Cost-effectiveness of strategies to prevent road traffic injuries in eastern sub-Saharan Africa and Southeast Asia: new results from WHO-CHOICE

As for the previous WHO-CHOICE analysis, a dynamic system modelled with a Haddon matrix [10] was used as a reference framework for identifying factors that have an impact on RTI. Each cell of the matrix allows opportunities for an intervention to reduce road traffic injuries. Factors in italics are those included in the analysis (see Table 1).

This analysis evaluates 13 individual and combination interventions. They are drawn from recommendations in the the World report on road traffic injury prevention [10] and are mainly focused on pre-event road safety measures, targeting change in human behaviour, due to the availability of robust evidence on their effectiveness and feasibility (see Table 2).

Key parameters in this analysis were the prevalence and distribution of RTIs, both fatal and non-fatal, the prevalence and distribution of risk factors for RTIs, the prevalence, distribution and effectiveness of interventions to reduce RTIs, the population size and mortality rates, and the health state valuations for the long-term sequelae of RTIs.

A literature review to give an overview of published data between 2006 and 2014 on fatal and non-fatal road traffic injuries, their risk factors and sequelae was conducted (see Additional file 1). The attributable fractions are calculated separately for all risk factors at the regional level based on the epidemiological evidence (e.g. exposure rates) from the countries in the region, weighted by population size. Key data on fatal and non-fatal injuries by road user type, sex and age group was provided by the International Injury Research Unit of the Johns Hopkins Bloomberg School of Public Health, which maintains and develops a global database of RTIs. Information collected with the literature review was used in triangulation of the attribution of the RTIs by road user group in combination of the data provided by the Johns Hopkins Bloomberg School of Public Health and the findings of the original literature review that informed the original model creation along with its attribution distribution (Figs. 1, 2, 3).

To measure the independent contribution of different risk factors to overall rates of RTIs in the population, we used the population attributable fraction (PAF), which can be defined as the fraction of incident cases attributable to the risk exposure:

Interventions are at first compared to a hypothetical scenario where the known effects of implemented interventions are removed, referred to as the null scenario. Then the marginal impacts of interventions are evaluated with reference to the null scenario. A multi-state population model [11] was used to estimate scenarios (see Fig. 4). Further details on the methods can be found in [5]. Non-fatal acute injuries of short term duration (e.g. bruises, cuts) were not considered in the analysis.

The same estimates of the effects of interventions as in the previous WHO analysis [5] were used (see Additional file 2). This is due to the fact that during initial literature scoping on the intervention effects in the regions modelled, no papers of suitable focus and/or quality were found to enable updating of the sub-model of the intervention effect estimates in the targeted countries. The estimates used in this analysis of the incidence, prevalence and case fatality rates of RTIs, as well as their associated levels of disability are also shown in Additional file 2. The impact of the selected interventions on population health were evaluated individually, and then as a combination by multiplying the effects of each individual intervention.

Costs of interventions were estimated at the health system level, and include the costs of all market-valued inputs required to deliver the intervention. For example, costs include those of the passage of legislation, the enforcement of legislation and programme management [12]. For “bicycle helmet use” and “motorcycle helmet use” interventions, the costs of equipping bicyclists and motorcyclists with helmets were included, since these costs represent an integral component of those interventions. For the “seatbelt use” intervention, the costs of installing driver and passengers seatbelts in cars not already so equipped were included. Costs are discounted at 3% per annum, assuming a 100 year implementation period. Capital costs are annualized over the lifetime of the asset. All prices are in 2010 International Dollars. 2010 was chosen as the baseline year in line with the 2010 Global Burden of Disease epidemiological data which forms the base of many of the disease models used in WHO-CHOICE. The main costing assumptions are shown in Additional file 2.

The effectiveness of interventions are reported in healthy life years (HLYs) gained due to the specific intervention (Tables 3 and 4).

Because the highest road fatalities are among car drivers and passengers in Southeast Asia (39% of all fatalities, Fig. 1), drink driving legislation and its enforcement via “random breath testing” at roadside checkpoints was found to be the most effective single intervention in this region. The legislation “motorcycle helmet use”, and its enforcement, was found to be the second most effective single intervention; this is consistent with the high proportion of motorcycles in this region and the percentage of road fatalities among this road user group (24%, Fig. 1).

In eastern sub-Saharan Africa, the enforcement of “speed limits” via mobile/handheld cameras at 80% coverage was found to be the most effective single intervention, probably reflecting the fact that pedestrians account for more than 50% of road fatalities among all road user groups in this region (see Fig. 1).

The legislation and enforcement of “bicycle helmet use”, at 80% coverage, was found to be the least effective single intervention in both regions.

Among the combination of interventions, a scenario that combined all five individual interventions was found to be the most effective in both regions.

The total costs estimated for motorcycle helmet use include not only the costs of the passage of legislation and its enforcement but also the costs to the household of purchasing safety equipment, which may explain why this intervention represents the most costly single intervention in both sub-regions. The household cost component is also added to the costs of “seatbelt use” and “bicycle helmet use”; the costs of “seatbelt use” is applied to cars that are not already equipped and “bicycle helmet use” targets only children aged 15 years or less (Tables 3 and 4).

Economies of scope are realised by combining individual interventions due to the synergies that exists between different enforcement strategies.

The cost effectiveness of individual interventions and their combinations are presented in Tables 3 and 4. Cost-effectiveness ratios are reported as costs (in international dollars) per HLY gained.

Among single interventions, “random breath testing”, at 80% coverage, was found to be the most cost-effective intervention in Southeast Asia, whereas in eastern sub-Saharan Africa, it was “speed limits”, at 80% coverage.

Combinations of individual interventions were found to be the most cost-effective: “speed limits + random breath testing + motorcycle helmet use”, at 90% coverage, in Southeast Asia and “seatbelt use + motorcycle helmet use + speed limits + random breath testing”, at 90% coverage, in eastern sub-Saharan Africa.

Figures 5 and 6 show the expansion path a decision maker could follow to achieve the maximum health gain for a given level of expenditure. The expansion path shows the order in which each intervention would be adopted based on its incremental cost-effectiveness ratio, compared to the previously adopted intervention, until no more health gain is possible [4].

Following the expansion path in Fig. 5, in Southeast Asia policymakers would first implement “speed limits + random breath testing + motorcycle helmet use”, at 90% coverage, and when additional resources become available, add “seatbelt use”, at 90% coverage, followed by “bicycle helmet use”, also at 90% coverage.

In eastern sub-Saharan Africa, after “seatbelt use + motorcycle helmet use + speed limits + random breath testing”, at 90% coverage, a policymaker could add “bicycle helmet use”, also at 90% coverage, to maximize health gain (see Fig. 6).