Attachment G - Oriiginal multi state report

Attachment G - Original NHTSA CODES Report

Benefits of Safety Belts
and Motorcycle Helmets
U.S. Department
of Transportation
National Highway
Traffic Safety
Administration

Report to Congress
February 1996

Based on Data from
The Crash Outcome Data Evaluation System
(CODES)

ACKNOWLEDGMENT
The National Center for Statistics and Analysis of the National Highway Traffic
Safety Administration wishes to acknowledge the outstanding collaborative effort by
the many individuals and organizations in the 7 CODES states -- Hawaii, Maine,
Missouri, New York, Pennsylvania, Utah, and Wisconsin -- in generating the data for
and assisting in the preparation of this report.

EXECUTIVE SUMMARY

This study was undertaken in response to Section 1031(b) of the Intermodal
Surface Transportation Efficiency Act of 1991 (ISTEA), which required the National
Highway Traffic Safety Administration (NHTSA) to conduct a study to determine the
benefits of safety belt and motorcycle helmet use in crashes.
The study employed methods whereby statewide data from police crash reports,
emergency medical services, hospital emergency departments, hospital discharge
files, claims, and other sources were linked so that those people injured in motor
vehicle crashes could be followed through the health care system.

Information for

both the injured and uninjured was then used to determine the benefits of protective
devices in motor vehicle crashes. The available financial information included
inpatient charges (acute care, rehabilitation, long-term care) and estimates of actual
costs using a charge-to-cost ratio. Through the cooperation of the highway safety and
medical communities, this was the first time these databases were linked using a
probabilistic computer algorithm. Grants were awarded to entities in Hawaii, Maine,
Missouri, New York, Pennsylvania, Utah, and Wisconsin to complete the linkage and
perform the analyses upon which this report is based. All of the states were able to
generate the linked data.
The Crash Outcome Data Evaluation System (CODES) study results revealed
that safety belts are highly effective in reducing morbidity (the occurrence of any injury)
and mortality. They also indicate that safety belts cause a downward shift in the
severity of injuries.

CODES Report to Congress

- i -

February 1996

The study results showed that the average inpatient charge for unbelted
passenger vehicle drivers admitted to an inpatient facility as a result of a crash injury
was more than 55 percent greater than the average charge for those that were belted,
$13,937 and $9,004 respectively. If, in the CODES states, all unbelted passenger
vehicle drivers had been wearing safety belts, it is estimated that inpatient charges
would have been reduced by approximately $68 million and actual inpatient costs
reduced by $47 million. Private insurance accounted for 69 percent of the inpatient
charges compared to 16 percent for public and 15 percent for other sources. In all
cases, the average inpatient charge was greater for drivers who were unbelted.
The study results also showed that motorcycle helmet effectiveness ranged
from 9 percent in preventing any kind of injury to 35 percent in preventing a fatality.
These results confirm previous NHTSA estimates. The average inpatient charge for
motorcycle crash victims receiving inpatient care was $14,377 for those who used
helmets, and $15,578 for those who did not, an 8 percent increase in charges for those
electing to not wear a helmet. Private insurance sources accounted for 63 percent of
inpatient charges compared to 23 percent for public and 14 percent for other sources.
For the private and public sources, average inpatient charges for motorcycle crash
victims were 15 percent and 5 percent higher, respectively, for the unhelmeted.
Helmets cannot protect the rider from most types of injuries. But further
analysis of the CODES data, possible because of the linked medical outcome, showed
that motorcycle helmets are 67 percent effective in preventing brain injuries. Thus, if all
motorcyclists had been wearing helmets, 67 percent of those unhelmeted motorcyclists
who received inpatient care for a brain injury would not have sustained the brain injury.
In other words, unhelmeted motorcyclists were over three times as likely to suffer a
CODES Report to Congress

- ii -

February 1996

brain injury as were helmeted motorcyclists.
Examination of the average inpatient charges revealed that the average charge
for inpatient care for a motorcyclist who sustained a brain injury is more than twice the
average charge for motorcyclists receiving inpatient care for other injuries. On
average, approximately $15,000 inpatient costs would be saved during the first 12
months for every injured motorcycle rider not sustaining a brain injury. Therefore, if all
injured motorcycle riders wore helmets, fewer victims would incur the high cost of
inpatient care associated with brain injury.
CODES demonstrated that linked, comparable data could be generated to
evaluate the benefits of belts and helmets in terms of medical and financial outcome.
Linkage enabled injury severity to be standardized among the CODES states. The
linked data represent a permanent data file. Besides the belt and helmet study, these
data were used to generate state-specific analyses and will continue to be used in the
future. The CODES states learned new linkage skills that can be applied to the linkage
of other types of records and also can be shared with other states interested in linkage.
Of even more importance is the fact that linkage identified previously unknown
problems with missing and inaccurate data. Correcting these problems for the study
improved the quality of the data in the permanent files, making the state data (linked or
unlinked) even more valuable for future uses.

