Effectiveness of transverse speed reducers and exploring factors contributing to road 1 traffic crashes on a rural two-lane highway: A mixed methods study

Background: In Ethiopia, there are an estimated 25.3 road traffic related deaths per 100,000 27 population, which is much higher than the global average road traffic fatality rate. Speed is the 28 most well-known risk factor influencing both the risk as well as the severity of the resulting 29 injuries. Although there is paucity of data from low-income countries, speed reducers have been 30 widely approved as an effective traffic calming countermeasure in high-income countries. We 31 aimed to (i) explore the effectiveness of transverse vertical speed reducers and, (ii) qualitatively 32 explore stakeholders’ perceptions of the factors that affect the risk of road traffic crashes. 33 Methods: Data on all crashes occurring from September 2010 to August 2015 were obtained. 34 Interrupted time series analysis using Poisson regression was used to estimate the effect of speed 35 reducers on the number of crashes per month before and after their installation in January 2012. 36 Focus group discussions and in-depth interviews were conducted with traffic police, drivers, 37 drivers’ training center owners, and community members to describe their perceptions about the 38 effects of the speed reducers. Quantitative and qualitative results were triangulated. 39 Results: There were 130 crashes during the study period. Of these, 45.4% were property damage 40 only, and 16.9% were fatal. After the speed reducers were installed, there was no statistically 41 significant difference (incidence rate ratio, IRR =1.17, 95% CI[0.60-2.30], p-value =0.644) in the 42 number of crashes per month, but there were changes in the distribution of crash severity (p- 43 value <0.001). Four core themes, with subsequent sub-themes, emerged as perceived 44 contributors to road traffic crashes. Of these core-themes, speedy and reckless driving, were 45 perceived as the strongest force perpetuating road collisions. Qualitative respondents disagreed on whether the speed reducers were effective and expressed concerns such as the lack of signage to warn drivers. Conclusions: Although speed reducers are proven to reduce collisions in high-income settings, this study in Ethiopia was inconclusive. Inappropriate design for the roadway type, sporadic placement, lack of signage and maintenance, and poor stakeholder coordination may This mixed-methods study used interrupted time series analysis to quantitatively evaluate the impact of the transverse speed reducers on the number of crashes occurring per month on a 45 km long section of asphalt two-lane highway in Hintalo Wejerat District. Qualitative methods including focus group discussions (FGDs) and in-depth interviews (IDIs) were used to explore community, traffic police, drivers’ training center owner, and driver perceptions regarding the speed reducers. The qualitative and quantitative results were triangulated to shed additional light on the speed reducers’ effectiveness or lack thereof.


3
Background 54 According to the WHO 2018 Global Status Report and the Global Burden of Injury study, 55 road traffic crashes (RTCs) continue to be the leading causes of mortality and morbidity 56 worldwide for people aged 5-29 years. 1 A meta-analysis from 15 African countries reported a 57 pooled crash injury rate of 65.2 per 100,000 population and pooled crash death rate of 16.6 58 deaths per 100,000 population. 2 In Ethiopia, a low-income country in Sub-Saharan Africa, there 59 are an estimated 25.3 road traffic related deaths per 100,000 population, translating to over 60 23,000 deaths annually 3 . This is much higher than the global average road traffic fatality rate of 61 17.5 deaths per 100,000 population. 3 Policies relating to drink-driving, speed limit, phone call, 62 and seat-belts use are included in the national legislation; however, because of resource 63 limitations and existence of competing national priorities, the enforcement of these laws is 64

poor. 3,4 65
Vehicle speed has been identified as a key risk factor for all kinds of RTCs, influencing both 66 the risk of a crash as well as the severity of the resulting injuries. It follows that controlling 67 vehicle speed can have a double effect of preventing crashes and reducing the severity of 68 resulting injuries. 3,5 "Traffic calming" refers to various road design modifications used for speed 69 management in areas where driver speeds are either excessive or inappropriate for the type and 70 road user mix of a given road. 5 Vertical traffic calming measures use forces of vertical 71 acceleration to discourage speeding. Horizontal measures use forces of lateral acceleration, while 72 audio-tactile measures use vibration, and narrowing measures use a psycho-perceptive sense of 73 enclosure to discourage speeding. 6 Some examples of vertical measures include speed bumps, 74 speed humps, speed tables, and speed slots/cushions. 6-9 75 There exists a considerable evidence base supporting the effectiveness and sustainability of 76 vertical speed control measures, mostly generated from high-income countries. 3,5 Logically, the 77 self-enforcing and cost-effective nature of these interventions make them an attractive option for 78 low-income settings, yet there is little reason to believe that findings from high-income settings 79 are direct to low-income contexts which tend to have a higher proportion of non-motorized 80 traffic, more lax vehicle maintenance and safety standards application, lower driver education 81 levels, insufficient police enforcement capacity, poorer quality roads, and weaker political 82 commitment to road safety. 3,10,11 All of these factors could affect the implementation of speed 83 The speed reducers in the study area are made of asphaltic concrete, and consist of 2.5 143 sinusoidal undulations, in contrast with other common vertical speed reducer designs which 144 typically have a single raised feature ( Figure 2). As shown in Figure 2 including focus group discussions (FGDs) and in-depth interviews (IDIs) were used to explore 156 community, traffic police, drivers' training center owner, and driver perceptions regarding the 157 speed reducers. The qualitative and quantitative results were triangulated to shed additional light 158 on the speed reducers' effectiveness or lack thereof. 159

