Advances in Transportation Studies: an international Journal 1 2 3 4 Modeling non-fatal road crash injuries for Pakistan using aggregate data 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 F. Subhana, H. Zhoub, S. Zhaoc, M. M. Naeemd, M. Sulaimane a Faculty of Management & Economics, Dalian University of Technology, Dalian, China. fsubhan@mail.dlut.edu.cn b School of Transportation and Logistics, Dalian University of Technology, Dalian, China. hzhou@dlut.edu.cn c School of Transportation and Logistics, Dalian University of Technology, Dalian, China. szhao@dlut.edu.cn d Department of Civil Engineering, Iqra National University, Peshawar, Pakistan. majid@inu.edu.cn e Department of Civil Engineering, University of Engineering and Technology, Peshawar, Pakistan. Muhammadsulaiman909@gmail.com Abstract Road crash injuries have emerged as a major health problem and have posed serious social and economic challenges around the globe. Traffic safety can be improved only through adequate safety measures if the underlying factors are well understood. Factors affecting crashes in developing countries, like Pakistan, are less studied in the literature. Also, the road crash injuries data reporting and recording systems in these countries are not well established. As such, this study by using the data from multiple sources including World Health Organization (WHO), International Road Federation (IRF) and World Bank (WB) compares the reported nonfatal road crash injuries of Pakistan using two different approaches. First, the road crash injury rates were compared with different groups of countries around the globe using two different indicators: (1) injuries per hundred thousand population and (2) injuries per thousand registered vehicles. Results indicated lower road crash injury rates Pakistan relative to other countries. Using the same indicators as response variables, linear regression models were estimated using Ordinary Least Square (OLS) regression. The total number of registered vehicles, maximum speed on rural roads, enforcement level of the seatbelt law, income level, and safety audit of new roads were found as significant explanatory variables. The average values of these variables were compared with those in the country. Finally, using the number of injuries as the response variables, count data models were developed and the number of road crash injuries for Pakistan were estimated and compared with the reported number of road crash injuries. The estimated road crash injuries were 4.5 times higher than the number of reported injuries. Keywords – Non-fatal Road Crash Injuries, Injury Rate, OLS Models, Count Data Models 40 -1- Advances in Transportation Studies: an international Journal 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 1. Introduction Annually, approximately 1.35 million fatalities and 20 to 50 million non-fatal injuries worldwide occur due to road traffic crashes, which pose a serious social and economic challenges globally [1, 2]. It is anticipated that the numbers of road crashes will continue to increase in the next decade or so. If such a situation is not properly addressed, road crash injuries will be the seventh leading cause of mortality by the year 2030. Road crash injuries are also the primary and the 8th leading cause of mortality for youngsters aged between 15 to 29 years and for people of all ages, respectively [1, 2]. Also, the deaths resulting from road crash injuries constitute 2.5% of the total mortality due to all causes around the globe [1]. Therefore, the implementation of effective countermeasures to improve road traffic safety around the globe is extremely necessary and urgent. Special concerns should be given to low and middle-income countries. The rate of mortality due to road crash injuries in lowincome countries is three times higher than in high income countries [1]. Figure 1, based on the data from World Health Organization’s Global Status Report on Road Safety-2018 [1], shows the proportion of population, registered vehicles, and road crash fatalities in high, middle, and lowincome countries. The figure depicts that both the population and number of registered vehicles are less in low-income countries compared to high-income countries, but, the rate of road crash fatalities is higher in low than in high-income countries. The ratio of the number of registered vehicles in low and middle-income countries is around 0.016, but the ratio of road crash fatalities between these countries is 0.163. Also, the ratio of road crash fatalities between low and high-income countries (1.857) is higher than their ratio of the number of registered vehicles (0.025). Similarly, the ratio of the number of registered vehicles between middle and high-income countries (1.475) is lower than the ratio of the number of road crash fatalities (11.428). Also, there has been no reduction in the number of road traffic deaths in any low-income country since 2013 [1]. It is roughly estimated that the lowincome countries experience a loss of 1% of their gross national product (GNP) owing to road crash injuries (both fatal and non-fatal), while in middle and high-income countries, the costs are 1.5% and 2% of the GNP, respectively. The possible reason for these low figures in low and middle-income countries can be the under-reported data provided by these countries [3]. -2- Advances in Transportation Studies: an international Journal Population % Vehicles % 1 9 15 40 59 76 69 Road Crash Fatalities % 13 7 Middleincome Low-income 80 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 High-income Figure 1. Proportion of population, road traffic deaths, and registered motor vehicles by country income category for the year 2016 [1] As a developing country, Pakistan has the sixth largest population of around 207.774 million in the world [4-6]. In the recent past years, the country’s economy has grown at a very slow pace and road infrastructure has been moderately improved. The total road length has increased from 2, 51, 661 km in the year 2002 to 2, 64, 212 km in year 2015 [5]. However, the motorized vehicle population has substantially grown from 5.3 million in the year 2002 to 11 million in year 2012 comprised of all vehicle types [5, 6]. Motorcycles, passenger cars, buses and trucks have been approximately increased by 110%, 150%, 30%, and 45%, respectively [6, 7]. The mushroom growth of motorization and the increase of vulnerable road users are the reasons for the higher number of road traffic crashes (RTC) in Pakistan. Figure 2 shows the change in gross national income (GNI), total road length, and motorization in the country with time. The GNI [8] and road infrastructure [9] show slow and moderate increasing trends, respectively. The number of registered vehicles shows a much higher increasing trend compared to road infrastructure and GNI up to the year 2009 [6, 7, 9], but a substantial increase afterwards [4]. The figure shows that number of road crashes and respective outcomes do not show a clear pattern of increase with the increase in GNI, road infrastructure, and registered vehicles. -3- Advances in Transportation Studies: an international Journal 20000 VKT (100 Millions) 18000 16000 GNI (US$) 14000 Road Crashes 12000 10000 Fatal Injuries 8000 6000 Non-fatal Injuries 4000 2000 Vehicles (1000) 0 Road Length (100 Km) 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 Figure 2. Increase in Motorization, Road Length, GNI, VKT, and Road Crashes with Time [5, 8-10] Road traffic crashes and injuries cause devastating effects on the country’s economy, resulting in a loss of approximately 6.48 million USD annually (1 USD = 154.4 PKR) [10]. In Pakistan, road traffic crashes are the second, fifth and eleventh leading causes for disability, overall healthy-life-year losses and premature fatality, respectively [11, 12]. Hyder et al. found that annually only 37 individuals per 10,000 registered vehicles face road crash injuries (RCIs) [13]. Ghaffar et al. found that each year approximately 1,500 RCIs occur per 100,000 populations [14]. Similarly, a study concluded that around 1700 out of 100,000 individuals experience RCIs annually in the country [15]. Furthermore, Shamim et al. found that the annual RCI rate is 184.3 per 100,000 populations [16]. While recent data suggested that the total number of RCI could be four to ten times higher than the officially reported figures [17]. The country’s policy-makers need immediate attention as these RCIs among all other injury categories result in more handicaps [15]. Besides, the country is facing an issue of crash data reporting as estimates of road crash injuries are not consistent according to the previous studies. For example, the National Road Safety Secretariat (NRSS) reported 2 million RCI, while the Social Indicator of Pakistan reported only 11,415 RCI for the year 2006 [5, 18]. The Global Status Report on Road Safety-2009 by World Health Organization (WHO) reported 12,990 road crash injuries in Pakistan for the year 2007 [17]. Similarly, Subhan et al. using aggregate data of 73 countries with vital registration records found that 655,171 individuals sustained RCI in the year 2016 [19]. While in -4- Advances in Transportation Studies: an international Journal 