key: cord-0952802-dkfr3gsu authors: Vandoros, S.; Papailias, F. title: Empty Streets, Speeding and Motor Vehicle Collisions during Covid-19 Lockdowns: Evidence from Northern Ireland date: 2021-01-05 journal: nan DOI: 10.1101/2021.01.03.21249173 sha: c3b5cf5f817e0a104d4c77b2a1130bd2cca9853d doc_id: 952802 cord_uid: dkfr3gsu Covid-19 and lockdowns have had spillover effects on other health outcomes. Motor vehicle collisions (MVC) are likely to have been affected by the pandemic due to, among others, less traffic volume and speeding on empty streets. This paper studies the impact of the pandemic on MVCs in Northern Ireland. Using monthly data on injuries and deaths, we find a steep decline in slight and serious injuries compared to what would have been expected in the absence of the pandemic. However, we find no effect on the number of deaths. Based on data from speeding tickets, a plausible explanation for the differential effect on the number of injuries and deaths is speeding on empty streets during the pandemic. Over 50 people lose their lives in Motor Vehicle Collisions (MVC) in Northern Ireland every year, and more than 700 are seriously injured. 1 As Covid-19 has been affecting nearly all aspects of everyday activities, there are reasons to believe that the number of MVCs has also been affected by the pandemic, as outlined by plausible mechanisms put forward in a previous study. 2 First, the impact on traffic volume may have an ambiguous effect on the number and severity of MVCs. People are likely to commute less due to lockdowns and due to being afraid of catching Covid-19. On one hand, less traffic volume means fewer vehicles at risk of collision. Higher unemployment rates observed during the pandemic 3 also are likely to lead to a reduction in the number of collisions, as the experience from previous recessions has shown. [4] [5] On the other hand, less congested streets give an opportunity for speeding, which may affect the likelihood and severity of collision. [6] [7] Second, factors associated with the pandemic have had an impact on driving behaviour. Distraction, due to worrying about family members who are ill 8 or due to economic uncertainty and financial worries, [9] [10] [11] has been associated with increased risk of MVC. Changes in drinking patterns have also occurred during the pandemic. Bars and restaurants were closed for certain periods, due to measures to tackle Covid-19, and had fewer customers when open as people were trying to reduce the risk of catching the virus. However, overall alcohol consumption may have increased. 12 Sleeping patterns (another risk factor) have also been affected by the pandemic. [13] [14] Empirical evidence on the impact of the Covid-19 pandemic on MVCs is scarce, as many countries publish data with a significant time lag. Early empirical evidence suggests that there has been a decrease in collisions during the pandemic. 2, [15] [16] [17] [18] Interestingly, in the USA there has been a reduction in MVCs in the first months of the Covid-19 pandemic, 15-16 but at the same time there has been an increase in deaths. 19 The objective of this study is to examine whether and to what extent MVCs, injuries and deaths were affected by the Covid-19 pandemic in Northern Ireland. We obtained monthly data on MVCs from the Police Service of Northern Ireland (PSNI). Data included the monthly number of slight injuries, serious injuries and deaths for the period from January 2015 to October 2020. Unfortunately, monthly data by cause of MVC were not available, but the PSNI provides monthly data on the number of traffic-related fines by type of violation. We investigated the lockdown effects employing Interrupted Time Series (ITS) analysis. The "intervention" started in March, when the first Covid-19 death occurred in Northern Ireland, and the first lockdown was introduced. This is also the period when reduced mobility was observed. 20 Separate analyses were conducted depending on the MVC outcome: number of slight injuries, serious injuries, and deaths. The model includes a time variable, a post-intervention dummy variable, and an interaction between the post-intervention dummy and the time variable, to capture any change in trend. In particular, for each target series we estimated the model presented in Equation (1). (1) where is the time trend, is a dummy variable which takes the value 0 in the pre-lockdown period and the value 1 in the post-lockdown period and is the error term assuming the standard assumptions hold. 0 can be interpreted as the overall (baseline) level at time = 0, 1 can be interpreted as the pre-lockdown trend, 2 indicates the change in the level after the lockdown and 3 indicates the change in the slope. The above model can be estimated assuming Gaussianity or Poisson (which is the standard approach in population health evaluations where the outcome is a count variable). The extended set of our results includes both approaches as well as the Quasi Poisson which accounts for overdispersion, even if this does not seem to be the case with our data. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint years (panel A). The number of slight injuries started increasing after the lockdown but remained below that of previous years throughout the study period. There is also a decrease in the number of serious injuries, although the difference compared to previous years is not as strong as in the case of slight injuries (panel B). Although in January 2020 the number of serious injuries was larger than that in any of the previous five years, this figure remained much lower than previous years' levels during the first lockdown, and below or at the lower end of the distribution in subsequent months. However, there is little or no evidence of any decrease in deaths during the first lockdown or in the following months (panel C). [Insert Figure 1 here] We naturally divided our sample in the pre-lockdown period (from January 2015 to February 2020) and the post-lockdown period (from March 2020 to October 2020) and started our exploratory data analysis (EDA) by investigating a simple time scatterplot for the three underlying target series. A first look reveals that there seems to be a slight, and possibly not statistically significant, break in the trend and a change in the direction of the slope for deaths ( 1 ). This is more evident for serious injuries ( 2 ) and slight injuries ( 3 ) where there is a downward shift in the entire line, with a simultaneous change in the slope ( Figure 2 ). [Insert Figure 2 here] We extended our EDA by calculating the descriptive statistics for the full sample, as well as the pre-and post-lockdown subsamples (Table 1) . Table 1 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. [Insert Table 1 here] Table 2 presents the ITS results for the three dependent variables. There is a simultaneous change in the trend as well as the slope in the post-lockdown periods for serious injuries (Y2) and slight injuries (Y3). In accordance with the results of the T-Test, there is no statistically significant difference for the Deaths ( 1 ) series. However, we have a negative change in the level after the lockdown ( 2 ) and an abrupt increase in the slope ( 3 ) for the Serious Injuries [Insert Table 2 here] Figure A1 in the Appendix illustrates the autocorrelograms of the residuals from the regression of each series. It is evident that there is no autocorrelation or obvious seasonality patterns. As an additional check, we also calculated an extended set of results including: Newey-West HAC t-values and confidence intervals; 21 Poisson distribution for the error term; and Quasi Poisson Distribution to account for overdispersion. Results are presented in Tables A1, A2 and A3 respectively, in the Appendix. In all cases, our qualitative conclusions about 2 and 3 still hold and confirm the results of the baseline ITS model. