Summary

Globally, motor vehicle crashes are the eighth-leading cause of death for all ages and the leading cause of death for people ages 15 to 29 (World Health Organization [WHO], 2013). In recognition of this fact, in 2010, WHO called for a Decade of Action for Road Safety 2011–2020. The problem is just as acute in the United States, where, in 2013, more than 32,700 people were killed and approximately 2.3 million were injured in motor vehicle crashes (National Highway Traffic Safety Administration [NHTSA], 2015). For 2010, the economic costs associated with motor vehicle crashes are substantial, estimated to be $242 billion in 2010 (Blincoe, Miller, et al., 2015). Fortunately, a wide range of evidence-based interventions, including both policies and programs, can help prevent motor vehicle–related injuries and deaths.

This report contains the technical documentation for the development of an online tool for decision makers—primarily state health, transportation, and safety officials—to use in determining the costs and effectiveness of various interventions to reduce injuries and deaths from motor vehicle crashes and in determining what interventions together generate the largest reductions in injuries and deaths for a given implementation budget. In addition to helping with intervention selection, this report contains detailed cost categories and data, which can be useful to state planners in determining the types and amounts of costs involved in the implementation of selected interventions. RAND experts developed the online tool for the National Center for Injury Prevention and Control at the Centers for Disease Control and Prevention (CDC). The tool is available at www.cdc.gov/motorvehiclesafety/calculator.

The impetus for the project was that, although much information has been collected and developed over many years about the effectiveness and costs of various interventions at locations where interventions have been implemented, it has been difficult to develop information systematically about the comparative effectiveness and costs for an individual state. Although states are major actors in the realm of motor vehicle policy and safety, it has been difficult to assess which interventions are the most cost-effective for a given budget and how the effectiveness compares with the costs. The project addresses this concern by collecting a wide variety of information in one report and allowing users to look at costs and effectiveness scaled to their states.

The tool does not contain information on every possible intervention.¹ Rather, the tool contains interventions that have not yet been widely implemented across states, so as to focus on those interventions that can yield the largest benefit. Our selection was based on four criteria. Interventions had to be

• intended to change driver or passenger behavior (as opposed to changes to roadway or vehicle engineering)
• implementable at the state level (or affected by state policy)
• demonstrated to be highly effective
• not already in widespread use.

Thus, the goal was to focus on those interventions that would bring the greatest possible effectiveness from implementation and that other states could adopt. We ultimately analyzed 14 interventions that met these criteria.

For each of these 14 interventions, we developed a set of cost components. Our final cost-estimating structure has ten cost components, divided into subcomponents. The cost of each intervention consists of one or more components. We obtained information about costs from a variety of sources, including journal articles; federal, state, and other organization reports; commercial sources; and interviews with state officials and stakeholders. Some costs, such as the cost of staff time for state personnel (e.g., highway patrols and department of motor vehicles [DMV] staff), were scaled to each state based on wage rates provided by the Bureau of Labor Statistics. For other cost components, we developed a “most common cost” that was used across all states, based on the range of costs we documented.

Costs fall into one of four categories: costs borne by the state (such as the cost of police time), costs paid by individuals to the state (such as a seat belt fine), costs borne by offenders but not paid to the state (such as purchase of an alcohol interlock), and costs borne by individuals to comply with the law (such as purchase of a motorcycle helmet). Because the tool considers only costs that directly affect the cost or revenue to the state, costs borne by individuals are not included in the cost calculation to prioritize and select interventions for implementation. To ensure that these costs do not impose an undue burden on the public, before implementing a new intervention, state decision makers might want to know the total costs that people will bear but not pay to the state for that intervention. Thus, in addition to detailed assembly and estimation of costs in the first two categories, we have provided some data on costs in the third and fourth categories.

We also developed estimates of the effectiveness of interventions. Effectiveness is defined as the reduction in injuries and deaths that implementing a particular intervention can create. We developed two types of information: the actual number of anticipated lives saved and injuries prevented and a monetized value of those two things. Our estimates are based on published articles and reports that document empirical studies of the effectiveness of interventions. To create our estimates, we prioritized studies using four criteria: They provided information on the primary outcomes of interest (i.e., crashes, injuries, and deaths), they used rigorous study designs, they were relatively recent, and they assessed interventions in the United States. We relied on meta-analyses and systematic reviews when available to identify studies accepted and cited in the field. Most of the studies, systematic reviews, or meta-analyses that we used in developing these estimates looked at deaths only, so, with one exception, we assumed that injuries were reduced at the same rate.

These reductions were scaled to each state, based on the number of injuries and deaths associated with particular crash causes in 2010 (using the Fatality Analysis Reporting System [FARS] and General Estimates System [GES] for injuries) for that state. This scaling is needed because interventions address particular crash types (for example, we expect a motorcycle helmet law to reduce motorcyclist deaths), and the mix of crash types varies between states (for example, Florida, California, and Texas have higher numbers of motorcycle crashes than other states have).