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 the same year, the Social Indicator of Pakistan reported only 11,544 RCI based on police records [5]. Figure 2 shows that the road crashes and injuries (both fatal and non-fatal) have decreased with the increase in motorization and road length which seems contradictory to the previous research findings. The inconsistency of reported and estimated number of injuries by previous studies and different national and international organizations is an indication of the limitation of road crashes and injuries data reporting systems in the country. Past studies have revealed that the system of road crash data collection of Pakistan based on police-reporting has serious limitations as police only collects the data which meet their legal requirements [13]. Prior to the application of any road safety strategies or measures, the authenticity of the reported data and the causes behind the road crashes and their outcomes must be thoroughly investigated. Previous studies considered a few contributing factors of the crashes and estimated the crashes nationally or locally. This paper aims to compare the road crash injury rates of Pakistan with different groups of countries and with the overall countries around the globe, and statistically explore the factors that influence the occurrence of road crash injuries. The reported road crash injury rates of Pakistan were compared with other countries around the world. The Ordinary Least Square (OLS) regression was used to explore the relationship between non-fatal RCI and explanatory variables. The values of the variables found significant in the models were averaged for 73 countries and compared with those in Pakistan. The number of non-fatal RCI for Pakistan were estimated using count data models (Poisson and Negative-binomial models) and compared with those reported by World Health Organization’s Global Stats Report on Road Safety -2007. 132 2. Literature review 133 2.1 Modeling and analysis methodologies 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 Many studies in the past used count data models for determining road crash frequencies and the corresponding factors. El-Basyouny and Sayed [20] used the a single count data model (poissonlognormal) to explore the factors responsible for road traffic crashes in different corridors. Caliendo, Guida and Parisi [21] estimated different count data models (poisson, negative binomial, and negative multinomial) for accident occurrence on multilane roads. Similarly, Anastosopolous and Mannering [22] used random-parameters count models (random-parameters negative binomial regression) to determine the road crash frequencies. To determine the severity road crash injuries, some studies used probabilistic aggregate and disaggregate models in the past. Savolainen and Mannering [23] used probabilistic models (nested logit and standard multinomial logit models) and Moore et al. [24] used mixed logit models to determine the severity road crash injuries among motorcyclists. Similarly, Chen and Chen [25] used mixed logit models to examine the difference between single-vehicle and multi-vehicle crashes. In addition, Abay [26] used disaggregate models (standard fixed-parameters ordered logit, the random parameters ordered logit, multinomial and mixed logit) to examine the factors associated with road crash injury severity. Some studies used non-linear regression for the purpose of estimating the number road traffic crashes and their respective outcomes. As such, Dogan and ANgüngör [27] applied non-linear -5- Advances in Transportation Studies: an international Journal 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 regression and artificial neural network (ANN) models to estimate the number of road traffic crashes, fatalities, and injuries. Many studies used the accident and injury rates as continuous variables. As such, Anastasopoulos, Tarko, and Mannering [28] used Tobit analysis by using accident rates as a continuous variable (number of accidents per 100 million vehicle miles travelled) to study accident rates on Indiana interstate highways. Some studies used linear regression analysis for the estimation and prediction of the outcomes of road traffic crashes and to examine their association with the explanatory factors. For example, Zargar et al. [29] used linear regression analysis to examine the factors associated with road crash injuries (RCIs). Haque [30] used regression analyses to investigate the effectiveness of the traffic legislations. Rakha et al. [31] used linear regression models for predicting road traffic crashes. Agyemang [32] used linear regression model to examine the relationship between road traffic crashes and population. Similarly, Zlatoper [33], Desai and Patel [34], and Cai, Zhu and Yan [35] used linear regression modesl to examine the relationship between road crash fatalities and different explanatory variables. Bener et al. [36] used both the regression and Smeed equations to estimate the number of fatalities. Ahmed et al. [37] used Ordinary Least Squares regression (OLS) model to estimate the annual road crash fatalities for Pakistan using the aggregate data for Asian countries. The number of fatalities per hundred thousand populations was used as the response variable (continuous variable). These studies highlighted the importance of linear regression analysis for the association of road crash injuries (both fatal and non-fatal) with certain explanatory variables. It is concluded from the above studies that the road crash injury rate is a continuous variable and can be associated with the explanatory variables using linear regression analysis. 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 Road crash rates are significantly affected by roadway geometrics [24, 28], pavement characteristics [22], pavement surface conditions [21], and technical design of road infrastructure [41]. In addition, the frequency of road traffic crashes and injury severity is significantly related to the road segment length, AADT, crosswalks density, business land use, un-signalized intersection density, number of lanes between signals [20, 21], traffic light indicators, driver behaviour, and vehicle and environmental characteristics [25]. Similarly, the road crash fatality rates in a country are significantly affected by population density, the number of registered vehicles, licensed drivers, and traffic fines, total road length, and economic conditions [35, 38, 39]. Also, the road crash injury severity is affected by traffic violations, collision type, traffic environment [22], and operating speed [40]. The road crash fatality rates in a country are significantly affected by population, the number of registered vehicles, total road length, population density, and economic conditions [38, 39]. According to World Health Organization’s world report on road traffic injury prevention [3], the increase in motorization and vulnerable road users is the primary cause of road crash injuries (RCI). Also, the increased urbanization, low traffic safety awareness, negligence in the implementation of traffic rules, vehicular overloading, bad road conditions, violations of road safety law, and low vehicle maintenance standards are responsible for the increased RCI. The road user characteristics and behavior, with traffic violation and negligence of safety devices (helmet and seat belt), alcohol consumption, road and vehicle type, and crash location increase the RCI rates [11, 23, 29, 42-44]. 2.2 Factors associated with road crashes and injuries -6- Advances in Transportation Studies: an international Journal 192 193 194 195 196 197 198 199 In Pakistan, the lack of driver’s training and driving experience, bad condition of roads, cell phone usage while driving, use of intoxicants, vehicular over loading, governmental mismanagement, inferior geometrics and signage, poorly designed pedestrian facilities, careless driving, drivers falling asleep at the wheel, burst tires, and brake failures are the major factors responsible for road crashes [45, 46]. Similarly, according to Ahmed et al. [37], the road density, indicator for lead road safety agency, vehicle child restraint law, and the maximum speed on rural and urban roads significantly affect the RCF rate. 200 3. Methodology 201 3.1 International comparison of RCI rates 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 A comparison of road crash injuries (RCI) in Pakistan with that of other countries was conducted, utilizing data sources including World Health Organization (WHO), International Road Federation (IRF) and World Bank (WB) [17, 47, 48]. The countries were divided into different income groups based on Gross National Income (GNI) defined by World Bank. The data on population, total number of RCI, and registered vehicles for the year 2007 were acquired from WHO’s Global Status Report on Road Safety – 2009 [17]. The reason for using the data from WHO – 2009 Global Status Report on Safety on Road Safety was that none of the recent reports contain the data on non-fatal RCI. The Injuries per hundred thousand population (IPHTP) and injuries per thousand registered vehicles (IPTRV) were computed as the normalized measures of relative road safety. Table 1 shows