A&I online - Motor Carrier Analysis and Information Resources Online

U.S. Dept. of Transportation, Motor Carrier Safety Advisory Council, Federal Motor Carrier Safety Administration

Link will open a new browser window

Print Version

SSDQ Continuous Improvement

Learn About Methodology

The Methodology to evaluate State safety data was developed by FMCSA to assess the completeness, timeliness, accuracy, and consistency of the State-reported commercial motor vehicle crash and inspection records in the Motor Carrier Management Information System (MCMIS). The evaluation is composed of nine performance measures and one indicator using FMCSA reporting standards to determine a performance rating for each measure. States receive an Overall State Rating of “Good,” “Fair,” or “Poor” each month based upon the reporting standards.

The Non-Fatal Crash Completeness (NFCC) measure is one of the nine measures in the evaluation. The calculation for determining the NFCC measure's rating has been modified. This website describes the modified evaluation process to the NFCC measure that will be incorporated into the SSDQ methodology.

View the current SSDQ methodology

View the What's Changing Interactive Guide

Non-Fatal Crash Completeness Measure: Modified Evaluation Process

All States are responsible for reporting fatal and non-fatal crash records to FMCSA that meet the standard reporting criteria. FMCSA uses these records to support its mission to reduce crashes, injuries, and fatalities involving large trucks and buses on our nation's roadways. Because of the complexity of State crash reporting systems, it is often difficult for States to determine if all reportable non-fatal crash records have been submitted to FMCSA. State-specific evaluations of crash reporting have determined significant underreporting and prompted the need to develop non-fatal crash record benchmarks for each State. The NFCC measure is intended to serve as a guideline to assess whether a State's non-fatal crash reporting falls within an expected range.

The NFCC measure's evaluation process has been refined to estimate the number of non-fatal crash records each State should be reporting to MCMIS. The NFCC measure determines a State rating by comparing the number of State-reported non-fatal crash records in MCMIS to an expected range of non-fatal crash records generated by a statistical model. The input data used by the model, how the data are processed through the model, and the evaluation and final determination of a State's NFCC rating are described in a 4-step approach.

Input State Fatal Data

The measure's algorithm utilizes a relationship between the numbers of fatal crash involvements to the number of non-fatal crash record involvements. State's fatal crash records reported to MCMIS are used as an input value to this measure. This record set represents interstate and intrastate motor carriers and includes large truck and bus vehicle types. The MCMIS fatal crash records cover a 12-month time period that ends six months prior to the MCMIS snapshot date.

Process Data Through a Statistical Model

It is hypothesized that a ratio of fatal crash involvements to non-fatal reportable involvements exists that is independent of any State’s data system and applies across all the States. Since the number of fatal crash involvements is generally well known, it is then possible to predict the number of non-fatal crash involvements that occur in a State. The NFCC measure utilizes a simple log-linear regression model that translates the number of fatal crash involvements in MCMIS to an expected range of non-fatal crash record involvements. The State fatal data is input to the statistical model to generate estimates of non-fatal crash involvements.

Two sets of prediction intervals are generated by the model to provide ranges of non-fatal crash involvements. The prediction intervals are expressed in terms of the probability that the number of reportable non-fatal crash involvements fall within the prediction interval. The model was programmed to output 99% and 90% prediction intervals. The prediction intervals are used to define expected data ranges, and each range has an upper and lower boundary.

The model incorporates an urbanization factor to account for the proportion of rural to urban commercial vehicle travel in a State. The purpose of the adjustment is to account for any variation in the ratio of fatal to non-fatal reportable crash involvements that may be due to variations in the proportion of rural or urban commercial motor vehicle travel in a State.

The document “A New Model of Crash Severities Reportable to the MCMIS Crash File” explains more about the statistical model used by this measure.

Compare State Non-Fatal Data to Model Results

The State-reported number of non-fatal crash records in MCMIS are compared against the expected data ranges output by the model. It is determined where the MCMIS non-fatal crash records fall within the expected data ranges. This determination is used to assign each State a base rating of 'good', 'fair', 'poor', or 'insufficient data'. The base rating assignments are one part of the final rating assignment process.

The diagram below illustrates the rating ranges based on the expected ranges generated from the statistical model. States with few fatal crash records may yield less predictable output and may be assigned an 'insufficient data' rating.

Rating Ranges
Under Reporting			Over Reporting
Poor *	Fair	Good	Fair
< Lower Boundary of 99% PI	>= Lower Boundary of 99% PI AND < Lower Boundary of 90% PI	>= Lower Boundary of 90% PI AND <= Upper Boundary of 99% PI	> Upper Boundary of 99% PI

* Assign a base rating of 'Insufficient Data' if Reported MCMIS Fatal Records is less than 15

The base rating assignments are defined as follows.

