Official Statistics

Linking police and fire service data on road collisions: an initial feasibility study

Published 28 July 2022

About this release

This document outlines work to establish the feasibility of linking police data on reported road casualties (collected via the STATS19 system) with data from the Incident Recording System (IRS) on road traffic collisions (RTCs) attended by fire and rescue services (FRSs).

This is the first phase of planned work to explore the potential of IRS data to add value to STATS19 data, by providing more detailed information on the type of RTCs that require fire service attendance.

The department welcomes comments on the methodology outlined, or ideas for maximising the value of the resulting linked dataset, via email to the road safety statistics team.

Main findings

This study establishes that a high linkage rate between STATS19 and IRS records can be achieved. Analysis on this linked dataset can be conducted to further understand the consequences of different RTCs on FRS activity.

Initial findings show:

• in 2020, 71.1% of IRS incidents were matched with a STATS19 accident record. Similarly, 72.7% of IRS incidents were matched in 2019
• matching rates vary between incident type, time of day, and fire and rescue territory. More serious fire and rescue incidents appear to have a higher matching rate
• the number of RTCs allocated certain fire and rescue incident types varies by RTC severity and road speed limit

Planned next steps include:

• continued development of the linkage methodology, including review of unlinked cases to explore if genuine matches are being missed or cases are outside the scope of STATS19
• improving the quality of linking by obtaining further variables such as age and sex of casualties involved in incidents (subject to agreement from the Home Office)
• further analysis to illustrate the usefulness of the linked dataset
• engagement with road safety stakeholders to assess potential wider use of the data outside Department for Transport (DfT)

1. Introduction

1.1 Background

STATS19

DfT collects and analyses data on road safety for Great Britain via STATS19 – a database comprising information on all personal injury RTCs occurring on public highways and reported to the police within 30 days.

IRS

The Incident Recording System (IRS) is a record of all incidents attended by Fire and Rescue Services (FRSs) in England maintained by the Home Office fire statistics team. DfT are linking STATS19 with a subset of the non-fire IRS incidents defined as ‘RTCs’. This category does not include road vehicle fires. The Home Office only has jurisdiction for fire policy in England and therefore data is for England only.

1.2 Aims and Objectives

When a RTC occurs, the fire service may attend the scene. DfT received an extract from the Home Office fire incident data in June 2021, to link with the DfT statistical dataset of RTCs reported to the police (STATS19).

The aim of linking to the data together was to learn more about the incidents that required the fire service.

2. Data for linking

2.1 STATS19 data

STATS19 data

STATS19 data is used as the basis for the department’s published road safety statistics, with an explanation of the data collection set out in detail, including in the STATS20 guide for reporting. The majority of the data is made available as an open dataset, with a small number of sensitive variables available on request for specified research purposes, including casualty home postcode.

IRS data

The department was provided a data extract of all recorded RTC incidents attended by fire and rescue serviced in England between year-ending March 2011 and year-ending March 2020. This comprised of 291,107 entries.

The dataset included the following information:

• the Fire and Rescue Service territory (area) in which the road vehicle fire took place
• time and date of the call to the fire and rescue service
• co-ordinates of where the road vehicle fire took place
• the type of location that the road vehicle fire occurred in
• the type of road vehicle fire incident (the action of the fire and rescue service at the scene)
• the number of fire service vehicles that attended the road vehicle fire
• the number of fire crew and officers that attended the road vehicle fire
• the time between the call being made and the first fire service vehicle attending the scene
• the time between the first fire service vehicle attending the scene and the incident being closed
• whether the incident involved a fatality or casualty

An incident is closed when the last FRS appliance, vehicle or officer left the scene of the incident. Both of these variables are usually collected through automatic systems.

More information can be found in the road vehicle fire dataset guidance published by the Home Office. Details of the variables provided are listed in the annex.

3. Linking methodology

3.1 Overview

Neither the IRS extract nor the STATS19 data contain a unique identifier that would allow for a direct and unambiguous linking of RTC records across the two datasets. Instead, an alternative strategy was required to identify as many high-quality matches as possible.

3.2 Linkage variables

In the absence of a common unique identifier, the following variables were used for linkage. Some tolerance was allowed in the degree of agreement on these variables.