the road crash injury rates and Pakistan’s ranking in regard to these RCIs among different groups of countries, including South Asian countries (SAARC – South Asian Association for Regional Cooperation), Asian countries, low-income countries, developing countries and in general, all the countries of the world. The ranking is based on descending order of road crash injury rates. Among the SAARC countries, Pakistan has the lowest road crash injury rates based on both IPHTP and IPTRV. Table 1 also shows that Pakistan is ranked the second and seventh safest country among Asian countries based on IPHTP and IPTRV, respectively. In low income countries, Pakistan is again ranked as the fourth and third last country based on IPHTP an IPTRV, respectively. Similarly, Pakistan is ranked as the third safest country among all the countries. Table 1. Pakistan’s Ranking among other Countries in Regard to Non-fatal RCI Rates Country Group Pakistan SAARC Countries Asian Countries Low-income Countries Developing Countries World IPHTP IPTRV 7.93 97.34 219.59 59.99 164.33 123.64 2.46 70.95 53.03 45.35 24.80 10.34 223 224 -7- Pakistan’s Ranking IPHTP IPTRV 7/7 37/38 38/41 116/120 163/166 7/7 32/38 39/41 113/120 163/166 Advances in Transportation Studies: an international Journal 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 Based on these results, it is evident that Pakistan has reasonable road crash injury rates, as compared to other countries in the world. It indicates that Pakistan’s road safety is in a good state. However, this may not be the true situation. On the contrary, it points out the issue of road crash data reporting in Pakistan. A comparison between different categories of countries based on average annual injuries was carried out that revealed high RCI rates for countries of high-income level, while lower RCI were noted for countries of low, middle-income, and Asian countries. Pakistan’s RCI rate as illustrated before in previous paragraphs does not evoke concern or worry. However, a country with a high population and a road infrastructure that suffers from blatant deterioration such as Pakistan cannot truly and fully match with the estimable outcomes pointed out by the figures. These results can also put the authenticity of the data in question. Hyder Ghaffar and Masood [13] using police recorded data found that police records only accounted for 14.3% of road traffic crashes. Similarly, Razzak et al. [49] used the surveillance system and concluded from their study that police records represented only 2-3% of RCI. According to Kayani, Fleiter and King [50], lack of police training, improper allocation of resources, departmental corruption, and inadequate data reporting and recording system are the main reasons of under-reporting of road crashes in the country. Furthermore, the definitions of road crash and injury used in Pakistan do not conform to international standards [51]. 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 To explore a relationship between road crash injury rates and influential factors, the data including socio-economic factors, national road safety policies and level of enforcement of traffic laws were obtained from WHO data sources for 73 countries having vital registration records. These countries were defined as the countries with vital registration records by the WHO’s Global Status Report on Road Safety – 2009 because they provided more than 85% of the data in the questionnaire used by the WHO for data collection [17]. The data of these countries was also used by the WHO’s Global Status Report as a reference in estimating the negative binomial model to estimate the data for the rest of the countries. They consisted of 35 European, 21 American, 15 Asian and 2 African countries. They were further divided into three groups based on Gross National Income (GNI) defined by World Bank. Group one consists of only one country with low-income level (GNI up to 935 USD), group two includes thirty-five countries with middle-income level (GNI between 936 – 11,455 USD) and group three entails thirty-seven countries with high-income level (GNI at least 11,456 USD) [48]. The data on total number of road crash injuries, GNI, and registered vehicles for the year 2007 were acquired from WHO’s Global Status Report on Road Safety – 2009 [17]. Similarly, the data on national road safety policies and level of enforcement of traffic laws including the effectiveness of overall enforcement of speed limits, information on vital registration records, audits of existing and new roads, funding of lead road safety agency, the speed limits on rural and urban roads along with their level of enforcement, national standards laws on drink-driving, seatbelt, motorbike helmet, and vehicle child restraint along with their enforcement levels from a scale of 0 – 10 (lowest to highest level), national policy for transportation investment, and national policy for promoting cycling and walking were also retrieved from the WHO report for the year 2007. Furthermore, the data on road density of these countries for the year 2007 were retrieved from International Road Federation’s database [47]. Table 2 presents the summary statistics of the variables used in the study. 3.2 OLS model estimation -8- Advances in Transportation Studies: an international Journal 267 268 Table 2. Summary Statistics of Variables Variable Mean SD Max. Min. 269 Injuries Per Hundred Thousand Population (IPHTP) 339.77 259.06 1523.24 Injuries Per Thousand Registered Vehicles (IPTRV) 10.21 9.27 62.16 Population (hundred thousand) 26.00 48.33 305.83 Number of Registered Vehicles ( ten thousand) 1258.81 325.54 25140.23 Road Density (km/square km of area) 1.09 1.32 7.04 Maximum Speed on Rural Roads (km/h) 78.70 15.89 110.00 Maximum Speed on Urban Roads (km/h) 54.41 10.36 100.00 Gross National Income (1000 US$) 19.94 18.01 76.45 National Speed Limits in Kilometres (1- Yes, 0 - No) 1.00 1.00 1.00 Enforcement level of Speed Limits on a scale of 0-10 5.64 1.72 8.00 National Drink-Driving Law (1 – Yes, 0 - No) 1.00 0.00 1.00 Enforcement Level of Drink-Driving Law on a Scale of 0-10 5.82 2.16 10.00 National Seatbelt Law (1- Yes, 0 - No) 1.00 0.00 1.00 Enforcement Level of Seatbelt Law on a scale of 0-10 6.47 1.92 9.00 National Motorbike Helmet Law (1 – Yes, 0 - No) 0.99 0.12 1.00 Enforcement Level of Motorbike Law on a Scale of 0-10 6.59 2.36 10.00 National Vehicle Child Restraint Law (1 – Yes, 0 - No) 0.79 0.40 1.00 Enforcement Level of Vehicle Child Restraint Law on a Scale of 0-10 4.34 3.05 9.00 Mandatory Installation of Seatbelt in Newly Manufactured Vehicles (1 0.44 0.50 1.00 - Yes, 0 – No) National Road Safety Strategy (1 – Yes, 0 - No) 0.90 0.29 1.00 National Policy for Promoting Walking (1 – Yes, 0 - No) 0.59 0.49 1.00 National Policy for Transportation Investment (1 if policy exists, 0 0.72 0.45 1.00 otherwise) Indicator variable for Safety Audit of New Roads (1 if audit exists, 0 0.59 0.50 1.00 otherwise) Indicator variable for Safety Audit of Existing Roads (1 if audit exists, 0 0.68 0.46 1.00 otherwise) Income Level (Low - 1, Middle – 2, High - 3) Low Middle 01 a 35 a Note: a Frequency of each level of income 270 271 272 273 274 275 276 277 278 The road crash injury rates offer a normalized measure of road safety. Exposure-based response variables such as number of injuries per hundred thousand vehicle kilometers travelled are more appropriate for estimating statistical models that use data from different countries with varying motorization levels. Since collecting reliable data on vehicle kilometers travelled is difficult especially in countries with not well established data reporting system, population based response variables can act as suitable surrogates for exposure to potential road crash situation. These response variables include the number of injuries per hundred thousand population (IPHTP), the number of injuries per hundred thousand registered vehicles (IPTRV), and the number of injuries per hundred thousand -9- 30.01 0.91 0.12 2.4 0.03 40.00 32.00 0.59 0.00 0.00 1.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 High 37 a Advances in Transportation Studies: an international Journal 279 280 281 282 283 284 285 286 287 288 289 290 291 kilometers of road network. The population based response variables are more appropriate to use in statistical models that use data from different countries for the same analysis year or against the past years. These values relatively are somehow static as they do not change radically over short periods of time [52, 53]. The present study used injuries per hundred thousand population (IPHTP) and injuries per thousand registered vehicles (IPTRV) as response variables for the cause-effect models. The number of injuries per hundred thousand populations does not depend upon vehicle usage or the total amount of travel. The number of injuries per thousand registered vehicles may partially reflect usage. These normalized measures of relative road safety are useful in quantifying overall risk on a comparative basis. The natural logarithm of the response variables was used in order to get positive outputs from the models. The independent variables along with their description are presented in Table 2. The model form is as follows: )) ∑ ) (1) 292 293 294 295 296 297 298 299 300 where, ln(y(i)) is the natural logarithm of the injury rate; β0 and βi’s are the model coefficients; xi represents the set of independent variables associated with road crash injuries; ε i is the error term. All the available explanatory variables in Table 2 were tried for models estimation. The models were estimated using step-wise OLS linear regression that develops the least squares regression equation in steps, adding one independent variable at a time and evaluating whether existing variables should remain or be removed. Due to high correlation among certain explanatory variables, the results of some variables were not intuitive, and so those variables were removed from the final models. 