5.1. Key Findings
The findings presented above show that road transport has a marked effect on public health in New Zealand, accounting for a net annual toll of ca. 17,815 years of life lost, and an estimated 24,736 years of healthy life lost (DALYs). In total, this represents at least 2.1 per cent of deaths in the country, and 3.3 per cent of the total years of life lost in 2012.
Road transport injuries, physical activity and traffic-related air pollution comprise the main agents of impact. On the basis of the data available here, road traffic noise appears to be of lesser importance, though our assessment may underestimate these effects.
In terms of mode, cars are responsible for about 52 per cent of the overall health burden, HGVs 21 per cent, LGVs 19 per cent, motorcycles 8 per cent and buses 1 per cent. Together, these demonstrate that HGVs and motorcycles make a disproportionate contribution to the burden of disease. Motorcyclists are also by far the most at-risk group, with death rates some 35 times those for car occupants, both per kilometre and per hour travelled. Cyclists, also, have a somewhat raised risk, for deaths twice that of car drivers and pedestrians, and for injuries (and DALYs) more than six-fold higher.
5.2. International Comparisons
In order both to evaluate the credibility of our results, and to put them into a wider context, the estimates of attributable disease burden from our study can be compared with those from other countries. In the case of traffic accidents, this is straightforward, because road safety data are routinely collected by most countries according to established, international protocols. For traffic noise, air pollution and physical activity, comparisons are much more difficult due to the limited number of published studies, and differences in the assessment dates, data quality and methodology of the studies that have been done. For all comparisons except physical activity, data are compared as DALYs per million people, in order to allow for differences in population size.
In terms of road accidents, it is evident that New Zealand performs rather poorly compared to other developed countries. In an environmental burden of disease study that closely mirrors our own, Kjellström
et al. [
43] produced estimates for Sweden in 2001 that translate to a figure of 2790 DALYs per million, considerably lower than ours (4752 DALYs per 1 million, for 2012). In the Netherlands, Knol and Staatsen [
68] estimated that road traffic accidents were responsible for 5000 DALYs per 1 million people in the year 2000. Given that road safety has generally improved in recent years, the burden of disease attributable to traffic accidents has probably declined in both countries, and is now likely to be lower than that in New Zealand. This is supported by data published by the International Transport Forum [
16]. In 2010, for example, New Zealand had the ninth highest road death rate from accidents (8.6 deaths per 100,000 people) out of 32 OECD member states. Expressed in terms of distance travelled, it was the third worst of 22 reporting countries (9.4 deaths per billion vehicle kilometres travelled). Data from the 2010 Global Burden of Disease Study [
69] likewise show that New Zealand ranks amongst the worst of the developed countries, with rates of death from road vehicle injuries similar to those of many East European countries, Egypt, India and the USA.
For noise, data are much sparser and almost all the available studies relate to European countries. Knol and Staatsen [
68] estimated a total of
ca. 1000 DALYs per 1 million people for the Netherlands in 2000, some five times our equivalent for New Zealand (205 per million). For Switzerland, Vienneau
et al. [
49] give a rate of 603 DALYs per million people for cardiovascular mortality due to road traffic noise in 2010, while for Sweden Kjellström
et al. [
43] reported a figure of 457 DALYs per 1 million people due to hypertension and ischaemic heart disease. In a six country study (covering Belgium, Finland, France, Germany, Italy and the Netherlands) by Hänninen
et al. [
70], estimates of the disease burden from traffic noise in 2004 ranged from 371 DALYs per million in Finland to 1483 per million in France. Stassen
et al. [
71] estimated a burden of 3420 DALYs per million people for traffic noise in Flanders, Belgium, for the year 2004, some seventeen times our estimates, though this analysis included a wider range of morbidity effects, including sleep loss, annoyance and hypertension, for all of which exposure-effect relationships are relatively uncertain.
While many studies have estimated the health impacts of air pollution at national level (e.g., [
68,
70]), relatively few have attempted to separate out the contribution from road traffic. In Sweden, however, Kjellström
et al. [
43] estimated a total of 3970 DALYs per million people from PM
2.5 and NO
2 for 2001, four times our figure for New Zealand (995 DALYs per 1 million). In Switzerland, Vienneau
et al. [
49] used dispersion modelling to estimate exposures to traffic-related PM
10. This implied a total burden of 1538 DALYs per 1 million people, about twice our equivalent for New Zealand (766 per million). An estimate can also be derived for Australia from the study of the health impacts of road traffic pollution by the Bureau of Transport and Regional Services [
72]. This yielded a total of 900–2000 deaths due to motor vehicle-related air pollution in the year 2000. Using a base population of 19.2 million, and assuming twelve years of life lost for every premature death, this gives 552–1248 DALYs per million, a range that encompasses our own for New Zealand.
