An example of the usefulness of joinpoint trend analysis for assessing changes in traffic safety policies

Highlights

• Driver deaths from traffic injury in Spain were analyzed with joinpoint regression.

• Traffic injury death rates related and unrelated to alcohol show similar declines.

• These declines accelerated in 2003 (alcohol-related) and 2006 (alcohol-unrelated).

• Changes in the trend of these rates match in time with changes in traffic safety regulations.

• Joinpoint regression may contribute to the assessment of traffic safety policies.

Abstract

Aim

To show the usefulness of joinpoint trend analysis for assessing changes in traffic safety policies.

Methods

Trends in driver death rates from traffic injury stratified by alcohol involvement in Spain during 2001–2011 were characterized with joinpoint regression to observe how changes in these rates match in time with the implementation of traffic regulations and interventions.

Results

Both alcohol-related (blood alcohol concentration >0.3 g/l) and non-alcohol-related rates decreased similarly (10–11% annually) during 2001–2011, although the former showed an earlier (2003 joinpoint) acceleration of the downward trend than the latter (2006 joinpoint); both joinpoints matched in time with relevant road safety interventions.

Conclusions

These results, which are consistent with results from previous analytical studies, show that joinpoint trend analysis, although not valid for causal inference, is useful for assessing changes in traffic safety policies. This methodology can be easily extended to other risk factors, such as excessive speeding or not wearing a seat belt.

Introduction

Policies aimed at reducing fatal traffic injuries are implemented in many countries (Connor et al., 2004, Elvik, 2012, Lardelli-Claret et al., 2009, Peden et al., 2004, Reynaud et al., 2002, Taylor et al., 2010, WHO, 2013). Previous studies suggest that some specific interventions, mainly focused on preventing reckless behaviours such as excessive speeding, alcohol-impaired driving (AID), or driving without helmets or seatbelts, have been effective (Brady et al., 2009, Castillo-Manzano et al., 2011, Elvik and Vaa, 2004, Fell and Voas, 2006, Goss et al., 2008, Ker and Chinnock, 2008, Liu et al., 2008, McMillan and Lapham, 2006, Novoa et al., 2009, Shults et al., 2004, Shults et al., 2009, Wilson et al., 2006). Single pre-/post intervention studies – e.g., interrupted time series – are perhaps the most common designs used for this purpose (Killoran et al., 2010, Novoa et al., 2010a, Novoa et al., 2010b, Novoa et al., 2011, Pulido et al., 2010, Wilson et al., 2010). However, it is difficult to assess the effectiveness of specific interventions due to the intrinsic complexity of the relationship between traffic injuries, risk factors, and policies, as well as the limitations of surveillance systems (Peden et al., 2004). One reason that often hinders this assessment is the implementation of various interventions concurrently or consecutively in a short space of time. This practical approach is justified because there is evidence that the combination of interventions is more cost-effective than single interventions (Bliss and Breen, 2009, Chisholm and Naci, 2008). For example, to reduce the frequency of AID, several interventions can be combined: new regulations increasing penalties for exceeding blood alcohol concentration (BAC) limits, stricter supervision of AID – e.g., gradually increasing the number of random breath tests, advertising campaigns informing about the dangers of AID, and interventions aimed at reducing average alcohol consumption in the general population – e.g., reducing the places and hours for sale as well as advertising of alcoholic beverages. Various measures focused on different risk factors (AID, speeding, non-use of seat-belts and other factors related to the driver, vehicle or road) can also be implemented concurrently. In these circumstances it is very difficult to assess the independent effect of a specific intervention, especially in the long term, because the length of post-intervention observation in most analytic studies is generally short in order to reduce the concurrent effect of other interventions and confounders (Killoran et al., 2010, WHO, 2013). Moreover, as many pre-post intervention studies use the overall rate of traffic injuries as the only outcome, the effect of an intervention focused primarily on a specific determinant of traffic accidents, such as AID (Goss et al., 2008), can be diluted or not detected. Finally, the degree of effectiveness of interventions may vary substantially by geographic area and context, (Bliss and Breen, 2009) therefore they should be evaluated whenever and wherever they are implemented.

In light of these problems, it is important to identify tools and methods that can contribute to the continuous monitoring and evaluation of the effectiveness of traffic safety intervention policies in the short and long term, allowing for adjustments in such interventions and helping to select those that lead to more sustainable changes. If possible, the assessment should focus on more specific outcomes than the overall rate of traffic injuries, for example, for interventions focused on reducing AID, the outcome should be the rate of alcohol-related traffic fatalities.

In this paper we propose a new assessment strategy to complement the conventional pre-/post-intervention studies. It is based on the use of joinpoint trend analysis to characterize significant changes in the time trends of road traffic fatalities stratified by levels of a specific risk factor at the time of death – for example BAC – and observing how these changes match in time with the implementation of traffic regulations and interventions. This strategy would provide insight into the differential impact of policies and interventions on the major risk factors, such as AID or excessive speed.

Although alcohol may contribute to fatal traffic injuries in people other than those who drive under the influence of this substance, such as passengers or pedestrians, often – as in this work – data on these individuals are not available or the figures are very small and highly affected by random fluctuations, as with pedestrian fatalities, so it is convenient to focus on fatal injuries in drivers. The specific study aim is to show the usefulness of joinpoint trend analysis for assessing changes in traffic safety policies by analyzing driver death rates from traffic injury (DDTI) stratified by alcohol involvement in Spain during 2001–2011.