301 4. Results 302 303 304 305 306 307 308 309 310 311 312 313 314 The detailed model results are shown in Tables 3 and 4. The coefficient of determination (R 2) was low for both the models and is 0.185 and 0.254, respectively and thus do not show a good fit for a highly varied data collected for 73 countries. However, the variables are found to be significant. Thus, the developed models helped to explore the factors associated with road crash injuries but could not be used for prediction [54-56]. The correlation between the independent variables was checked thorough the variance inflation factor (VIF) which quantifies the severity of multi-collinearity in an ordinary least squares regression analysis. It provides an index that measures how much the variance of an estimated regression coefficient (the square of the estimate's standard deviation) is increased because of collinearity. The VIF of all the independent variables were in the acceptable range (VIF < 10, no severe collinearity). Also, the Durbin-Watson Statistic less than 4 indicated no sign of auto-correlation of the residuals. As shown in Table 3, the number of registered vehicles is significantly associated with IPHTP (correlation coefficient (r) = 0.318). All else being equal, the higher the number of registered vehicles - 10 - Advances in Transportation Studies: an international Journal 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 in a country, the higher the expected injuries per hundred thousand population. This finding is reasonable as the higher number of registered vehicles represents a higher exposure to road crash injuries [3]. The maximum allowable speed on rural roads is found to be positively associated with IPHTP (r = 0.291), indicating that high speeds pose higher risks of road crash injuries. This finding is quite intuitive and consistent with past research [17]. The RCI rate is expected to be higher in countries that have high speed limits for their highway networks compared to countries with comparatively lower speed limits. Over speeding has been identified as the most important risk factor influencing road crash risk and injury severity [57-59]. Past research has also revealed that higher speed limits are associated with higher road crash severity [60, 61]. The allowable speed limits are mostly higher on rural roads which make the curves and intersections dangerous. Also, the marked speed differences between various vehicle classes increase the risk of road crashes on these roads. Additionally, rural roads have a much lower rate of traffic law enforcements and the drivers are more pertinent to have higher speeds as the chances of being caught and get a ticket are very less. All countries, particularly the low and middle-income countries with poor condition of road infrastructure, need to have adequate legislation that ensure lower speed limits on their highway networks and allow authorities to manage speed limits in their jurisdiction [59]. Although, most countries have enacted national speed limit laws but their enforcement is equally essential for reducing road crash risk and injury severity. Enforcement of speed limit laws needs to be given increasing emphasis in low and middle-income countries that have high RCI rates. Enforcement level of seatbelt law refers to the effectiveness of enforcement on a scale of 0 -10 (0 not at all effective, 10 - highly effective). Effectiveness of overall enforcement level of seatbelt law is negatively related with the IPHTP. This finding is intuitive as well. Strict enforcement level of seatbelt law results in lesser RCI as seat belts reduce the likelihood of being injured during a road traffic crash [62]. It is a well-established fact that vehicle occupants not wearing seatbelts are at increased risk of injury and fatality in the event of a road crash, therefore any resources dedicated for enforcement shall benefit the community by lowering the number of injuries and fatalities. Mandatory seatbelt legislation is highly effective in promoting seat-belt wearing and is a cost-effective means of reducing road crash injuries and fatalities, especially in rapidly motorizing low and middle income countries [63]. The use of seat belts requires multi-sectoral action beyond appropriate legislation including publicity along with the provision of in-vehicle seat belt reminders, which has shown to be highly effective [64]. 348 - 11 - Advances in Transportation Studies: an international Journal 349 350 Table 3. Model Estimation Results (IPHTP) Variable Coeff. Constant 4.644 Std.Err. t-stats P-value 0.417 RV (Number of Registered Vehicles in 6.455E-03 0.000 Millions) MAXSR (Maximum Speed on Rural Roads) 0.012 0.005 High-SBLENF (High Enforcement of Seatbelt -0.324 0.178 Law) 2 R Adjusted R2 Durbin-Watson Number of observations VIF 95 % CI LB UB 11.131 0.000 3.812 5.476 2.480 0.016 1.098 0.000 0.000 2.238 0.028 -1.815 0.074 1.115 0.001 0.023 1.063 0.032 0.679 0.185 0.149 2.311 73 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 Table 4. Model Estimation Results (IPTRV) Std.Err. t-stats P-value VIF 95 % CI LB UB Variable Coeff. Constant 2.983 0.250 11.948 0.000 2.485 3.481 High-ILVL (High Income Level) AVE-SBLENF (Average Enforcement Level of Seatbelt Law) High- SBLENF (High Enforcement Level of Seatbelt Law) SANR (Safety Audit New Roads) R2 Adjusted R2 Durbin-Watson Number of observations -0.672 0.179 -3.753 0.000 1.106 -1.029 -0.315 -0.509 0.242 -2.102 0.039 2.018 -0.991 -0.026 -0.518 0.262 -1.976 0.052 2.052 -1.042 0.005 -0.350 0.176 -1.989 0.051 0.254 0.210 2.173 73 1.034 -0.701 0.001 The model results in Table 4 revealed that income level was inversely related to the response variable and significant at 95% confidence level. The higher the level of income in a country, the lower the IPTRV. There is a strong association between the risk of fatality in a road crash and the income level of a country. The risk in high-income countries is 10 times less than that of low-income countries [1]. A high-income country can ascertain the existence of effective road safety policies, well-maintained road infrastructure and pavement condition, signalized at-grade intersections, alternatives for congestion control, advanced public transport system, superb vehicle conditions, and educated drivers and pedestrians. The model results also revealed that countries with effective safety audits for their roads have lower road crash injuries. The road safety audit aims at identifying the potential safety problems - 12 - Advances in Transportation Studies: an international Journal 366 367 368 369 370 371 372 during the planning and design of new roads. It helps in identifying potential safety hazards and highlights road crash prone locations [65]. When safety is taken into consideration during the planning, design and operation of roads, substantial contributions can be made to reducing RCF and RCI [1]. Road safety inspections can provide mechanisms to identify failings in infrastructure that are responsible for affecting likelihood of road crashes and the severity of resulting injuries [1], and thus helps to eliminate the road safety deficiencies in the country. 