Base Rating Assignments:

Good rating range: Reported Non-Fatal crash records (from MCMIS) is greater than or equal to lower boundary of 90% prediction interval AND less than or equal to upper boundary of 99% prediction interval

Fair rating range Reported Non-Fatal crash records (from MCMIS) is greater than or equal to lower boundary of 99% prediction interval AND less than lower boundary of 90% prediction interval (under reporting) OR

Reported Non-Fatal crash records is greater than upper boundary of 99% prediction interval (over reporting)

Poor rating range: Reported Non-Fatal crash records (from MCMIS) is less than lower boundary of 99% prediction interval AND MCMIS fatal crash records is greater than or equal to 15 (under reporting)

Insufficient Data rating range: Reported Non-Fatal crash records (from MCMIS) is less than lower boundary of 99% prediction interval AND MCMIS Fatal crash records is < 15

Determine Rating Results

The NFCC rating is determined by reviewing the base rating results and the most recent fatal crash completeness (FCC) measure rating for each State. Both of these ratings are important for the assignment of the final NFCC rating.

The FCC measure rating is considered since the model's output is using State-reported fatal crash records to generate estimated non-fatal crash record values. The FCC measure rating depicts how well States are reporting MCMIS fatal crash records. If a State's FCC rating is 'good' or 'fair', it is assumed their MCMIS fatal crash reporting is sufficient to estimate the expected number of non-fatal crash records. In those cases, the NFCC base rating is regarded as the final NFCC measure rating.

If a State's FCC rating is either 'poor' or 'insufficient data', it is assumed their MCMIS fatal crash reporting is insufficient due to inadequate reporting or too few fatal crash involvements occurring in the State. An “override” is administered, meaning, the base rating cannot be used. The override will result in an NFCC rating of either 'poor' or 'insufficient data' due to incomplete MCMIS fatal reporting.

The final Non-Fatal Crash Completeness rating is determined as follows:

Rating	Criteria
Good	Reported MCMIS non-fatal records fall within Good rating range AND FCC rating is Good or Fair
Fair	Reported MCMIS non-fatal records fall within Fair rating range AND FCC rating is Good or Fair
Poor	Reported MCMIS non-fatal records fall within Poor rating range AND FCC rating is Good, Fair or Poor
Insufficient Data	Base rating is assigned Insufficient Data OR Reported MCMIS non-fatal records fall within Good or Fair rating range AND FCC rating is Poor OR FCC rating is Insufficient Data

Definitions

Bus: A motor vehicle with seats for at least nine (9) people, including the driver's seat.

Fatal Crash: A crash where one or more persons die within 30 days of the crash. The fatality does not have to occur at the scene of the crash. It includes any person involved in the crash, including pedestrians and bicyclists, as well as occupants of the passenger cars and trucks.

Interstate Carrier: Carriers that transport a commodity outside the State of its place of business.

Intrastate Carrier: Carriers that transport a commodity only within the State of its place of business.

Large truck: Any truck having a gross vehicle weight rating (GVWR) of more than 10,000 pounds or a gross combination weight rating (GCWR) over 10,000 pounds.

Log-linear regression model: A statistical tool where existing data are used to define a relationship between two sets of numbers, so that in the future, one set can be reliably predicted from the other. In reference to the NFCC measure, verified fatal and non-fatal crash record data were used to define a relationship for predicting an expected range of non-fatal crash records from a known quantity of fatal crash records.

NFCC Base Rating: The calculation of a State's NFCC measure rating is determined by reviewing its Base Rating to its fatal crash completeness measure rating. The Base Rating establishes a good, fair, poor or insufficient data rating range for each State. Each designation is determined by where a State's non-fatal crash reporting falls.

Non-fatal crash: A crash where one or more persons has non-fatal injuries requiring transportation by a vehicle for the purpose of obtaining immediate medical attention; or one or more of the vehicles were towed away from the scene due to “disabling damage”. The towed vehicle need not be the commercial motor vehicle involved in the crash.

Prediction Interval: An estimated range of values, as defined by a lower and upper value, is determined by a mathematical model and the probability that another number will fall between them. In reference to the NFCC measure, the log-linear regression model generates a unique interval of non-fatal crash records.

Urbanization factor (RU factor): A ratio of rural to urban commercial motor vehicle (CMV) travel in the State as calculated from the Federal Highway Administration's data in the Annual Highway Statistics publication. For each State, the RU factor is calculated by averaging the most recent three years of annual rural CMV travel miles by the average of the most recent three years of annual urban CMV travel miles.

FMCSA DATA QUALITY
PROGRAM MANAGER

Betsy Benkowski: 202-366-5387

For additional information, contact your FMCSA Technical Assistant:

EASTERN SERVICE CENTER STATES:

Janet Bibinski: 617-494-2308

SOUTHERN SERVICE CENTER STATES:

Deirdre Hering: 617-494-3535

MIDWESTERN SERVICE CENTER STATES:

Kevin Berry: 617-494-2857

WESTERN SERVICE CENTER STATES:

Kevin Berry: 617-494-2857

Send us your questions or comments on the SSDQ methodology by using the A&I Online Feedback Form. Federal and State users may contact their TA directly.

Federal Motor Carrier Safety Administration
1200 New Jersey Avenue SE, Washington, DC 20590 • 1-800-832-5660 • TTY: 1-800-877-8339 • Field Office Contacts