Research on Control and Prediction of Alcohol Impaired Driving with Ignition. Interlocks

Transcription

1 Research on Control and Prediction of Alcohol Impaired Driving with Ignition SAE International Meeting February 6, 2009 Interlocks Paul R. Marques, Ph.D. Senior Research Scientist

2 Important Distinction Interlock Devices the hardware Early devices were non-specific Tagucci semiconductors, currently most are alcohol specific fuel-cell devices. Interlock Programs the regulations Laws (court or licensing programs), sanctions, reporting protocols, information flow, treatment links, consequences etc

3 Four Key Elements (of alcohol interlock devices) (1) an alcohol sensor that prevents engine starts if breath alcohol is detected above some minimal level such as BAC.02 g/dl (e.g.,.02%) (2) a running retest feature, which requires at least one retest after the car has started, typically retests are every 15 to 45 minutes while driving (3) a tamper-detecting installation in the engine, which requires inspection every 30 to 60 days to preclude bypass wiring (4) a data-recording system that logs the time of all BAC tests, starts, stops, procedural violations or circumvention attempts.

4 Topics Today Interlock Penetration / Growth Effectiveness Evidence Patterns in the Data: Drinking, Vehicle Use, High Risk Offenders First Offender Risks

5 Events in interlock history California law Alberta law NHTSA Model Specs O ICADTS Interlock WG FL program MADD Campaign Year Voas et al. AB study Beck et al., MD study

6 FBI estimate of 1.4 million DUI arrests annually ,4 million DUI arrests annually (in 2007: FBI) MADD Campaign to End Drunk Driving will seek 600,000 more interlocks in 5-7 years. What might accomplish this? First and multiple offenders, More enforcement, Better monitoring, Program tied to the BAC tests and other objective indicators, Stronger (less appealing) alternatives for the opt outs Total Convictions? MADD projection

7

8 New Mexico interlock rate represents 49% of all convicted DUI offenders there

9 Topics Today Interlock Penetration / Growth Effectiveness Evidence Patterns in the Data: Drinking, Vehicle Use, High Risk Offenders First Offender Risks

10 Effectiveness We want to reduce alcohol crashes, crash injuries and fatalities but we mostly measure drunk driving recidivism as the target outcome measure. Arrest/conviction for DUI recidivism is a proxy for alcohol crashes. There are ~160+ DWI arrests for each fatality in which a driver had a BAC >=.08 g/dl. Soon there will be enough interlocks to estimate the extent of alcohol crash risk reduction.

11 New Mexico

12 Willis et al., 2004 Interlock ON 13 study samples Maryland RA study OVERALL =.36 (i.e., 64% reduction)

13 Willis et al., 2004 Interlock OFF 10 study samples Maryland RA study

14 Topics Today Interlock Penetration / Growth Effectiveness Evidence Patterns in the Data: Drinking, Vehicle Use, High Risk Offenders First Offender Risks

15 Interlocks: first and multiple offenders give median 6-8 BAC tests every day* Quebec BAC Tests per day * 7x 365 = 2555 tests/year Note: Start-up tests are about half of all tests Frequency Std. Dev = 4.63 Mean = 8.8 N =

16 Predicting Future DUI with the RATE of Interlock BAC test positives 24 Mo. AFTER INTERLOCK N=720 Offenders in Each Bar Data from Marques, Voas, Tippetts (2003) Addiction, Quebec Prediction Study Support from NIAAA top 40% DURING INTERLOCK BAC Rate.02%

17 Recidivism Rates During and After Interlock by RATES of Failed BAC tests (NM) Total N~5000 ~ 500 / line Zeros Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Total top 30% mo 12mo 6mo 12mo 18mo 24mo DURING Interlock AFTER Interlock

18 How Predictable are People? Québec BAC.02% Weekday morning 7-8 am All BAC tests Alberta

19 Morning high BAC most frequent 7-8 am Weekdays - Max positive BAC Confirmed in 4 jurisdictions (TX,AB, QC, NM) based on more than 40 million BAC tests Hour of Day Quebec All Tests Quebec Tests>=.02% Alberta All Tests Alberta Tests >=.02 Texas All Tests TX tests>.02%

20 Recidivism Rates During and After Interlock by RATES of Failed Morning BAC tests top 30% Zeros Decile 7 Decile 8 Decile 9 Decile 10 Total mo 12mo 6mo 12mo 18mo 24mo DURING Interlock AFTER Interlock

21 Z scores 10 alcohol biomarkers by 3 groups of failed interlock BAC Alberta 2008 Low None High *** *** *** *** *** *** ns * *** *** *** P<.0005 top 20.5% of fail rate

22 Topics Today Interlock Penetration / Growth Effectiveness Evidence Patterns in the Data: Drinking, Vehicle Use, High Risk Offenders First Offender Risks

23 First and Multiple Offenders by the 3 Interlock Risk Groups None Low High

24 What about First Offenders? Alcohol Mentions/1000 drivers in Maryland 5 million licensees The distinction between first and second offender is small compared to first and nonoffender. A first offender has ten times more alcohol mentions than a non-offender. A second offender has only 25% more than a first time offender x higher Prior DUI Rauch et al., TRB 2002

25 Similar evidence from New Mexico % Re-arrested within 3 Years 25% 20% 15% 10% 5% 0% 5x higher None 1st 2nd 3rd 4th 5th+ Conviction Number in 2003 Relative to Non-Offenders, First Time DUI Offenders are 5x more likely to be arrested within 3 years for DUI

26 If attainable: why alcohol sensing in all cars would be useful to road safety Interlock use by high risk drinking drivers is low Risk of arrest is low about episodes of impaired driving for each arrest (1.4 million arrests/yr) Estimate ~ 60-65% of arrestees convicted Of those convicted and eligible (only 5-50% install interlocks depends on state) Of those who install, about 30% rack up high rates of failed BAC tests (and/or procedural violations) Typical installation is 6-18 months Those who have high fail rates are persistent drinking drivers: re-arrest rate in 2 years (about 15% - 25) Most who install use the interlock car appropriately but they may not be the highest risk offenders

27 Drunk Driver Plows into Mexican Bike Race One Dead, 10 Injured, June 1, 2008 Slide from Dick Roth