373 5. Comparison of models significant variables means with Pakistan 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 A comparison of average values of significant variables in both the models for the 73 countries with those of Pakistan is presented in Table 5. The data for the years 2010, 2013, and 2016 were collected from the WHO’ Global Status Reports on Road Safety for the years 2013, 2015, and 2018 respectively. The mean of the enforcement level of seatbelt law for the countries under observation has increased since 2007. According to World Health Organization’s Global Status Report on Road Safety-2018 [1], 105 countries have seatbelt laws that align with best practice. Since 2014, seven countries have made changes to their seatbelt legislation of which five countries have laws that meet best practice. The net increase of these countries accounts for an additional 113 million people covered by best practice seatbelt laws. Of the 161 countries with national seatbelt laws, 105countries representing 71% of the world’s population have adopted the best practice of mandating the use of seat belts by both front and rear occupants [1]. Statistical analyses show that a significant increase in seat belt usage rates among both drivers and passengers for both genders resulted from the accompanying media and enforcement campaigns [66]. The enforcement level of seatbelt law in Pakistan was much lower than the mean level for the countries under observation in the study year. The enforcement level of seatbelt law remained unchanged up to the year 2013 and has been increased in the year 2016 but is still lower than the mean of the countries with vital registration records. The Road Safety Audit for New Roads was not present in the country in 2007 but was present in most of the countries (average value greater than half). Road infrastructure is strongly associated to road crash injuries during road crashes. Improving road infrastructure is critical in the improvement of overall road safety [67]. One hundred forty-seven countries carry out road safety audits for new roads while 114 countries reported doing safety assessments on existing roads around the globe [1]. The aim of carrying out road safety audits is to create a safe road environment instead of placing the main responsibility on road users who likely fail to deal with intrinsic dangers of the road. The country’s rural roads have higher speed limits than most of the other countries. In the past few years, the speed limit has drastically increased from 95 km/h to 110 km/h [1, 2, 59]. The country’s income level was lower than the mean of the income levels for the countries under observation. Only the number of registered vehicles which is positively associated with the RCI rate was lower than the mean of the other countries. The number of registered vehicles has also increased from the mean of the countries under observation in the year 2016. The comparison of the means of these factors with those in Pakistan depicts the data reporting issues. Table 6 shows the average values of the significant variables in the models without the outliers. The outlier is a data-point that lies outside the overall pattern in a distribution of a dataset. The outliers were identified using the interquartile range (IQR) method i.e., 1.5 IQR. This method states - 13 - Advances in Transportation Studies: an international Journal 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 that a data-point is an outlier if it is more than 1.5 times the IQR above the third quartile (> Q 3 + 1.5*IQR). Similarly, a data-point less than the 1.5 times the IQR below the first quartile (< Q1 1.5*IQR) is also identified as an outlier. Each variable in the data was investigated using this method. The data-points that were identified as outliers were removed from the dataset and the means were recalculated and compared with the values of Pakistan. Only the number of registered vehicles and the maximum speed limit on rural roads had outliers. The income level, enforcement level of seatbelt law, and the safety audit of new roads did not contain any outliers. In addition, the injuries per hundred thousand population and the injuries per thousand registered vehicles (the dependent variables) also had outliers. The removal of the outliers significantly reduced the means and variances of the variables. The comparison in Table 6 shows that Pakistan again has low RCI rates than the mean RCI rates of the countries under consideration. The number of registered vehicles in Pakistan is higher than the mean number of registered vehicles in 73 countries. Also, the maximum speed limit on rural roads was higher than the average of the speed limits for the countries under observation. A comparison of average values of significant variables in both the models for low-income countries with those of Pakistan is presented in Table 7. The number of registered vehicles and the maximum speed limit on rural roads in Pakistan are higher than the mean of the low-income countries. The enforcement level of seatbelt law is higher than the mean of the low-income countries in the year 2016. The comparison with Asian countries (Table 8) shows that Pakistan again has higher number of registered vehicles and maximum speed limits on rural roads. Also, the enforcement level of seatbelt laws is lower than the mean of the enforcement levels for the Asian countries. Similarly, Table 9 shows the comparison of the mean values of the models significant variables for developing countries with those of Pakistan. The number of registered vehicles and maximum speed on rural roads are higher in Pakistan. The enforcement level of seatbelt is again lower than the mean of the enforcement levels for the developing countries before 2016. The enforcement level of seatbelt law is recently doubled in the country and is higher than the mean of the developing countries. 433 - 14 - Advances in Transportation Studies: an international Journal 434 435 436 Table 5. Comparison of Models Significant Variables Means of Countries having Vital Registration Records with Pakistan 2007 2010 2013 2016 Variable X1 Pakistan X1 Pakistan X2 Pakistan X3 Pakistan 1258.80 1374.13 1546.55 1661.20 528.71 785.30 908.04 1835.25 (3255.42) (3382.13) (3608.55) (3801.97) Max Speed on Rural 78.70 89.92 89.31 91.10 95 110 110 110 Roads (15.89) (14.31) (14.73) (14.83) 2.49 2.53 2.63 2.97 Income Level 1.00 2.00 2.00 2.00 (0.53) (0.53) (0.49) (3.34) Enforcement Level of 6.47 6.45 7.06 6.70 3.00 3.00 3.00 6.00 Seatbelt Law (1.92) (1.87) (1.53) (1.49) Safety Audit of New 0.59 0.833 0.884 0.84 0.00 1.00 1.00 1.00 Roads (0.50) (0.37) (0.32) (0.37) Note: Standard Deviations in brackets; X1 denotes mean of 73 countries; X2 denotes mean of 70 countries; X3 denotes mean of 69 countries; RV denotes number of registered vehicles in unit of ten thousands RV 437 438 439 440 441 442 Table 6. Comparison of Models Significant Variables Means of Countries having Vital Registration Records with Pakistan (Outliers Excluded) 2007 2010 2013 2016 Variable X1 443 444 Pakistan X1 Pakistan X2 Pakistan X3 Pakistan 312.67 IPHTP 7.93 (201.07) 9.08 IPTRV 2.46 (6.13) 484.24 562.79 682.89 765.93 RV 528.71 785.30 908.04 1835.25 (675.53) (784.16) (969.66) (1060.19) Max Speed on Rural 78.70 90.58 89.95 91.55 95 110 110 110 Roads (km/h) (15.89) (13.41) (13.89) (11.55) 2.49 2.53 2.63 2.97 Income Level 1.00 2.00 2.00 2.00 (0.53) (0.53) (0.49) (3.34) Enforcement Level of 6.47 6.45 7.06 6.70 3.00 3.00 3.00 6.00 Seatbelt Law (1.92) (1.87) (1.53) (1.49) Safety Audit of New 0.59 0.833 0.884 0.84 0.00 1.00 1.00 1.00 Roads (0.50) (0.37) (0.32) (0.37) Note: Standard Deviations in brackets; X1 denotes mean of 73 countries; X2 denotes mean of 70 countries; X3 denotes mean of 69 countries; RV denotes number of registered vehicles in unit of ten thousands 445 - 15 - Advances in Transportation Studies: an international Journal 446 447 Table 7. Comparison of Models Significant Variables Means of Low-income Countries with Pakistan 2007 2010 2013 2016 Variable X Pakistan X Pakistan X Pakistan X Pakistan 304.02 417.27 657.02 1032.21 528.71 785.30 908.04 1835.25 (1191.79) (1886.40) (2670.34) (3762.39) Max Speed on Rural Roads 72.87 93.00 92.65 95 110 110 110 (km/h) (19.35) (13.43) (12.25) Enforcement Level of 2.85 4.41 5.29 5.07 3.00 3.00 3.00 6.00 Seatbelt Law (2.86) (2.53) (2.34) (2.12) Safety Audit of New 0.63 0.76 0.79 0.86 0.00 1.00 1.00 1.00 Roads (0.49) (0.43) (0.41) (0.35) Note: Standard Deviations in brackets; X denotes mean of 40 countries; RV denotes number of registered vehicles in unit of ten thousands RV 448 449 450 451 452 Table 8. Comparison of Models Significant Variables Means of Asian Countries with Pakistan 2007 2010 2013 2016 Variable X Pakistan X X Pakistan X Pakistan 1242.11 2161.74 2069.85 3147.85 528.71 785.30 908.04 1835.25 (2811.14) (3826.51) (4794.19) (6875.62) Max Speed on Rural 75.22 90.71 90.19 95 110 110 110 Roads (km/h) (20.36) (16.35) (14.63) 1.98 0.50 2.16 2.19 Income Level 1.00 2.00 2.00 2.00 (0.71) (0.66) (0.65) (0.55) Enforcement Level of 4.76 2.08 5.97 6.30 3.00 3.00 3.00 6.00 Seatbelt Law (2.83) (2.47) (2.28) (2.03) Safety Audit of New 0.59 0.22 0.86 0.98 0.00 1.00 1.00 1.00 Roads (0.50) (0.70) (0.35) (0.15) Note: Standard Deviations in brackets; X denotes mean of 46 countries; RV denotes number of registered vehicles in unit of ten thousands RV 453 454 Pakistan 455 456 - 16 - Advances in Transportation Studies: an international Journal 457 458 Table 9. Comparison of Models Significant Variables Means of Developing Countries with Pakistan 2007 2010 2013 2016 Variable X Pakistan X Pakistan X Pakistan X Pakistan 547.92 773.1 957.98 1488.91 528.71 785.30 908.04 1835.25 (1703.64) (2462.82) (3117.79) (4591.83) Max Speed on Rural 73.74 89.83 89.34 95 110 110 110 Roads (km/h) (18.85) (15.26) (15.67) 1.69 1.79 1.86 1.88 Income Level 1.00 2.00 2.00 2.00 (0.51) (0.49) (0.53) (0.52) Enforcement Level of 4.63 5.42 5.97 5.81 3.00 3.00 3.00 6.00 Seatbelt Law (2.90) (2.33) (2.13) (1.99) Safety Audit of New 0.60 0.81 0.81 0.86 0.00 1.00 1.00 1.00 Roads (0.49) (0.58) (0.39) (0.35) Note: Standard Deviations in brackets; X denotes mean of 120 countries; RV denotes number of registered vehicles in unit of ten thousands RV 459 460 461 462 6. Estimation of Non-fatal Road Crash Injuries 463 6.1 Modeling RCI counts 464 465 466 467 468 469 The number of road crashes and their outcomes (injuries and fatalities) are commonly estimated using count data models because the number of crashes and the resulting injuries (fatal and non-fatal) is a non-negative integer. The number of RCI (count data) is generally modeled and estimated with a Poisson regression or its derivatives including the negative binomial and zero-inflated models [68]. For the basic Poisson model, the probability P(ni) of a RCI is 470 471 472 473 474 475 476 477 478 479 480 481 ) ) (2) where i is the Poisson parameter, which is the expected number of RCI, E[ni]. Poisson regression specifies the Poisson parameter i (the expected number of RCI) as a function of explanatory variables by typically using a log-linear function: ) or, equivalently LN( i) = βXi (3) where Xi represents the set of explanatory variables and β’s are model coefficients [68]. However, a Poisson model may not always be appropriate because its distribution assumes the mean and variance to be equal (E[ni] = VAR[ni]). The data under analysis are said to be overdispersed (E[ni] < VAR[ni]) or under-dispersed (E[ni] > VAR[ni]) if the equality of mean variance - 17 - Advances in Transportation Studies: an international Journal 482 483 484 485 486 does not hold. This will result in incorrect standard errors of the estimated parameters and ultimately incorrect inferences could be drawn. In order to account for the inequality of the mean and variance, the negative binomial model is derived by rewriting ) (4) 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 where EXP(ɛi) represents a gamma-distributed error term with a mean and variance of 1 and α, respectively. The addition of the error term (EXP(ɛi)) which is gamma-distributed with a mean and variance of 1 and α, respectively, allows the variance and mean to differ as VAR[ni] = E[ni][1 + αE[ni]] = E[ni] + αE[ni]2. The Poisson regression is a restricted model of the negative binomial regression as α approaches 0. Thus, if α is significantly different from 0, the negative binomial is appropriate, and the Poisson model otherwise [68]. To describe the relative magnitude between the dependent and explanatory variables based on parameter estimates, marginal effects can be estimated. In the case of the number of RCI, marginal effects give the change in the number of RCI with a unit change in any explanatory variable, x, and are simply calculated as the partial derivative, ∂ i/∂x, where i is defined as in Equations (3) and (4) for Poisson and negative binomial models, respectively. The number of RCIs in a country depend on certain characteristics such as population, number of registered vehicles, road density, maximum allowable speeds on their road sections, and the existence and enforcement of traffic laws. Thus, the objective of statistical modeling is to estimate the expected number of RCIs in a given region (country) as a function of its characteristics. In other words, it is assumed that there exists a “systematic”, i.e. causal, component in the number of RCIs and “explanatory variables” may account for this non-random component. This implies defining a “regression model” where the explanatory variables (and possibly combinations thereof) act as “covariates”. 509 6.2 Selecting the regression model 510 511 512 513 514 Given the data set of the number of RCI and explanatory variables, the first step was to check the presence of “over-dispersion”, to discriminate between the two count data models (Poisson and NB models). On the basis of the ML estimates βˆ of the β regression coefficients under the Poisson model, the null hypothesis that α= 0 against α > 0 can be tested by the statistic: 515 516 517 518 519 520 ∑ ∑ ) ⁄ (5) which is asymptotically distributed as a standard Normal random variable [69]. Large positive values of Z indicate over-dispersion, whereas large negative values indicate under-dispersion. Once the model type was chosen, a stepwise forward procedure based on the Generalized Likelihood Ratio Test (GLRT) was used to decide which explanatory variables should be included in - 18 - Advances in Transportation Studies: an international Journal 521 522 523 524 525 526 527 528 the regression model. It is recognized that the effect of variables that affect the probability of road crash occurrence can be effectively represented by multiplicative terms, whereas the effect of variables that act as point hazards can be effectively represented by additive terms [70, 71]. Thus, the regression model should have both a multiplicative and an additive portion in general. For the multiplicative component, the exponential choice appears to be a natural one as it ensures that the expected number of RCI is always positive. Therefore, the log-linear regression model λi(xi; β) = exp (∑ ) was assumed for the expected number of RCI. 529 6.3 Total RCI counts 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 The Poisson model was estimated at first using the stepwise procedure based on GLRT. The data set consisted of the total number of non-fatal RCI reported for each country. The candidate set of explanatory variables was the same as used in the estimation of OLS regression models (Table 2). The residual deviance which is the difference between the deviance of the estimated model and the maximum deviance of the ideal model where the predicted and observed values are identical was used to perform the goodness of fit of the overall model. The large residual difference and therefore the significance of the goodness of fit indicated that the model does not fit the data well. The test for over-dispersion was performed and the value of the Z statistic (Equation (5)) was calculated at convergence of the stepwise procedure. Since Z = 2.8873 (p = 0.0019), there was a clear evidence that over-dispersion was present and, consequently the NB model was estimated. For both the regression models, different combinations of the explanatory variables were tested and in the final models, population, the number of registered vehicles, the square of the road density, the enforcement level of motorbike helmet law (high-enforcement), and the safety audit of new roads were significant explanatory variables at the 1% level. Also, the estimated coefficients have the expected signs for both the models. The number of RCIs increases with the increase in the population and the number of registered vehicles. In contrast, the number of these RCIs decreases with the square of road density, the enforcement level of motorbike helmet law, and the safety audit of new roads. Tables 10 and 11 respectively show the parameter estimates at convergence of the stepwise procedure for both the Poisson and NB models. - 19 - Advances in Transportation Studies: an international Journal 551 552 Table 10. Model Estimation Results (Poisson) Variable Coeff. Constant 11.172 Std.Err. POPM (Population in Millions) 7.151E-05 RVM (Number of Registered 0.016 Vehicles in Millions) RDENSQUARE (Square of Road -4.638E-03 Density) High-HLENF (High Enforcement -0.8156 level of Motorbike Helmet Law) SANR (Safety Audit New Roads) -0.6608 Restricted Log Likelihood (constant only) Log Likelihood at Convergence Chi-square (p-value) AIC BIC Z-stats P-value VIF 95 % CI LB UB 5.277E-04 21170.7 0.000 3.691 5.328 9.314E-07 0.000 1.101 0.000 0.000 3.111E-06 5077.71 0.000 1.116 0.002 0.023 3.607E-05 -128.59 0.000 1.125 -0.772 -0.061 1.046E-03 -779.38 0.000 7.427E-04 -889.83 0.000 1.059 0.014 553 - 20 - 76.78 -24164272 (df=1) -9368658 (df=6) 29591227 (0.000) 18737328 18737347 0.258 Advances in Transportation Studies: an international Journal 554 555 Table 11. Model Estimation Results (Negative-Binomial) Z-stats P-value VIF 95 % CI LB UB Variable Coeff. Std.Err. Constant 9.925 0.187 53.117 0.000 3.691 5.328 POPM (Population in Millions) RVM (Number of Registered Vehicles in Millions) RDENSQUARE (Square of Road Density) High-HLENF (High Enforcement level of Motorbike Helmet Law) SANR (Safety Audit New Roads) Restricted Log Likelihood (constant only) Log Likelihood at Convergence Chi-square (p-value) AIC BIC Theta (Std. Err) 0.004 0.001 4.332 0.000 1.101 0.000 0.000 0.075 0.005 14.535 0.000 1.116 0.002 0.023 -0.034 0.009 -3.608 0.000 1.125 -0.772 -0.061 -0.391 0.235 -1.670 0.1 -0.630 0.215 -2.928 0.001 1.059 0.014 0.258 -1882.031 -1806.728 (df=7) 150.61 (000) 3627.457 3649.156 1.1671 (0.0528) 556 557 558 559 560 Using the NB model (Table 11), 58,655 RCI were estimated for Pakistan. The estimated number of RCI was much higher (approximately 4.5 times) compared to the reported annual RCI of 12,990 [5, 17]. 