DOES A TOWBAR INCREASE THE RISK OF NECK INJURY IN REAR-END COLLISIONS? A STUDY BASED ON A MERGER OF DANISH ADMINISTRATIVE REGISTRIES ANNE VINGAARD OLESEN H A R R Y L A H R M A N N T H E T R A F F I C RESEARCH GROUP AT AALBORG UNIVERSITY D E N M A R K
The tow-bar A common sight in Denmark especially in rural areas Prevalence of 60% in the Motor Vehicle Registry in the period 2003-12 From seller s car database on the internet 37% in all of Denmark (2017) A count in typical parking lot gave a prevalence of 48% with 90% towbars actually on the car (2017) Prevalence has dropped recently because of an increase in small cars (as cars number two and else if mounting is not possible)
The study aim To investigate the association between the risk of neck injury in occupants of the struck car in a rear-end collision
The registry merger The National Registry of Motor Vehicles: registration plates, towbar status Police reports: personal identification numbers, registration plates, accident type The Central Person Registry: personal identification numbers, vital status, nationality The National Hospital Discharge Registry: personal identification numbers, diagnoses Socio-economic databases at Statistics Denmark: personal identification numbers, income, highest attained education, socioeconomic status All information available at date of accident The personal identifier and the registration plate are the central merger variables
The modern collision zone curls in rear-end collisions In order to absorb the force and protect occupants in the cabin Mounting of a stiff tow-bar frame is highly likely to inflate the curling mechanism May increase the risk of neck injury in the struck car with a towbar Several ways to mount the tow-bar And now possible to demount when not using it All of which may mix the impacts of a towbar registered in a general Motor Vehicle Registry
Strong experimental evidence: Krafft et al. (2000). AAP 32: 187-95. Two Volvo 240s one with and one without a towbar Car acceleration higher in towbar equipped car with peak at 9.6g Car acceleration in car without towbar had peak at 8.0g Dummy in the car with towbar experienced 33% higher peak acceleration in lower neck, 8.9g Compared with 6.7g for dummy in the car without towbar Increased risk of 22% for long-term consequences of neck injury in struck cars with a towbar (p<0.001) One other study found no effect Linder et al.?? Few other sources cite Krafft et al.(2000)??
Negative finding: no effect of tow-bar on neck injury Could be explained by measurement error on the information on tow-bar in the Motor Vehicle Registry: imprecise exposure erases the effect if present Counts, person-years, and neck injury rates. Public registry data from Denmark, 2003 2012 N N P-years IR Variable Variable Persons Events in 1000s per 1000 95% CI significance All 9370 1519 7.810 194.5 185.0 204.5 1-year prevalence of 16% Tow-bar on struck car Yes 5019 827 4.168 198.4 185.3 212.4 p < 0.0001 No 3244 575 2.652 216.8 199.8 235.3 Unadjusted IRR=0.92; p=0.11 Unknown 1112 117 0.990 118.2 98.6 141.6
Estimated hazard ratios in multivariate Cox proportional hazard regression. Hazard ratios can be interpreted as incidence rate ratios. Public registry data from Denmark, 2003 2012 Multivariate analysis Hazard Robust Variable Variable ratio SE 95% CI P-value significance Tow-bar on struck car Yes 0.915 0.0523 0.818 1.024 0.122 0.111 No 1.000 Unknown 0.820 0.0892 0.662 1.015 0.068 Adjusted for: gender, age, seat, income, education, accident type, calendar year, weight of car, first registration year
Sex Male 1.000 p < 0.0001 Female 1.538 0.0852 1.380 1.714 0.000 Age in years 0 17 0.936 0.2096 0.604 1.452 0.769 p < 0.0001 18 29 1.022 0.0766 0.882 1.183 0.775 30 39 1.000 40 49 0.904 0.0665 0.783 1.045 0.172 50 59 0.688 0.0613 0.577 0.819 0.000 60 69 0.546 0.0677 0.428 0.696 0.000 70+ 0.285 0.0659 0.181 0.448 0.000 Females have increased risk Older persons tend to underreport?