CODES Report to Congress

- iii -

February 1996

CODES Report to Congress

- iv -

February 1996

INTRODUCTION
The Intermodal Surface Transportation Efficiency Act of 1991 (ISTEA), required
the Department of Transportation and NHTSA to study the benefits of using safety
belts and motorcycle helmets. This requirement appears in Section 1031, USE OF
SAFETY BELTS AND MOTORCYCLE HELMETS,
HELMETS part (b) STUDY:
STUDY
"(1) In general.
general -- The Secretary shall conduct a study or studies to determine
the benefits of safety belt use and motorcycle helmet use for individuals involved
in motor vehicle crashes and motorcycle crashes, collecting and analyzing data
from regional trauma systems regarding differences in the following: The
severity of injuries; acute, rehabilitative and long-term medical costs, including
the sources of reimbursement and the extent to which these sources cover
actual costs; government, employer, and other costs; and mortality and
morbidity outcomes. The study shall cover a representative period after
January 1, 1990."

Funding was made available through the ISTEA, and the legislation called for a
report to be submitted to Congress within "... 40 months after the funds for such a
study were made available...." This report contains the results of that study.

CODES Report to Congress

- 1 -

February 1996

STUDY DESIGN
Population Coverage
In designing the study, NHTSA considered the requirements specified in the
legislation, the types of information which would comply with those specific analytic
requirements, and the availability of such information. Consequently, the study results
presented in this report are based on statewide databases in the participating states
which include all persons involved in police-reported crashes -- those who were injured
or who died and those who were not injured. In this manner, comparisons between
those using and not using safety belts or motorcycle helmets could be made by
identifying and contrasting the characteristics of the injured and uninjured occupants
within each of the use groups. Exhibit 1, generated from previous research

1

(all

footnotes are on page 35), shows the approximate distribution of persons involved in
motor vehicle crashes by the severity of injury and emphasizes the fact that the
majority of persons involved in crashes are not injured.

0.2 % Fatalities

Exhibit 1.

Distribution of Persons Involved in Motor Vehicle
Crashes by Injury Outcome

CODES Report to Congress

- 2 -

February 1996

In an individual state, the comprehensive source of information about all people
involved in motor vehicle traffic crashes (including those involving passenger cars,
vans, light trucks and motorcycles) is the statewide database created from crash
reports filed by police agencies in the state. Police officers, who investigate the crash
at the scene, complete a report which includes information about the crash, the
vehicles, and the persons involved. Selected data from these reports are entered into
an electronic database from which most states produce an annual report describing the
crashes occurring in that state. The availability of these databases in electronic format
make the data readily available for statistical analysis.
Statewide crash databases have some limitations, however. The individual data
items collected by states vary widely. The training of the police officers in crash
investigation and data collection, the data elements collected by police officers, and the
actual data coded by each state onto its database, are different. Many states include
information about all vehicle drivers and passengers, injured or uninjured, while others
include information for passengers only if they are injured in the crash. Not all motor
vehicle crashes are reported to the police. Every state has a minimum reporting
threshold which excludes some or all of those crashes causing only minor property
damage and no injuries. Thus, there may be even more successes resulting from the
use of safety measures, currently undocumented and unavailable for analysis. Despite
these shortcomings, statewide police-reported motor vehicle crash data are the only
comprehensive source of data on a definable set of crashes within a given state.
Police-reported crash data also are limited in providing information about the
medical and financial outcomes of crash victims. Police officers are not trained

CODES Report to Congress

- 3 -

February 1996

diagnosticians and, in most states, do not code specific injuries to crash-involved
occupants, nor are they able to obtain information on the financial consequences to the
injured victims. Police crash reports typically contain a field for "injury severity". In
most states the police use a standard scale similar to the "KABC0" scale, where:
K = killed;
A = severe or incapacitating injury;
B = non-incapacitating injury;
C = possible injury; and,
0 = not injured.
The application of this scale depends on a police officer's evaluation at the
scene. Persons with different medical severities are often included within the same
class. Frequently, transport by emergency medical services (EMS) of a crash victim
for treatment is enough for the police officer to code "incapacitating injury." On the
other hand, some injuries are not immediately evident at the scene of a crash, and a
victim who is later diagnosed with a serious injury can be initially classified as not
injured.
Therefore, police-reported crash data alone are not sufficient to satisfy the study
requirement to examine the benefits of safety belts and motorcycle helmets with
respect to "the severity of injuries," the "... medical costs, including the sources of
reimbursement and the extent to which these sources cover actual costs," and "...
government, employer, and other costs." Information to address these requirements
must be found in other data sources.
CODES Report to Congress