561 7. Discussion 562 563 564 565 566 567 568 569 570 571 572 573 574 In Pakistan, inappropriate training and corruption of police and the absence of devotion to national road safety are the main reasons of inadequate data reporting and recording systems [50]. Under such serious inadequate crash data reporting systems, it is difficult to explore the factors associated with road crash injuries at the national level and the country cannot provide appropriate policy interventions. The enforcement level of seat belt law remained lower till the year 2013. The enforcement level has been doubled in the year 2016 and is higher than the mean of the low-income and developing countries but is lower than the mean of the Asian countries and the countries with vital registration records. This highlights that the safety agencies are not active enough to efficiently enforce their laws. The enforcement of seatbelt law only applies to the vehicle driver and not to all the occupants of the vehicle. Also, the use of seat belt is only limited to specific roads and vehicles. The highest rate of seat belt is only on motorways (53%) and the lowest rate is on the rural roads (5%). The lowest rate of seat belt usage is due to the unawareness of traffic laws and usefulness of its use, forgetfulness, and - 21 - Advances in Transportation Studies: an international Journal 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 careless attitude of drivers and traffic wardens in the country [72]. Also, the country has failed to evaluate the national seatbelt wearing rate which reveals the fact of unsatisfactory enforcement and ignoring its importance in the road safety [17, 73]. The Road Safety Audit for New Roads was not present in the country in the year 2007, but has been formed since 2010. Road Safety Audit is highly significant in the overall road safety and making long-term road safety policies and interventions for improvement. Having a policy, clear roles & responsibility and proper road safety audit management is vital in ensuring adequate road safety [74].The country’s rural roads have higher speed limits than most of the other countries. In the past few years, it has been drastically increase from 95 km/h to 110 km/h [1, 2, 59]. A general principle is that a 1% increase in the mean speed increases the risk of fatality and serious injury by 4% and 3%, respectively, during a road crash [57]. On the contrary, a 5% reduction in average speed can reduce the number of fatalities by 30% [75]. Similarly, small reductions in speed also influence the thresholds for death and serious injury risk in different crash scenarios [1]. Also, the number of registered vehicles has been doubled since 2013 which also shows the increase in exposure to road crashes. Road crash injuries are expected to decrease by a strict enforcement of seatbelt laws, reduction of speed limits, and the existence of effective road safety audits. 592 8. Conclusion 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 RCI have turned up as a moral challenge for the society. Consistent information regarding the total number of RCI and the factors contributing to high RCI rate is the primary step for initiating adequate countermeasures. This study collected the aggregate data on non-fatal RCIs from the WHO’s Global Status Report on Road Safety-2009 and compared the RCI rates based on two normalized measures for road safety. The study also collected the data of 73 countries having vital registration records, and estimated the OLS models that helped explore the factors associated with the RCI rates. The models present a reliable tool for exploring the factors affecting RCIs, especially in countries having inadequate road crash data reporting and recording systems. RCIs were estimated using count data models (Poisson and Negative-Binomial models). The estimated RCIs were found to be much higher than the reported data. The importance of road crash data reporting and recording systems at the country level is also highlighted by comparing the means of the significant variables for the countries under observation with those in Pakistan. The findings provide an insight for applicable legislation and reasonable investment in road safety for reducing the RCI rate. Based on the findings, RCIs are expected to decrease by a strict enforcement of seatbelt laws, reduction of allowable speed limits, and the existence of effective road safety audits. Furthermore, an extensive road crash data reporting and recording system is suggested for Pakistan and other developing countries. This study also has some limitation. The present research used data from the World Health Organization’s Global Status Report on Road Safety -2009 which was limited to police-reported road crash injuries. The WHO provides no data on non-fatal road crash injuries at the country level except the WHO-2009 report. So, this study used only the available data. The data from the WHO has certain limitations due to non-standardized measures used for the level of enforcement of traffic laws and the - 22 - Advances in Transportation Studies: an international Journal 616 617 618 619 620 road safety definitions. Data that are more detailed than police records would open up additional analysis possibilities and allow more precise model estimation. Furthermore, future research should focus on the effectiveness of the levels of traffic law enforcement and the behavior of the road users towards these traffic laws along with their evaluation. - 23 - Advances in Transportation Studies: an international Journal 621 Notes: 622 Conflict of interest 623 624 The authors have no conflict of interest regarding this study. 625 Acknowledgments 626 627 628 629 630 The authors would like to thank Dr. Yasir Ali for his feedback on the paper. The authors also acknowledge the World Health Organization (WHO) for publishing such useful book and the aggregate data regarding the situation of road safety around the globe. 631 632 633 This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Funding - 24 - Advances in Transportation Studies: an international Journal 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. World Health Organization (2019). Global status report on road safety (2018). Geneva (Switzerland): WHO. World Health Organization (2015). Global status report on road safety 2015. World Health Organization. Peden, M. (2004). World report on road traffic injury prevention, World Health Organization, Geneva. SIP (2020). Social Indicator of Pakistan. Pakistan Bureau of Statistics, Islamabad: Government of Pakistan, Statistics Division. (http://www.pbs.gov.pk/content/population-census. Accessed 20 January 2020). Accessed: 20 January 2020. SIP (2016). Social Indicator of Pakistan. Pakistan Bureau of Statistics, Islamabad: Government of Pakistan, Statistics Division. Shoaib, M. (2013). Pakistan Economic Survey 2012–13. Ministry of Finance, Government of Pakistan, 2013. National Transport Research Center. (2003). Accidents in Pakistan - a Review. Islamabad: Government of Pakistan, Ministry of Communication. 2003. Pakistan, Pakistan - GNI per capita GNI per capita, Atlas method (current US$). 2018. (https://www.indexmundi.com/facts/pakistan/gni-per-capita. Accessed 15 May 2020). National Transport Research Center. (2011). Road and Road Transport Statistics. Islamabad: Government of Pakistan, Ministry of Communication. 2011. Ahmed, A. (2007). National Road Safety Plan 2007-2012. Islamabad: National Road Safety Secretariat, Ministry of Communications, 2007a. Hyder, A.A., Amach, O.H., Garg, N., Labinjo, M.T. (2006). Estimating the burden of road traffic injuries among children and adolescents in urban South Asia. Health Policy, 77, pp. 129–139. Hyder, A.A., Morrow, R.H. (2000). Applying burden of disease methods in developing countries: a case study from Pakistan. American Journal of Public Health, 90(8), pp. 1235–1240. Hyder, A.A., Ghaffar, A., Masood, T.I. (2000). Motor vehicle crashes in Pakistan: the emerging Epidemic. Injury Prevention, 6, pp. 199–202. Ghaffar, A., Hyder, A.A., Masud, T.I. (2004). The burden of road traffic injuries in developing countries: the first national survey of Pakistan. Public Health, 118(3), pp. 211-217. Fatmi, Z., Hadden, W.C., Razzak, J.A., Qureshi, H.I., Hyder, A.A., Pappas, G. (2007). Incidence, patterns and severity of reported unintentional injuries in Pakistan for persons five years and older: results of the National Health Survey of Pakistan 1990-94. BMC Public Health, 7(1), pp. 152. Shamim, S., Razzak, J.A., Jooma, R., Khan, U. (2011). Initial results of Pakistan's first road traffic injury surveillance project. International Journal of Injury Control and Safety Promotion, 18(3), pp. 213-217. World Health Organization. (2009). Global status report on road safety: Time for action, Geneva: World Health Organization. Ahmed, A. (2007). Road Safety in Pakistan. National Road Safety Secretariat, Islamabad, 2007b: pp. 142. Subhan, F., Kanwal, H., Sulaiman, M., Naeem, M.M., Shafiq, M.M.I., Sajjad, U., Ajwad, A., Aqdas, A. (2017). National Road Crash Injuries – An Estimation and Comparison with Previous National Studies. The Nucleus, 54(4), pp. 210-213. El-Basyouny, K., Sayed, T. (2009). Accident prediction models with random corridor parameters. Accident Analysis & Prevention, 41(5), pp. 1118-1123. - 25 - Advances in Transportation Studies: an international Journal 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. Caliendo, C., Guida, M., Parisi, A. (2007). A crash-prediction model for multilane roads. Accident Analysis & Prevention, 39(4), pp. 657-670. Anastasopoulos, P.C., Mannering, F.L. (2009). A note on modeling vehicle accident frequencies with random-parameters count models. Accident Analysis & Prevention, 41(1), pp. 153-159. Savolainen, P., Mannering, F. (2007). Probabilistic models of motorcyclists’ injury severities in singleand multi-vehicle crashes. Accident Analysis & Prevention, 39(5), pp. 955-963. Moore, D.N., Schneider, IV W.H., Savolainen, P.T., Farzaneh, M. (2011). Mixed logit analysis of bicyclist injury severity resulting from motor vehicle crashes at intersection and non-intersection locations. Accident Analysis & Prevention, 