CHAPTER 4
Supplemental Data Sources and Linkages

The research team identified supplemental data sources and linkages states and jurisdictions may use to identify and mitigate the under- or overreporting of impaired and distracted driving crash data. It was recognized that states and jurisdictions may not have access to all data sources, and some data sources may be incomplete, costly, or untimely to utilize. Using multiple data sources beyond crash data can help give a clearer picture of the extent of misreporting on impairment and distraction-related crashes.

4.1 Linked Hospital and Crash Data

One common accessible data source and method states and jurisdictions can use to identify the misreporting of impaired driving is linking hospital data to crash data. Previously initiated as a federally funded program, the CODES program was transitioned to state-funded programs, and several states still refer to linked crash and hospital records as CODES. The CODES program links person-level crash data with the crash victimʼs medical outcomes resulting from the crash. Linked hospital data can be inpatient, emergency department, or trauma registry and may give insight into impairment law enforcement missed, alcohol use below the legal limit that was not reported on the crash report forms, or cases of polydrug (i.e., when alcohol is used in conjunction with other intoxicants) in which one or more drugs is not reported. Linked hospital data contain International Classification on Disease (ICD) codes that contain diagnostic information for each disease, injury, symptom, and so on, that might be medically classified. For this research effort, the research team used ICD version 10 codes (ICD-10-CM) to develop methodologies that address the misreporting of impaired driving. The World Health Organization finished work on ICD version 11 in 2019 and approved it for implementation on January 1, 2022. The United States has not yet adopted the new version of ICD at the time of this report.

One point of information related to linked hospital data provided by subject matter experts significantly altered the anticipated findings. Specifically, the effect alcohol exclusion laws have on the reliability of hospital data when looking at the reporting of impairment. Alcohol exclusion laws allow health insurers to deny individuals coverage if impairment was present and played a role in the events leading to their injury. Insurance companiesʼ refusal to cover medical expenses may result in hospitals covering the treatment costs if the patient cannot pay. Therefore, hospitals choosing not to test for alcohol to avoid this potential issue would affect the availability and reliability of hospital data. A traffic safety fact sheet published by NHTSA (2008) titled “Traffic Safety Facts: Alcohol Exclusion Laws” encouraged states to investigate the existence of alcohol exclusion laws and to support legislation that would prohibit them. At the time of the fact sheet, 29 states, including the District of Columbia, had alcohol exclusion laws in place with an additional seven states not having alcohol exclusion laws but at the same time not having laws that prohibited them.

A study by Azagba, Ebling, and Hall (2023) reported that 18 states still have alcohol inclusion laws. Reported states are Alaska, Arkansas, Delaware, Florida, Georgia, Hawaii, Kansas, Kentucky, Louisiana, Mississippi, Missouri, Nebraska, New Jersey, New York, Pennsylvania, South Carolina, Virginia, and West Virginia. Fifteen states specifically prohibit alcohol exclusion laws, and 17 states do not have laws permitting or prohibiting alcohol exclusion laws.

4.2 Citation Data

Another source states and jurisdictions can use to better identify the prevalence of impaired or distracted driving is law enforcement citation data. Specific citations related to impairment or distraction while driving (i.e., cell phone use) can be used to better understand the prevalence of misreported crashes. Citation data in general would give a better sense of how often these types of citations are issued and could be compared with roadside surveys or self-reported surveys of driver behaviors. When citation data are linked to specific crashes, these data may be used to understand issues with incomplete or inaccurate crash report data fields. Comprehensive citation data required for a scientific analysis were not available at the time of this research.

4.3 Crash Reconstruction

Crash reconstruction reports—specifically, how often the reconstruction experts suspect, request, and download driversʼ cell phone data—can be useful, particularly in identifying distracted driving. These data can provide insight into the prevalence of distraction and specific information on the type of distraction before the crash. However, crash reconstructions are infrequent and are often limited to crashes involving fatalities or serious injury crashes, after which criminal or civil legal proceedings may result.

While practices may vary between states, most fatal and serious injury crashes that include a post-crash reconstruction will include a download of the vehicleʼs event data recorder (EDR) and a capture of driversʼ cell phone records. For example, the Wisconsin State Patrol reconstructs all fatal crashes within Wisconsin. Iowa, Minnesota, Illinois, and other states follow a similar practice. Reconstructions can provide additional crash data, including data on “probable” distraction, when phone records or other post-crash data provide sufficient evidence, or a lack of driver reaction to a hazard correlating to distraction. A post-crash data reporting methodology is required to incorporate these new data into the crash record. Lastly, EDRs within vehicles may show cases in which cell phones were used before a crash, provided cell phones were docked to the vehicle or connected via Bluetooth.

4.4 Laboratory Tests

Toxicology data that can be linked to crash reports are available for crashes in which alcohol or drug impairment was suspected by the reporting officer, and a human sample was collected for testing. Toxicology data may also be available for fatal and serious injury crashes as part of the post-crash investigation. Because of this availability, toxicology data may be linked to crashes in which the reporting officer initially reports suspected alcohol or drug use and to crashes already receiving increased attention due to the severity of the injuries. As a result, linked toxicology data will primarily show overreporting if the test results come back negative. Test result data that arrive after the crash report is submitted may, of course, be used to correct the earlier data (including indications that the officer suspected impairment). Data elements for suspected impairment may be left unchanged, in which case overreporting can happen if the data elements analysts use to define impaired driving crashes use a criterion that relies on “any indication” of

alcohol or drug involvement—this was the criterion cited by several survey respondents. If the criterion is based on test results, the analysts will need to account for delays in obtaining data from labs or, perhaps, never getting test result data from hospitals. Therefore, one concern with using toxicology data from the crash database is that they may be incomplete for an indefinite period. Using a partial data source to gauge misreporting adds uncertainty.

As an example, the research team faced difficulty obtaining toxicology data and received information from state agencies indicating data availability as a problem for them as well. If a state agency were able to access toxicology data, it would still need to develop a process for linking the data to crash reports if one were not already in place. This would require probabilistic linkage and result in a partial data set including only the matches that met a threshold of certainty.

The research team also queried the value of toxicology data for estimating the misreporting of drugs due to stop testing policies that may be in place. The Center for Forensic Science and Education (2024) asked laboratories whether they make an administrative decision to stop testing if a BAC result is at or above a certain concentration. Of the 80 laboratories that responded, 51 percent reported that they do have a policy in place. States need to work with the laboratories they are getting data from to determine whether a stop testing policy is in place, what criteria result in stop testing, and how those policies will influence their analysis of misreporting. Because of the costs associated with drug testing, it is also of concern that labs will generally test for selected classes of drugs based on the information provided by law enforcement (or based on other criteria). For this reason, the toxicology data are not a suitable source to assess the relative prevalence of all classes of drugs—not all labs can afford to run tests of all possible drugs. Some cases of impaired driving will likely be missed, even if toxicology results are available. In other words, a negative test result for drug-impaired driving will only reliably indicate that the drugs tested for (not all drugs) were not detected in the sample.