Utilizing Crash Data Performance Metrics to Drive Improvement Richie Frederick, Program Manager Florida Department of Highway Safety and Motor Vehicles Thomas Austin, Management Analyst Florida Department of Highway Safety and Motor Vehicles
Establishing Crash Data Performance Metrics Establishing Crash Data Performance Metrics 1
Objectives Identify the need for crash data performance metrics Identify the approach for establishing measures for accuracy and completeness Establishing Crash Data Performance Metrics Using the measures to make improvements Lessons learned and next steps for expanding the program 2
Motoring Environment The Department of Highway Safety and Motor Vehicles is the official custodian for driver license, motor vehicle, and crash record data. In 2016, Florida had the following: 20,148,654 Estimated Population 16,568,874 Licensed Drivers 20,472,415 Vehicles Registered 3,133,609 Citations Issued 50,597 Sworn Law Enforcement Officers 3
Motoring Environment In calendar year 2016, Florida had the following: 711,666 Crashes Reported 254,245 Injuries Reported 3,176 Total Fatalities 10 Approved e Crash Vendors 92% of Crash Reports Received Electronically 4
Background In 2007, Electronic crash reporting started as a pilot project with the Florida Highway Patrol In 2008, the program was rolled out statewide Florida used TRCC grants to make improvements on specific data elements, which proved beneficial 5
Background Florida lacked formal performance measures, which was identified during records assessments Identified the need to create more comprehensive measures to: Track accuracy and completeness over time Identify areas of ambiguity on the report and in the crash manual Determine where to focus program resources Setting goal of 5% improvement per year 6
Background 7
Baselines REPORTING PERIOD TYPE ITEM DESCRIPTION SCORECARD 10/01/2015 09/30/2016 Accuracy Event 93.40% 10/01/2015 09/30/2016 Accuracy Person 97.83% 10/01/2015 09/30/2016 Accuracy Vehicle 97.86% 10/01/2015 09/30/2016 Completeness Event 95.92% 10/01/2015 09/30/2016 Completeness Person 96.44% 10/01/2015 09/30/2016 Completeness Vehicle 97.15% 8
Baseline Goals TYPE 2017 2018 2019 Final 5% Goal 5% Goal 5% Goal Increase A E 93.73% 94.05% 94.35% 0.94% A P 97.93% 98.04% 98.14% 0.31% A V 97.97% 98.07% 98.17% 0.30% C E 96.12% 96.32% 96.50% 0.58% C P 96.62% 96.79% 96.95% 0.51% C V 97.29% 97.42% 97.55% 0.41% 9
Approaching Accuracy Measures 10
Accuracy Approach Evaluated 226 fields on crash form to quantify accuracy Identified relational fields to judge accuracy based on the consistency of the data elements 32 checks for accuracy 11 checks for event page 10 checks for person page 12 checks for vehicle page 11
Accuracy Example If First Harmful Event is 14 Motor Vehicle in Transport, then 12
Accuracy Example the crash report should have at least two vehicle sections 1 2 13
Approaching Completeness Measures 14
Completeness Approach Evaluated 226 fields on crash form to quantify completeness Identified principal and relational fields, which would require values in a corresponding field to judge completeness 53 checks for completeness 19 checks for event page 18 checks for person page 16 checks for vehicle page 15
Completeness Example If Injury Severity is 2 (possible), 3 (nonincapacitating) 4 (incapacitating), 5 (fatal), or 6 (non traffic fatality), then 16
Completeness Example Source of Transport to Medical Facility cannot be null 17
Making Improvements Conducted 11 train the trainer workshops around the state on improving crash reporting, with specific emphasis on data elements identified by the measures Attended by 618 law enforcement officers representing 144 agencies 18
Intersection to Junction If Type of Intersection is 2 (four way intersection), 3 (T intersection), 4 (Yintersection), 5 (traffic circle), 6 (roundabout), or 7 (five point or more) then 19
Intersection to Junction First Harmful Event Relation to Junction cannot be 1 (non junction), 4 (driveway/alley access related), or 5 (railway grade crossing) 58.34% accuracy rate statewide 20
Motorcycle Endorsements If Person is 1 Driver and Vehicle Body Type is 11 Motorcycle then 21
Motorcycle Endorsements Required Endorsements cannot be 3 No Required Endorsement 73.13% accuracy rate statewide 22
Data Visualizations 23
Data Visualizations 24
Data Visualizations 25
Data Visualizations 26
Data Visualizations 27
Data Visualizations 28
Data Visualizations 29
Next Steps Continue the review of state approved data collection software applications to recommend improved data validation capabilities to improve the accuracy, completeness, uniformity, and timeliness of crash data Conducting train the trainer workshops with content tailored to issues identified by performance metrics 30
Next Steps Create a crash uniformity measure comparing the current Florida crash data elements and attributes to the new MMUCC 5 th Edition standard Develop a crash dataset completeness measure to perform trend forecast analysis for the total crashes expected from counties and agencies Develop accuracy and completeness metrics for Uniform Traffic Citations 31
Next Steps 32
Conclusion Performance metrics allow tracking over time and establishing goals to continually improve Compared fields on crash form to quantify accuracy and completeness scores Used measures to create improvements; clearer guidance to reporting officers and tailored training content Areas of focus improved 33
Thank you for participating in this session. Richie C. Frederick, Program Manager Bureau of Records, Crash Data & Record Systems Support Email: richiefrederick@flhsmv.gov Thomas Austin, Management Analyst Bureau of Records, Crash Data & Record Systems Support Email: Thomasaustin@flhsmv.gov 34