We used an existing protocol to determine the economic benefits of a life saved or an injury prevented. This was developed by Blincoe, Miller, et al., 2015, specifically to monetize the benefits from reducing crashes.² The protocol includes nine cost categories; we scaled three of them at the state level and used national estimates (updated to 2012 dollars) for the other six. This provided us with a monetary value for each life saved or injury prevented specific to each state.

Using these costs and effectiveness or its monetized benefits, we then developed and programmed the online tool using SAS software for compatibility with CDC website requirements. The tool has two modes of analysis:

• Cost-effectiveness analysis assesses the costs and effectiveness of each intervention separately, without accounting for any interdependencies between interventions.
• Portfolio analysis takes interdependencies between interventions into account.

The cost-effectiveness analysis uses the information described above to provide the total annual monetized benefits and annual costs for each separate intervention selected. The attractiveness of an intervention is measured by the ratio of effectiveness to costs. The higher the ratio of effectiveness to costs, the more cost-effective the intervention. The portfolio analysis accounts for three sets of interdependencies, defined by the crash cause:

• alcohol interlock, license plate impoundment, limits on diversion and plea agreements, and vehicle impoundment
• saturation patrols and sobriety checkpoints
• primary enforcement of seat belt laws and seat belt enforcement campaign.

In each case, the new number of injuries and deaths averted is less than the sum of the individual interventions. This effect is easiest to see in an extreme hypothetical example in which two interventions could each reduce deaths by 100 percent. Implementing both interventions would still reduce deaths by a total of 100 percent, not the sum of the individual interventions or 200 percent. In a case in which interdependencies exist among three interventions (alcohol interlock, limits on diversion and plea agreements, and vehicle impoundment), we have provided detailed calculations in this report (see Chapter Five) to show that the implementation of these three interventions would reduce the number of injuries and deaths by 53 percent. In contrast, if interdependencies are ignored, a larger reduction of 65 percent would result. The latter is larger than the former by 24 percent. If license plate impoundment were also implemented, the latter would be larger by 39 percent.

For both types of analysis, there are two ways to conduct model runs:

• Fines included means that the analysis takes into account the revenue that the state receives from those interventions for which offenders pay fines. This revenue is assumed to be available to fund the implementation of the interventions from which the revenue is generated, as well as other interventions.
• Fines excluded means that the analysis excludes any revenue received from fines. Without fines to defray some of the costs, the total cost to the state in implementing the selected interventions can be substantially higher.

In addition, users can conduct sensitivity analysis under the portfolio analysis option. For these analyses, the user can change any of the top-level cost and effectiveness assumptions and see the results.

The tool was developed, programmed, and tested by RAND experts. In accordance with CDC direction, we used .NET and SAS software. Once the tool was completed, we turned it over to be hosted by CDC.

There are many challenges associated with developing the cost and effectiveness estimates. In particular, numerous assumptions are needed to generate these estimates. We note a few examples here. First, the cost-effectiveness estimates reflect assumptions about the level and characteristics (e.g., how much enforcement is done, whether there was a publicity campaign) of implementation of the successful intervention. As a specific example, we estimated the number of cameras for each state’s red-light and speed camera enforcement. Second, the effectiveness estimates from the literature are based on conditions in a specific jurisdiction, which might not reflect the conditions in others. Third, in many cases, there was no evidence for a specific parameter (e.g., the effect that an intervention could have on crash-related injuries), and we had to make assumptions. For example, effectiveness estimates for injuries were not available for most interventions, so, in the absence of more-specific information, we assumed that the reduction was the same as for fatality reductions. Finally, we used data for scaling by states from national databases, which we assumed accurately reflected conditions across states.

We have tried to mitigate this problem in several ways. We have worked to find the best available evidence on which to build the assumptions. We have also described our assumptions and calculations in detail in this report, so the reader can assess the assumptions for him- or herself. Finally, those who disagree with the assumptions can conduct sensitivity analyses with the tool by adjusting model parameters and use that analysis to inform their selection of the most cost-effective interventions.

The estimates provided by the tool are approximations. They are meant to give decision makers a sense of the relative costs and effects of different interventions under consideration. There may be other costs and benefits not captured by the tool that should be considered (e.g., the improved employment or quality of life among people who are deterred from driving while drunk, effects on civil liberties) or political issues that make some interventions more feasible than others. In essence, the estimates are designed to be one category of information in the larger policy debate.

Despite the necessary reliance on assumptions to build the model, we believe that the tool will be of great use to state decision makers. Although information about which interventions are effective has been generally available, this is the first effort to estimate the implementation costs across a broad array of interventions and to translate these costs to the state level according to a specific state’s demographics and traffic crash profile. States need information on both the potential costs and effects of interventions to make informed resource allocation decisions.