Although a large number of exploratory studies and prospective assessments have been done, comparable analysis of the health impact of physical activity associated with prevailing road transport use in real-world settings are sparse. However, relatively large savings in the burden of disease, especially from ischaemic heart disease and stroke, have been claimed for a switch from car to cycling use, in prospective assessments of road transport policies. Using the HEAT model, for example, Lindsay
et al. [
20] estimated a saving of 20.5 deaths annually through increased physical activity if 1 per cent of short car trips (
ca. 45 million km) in urban New Zealand were switched to cycling, and 116.5 deaths from a 5 per cent switch (223 million km). This represents 1 death saved per 2.2 million cycling km and 1 per 1.9 million km, respectively. In Adelaide, using a comparative risk assessment approach similar to our own, but including a wider range of health outcomes over the age range 15+, Xia
et al. [
73] estimated a saving of 1 death per 4 million cycling km for a scenario yielding a 5 per cent switch to cycling and 1 per 3.75 million km for a 10 per cent switch. In sensitivity analyses, however, the number of deaths saved more than halved when the 70+ age range was excluded (implying a saving of approximately 1 death per 8 million km). In Barcelona, Rojas-Rueda
et al. [
74] estimated a saving of 12.46 deaths as a result of the introduction of a bicycle sharing scheme, which attracted 25,426 new users, cycling a total of 45 million km annually. This translates to 1 death saved per 3.5 million km cycled. In our study, by comparison, 40 deaths are estimated to have been avoided from 346,000 cyclists and walkers, travelling approximately 173 million km—equivalent to 1 death saved per 8650 people using active transport, and 1 per 4.3 million km walked or cycled.
These comparisons show distinct differences between our results in New Zealand and those from other studies and countries. In terms of road accidents, the differences are small, but the attributable burden in New Zealand is relatively high. In terms of air pollution and noise (and perhaps physical activity), the differences are larger, and the estimated impacts lower in New Zealand than the comparison countries.
These differences are no doubt due in part to differences in methodology between those used in this rapid assessment and those applied in more detailed studies. For example, our study used relatively simple modelling techniques for air pollution and noise, rather than the dispersion or geostatistical models used in many other assessments. Monitoring data in New Zealand are also sparser than those available to many other studies. Likewise, in assessing the effects of active transport, we used a simple assumption about changes in activity levels, rather than modelling energy expenditure, as often used in studies of cycling, and only considered regular commuting. Together, therefore, these comparisons confirm our suspicion that, though our results are of the correct order of magnitude, they tend to under-estimate the true health impact of road traffic in New Zealand, and should therefore be interpreted as low-end estimates of the attributable health burden.
Nevertheless, some of the differences may be real and reflect New Zealand’s specific geography. Compared to most of the other countries for which data are available, for example, New Zealand has a small and sparsely distributed populated. While traffic volumes are relatively high on a per capita basis, therefore, both traffic and population densities (per km2) are low, and relatively few people live close to busy roads. The potential for exposure to air pollution and traffic noise is, therefore, less. In the context of these differences, the relatively large burden in New Zealand from road traffic accidents presents an anomaly in our results. Given that data in this area are internationally consistent, this cannot be dismissed as an artefact of the methodology. Instead it implies something particular about New Zealand roads, vehicles or driving behaviour that raises accident risks above those of most other developed countries.
5.3. Policy Implications
Notwithstanding their inherent uncertainties, the results of this assessment clearly have implications for policy on transport and health in New Zealand. They not only show that road traffic creates a substantial public health impact, but also that road traffic accidents make a major contribution to the burden of injury and disease. Within this burden, motorcyclists and, to a somewhat lesser extent, pedal cyclists face particularly high risks. The results also suggest that heavy goods vehicles bear a disproportionate source of responsibility for the associated health impacts and deserve special attention.
The integrated framework used in this assessment also indicates the close interdependence of these health impacts, and hints at the direction in which policy needs to move. Piecemeal solutions, aimed at individual elements of the system, are rarely likely to be effective in totality. In some cases, they may even be counter-productive by transferring risks from one area to another, or by increasing the impacts of one pathway in order to reduce those from another. In the face of a growing population and rapid urban development (especially in the Auckland area), the real need is therefore for more integrated actions, which work coherently across the system.
The best way of achieving this is likely to be through interventions that shift a substantial proportion of the population towards walking, cycling and public transport, and away from reliance on car use. This would not only reduce emissions of air pollution and noise but also help to raise levels of physical activity, and thereby save further lives. The opportunity benefits of such policies have already been demonstrated by Lindsay
et al. [
20]. They suggested that a 5 per cent reduction in vehicle kilometres as a result of moving short trips (<7 km) from car to cycling could save as many as 117 deaths annually, as a combined result of reduced air pollution and increased levels of physical activity—and allowing for a small increase in deaths of cyclists in road accidents. Large savings in hospital admissions and restricted activity days due to injuries and pollution-induced illness were also indicated. Notably, the benefits in terms of reduced air pollution were small; the main health savings come from changes in physical activity. It is also apparent that additional benefits could be achieved by encouraging people to take public transport, as this usually involves some walking to access the system. Since New Zealand has a relatively high proportion (30 per cent) of people defined as obese [
55], and tackling obesity is a major policy goal, such changes would have important social, as well as health consequences, far beyond the transport sphere.