Data collection and statistical methods 160
Quantitative 161 Local traffic police report all road collisions involving property damage or injuries using a 162 is a scaled interaction term between time and intervention that models the slope change post-178 intervention (coded 0 up to the last point before the intervention phase and coded sequentially 179 from 1 thereafter). 20,21 Fourier terms were included to adjust for seasonal patterns: k is the 180 number of pairs of Fourier terms included in the model (k=1 for annual seasonality, k=2 for 6 181 monthly seasonality, etc.), T is the number of time periods described by each sinusoidal function 182 (T= 12 for 12 months in a year). 22,25 Lastly, is the error term at time t. A lag term was not 183 included in the a priori model since the evidence shows that speed reducers tend to have an 184 immediate, rather than gradually occurring, effect on reducing vehicle speeds. 6,7 185 Incidence rate ratios (IRR) were derived with their 95% Confidence Intervals (CIs) from the 186 The principal investigators (Nigus Asefa and Hannah Yang) went to the study site and conducted 211 both the FGDs and the IDIs in the local language, Tigrigna. Informed written consent was 212 obtained from all participants. All responses were audiotaped and verbatim transcripts were 213 generated from the audio recordings. The results were then coded in Atlas.ti software using an 214 inductive approach. Similar codes were grouped and re-grouped to identify all non-repetitive 215 core themes that emerged from the data. 216 Results 217

Quantitative 218
The study investigators visited the site several times to abstract data from the traffic police 219 reports, conduct FGDs and in-depth interviews, and observe the design and placement of the 220 speed reducers. The photograph in Figure 3 was taken by the study investigators during one of 221 the site visits. 222 were 2.10 crashes p er month at the beginning of the period ( ). 259

Qualitative 260
The mean age of the FGD participants was 35.7 ± 8.4 years, with a minimum of 25 and a 261 maximum of 50 years. Table 2 shows the themes that emerged from discussions and interviews 262 with police officers, community members, and transport drivers regarding their perceptions 263 about the causes of RTC and the speed reducers' effects in mitigating collisions, their effects on 264 drivers and vehicles, and the suitability of their design and application. 265 Four core themes with associated sub-themes emerged from the inductive analysis of the data 266 (Table 2). According to the current study, the main facilitators contributing to RTCs can be 267 categorized as factors related to drivers, pedestrians and community matters, aspects of road 268 design, and administrative and policy issues. 269 The participants repeatedly stressed that driver-related factors, such as speeding, failing to yield 272 for pedestrians, and passing into oncoming traffic were the most common causes of RTCs (Table  273 1). Among these factors, participants stated that speedy and reckless driving is the major factors 274 contributing to RTCs (Figure 7) and also explained that it is a common phenomenon to see 275 drivers speeding dangerously in residential and business districts where there is increased 276 pedestrian traffic. One participant explained the situation as: 277 "First of all, accidents are really the drivers' fault. There are also some bad places 278 [roads], but, most of the accidents are caused by speeding, breaking the speed limit, 279 passing when there is opposing traffic, and not knowing the vehicle's maximum load. So, 280

generally, I say it's the drivers' problem." Driver, in his 30's 281
Some of the respondents mentioned that a higher frequency of collisions is observed among the 282 younger drivers and those with less driving experience ( Figure 6). 283

"The younger drivers and most of the drivers with new licenses tend to drive recklessly 284
with high speed whereas those who are experienced tend to drive slowly". Local 285

resident, in his 40's 286
On the contrary, some participants explained that age and experience do not matter according to 287 their observations, citing that many of the crashes occurring on rural, mountainous roads involve 288 trailer trucks which are typically driven by drivers holding level 5 licenses (presumably older and 289 more experienced). 290 A substantial share of RTCs can also be attributed to driver's impatience and failure to let 291 pedestrians cross the road first. The participants reflected that negligence towards pedestrians in 292 combination with excessive speed are the main problems that lead to RTCs, especially in 293 town/urban areas. 294

Theme 2 Community and pedestrian-related factors 295
The current study also found that the community's low awareness about road safety is an 296 important contributor for RTCs, putting rural pedestrians at, particularly increased risk. Some and on market days, thus reaching adult audiences. However, study participants believed that 315 safety education would be more effective if targeted towards youths. Respondents also 316 forwarded that resource constraints hinder road safety awareness, especially in rural 317 communities. Due to the lack of funding and transportation, traffic police are only able to deliver 318 safety messages to rural dwellers when they come to town for a market day or other events. 319 According to the participants, on-the-spot verbal instruction alone is not enough to bring 320 consistent behavior change concerning road safety. Traffic police officers agreed that community 321 awareness of road safety is still in its infancy stage 322