43(3), pp. 621-630. Chen, F., Chen, S. (2011). Injury severities of truck drivers in single-and multi-vehicle accidents on rural highways. Accident Analysis & Prevention, 43(5), pp. 1677-1688. Abay, K.A. (2013). Examining pedestrian-injury severity using alternative disaggregate models. Research in Transportation Economics, 43(1), pp. 123-136. Doğan, A.A.E., ANgüngör, A.P. (2008). Estimating road accidents of Turkey based on regression analysis and artificial neural network approach. Advances in transportation studies, 16, pp. 11-22. Anastasopoulos, P.C., Tarko, A.P., Mannering, F.L. (2008). Tobit analysis of vehicle accident rates on interstate highways. Accident Analysis & Prevention, 40(2), pp. 768-775. Zargar, M., Roudsari, B.S., Shadman, M., Kaviani, A., Tarighi, P. (2003). Pediatric transport related injuries in Tehran: the necessity of implementation of injury prevention protocols. Injury, 34(11), pp. 820-824. Haque, M.O. (2011). Road casualty in Brunei Darussalam: country commentary. Advances in transportation studies, (25). Rakha, H., Arafeh M., Abdel-Salam, A.G., Guo F.F., Flintsch, A.M. (2010). Linear regression crash prediction models: issues and proposed solutions. Efficient Transportation and Pavement Systems: Characterization, Mechanisms, Simulation and Modeling, pp. 241-256. Agyemang, B. (2013). Regression Analysis of Road Traffic Accidents and Population Growth in Ghana. International journal of business and social research, 3(10), pp. 41-47. Zlatoper, T.J. (1984). Regression analysis of time series data on motor vehicle deaths in the United States. Journal of Transport Economics and Policy, pp. 263-274. Desai, M.M., Patel, A.K. (2011). Road accidents study based on regression model: a case study of Ahmedabad City. In National Conference On Recent Trends In Engeneering & Technology, pp. 1-8. Cai, H., Zhu, D., Yan, L. (2015). Using multi-regression to analyze and predict road traffic safety level in China. In 2015 International Conference on Transportation Information and Safety (ICTIS), pp. 363369 IEEE. Bener, A., Yousif, A., Al-Malki, M.A., El-Jack, I., Bener, M. (2011). Is road traffic fatalities affected by economic growth and urbanization development? Advances in transportation studies, 23, pp. 89-96. Ahmed, A., Khan, B.A., Khurshid, M.B., Khan, M.B., Waheed, A. (2016). Estimating national road crash fatalities using aggregate data. International Journal of Injury Control and Safety Promotion, 23(3), pp. 249-254. Bener, A., Crundall, D. (2005). Road traffic accidents in the United Arab Emirates compared to Western countries. in Advances in Transportation Studies, Section A 6. Citeseer. Bener, A., Bener, O.F. (2007). Mobile phone use while driving and risk of road traffic injury: applying the Lorenz Curve and associated Gini Index. Advances in Transportation Studies, 13, pp. 77-82. Harnen, S., Umar, R. R., Wong, S. V., Hashim, W. W. (2006). Motorcycle accident prediction model for junctions on urban roads in Malaysia. Advances in Transportation Studies, an International Journal Section A, 8. - 26 - Advances in Transportation Studies: an international Journal 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. Belwal, R., Belwal, S., Quraini, A.A. (2015). Road Traffic Accidents (RTAs) and road safety in Oman: an analysis of people's perception towards the causes. Advances in Transportation Studies, 37. Burgut, H.R., Bener, A., Sidahmed, H., Albuz, R., Sanya, R., Khan, W.A. (2010). Risk factors contributing to road traffic crashes in a fast-developing country: the neglected health problem. Turkish Journal of Trauma and Emergency Surgery, 16(6), pp. 497-502. Schmucker, U., Dandona, R., Kumar, G.A., Dandona, L. (2011). Crashes involving motorised rickshaws in urban India: Characteristics and injury patterns. Injury, 42(1), pp. 104-111. Afukaar, F.K., Antwi, P., Ofosu-Amaah, S. (2003). Pattern of road traffic injuries in Ghana: implications for control, Injury Control and Safety Promotion, 10(1-2), pp. 69-76. Shah., S.A.R., and Khattak, A. (2013). Road traffic accident analysis of motorways in Pakistan. International Journal of Engineering, 2(11), pp. 2340-2354. Khan, A.M., Tehreem, A. (2012). Causes of road accidents in Pakistan. Journal of Asian Development Studies, 1(1), pp. 22-29. Gonzalez, C., Perez, ME. (2012). IRF World Road Statistics 2012: Data 2005-2012. World Road Statistics, 2012. WB (2007). International Bank for Reconstruction and Development/The World Bank, World Development Indicators, Washington DC. Razzak, J.A., Shamim, M.S., Mehmood, A., Hussain, S.A., Ali, M.S., Jooma, R. (2012). A successful model of Road Traffic Injury surveillance in a developing country: process and lessons learnt. BMC Public Health, 12, pp. 357. Kayani, A., Fleiter J.J., King, M.J. (2013). Under-reporting of road crashes in Pakistan and the role of fate. Traffic Injury Prevention, 15(1), pp. 34-39. Khoso, A. K., Ekman, D. S., Bhatti, J. A. (2011). Comparison of Highway Crash Reporting in Pakistan With the World Health Organization Injury Surveillance Guidelines. Traffic Injury Prevention, 12(3), pp. 279-282. Bester, C.J. (2001). Explaining national road fatalities. Accident Analysis & Prevention, 33(5), pp. 663672. Roess, R.P., Prassas, E.S., W.R. McShane, W.R. (2004). Traffic engineering. Pearson/Prentice Hall. Andersson, H. (2007). Willingness to pay for road safety and estimates of the risk of death: evidence from a Swedish contingent valuation study. Accident Analysis & Prevention, 39(4), pp. 853-865. Svensson, M. Johansson, M.V. (2010). Willingness to pay for private and public road safety in stated preference studies: why the difference? Accident Analysis & Prevention, 42(4), pp. 1205-1212. Persson, U., Norinder, A., Hjalte, K., Gralén, K. (2001). The value of a statistical life in transport: findings from a new contingent valuation study in Sweden. Journal of Risk and Uncertainty,. 23(2), pp. 121-134. Finch, D. J., Kompfner, P., Lockwood, C. R., Maycock, G. (1994). Speed, speed limits and accidents. Safety Resource Centre. Moradi, A., Motevalian, S. A., Mirkoohi, M., McKay, M. P., Rahimi-Movaghar, V. (2013). Exceeding the speed limit: prevalence and determinants in Iran. International journal of injury control and safety promotion, 20(4), pp. 307-312. World Health Organization (2013). Global status report on road safety: Supporting a decade of action, World Health Organization, Geneva, Switzerland. Malyshkina, N.V., Mannering, F. (2008). Effect of increases in speed limits on severities of injuries in accidents. Transportation research record, 2083(1), pp. 122-127. Eustace, D., Almutairi, O.E., Hovey, P.W. (2016). Modeling factors contributing to injury and fatality of run-off-road crashes in Ohio. Advances in Transportation Studies, Section B, 40. - 27 - Advances in Transportation Studies: an international Journal 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. Elvik, R., Høye, A., Vaa, T., & Sørensen, M. (2009). Driver Training and Regulation of Professional Drivers', The Handbook of Road Safety Measures. Emerald Group Publishing Limited. Stevenson, M., Yu, J., Hendrie, D., Li, L. P., Ivers, R., Zhou, Y., Su, S., Norton, R. (2008). Reducing the burden of road traffic injury: translating high-income country interventions to middle-income and low-income countries. Injury prevention, 14(5), pp. 284-289. Farmer, C.M., Wells, J.K. (2010). Effect of enhanced seat belt reminders on driver fatality risk. Journal of safety research, 41(1): pp. 53-57. Eric, H., Frank, W. (1999). Road Safety Audit Guidelines. Fredericton, New Brunswick: University of New Brunswick. Vasudevan, V., Nambisan, S.S., Singh, A.K., Pearl, T. (2009). Effectiveness of media and enforcement campaigns in increasing seat belt usage rates in a state with a secondary seat belt law. Traffic injury prevention, 10(4), pp. 330-339. Turner, B., Breen, J., Howard, E. (2015). Road safety manual: a manual for practitioners and decision makers on implementing safe system infrastructure. Washington, S., Karlaftis, M.G, Mannering, F.L. (2003). Statistical and Econometric Methods for Transportation Data Analysis. Chapman & Hall/CRC New York. Dean, C., Lawless, J.F. (1989). Tests for detecting overdispersion in Poisson regression models. Journal of the American Statistical Association, 84(406), pp. 467-472. Hauer, E. (2004). Statistical road safety modeling. Transportation Research Record, 1897(1), pp. 8187. Hauer, E., Council, F.M., Mohammedshah, Y. (2004). Safety models for urban four-lane undivided road segments. Transportation research record, 1897(1), pp. 96-105. Klair. A. A., Arfan. M. (2014). Use of seat belt and enforcement of seat belt laws in Pakistan. Traffic Injury Prevention, 15(7), pp. 706-710. Bhatti, J.A., Ahmed, A. (2014). Evaluating performance of a Lead Road Safety Agency (LRSA) in a low-income country: A case study from Pakistan. International Journal of Injury Control and Safety Promotion, 21(2), pp. 136–143. Mansor, S. A., Saman, M. A., Razman, T. T., & Masnel, H. (2019, April). Road safety audit–what we have learnt? in IOP Conference Series: Materials Science and Engineering. (Vol. 512, No. 1, p. 012023). IOP Publishing. World Health Organization. (2017). Managing speed (No. WHO/NMH/NVI/17.7). World Health Organization. - 28 -