Data Sources for Risk Assessment

Introduction

The data sources required for risk assessment are as follows:

  • the number of people who suffered fatal injuries in a given mode of travel per year.
  • the amount of time, or distance travelled, by the total population in that mode of travel per year.

As with all routinely collected national information, it is important to be aware of how the information is collected, and how reliable it is.

Fatal Injuries

STATS19

Fatal injuries in road crashes are reported by the police in the form of what is known as STATS19 data. The STATS19 system will include all people injured in a road crash on a public highway who died within 30 days of the incident. It is highly unlikely that fatally injured casualties will not be known to the police and reported in STATS19. However, pedestrians killed in falls are never recorded as road fatalities, whereas cyclists killed in falls (on a public highway) usually are. STATS19 provides great detail about the circumstances of crashes.

International Classification of Disease

There is a separate record of fatalities provided by coroners’ records. This information can be extracted from Mortality Statistics. The information is coded using a diagnostic tool called International Classification of Disease (ICD), currently in its 10th edition (i.e. ICD-10). (Update: as of May 2019 the  11th edition is available).  A major advantage of ICD data is that they include pedestrians killed in “falls in the highway”, which are not recorded in STATS19. It also distinguishes between cyclists injured off-road, as against on-road falls. Unfortunately, the completion of codes by relevant health institutions leaves a great deal to be desired. An unknown number of pedestrian falls in the highway are not recorded as such even in ICD-10. This means that only limited information about the circumstances of crashes is available from the ICD-10 system.

Discussion and Conclusions

A significant problem is that STATS19 and ICD-10 are not readily compatible.  Misinterpretation of comparisons has led to incorrect conclusions about the reporting of cyclist casualties (read more here [to be drafted]). Considerable technical knowledge of the coding is required to make comparisons between STATS19 and ICD-10. My experience is that such technical knowledge is more or less universally lacking in the authors of official reports.

In summary, the best source of information on fatal injuries is the STATS19 police reported database. However, this system does not include pedestrians killed in falls and may not include all cyclists killed in falls.

Exposure

There are two main sources of information on travel by individuals in Britain:

  • National Road Traffic Survey (NRTS)
  • National Travel Survey

National Road Traffic Survey

These data come from counting passing traffic at numerous points around the road network of Britain. The UK Department for Transport (DfT) operates a national system of 180 automatic traffic counters (ATCs), supported by ten thousand manual counts per year. Local authorities also carry out traffic counts based on both automatic and manual counts. If you want to know what an ATC looks like, this is one in “action”:

Now you will notice them everywhere…

The raw count information is then subjected to what are known as “expansion factors” to correct for the  many potential confounding variables.

The NRTS includes counts of cyclists, but only on the public highway. Cycling on canal tow paths or national cycle routes off-highway will not be included. The DfT is aware that cyclists prefer quiet roads to busy ones. In recent years more effort has been made to count traffic on lesser roads. This means that historic figures may understate to some extent the amount of cycling, although it should be stressed that the same basic principle of 12 hour manual counts has been applied consistently for many decades. Also, cyclists were commoner on main roads (especially rural ones) in the past than is the case today. Therefore, my hunch is that data from the 1970s and 1980s are reasonably compatible with more recent counts.

The NRTS does not include pedestrian traffic.

National Travel Survey

The National Travel Survey (NTS) is also administered by the DfT. Every year, 8,000 households are randomly selected, and the householders (typically about 15,000) are interviewed face-to-face and asked to complete a seven day travel diary recording all of their travel habits. The response rate is 55-65%. Results are adjusted against a standard population to correct for non-responses.

The NTS does include walking. It includes all travel on paved rights of way, including canal tow paths etc. However, it excludes travel for work, by visitors to Britain, and by those in hostels or halls of residence. The basic unit of travel in the NTS is a trip, defined as a one-way course of travel with a single main purpose. A trip consists of at least one stage; a new stage is defined when persons change travel mode.

The great benefit of the NTS is that it provides detailed information on personal travel, including sex, age and social class. The main drawback of the NTS is that it is comparatively recent. Early NTSs in the 1970s and 1980s were occasional, had poor response rates and used methods that have been much developed. It is only over the last fifteen years or so that the NTS has produced a reasonably consistent record.

In recent years, the NTS has reduced its scope to include England only. Equivalent data for Scotland are collected by the Scottish Households Survey.

Conclusions: Comparison of NRTS and NTS For Cycling

Presented here is a comparison of the NTS and NRTS during the years 2002-2016 for England (as noted above, in recent years the NTS has only included England). Granted that the two datasets do not measure exactly the same scope of cycling, it is nonetheless interesting to see that both show similar levels of cycling and a similar trend in recent years:

NTS vs NRTS England

Chart 1: Comparison of National Road Traffic Survey and National Travel Survey data for cycling, 2002-2016, Cycling mode only.