Theme 3 Road-related problems 323
Although study participants believed that drivers are the main cause of RTCs, they also revealed 324 how the difficult terrain in the study area, natural factors such as weather, and poor road 325 maintenance could increase the risk of RTC. A public transport driver explained: There were conflicting views on how the speed reducers affected crash incidence in the study 332 area. Some participants strongly believed the speed reducers decreased the incidence of 333 Despite the benefits, the participants also reported drawbacks regarding the design and 342 placement of the speed reducers. Participants reported that the bumps were too tall when they 343 were first constructed and hindered the movement of small vehicles, though they have eroded 344 considerably since then due to poor maintenance. The absence of signage before reaching the 345 speed bumps was also believed to expose unfamiliar drivers to increased risk of RTCs.

30's 351
Discussants also noted that drivers swerve off the road or out of their lane to avoid driving over 352 the bumps, which could also increase collisions. 353 Discussants also noted that drivers swerve off the road or out of their lane to avoid driving over 354 the bumps, which could also increase collisions. 355

Theme 4 Administrative and policy-related issues 356
Administrative and policy-related problems were reported as additional underlying causes for 357 RTCs. Traffic police struggle to enforce traffic laws is limited due to personnel shortages, budget in the area. All drivers are supposed to undergo training before obtaining their driving license, 369 but according to the study participants that is not always the case. Bribery, forgery, dishonesty, 370 or personal favors allow certain drivers to gain their license without undergoing any training at 371 all. Even drivers who legitimately graduate from a training center tend to lack meaningful 372 experience on realistic road layouts due to the generally weak and non-standardized curriculum 373 ( Figure 6). 374 The participants repeatedly mentioned that the training centers do not provide training in real life 375 driving situations and hence drivers graduating from training centers in the region lack adequate 376 skills to mitigate the possible challenges of a real situation. One of the FGD participants stressed 377 how the government sanctioned training curriculum is ineffective in preparing drivers for work: 378 When the licensing process is thorough like this, drivers can really learn the profession. 385 They shouldn't train only in the town and during the day because they will not only be 386 driving in towns and daytime." Traffic police, in his 40's 387 Next, there is no efficient way for the traffic police to verify the validity of someone's license. 388 Every Regional State in Ethiopia has the power to issue driving licenses that are valid in all parts 389 of the country. But when a traffic police officer stops a driver in the field there are no quick 390 means of checking the validity of their license or the drivers' past traffic infractions. The 391 participants also suggested that lifesaving first aid training should be provided for traffic police 392 staffs, as they are almost always the first to arrive at the scene of an RTC. 393 The discussants also indicated that the insurance claim process is another problem that is 394 contributing to increasing RTCs. the current system requires vehicles and wreckage to stay in 395 their position after a collision until insurance company representatives arrive and get their 396 required documentation. This is very serious when the crash happens in a narrow section of the 397 road or a place with low visibility. 398 20 expected number of crashes that would have occurred without the intervention. However, the 519 increased time, monetary, and computational costs were problematic for the scope of this study. 520 Lastly, reliance on routine data can also introduce bias, because it may not always be complete or 521 may be affected by changes in data collection requirements. 522

Conclusion and recommendations 523
Ethiopian Roads Authority should establish evidence-based design and placement standards 524 for speed reducing road modifications that are specific to the roadway functional class and 525 empirically measured traffic characteristics such as vehicle speeds, traffic volume, road user mix, 526 and emergency vehicle access. These standards should be followed both by governmental 527 workers and private contractors. An established requesting process, a high level of community 528 involvement, and an intensive needs assessment including baseline measurement of the problem 529 and an expert engineering study should all precede the construction of any speed reducing 530 device, while regular maintenance and evaluation should follow it. On the other hand, increasing 531 the number of traffic police, providing patrol vehicles, and procuring speed radars and 532 breathalyzer machines could improve the capacity of traffic police to enforce existing laws and 533 control driver speeds. Behavioral change campaigns addressing the drivers as well as community 534 members (pedestrians and passengers) could also be effective if tailored to the local society and 535

culture. 536
Overall, better coordination between contractors, local police, and different governmental 537 sectors is needed to attain ownership and sustainability of road design interventions. Finally, the 538 findings of this study demand further investigation on how road modifications and traffic 539 calming technologies can be effectively adopted in low-and middle-income settings to reduce 540

Ethical consideration 543
The Institutional Review Board of Mekelle University College of Health Sciences approved this 544 study. A supporting letter was also obtained from the Tigray Traffic Police Commission. The 545 collision data was received and analyzed in a de-identified format. Informed written consent was 546 obtained from each discussion and interview participant. The investigators explained the study 547