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint Interestingly, while there was a steep decline in the number of injuries, there didn't seem to be any change in the number of deaths. Could this be a result of greater collision severity, due to speeding on empty streets? [6] [7] Or could it be a result of increased alcohol consumption? 12 Figure A2 in the Appendix shows the trends in speeding fines and fines relating to driving under the influence of alcohol or drugs in 2020 and previous years (before 2017 there did not seem to be much testing for alcohol). There is a steep increase in speeding fines in May and June 2020 compared to years 2017-2019. This could be a result of increased traffic violations, or a result of a greater police crackdown as a result of more speeding-related accidents. There didn't seem to be any relative increase in the number of alcohol and drug violations, possibly because people were drinking more at home, as bars and restaurants were shut, or because they were afraid of catching the virus there after they opened. This study shows that there has been an increase in MVC-related injuries in Northern Ireland is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint This differential impact on the number of injuries and the number of deaths is likely to be a result of speeding in less congested streets, due to reduced traffic volume during the lockdown and the pandemic in general. Unfortunately, monthly data by cause of MVC were not available. However, an increase in fines for speeding in May and June 2020 points in that direction. Of course, the number of fines is affected by policing intensitybut even if there was a crackdown on speeding particularly in May-June 2020, this would likely be a response to increased speeding violations. Speeding can lead to more severe crashes, thus resulting in death. Therefore, it appears that the negative impact of the reduced number miles travelled was counterbalanced by higher speeds. Another factor that may have not allowed the number of deaths to decrease is health system capacity during the pandemic. As ambulances and hospitals were responding to large numbers of Covid-19 patients, this may have led to a delay in treatment of MVC injuries, that could lead to more fatalities that otherwise. However, it is important to note that this is just a plausible explanation and this does not follow directly from the findings of this study. Future research can examine whether health system capacity was in any way related to MVC fatalities during the pandemic. An increase in alcohol consumption reported elsewhere during the pandemic could also have been a factor, 12 but there was no evidence of an increase in fines relating to drunk driving. It is possible that increased alcohol consumption was counterbalanced by drinking at home rather than in pubs or restaurants, thus avoiding drink driving. It is not possible to examine the role of any change in sleep duration and patterns with the data currently available. This can be examined by future research. Results of this study add to existing research on how the pandemic affected MVCs. While there is consensus on a reduction in collisions, [15] [16] [17] [18] the results on fatalities are not always in the same direction. 2, 19 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint Findings also provide additional evidence on the spillover effects of Covid-19 on other health outcomes in Northern Ireland and elsewhere. [22] [23] [24] [25] [26] [27] Overall, this study highlights yet another indirect consequence of the pandemic in the short run, but as people adapt their behaviour, additional evidence will be needed to study the long-term effects on motor vehicle collisions. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted January 5, 2021. ; https://doi.org/10.1101/2021.01.03.21249173 doi: medRxiv preprint Road Traffic Collision Statistics Covid-19, Lockdowns and Motor Vehicle Collisions: Empirical Evidence from Greece Economic consequences of Covid-19: A counterfactual multi-country analysis Are recessions good for your health? Effects of the 2008 recession on health: a first look at European data COVID-19 lockdown and fatal motor vehicle collisions due to speed-related traffic violations in Japan: a time-series study The Guardian, 2020. UK motorists caught at 'extreme speeds' on quiet lockdown roads Emotional stress and traffic accidents: the impact of separation and divorce Greasy roads: the impact of bad financial news on road traffic accidents The short-term impact of economic uncertainty on motor vehicle collisions When the market drives you crazy: Stock market returns and fatal car accidents Changes in adult alcohol use and consequences during the COVID-19 pandemic in the US Changes in sleep pattern, sense of time and digital media use during COVID-19 lockdown in Italy Effects of lockdown on human sleep and chronotype during the COVID-19 pandemic Vehicle related injury patterns during the COVID-19 pandemic: What has changed Initial impact of COVID-19's stay-at-home order on motor vehicle traffic and crash patterns in Connecticut: an interrupted time series analysis Impact of COVID-19 lockdown policy on homicide, suicide, and motor vehicle deaths in Peru COVID-19 lockdown and reduction of traffic accidents in Tarragona province COVID Lockdowns, Social Distancing, and Fatal Car Crashes: More Deaths on Hobbesian Highways Google COVID-19 Community Mobility Reports A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix Estimating excess mortality in people with cancer and multimorbidity in the COVID-19 emergency The impact of the COVID-19 pandemic on cancer deaths due to delays in diagnosis in England, UK: a national, population-based, modelling study. The lancet oncology Mental health before and during the COVID-19 pandemic: a longitudinal probability sample survey of the UK population Trends in suicide in Japan by gender during the COVID-19 pandemic Excess mortality in the United States during the first three months of the COVID-19 pandemic Excess Mortality during the Covid-19 pandemic: Early evidence from England and Wales