The variables relevant for linking were:

• time and date of call (compared to RTC date and time in STATS19)

• X and Y co-ordinates (compared to the Easting and Northing recorded in STATS19)

3.3 Linkage approach

Data subset

A subset of the data was made based on filters on the following variables:

• year – the initial matching and analysis was focused on 2019 and 2020 records only

• whether the incident involved a fatality or casualty (STATS19 refers to RTCs with at least one casualty)

• location type – the initial matching and analysis included only ‘road vehicles’ to exclude any incidents that occurred when the vehicle was off-road

Location plotting

The co-ordinates of the IRS records were plotted on a map to check that they were in Great Britain. For 2020, this was true for all but 3 incidents which were located in the North Sea (the data recorded these as being in the Derbyshire territory). For comparison, plotting was conducted for 2019 incidents and there were a similar number of unexpected locations. Because there were only a small number of these incidents, no further investigation was conducted.

Matching process

A dataset of potential matches was created where all records from the IRS dataset were matched to all RTCs in STATS19 which occurred within the same day in 2020. The difference for both the location and time variables between the IRS and STATS19 dataset was calculated for every potential match.

To exclude unlikely matches, only linked records with the following criteria were classed as possible matches:

• a difference in distance of under 5 miles
• a time difference of under 90 minutes

No adjustments were made to account for incidents which occurred near midnight potentially matching with a STATS19 record for the following or previous date – this is a limitation of the approach but is not considered to have a large impact on the results.

3.4 Definition of ‘best’ matches

Where an IRS incident matched onto several STATS19 records, the ‘best’ matches were defined as the records with the closest distance.

4. Matching results

For 2020, there were 7,090 fire incidents which matched with a RTC record within the matching tolerance and were the closest match on both time and distance. A further 308 fire incidents met the filtering requirements but did not have a linked RTC which was the closest match on time and distance. For these 308 incidents, the best match was assumed to be the match with the smallest difference in distance. Taking these 7,398 ‘matched’ incidents, this gave an overall 71.1% matching rate for 2020.

In 2019, there were 9,395 fire incidents which matched with a RTC record within the matching tolerance and were the closest match on both time and distance. This is a slightly higher number than in 2020, which is as expected due to the reduction in RTCs in 2020 associated with the COVID-19 pandemic. A further 526 fire incidents met the matching requirements but did not have a linked RTC which was the closest match on time and distance. As for 2020 data, the best match was determined by the match with the smallest difference in distance, which provided a total of 9,925 ‘matched’ incidents and a matching rate of 72.7%.

Best matches

The average time difference between the fire and RTC record was 13.6 minutes in 2020 and 14.1 minutes in 2019. The average distance difference was 0.4 miles in 2020 and 0.5 miles in 2019. The charts below show the distribution of best matches on the difference in distance and time between the fire incident and STATS19 data for 2020. The same charts were produced for 2019 matches and showed a very similar distribution.

Chart 1: Number of best matches in distance (miles) between fire incident and STATS19 data (best matches only) (2020)

Chart 2: Number of best matches by difference in time (minutes) between fire incident and STATS19 data (2020)

The matching process relies on the fact that there are only a small number of RTCs in a certain area in given day, which would reduce the likelihood of incorrect matches (that is, 2 non-matching records being linked together). There has not yet been validation of the matches, so that it is possible the above includes some incorrect matches. However, this is an area of planned further work.

5. Initial analysis of best matched records

Initial analysis explores the association between matching rates and available variables, and relates to best matches only (those that are the closest match on time and distance).

5.1 Type of incident

The matching rate is generally higher for more serious categories, for example ‘extrication of person’.

Table 1 shows the matching rate for different incident categories for 2020 (data for 2019 shows a very similar pattern so is not shown).

Table 1: Percentage of matched IRS records and percentage point difference between matching rate and all records matching rate by type of incident (2020)

Type of incident	Number of incidents	Number of matched incidents	% matched to STATS19	% point difference between type of incident matching rate and average matching rate for all categories
Extrication of person(s)	3016	2405	79.7	8.65
Wash down road	42	31	73.8	2.75
Medical assistance only	801	590	73.7	2.65
Release of person(s)	1099	808	73.5	2.45
Make scene safe	2207	1526	69.1	-1.95
Other	271	182	67.2	-3.85
Make vehicle safe	2736	1715	62.7	-8.35
Stand by - no action	168	101	60.1	-10.95
Advice only	72	40	55.6	-15.45

5.2 Time of day

Chart 3 shows slightly lower matching rates at the beginning and end of the day for 2020 records. Again, a similar pattern was demonstrated within the 2019 records. This is expected as only records with the same day in STATS19 and IRS were matched (for example a record with an IRS time of 11.55pm and STATS19 time of 12.05am the following day would not match despite having 10 minutes difference due to the chosen matching method). Aside from this and slightly lower matching rates in the early hours of the morning, there does not appear to be a pattern in matching rates by hour. This could be explored further in a follow up to this initial analysis.