Seat in struck car Driver 1.000 p < 0.0001 Passenger 1.524 0.1302 1.289 1.801 0.000 Household income <1st quintile 0.827 0.0740 0.694 0.986 0.034 p < 0.0001 1st 2nd quintiles 0.990 0.0782 0.848 1.155 0.895 2nd 3rd quintiles 1.000 3rd 4th quintiles 0.974 0.0785 0.832 1.141 0.746 >4th quintile 0.843 0.0774 0.705 1.010 0.063 Unknown 0.076 0.0363 0.030 0.194 0.000 Passenger seat associated with increased risk of neck injury Poor and wealthy have reduced risk
Highest level of education Primary education 1.019 0.1418 0.776 1.339 0.891 0.497 Upper secondary education 0.992 0.1563 0.728 1.350 0.957 Vocational Education and Training (VET) 0.948 0.1279 0.728 1.235 0.694 Qualifying educational programs 2.703 1.9514 0.657 11.127 0.168 Short-cycle higher education 1.000 Vocational bachelors programs 1.096 0.1569 0.828 1.451 0.522 Bachelors programs 0.929 0.2552 0.543 1.592 0.790 Masters programs 0.866 0.1586 0.605 1.240 0.431 PhD programs 0.592 0.4267 0.144 2.430 0.467 Unknown 0.812 0.1716 0.536 1.228 0.323 Educational effect explained by income level etc.
Accident type Hit directly from the back when driving straight ahead 1.000 0.069 Hit from the back when turning right 0.730 0.1087 0.545 0.977 0.035 Hit from the back when turning left 0.943 0.0658 0.822 1.081 0.397 Calendar year 2003 2004 1.000 0.481 2005 2006 0.988 0.0770 0.848 1.151 0.880 2007 2008 0.899 0.0728 0.767 1.054 0.188 2009 2010 0.979 0.0835 0.828 1.157 0.806 2011 2012 0.882 0.0817 0.735 1.058 0.175 Reduced risk when turning right No significant effect of calendar year
Weight of struck car in kilograms 500 1000 0.996 0.0681 0.871 1.139 0.950 0.985 1000 1500 1.000 1500 2000 1.019 0.1100 0.825 1.259 0.861 2000+ 0.853 0.2513 0.479 1.519 0.589 Unknown 1.029 0.6233 0.314 3.373 0.962 First registration year of struck car 1966 1989 0.946 0.1076 0.757 1.182 0.623 0.495 1990 1994 1.035 0.0940 0.867 1.237 0.701 1995 1999 0.962 0.0739 0.828 1.119 0.618 2000 2004 1.000 2005 2009 0.839 0.0784 0.699 1.008 0.060 2010 2015 0.827 0.1608 0.565 1.211 0.328 Unknown 1.079 0.6575 0.327 3.562 0.900 No effect of weight and first registration year of struck car
Conclusion 16% of occupants of struck cars were diagnosed with neck injury by doctors in hospitals/emergency rooms after rear-end collisions No effect of tow bar on the struck car in large study of 9370 occupants Measurement error might induce result Count of cars in typical parking lot demonstrated that 90% of tow-bars were actually on the car (and not taken off) More crash test needed in order to corroborate registry finding Could the tow-bar be associated with neglectible risk of neck injury in modern cars Law of physics does somehow rule
What we learnt from this merge Remarkable possibilities for registry merge in Denmark and Scandinavia which is of course well-known Merge by registration plate in police reports and Motor Vehicle Registry possible Merge by personal identifier in police reports and National Discharge Registry (hospital registry) possible Denmark is only a novice regarding registry studies
Strengths and limitations of registry studies Complete populationbased studies with many cases (likely a high power) Facilitate assessment of various hypotheses such as mix between different personal injuries and diagnoses and hospital courses But - Measurement error such as on information of tow-bar status in the Motor Vehicle Registry Lack of information on confounders such as (precise) personal characteristics, attitudes and psychological profiles Missing values in some variables unknowns
Experiences with registry data in Scandinavia? Which kinds of merges are possible in the different countries? Experiences with different sources such as police reports with registry data? Experiences with various sorts of sources of error? How to overcome the limitations: sensitivity analyses and multiple imputation or? Other issues?