Estimating MCMIS Crash File Underreporting from Carrier Records - PowerPoint PPT Presentation

Estimating MCMIS Crash File Underreporting from Carrier Records March 22, 2017 Office of Research and Information Technology

Background  Anecdotally, carriers believer crashes are underreported to MCMIS  Common estimate of 15% underreporting  Study conducted by UMTRI to get a sense of the level of underreporting  Not a statistically valid sample Office of Research and Information Technology 1

Sample 6 carriers agreed to participate:  3 national for-hire with straight & combination trucks  1 regional with primarily combination trucks  1 private with straight and some combination trucks  1 private with only straight trucks Accident Files:  4 provided full accident files but identified crashes they thought were reportable  2 provided only crashes they thought were reportable Office of Research and Information Technology 2

Reportable Crash Criteria Part 390.5 provides definitions of commercial motor vehicles and accidents. Truck with GVWR over 10,000 or GCWR over 10,000, or Bus with seating for at least 9, including the driver, Vehicle or Vehicle displaying a hazardous materials placard. Fatality, or Injury transported to a medical facility for immediate Crash medical attention, or Vehicle towed due to disabling damage. Office of Research and Information Technology 3

The Carrier Data, 2012-2014  Carriers supplied 58,333 event records  They identified 8,392 meeting MCMIS crash reporting criteria.  4,777 crash records in MCMIS for the 6 carriers  3,885 records (81.3%) in MCMIS matched carrier IDed reportable crashes  303 records (6.3%) in MCMIS that carriers IDed as not reportable  589 MCMIS crash records (12.3%) not in carrier files • Insufficient information in carrier files to link the records to the correct crashes in MCMIS. • Incorrect information in either the MCMIS file or carrier files as to times, places, and vehicles. • Crashes erroneously attributed to a carrier.  That leaves 4,507 crashes as potentially underreported Office of Research and Information Technology 4

The State Data, 2012-2014  Crash Files from 15 States  FL, GA, ID, LA, MD, MI, MO, NE, NJ, NM, NY, OH, OR, UT, WA  These States accounted for 1,462 of the 4,507 missing crashes  1,462 crashes carriers IDed as reportable searched for  564 crash records found in State files  710 not found  188 with insufficient data in State file to match (excluded from further analysis)  That leaves 1,274 crashes to investigate Office of Research and Information Technology 5

Estimating 564 Records in State Files not in MCMIS  Combining sorting by crash severity and vehicle type:  197 (35%) met FMCSA crash reportability criteria  276 (49%) did not  91 (16%) unknown (towed but PAR didn’t capture disabled)  Estimating proportion of unknowns that qualify:  GES – 84% of PDO crashes involved disabling damage  84% of 91 unknowns = 76 potentially reportable crashes  Final estimates for the 564 records in State files but not in MCMIS  273 (48%) were reportable crashes  291 (52%) were not reportable Office of Research and Information Technology 6

Possible Reasons for not Uploading to MCMIS  Why were 48% of reportable crashes in State files not reported to MCMIS?  Misclassification: 25% were misclassified as light trucks  Incorrect DOT number: 16%  Missing DOT number: 36% Office of Research and Information Technology 7

Crashes in Carrier Files but not State Files  What about the 710 crashes carriers considered reportable but weren’t in State files? No PAR was filed. Met MCMIS reporting criteria? Number % Yes 28 3.9 Likely yes 88 12.4 Maybe, if towed 261 36.8 Likely no 109 15.4 No 104 14.6 No, light duty 24 3.4 Road debris, damage 18 2.5 Deer/animal-minor 71 10.0 Unknown 7 1.0 Total 710 100.0  Estimated 56% of missing crashes were reportable crashes. Office of Research and Information Technology 8

Conclusions  What is the implied level of underreporting to the MCMIS crash file, based on the carrier files?  For the carriers studied, 4,777 crashes were reported to MCMIS.  An estimated additional 2,468 should have been reported.  Underreporting was estimated to be about 34 percent  How reliable was carrier-identification of MCMIS- reportable crashes?  Carriers correctly identified about two-thirds of the crashes as MCMIS-reportable.  But about 25 percent of the crashes they identified as reportable were not.  And they missed about 6 percent of their reportable crashes. Office of Research and Information Technology 9

Recommendations  Encourage State and local police to prepare PARs for reportable large truck and bus crashes  Will involve training on what’s a reportable crash.  Emphasize reporting DOT numbers  Will involve training on how to identify the correct DOT number.  Improve identification of truck and bus types  Use NHTSA’s vPIC to decode VINs.  Available at https://vpic.nhtsa.dot.gov/decoder/Decoder . Office of Research and Information Technology 10