5.4. Developing Environmental Health Indicators
Rapid assessments such as the one done here are not only useful in highlighting priorities and opportunities for policy action. They also help to show the information that is needed to support such actions in the longer term. Environmental health indicators are an important tool in this context [
75,
76], and over recent years a number of environmental health indicator sets have been established, both nationally and internationally [
77]. To be truly effective, indicators need to be both comprehensive and balanced, in order to avoid biasing policy towards specific areas. If they are to guide policy in ways that minimise adverse health impacts, they also need to be focused on, and give direct estimates of, the scale of health effects. There is, however, little evidence that any of the indicator sets developed to date really match these requirements; most seem to be somewhat arbitrary, and driven largely by practical constraints of data availability and past or prevailing policy concerns. In the road transport domain, for example, indicator sets tend to focus on source activity (e.g., vehicle numbers, flows) and environmental pressures and proximal effects (e.g., concentrations of air pollutants) rather than health impacts. Where health outcome indicators have been included, it has often been in the form only of traffic accidents, or general measures such as overall respiratory mortality. The European Environment Agency [
78] lists 17 transport-related indicators, of which five relate to atmospheric emissions and air pollution, and one to wellbeing (exposure to traffic noise). In New Zealand, the Ministry of Transport [
79] maintains
ca. 175 indicators, but of these only three are categorised as referring to public health: annual nitrogen dioxide concentrations, Auckland light vehicle emissions and noise measurements—and the last of these is void. In neither case, therefore, are any measures of health outcome reported. Innovatively, however, the recently published Air Domain Report for New Zealand [
19] does include an indicator on estimated health impacts from human-made PM
10.
The analysis done here shows the way in which a more rigorous, health-focused approach to indicator development can be achieved, by linking them to an integrated health impact assessment. On this basis, the conceptual model used to define the assessment (
Figure 1) gives a framework within which to identify the indicators that are needed. The results likewise help to interpret the indicators, by showing the overall health burden that they represent, and the relative contribution attributable to each pathway and source.
This approach may seem daunting. In the case of road transport in New Zealand, for example, it implies indicators for each of the boxes in
Figure 1—far beyond the small collection of “core” indicators often anticipated by policy-makers. The benefits, however, are substantial, for it ensures that the indicators provide a complete and balanced description of the system as a whole, and helps to reduce biases in decision-making. By structuring the indicators within a coherent framework, it also makes it easy to aggregate the information to a smaller number of “headline” indicators (e.g., total health burden from road transport) without distorting the information, or equally to drill down from any higher level of indicator to reveal the underlying causes and impacts.
Like other forms of information, however, indicators require data, and this analysis has also shown some of the problems that exist in this respect. Much of the monitoring needed to carry out the assessment, and to track health impacts from road transport, are currently deficient. Examples include the lack of measured data on ambient traffic noise, and the sparse and inconsistent coverage provided by the air pollution monitoring networks (including the lack of data on PM
2.5). Opportunities for modelling, too, are limited by lack of relevant input data. In the case of noise, for example, modelling could be used to produce noise maps like those available for large urban areas in Europe, under the Environmental Noise Directive [
80]. Similarly, dispersion modelling could be used to estimate exposures to air pollution. To allow this improvements are needed in the availability of georeferenced data, especially on traffic flows, building heights and emissions.
The issue of georeferencing also deserves emphasis. Geographically aggregated (e.g., national) indicators are rarely sufficient as a basis for assessing and tracking health impacts, as marked variations in exposure and risk typically occur across the community, often at a local level. Without information on where risks occur, it is impossible to identify who is at risk and thus to allow for the social factors that may be involved, either in assessing impacts or in developing effective policy responses. For example, evidence suggests that areas of higher socioeconomic deprivation bear the brunt of many of the impacts considered here—the so-called environmental justice effect [
81]. Many of these effects are currently difficult to quantify because of a lack of suitably localised (or population-specific) data on the risk factors of concern.
In summary, a major constraint on indicator reporting and policy support in relation to road transport and health in New Zealand is the unavailability or limited quality of the relevant data. As is often argued, jurisdictions measure what they manage, and manage what they measure. The effect is to reinforce existing policy concerns and risk neglect of new and emerging issues. If policy is to be proactive and preventative, as is needed to avoid unwanted public health impacts, then it must have access to much more insightful information than is currently the norm—and this requires an underpinning of reliable, routinely monitored data.
Herein we see the value of the type of rapid assessment done here. Whilst completed quickly and at low cost (the total staff time for conceptualisation, data collation and analysis was ca. 300 hours), it provides a first approximation of the health impacts, attributed to source, which of itself can inform the policy debate. It provides, therefore, information for action in a timely manner, and gives a framework within which to identify the information needed for continuing policy support, and the indicators that should be developed for this purpose.