- 4 -

February 1996

Data relating to the type and cost of medical treatment provided to crash victims
could come from several different medical service providers: emergency medical
service (EMS) treatment and transport, hospital emergency department (ED)
treatment, inpatient hospital treatment, nonhospital-based physician services,
rehabilitative/long-term care, etc. Each one of these providers may have its own
system for record keeping and billing. Some of these systems are statewide and some
are limited to the facility or office where the treatment occurs. None of the data
systems, however, are designed to be linked with other systems.
One statewide medical outcome database, a centralized database of inpatient
hospital discharge information, is available in most states. This file usually includes
diagnosis codes, procedure codes, and charges relating to the patient's hospital
confinement. This last item, hospital charges, includes only the institutional
component, not the professional fees, incurred at any of these facilities and does not
represent actual costs.
The availability at the state level of other medical outcome information for motor
vehicle crash victims is limited. Some states support statewide databases such as
EMS run reports, hospital emergency department records, or rehabilitative and longterm care facility treatment summaries. Not all states collect or build statewide files
from all these sources of information. But where they are available, these databases
and the hospital discharge database are a valuable information source for highway
safety research, particularly when an individual patient record can be associated with
data on a police crash report.
CODES Report to Congress

- 5 -

February 1996

NHTSA, therefore, elected to meet this study's data needs using police-reported
crashes as the study population. The police reports were used to identify crashes,
circumstances about the crashes, the vehicles, and the people involved. Patient
information from the available data sources, usually EMS, ED, hospital discharge, and
long-term and rehabilitative care databases, was used to identify injury outcomes and
charges for those injured. The different databases were linked to obtain populationbased occupant-specific outcome data for injured crash victims. Linking the databases
enabled the information about the injury-causing event (the motor vehicle crash) to be
directly related to specific medical and financial consequences for each person
involved in the crash. This detailed information was used to evaluate the failures and
successes resulting from the use or non-use of safety devices. Exhibit 2 shows
schematically the linking of databases in a state where all such databases are
available.

Traffic Records

Hospital
Discharge

Outpatient

Exhibit 2.

Insurance
Claims

Rehab/
Long
Term
Care

Death
Certificate

Crash and Injury Data Sources

CODES Report to Congress

- 6 -

February 1996

Grants
NHTSA elected to fulfill the legislative mandate by providing grants to several
states to link available databases and perform analyses using the linked data. Access
to the needed databases would be obtained by the grantee from the owner agencies.
The grantee would link the medical outcome databases to the police crash report file.
NHTSA decided not to fund any states which would need to create new databases and
instead focused on those that could guarantee delivery of the data for the analyses.
Additionally, to promote cooperation between owners and users of the required
databases, the grants would be made to a single applicant in a state who would be
responsible for obtaining cooperation from the owners of the data.
In a May 5, 1992, Federal Register announcement, NHTSA published a grant
solicitation requesting applications from states. The announcement specified that
applicants should have an available statewide crash report database, computerized
EMS, ED, hospital discharge, and rehabilitative/long-term care databases, and further,
be willing to collaborate with the owners of these databases, link the databases,
perform analyses as specified by NHTSA, and provide NHTSA with the linked data and
results of the analyses. Any state agency, nonprofit organization, or educational
institution capable of setting up a coalition of data owners and users to perform the
linkage was eligible to apply. Any representative period after January 1, 1990, was
allowed, to obtain the most recent data each state had available without unnecessarily
limiting the number of possible applicants. Twenty states responded to the solicitation.
Selection of the grantee states was independent of safety belt and helmet use rates
and current statutes mandating use. NHTSA titled the project “CODES”.

CODES Report to Congress

- 7 -

February 1996

On September 30, 1992, NHTSA awarded grants to conduct the study to
entities in 7 states: the Departments of Health in Missouri, New York, and
Pennsylvania; the Department of Public Safety in Maine, the Department of
Transportation in Wisconsin; and state university systems in Hawaii and Utah.
Besides having the needed data, the 7 states had other characteristics useful for this
study. All of the states except Maine had enacted mandatory safety belt use laws.
During the study period, helmet use laws, at various times adopted by all 7 states,
were in force in 3 states: Missouri, New York, and Pennsylvania. Each state had a
statewide police crash report database and a statewide hospital inpatient discharge
file. All states except Wisconsin had a statewide EMS database. Wisconsin had date
of birth and zip code of residence for drivers and injured passengers on its crash file,
which facilitated linkage, in the absence of EMS data, to the hospital discharge and
insurance claims files. Only Missouri and Utah had access to statewide emergency
department data. New York obtained emergency department data for New York City.
The other states attempted to obtain information for specific population groups from
insurance claims databases. Hawaii and New York were no-fault insurance states.
New York had access to statewide vehicle insurance claims information, and Hawaii
had access to statewide health insurance claims information.
To promote collaboration, each grantee formed an advisory committee of the
owners and users of the databases they intended to link. In most of the states, this
was the first time these diverse groups collaborated on highway safety issues. These
advisory committees addressed the problems of data accessibility, confidentiality,
uniformity, and quality. They worked to define potential uses for the data and to set up

