effect. In order to address the annual variability in the reported data, this paper has used 3-year average values for all analyses. Subgroup (states and union territories) and sensitivity analyses have also been done in an effort to understand the effect of some of these limitations.

## Conclusion

The cross-sectional association between economic development (as measured by NDP) and road traffic injury rates for states and union territories in India correlates well with cross-national studies and reflects the potential for state investments in road safety in addition to any overall national efforts. Moreover, this study also highlights the diversity in the health impact of road transport within a developing country.

Kopits and Cropper4 projected that death rates from road crashes would not decline until 2042 in India. Such analysis appropriately reflects the historical trend that investments for road safety increase with national income. However, developing nations such as India do not need to wait for rising fatality rates before adopting road safety measures. The more economically developed states should consider immediate investment in road safety measures to halt the rising mortality from road crashes. Measures such as traffic law formulation, implementation, and traffic segregation, as suggested by WHO,2 should be undertaken. States in earlier stages of development need to recognize the potential loss of health that will occur, and should phase in interventions to prevent traffic injuries. At the same time, studies to understand the state-specific determinants to guide effective measures should be conducted. It should be an imperative of the Indian government to foresee the future and invest in road safety measures now to curb the human and economic losses occurring daily from RTI.

Key points

• This study was conducted with an aim to study the trends of RTI and their relation to economic development using public sector data in a developing country.

• Morbidity and mortality from RTI have been rising over the past three decades in India, although it is not always reflected in commonly used health indicators.

• Mortality from RTI varies widely between regions within India and suggests a rising burden of disease.

• Fatality from RTI demonstrates an inverse U-shaped relationship with economic growth; one of the first such reports from a developing country.

• Countries like India can take concrete steps now to curb the rising loss of life from RTI well before the natural point of decline expected with increasing economic development in the future.

At the time of the study Nitin Garg was a research assistant in the Department of International Health, Johns Hopkins Bloomberg School of Public Health Baltimore, MD, USA.

## References

1
WHO. Mental Health: new understanding, new hope: World Health Report 2001. Geneva: World Health Organization,
2001
.
2
Peden M, Scurfield R, Sleet D, et al. World report on road traffic injury prevention: Geneva: WHO,
2004
.
3
Jacobs G, Aeron-Thomas A, Astrop A. Estimating global road fatalities: Crowthorne: Transport Research Laboratory,
2000
.
4
Kopits E, Cropper M Traffic fatalities and economic growth.
Accid Anal Prev

2005
;
37
:
169
–78.
5
Mohan D Accidental death and disability in India; a stocktaking.
Accid Anal Prev

1984
;
16
:
279
–88.
6
UNICEF. The Official Summary of The State of the World's Children 2004. New York: UNICEF, 2004. Available at: http://unicef.org/infobycountry/india.html (23 July 2004, date accessed).
7
Grossman G, Krueger A. Economic growth and the environment.
Q J Econ

1995
;
110
:
675
.
8
Road safety cell. Motor Transport Statistics. Ministry of Road Transport and Highway, Government of India. Available at: http://morth.nic.in/mts.htm (18 December 2003, date accessed).
9
Census Bureau of India. Population Census Data. Ministry of Home Affairs, Government of India. Available at: http://www.censusindia.net (8 October 2003, date accessed).
10
Per Capita Net State Domestic Product at current prices 1985–86 to 2001–02 (Rs). India Infoline Ltd. Available at: http://www.indiainfoline.com/econ/andb/nia/nia10.html (18 December 2003, date accessed).
11
Web Finance Inc. Net Domestic Product. Web finance definitions. Available at: http://www.investorwords.com/3243/Net_Domestic_Product.html (15 April 2004, date accessed).
12
Merchant AT, Pitiphat W. Directed acyclic graphs (DAGs): an aid to assess confounding in dental research.
Community Dent Oral Epidemiol

2002
;
30
399
.
13
Odero W, Garner P, Zwi A. Road traffic injuries in developing countries: a comprehensive review of epidemiological studies.
Trop Med Int Health

1997
;
2
:
445
–60.
14
Ramachandran V. A study of medico-legal deaths in a rural area of Pondicherry (South India).
Indian J Public Health

1982
;
26
:
229
–33.
15
Maheshwari J, Mohan D. Road traffic injuries in Delhi: A hospital based study.
J Traffic Med

1989
;
17
:
23
–7.
16
Sathiyasekran B.W.C. Study of injured and injury pattern in road traffic accident.
Indian J Forensic Sci

1991
;
5
:
63
–8.