Chart 3: Percentage of matched IRS records and percentage point difference between matching rate and all hours matching rate by hour (2020)

5.3 Fire and rescue territory

The proportion of fire incident records matched varies considerably by FRS area. The reason for this is currently unknown, however it is an area of interest for future analysis. Table 2 shows the matching rate by FRS area for 2020, which showed a similar pattern to 2019. The 5 FRS areas with the highest matching rates were the same in both 2019 and 2020.

Table 2: Percentage of matched IRS records and percentage point difference between matching rate and all territory matching rate by fire and rescue territory (2020)

FRS territory	Number of fire incidents	Number of matched fire incidents	% matched to STATS19	% point difference between type of incident matching rate and average matching rate for all categories
Greater London	836	763	91.3	20.3
East Sussex	119	108	90.8	19.8
West Sussex	151	137	90.7	19.7
Isle Of Wight	27	24	88.9	17.9
Surrey	266	235	88.3	17.3
Hampshire	317	267	84.2	13.2
Cumbria	69	57	82.6	11.5
Kent	316	259	82.0	11.0
Cornwall	99	80	80.8	9.8
Cheshire	127	102	80.3	9.3
Lincolnshire	234	188	80.3	9.3
South Yorkshire	175	135	77.1	6.0
Bedfordshire	130	99	76.2	5.2
Nottinghamshire	188	141	75.0	4.0
Cambridgeshire	156	116	74.4	3.4
Northamptonshire	166	123	74.1	3.0
North Yorkshire	112	82	73.2	2.2
Humberside	128	93	72.7	1.7
West Yorkshire	255	185	72.5	1.5
Hertfordshire	232	168	72.4	1.4
Essex	346	249	72.0	1.0
Gloucestershire	99	71	71.7	0.7
Lancashire	324	231	71.3	0.3
Dorset and Wiltshire	218	155	71.1	0.0
Warwickshire	129	91	70.5	-0.5
Norfolk	185	130	70.3	-0.8
Suffolk	132	92	69.7	-1.3
Merseyside	258	179	69.4	-1.6
Tyne and Wear	131	90	68.7	-2.3
Cleveland	83	56	67.5	-3.5
West Midlands	1235	822	66.6	-4.5
Derbyshire	267	176	65.9	-5.1
Leicestershire	309	198	64.1	-7.0
Northumberland	77	49	63.6	-7.4
Shropshire	128	80	62.5	-8.5
Buckinghamshire	270	168	62.2	-8.8
Hereford and Worcester	241	150	62.2	-8.8
Oxfordshire	179	110	61.5	-9.5
Berkshire	145	89	61.4	-9.6
Durham	123	75	61.0	-10.0
Devon and Somerset	345	209	60.6	-10.4
Avon	140	81	57.9	-13.2
Greater Manchester	687	377	54.9	-16.2
Staffordshire	258	108	41.9	-29.2

5.4 Comparison of matched and unmatched incidents: 2020

An initial comparison was conducted on fire incident details between (best) matched and unmatched incidents for 2020. Table 3 shows that compared to unmatched data, records in the 2020 matched data had a:

• slightly higher number of vehicles attending the scene
• higher number of personnel attending the scene
• longer time at the scene
• shorter total response time

Table 3: FRS response average in matched data and unmatched data by variable (2020)

Variable	Average in matched data	Average in unmatched data
Vehicle   count	2.6	2.3
Total   personnel	9.3	8.5
Time at   scene	73.7	61.6
Total   response time	9.3	9.5

This suggests the records matching to STATS19 are the more serious incidents, which is further supported by table 4. However, the majority of incidents attended by the fire service do not involve serious injury as defined in STATS19.

Table 4: Percentage of collisions of each severity for IRS best matched data and for all STATS19 data (2020)

Dataset	Fatal	Serious	Slight
Matched   IRS data (best matches)	3.9	33.9	62.2
STATS19   only	1.5	20.1	78.3

6. Initial illustrative analysis of linked dataset

A small number of initial analyses were conducted on the 7,398 best matches between the IRS and STATS19 within 2020. These are examples of the types of analyses that could be conducted in the future, to widen the understanding of the consequence of RTCs and the response of the fire and rescue services.

Table 5 shows the percentage of RTCs that were fatal, serious and slight in STATS19, for each different type of fire and rescue incident. Extrication of person(s) had the highest percentage of fatal and serious incidents (7% and 43% respectively. There were no matched incidents involving fatalities requiring advice only or washing down the road, with advice only having the greatest proportion of incidents categorised as slight (78%).

Table 5: Percentage of fatal, serious and slight RTCs for each incident type (2020)

Type of incident	Number of incidents, all severities	Fatal (%)	Serious (%)	Slight (%)
Extrication of person(s)	2405	7	43	50
Medical assistance only	590	6	38	56
Other	182	5	30	65
Release of person(s)	808	4	32	64
Make scene safe	1526	2	30	68
Stand by - no action	101	2	37	61
Make vehicle safe	1715	1	24	75
Advice only	40	0	23	78
Wash down road	31	0	35	65

Table 6 shows the percentage of RTCs occurring on roads of different speed limits, between different types of fire and rescue incidents. For all incident types, the largest percentage of RTCs occurred on roads with a speed limit of 30mph, generally followed by 60mph. Incidents requiring extrication of person(s) had a particularly high proportion occurring on 60mph roads and a lower proportion occurring on 21 to 30mph roads compared to other incident types.

Table 6: Percentage of RTCs by speed limit (STATS19) and incident type (IRS) (2020)

Type of incident	Number of incidents, all speed limits	20 mph (%)	30 mph (%)	40 mph (%)	50 mph (%)	60 mph (%)	70 mph (%)
Advice only	40	5	63	15	0	8	10
Extrication of person(s)	2405	2	36	12	9	30	11
Make scene safe	1526	6	49	12	6	16	11
Make vehicle safe	1715	4	58	10	5	14	8
Medical assistance only	590	6	50	11	5	19	9
Other	182	7	63	5	4	13	7
Release of person(s)	808	4	43	9	8	24	12
Stand by - no action	101	4	45	12	3	22	15
Wash down road	31	13	58	6	6	13	3

As noted above, this is a simple illustrative analysis and further analysis of the linked dataset would be possible, including any of the variables captured in STATS19. This could help to learn more about the incidents that required the fire service, such as:

• are there certain vehicle modes or road types that are more likely to involve the fire service?
• are there any trends in the factors associated with RTCs involving the fire service?
• what is the average duration of the fire service involvement in different types of collision?

7. Next steps

This work has established that linking the two datasets is feasible, and as a result further development of the approach is likely to be worthwhile subject to sufficient user interest. The proposed next steps, subject to agreement from the Home Office, include:

• continued development of the linkage methodology, including review of unlinked cases to explore if genuine matches are being missed or cases are outside the scope of STATS19
• improving the quality of linking by obtaining variables such as age and sex of casualty, and time that the fire service arrived at scene
• further analysis to illustrate the usefulness of the linked dataset
• engagement with road safety stakeholders to assess potential wider use of the data outside DfT

Feedback

We welcome further feedback on any aspects of the department’s road safety statistics including content, timing, and format, via email to the road safety statistics team.

Annex: IRS variables

Variable name	Variable description
PUB_INCIDENT_ID	Unique incident identifier
territory_frs	Fire and rescue territory where the incident took place
CALENDAR_YEAR	Year of incident
AT_CALL_TIME_OF_CALL	Time and date of call to fire and rescue service
property_type_t0102	The type of location the incident occurred in
X_COORDINATE	Co-ordinates where the incident took place
Y_COORDINATE	Co-ordinates where the incident took place
fire0902_category_detailed_d	The type of incident (action of fire and rescues service at the scene)
vehicle_count_d	The number of fire service vehicles that attended the incident
total personnel	The number of fire crew and officers that attended the incident
total_response_time_d	The time between the call being made and the first fire service vehicle attending the scene
time_at_scene_d	The time between the first fire service vehicle attending the scene and the incident being closed
fatality_casualty_d	Whether the incident involved a fatality or casualty

Instructions for printing and saving

Depending on which browser you use and the type of device you use (such as a mobile or laptop) these instructions may vary.

How to search

Select Ctrl and F on a Windows laptop or Command and F on a Mac

This will open a search box in the top right-hand corner of the page. Type the word you are looking for in the search bar and press enter.

Your browser will highlight the word, usually in yellow, wherever it appears on the page. Press enter to move to the next place it appears.

Tablets and mobile devices normally have the option to “find in text” and “print or save” in their sharing or quick options menu of their browser, but this will vary by device model.

Contact details

Road safety statistics

Email roadacc.stats@dft.gov.uk

Public enquiries 020 7944 6595

Media enquiries 020 7944 3021