CODES Report to Congress

- 8 -

February 1996

and review research projects whose results were included as part of the grantees' final
reports to NHTSA. In short, these committees worked to coordinate the components
which would promote the use of linked data for supporting injury control in the highway
safety environment for the current study and in the future.
As discussed in the preceding section, states differ in the amount and type of
data collected on police crash reports and coded onto their crash databases. Some of
these differences affect the CODES analyses. Of the CODES states, Missouri and
Wisconsin do not collect or code information about vehicle passengers who were not
injured in the crash. Thus, the safety belt benefit analyses presented in this report
were restricted to drivers. However, a comparison of the results for the belt analyses
for drivers compared to all occupants in the 5 states which had the data found no
significant differences. All of the states were able to generate the linked data.

Overview of Data Linkage
The CODES grantees linked their available databases using probabilistic
linkage, a computer algorithm which makes use of a combination of indirect identifiers
(age, date of birth, date of event, sex, etc.) and, when available, direct identifiers
(name, unique number, etc.). After completing the linkage, each state had a database
in which crash-involved occupants were linked to the available medical outcome
databases (EMS, emergency department, hospital discharge, and/or
rehabilitative/long-term care). Those occupants reported by the police as not injured
were less likely to be linked to a medical outcome record, while those reported as
having incapacitating injuries were more likely to be linked. Those in the middle (i.e.,

CODES Report to Congress

- 9 -

February 1996

occupants who were reported with possible or non-incapacitating injuries) had varying
linkage rates, since they were less likely than those with serious injuries to be
transported or treated at an inpatient facility, but very likely to receive outpatient
treatment. Thus, the final linkage rates varied according to the availability of outpatient
or claims data.
The success of the linking process was influenced by 2 factors. First, although
the probabilistic linkage algorithm increased the likelihood of accurate matches, some
databases lacked sufficient information to discriminate among the events and persons
involved. Because of privacy considerations, some of the data systems which collect a
person's name do not code it on their electronic database, and, therefore, it is not
available for linking. Having a victim's name or a common identifying number on each
record enhanced the accuracy and helped to increase the number of matches obtained
from the various files. Second, information about the external cause of injury (Ecodes) was not uniformly collected and coded by the hospitals. Thus, injury records
selected for linkage could not be restricted to only those with an E-code indicating
cause of injury as a motor vehicle crash. E-codes documented in the inpatient record
make it possible to exclude from the linkage those individuals not injured in motor
vehicle crashes.

Outcome Measures
To estimate the benefits of safety belts and motorcycle helmets with respect to
different categories of injury severity, outcome measures for crash-involved occupants
were established. A precursor to the actual outcome measures was a combined scale

CODES Report to Congress

- 10 -

February 1996

of "injury severity" and "treatment given" for each person in the linked files. This
preliminary scale was based on information generated from the linked data and is
shown in Exhibit 3 in increasing severity. Each crash-involved motor vehicle occupant
or motorcycle rider was coded into one of the 5 mutually exclusive categories.

Exhibit 3.
Severity/Treatment Definitions Used in the CODES Analysis of
Effectiveness of Safety Belts and Motorcycle Helmets
Severity/
Treatment

Definition

Not Injured

Reported by the police either as possible injury or not injured and
did not link to a medical outcome record

Slightly
Injured

Reported by the police as injured (except possible injury) but did
not link to a medical outcome record or as possible injury and
linked to an insurance claim record

Transported

Linked to an EMS and/or Emergency Department record but was
not linked to a hospital inpatient record

Inpatient

Linked to medical outcome record indicating inpatient treatment
(acute, rehabilitative and/or long-term care)

Died

Police-reported killed or linked to a medical outcome record
indicating death within 30 days after the crash as a result of the
crash

The actual outcome measures derived from this scale were:
(1) -- Died;
(2) -- Died or inpatient;
(3) -- Died, inpatient, or transported;
(4) -- Any injury (Died, inpatient, transported, or slightly injured).
These dichotomous measures permitted the grantees to use logistic regression
models for the analysis.

CODES Report to Congress

- 11 -

February 1996

The linking of the various databases in the grantee states produced a large
number of crashes which contributed 879,670 passenger vehicle drivers and 10,353
motorcyclists for this study. Exhibit 4 shows the distribution of crash-involved
passenger vehicle drivers and motorcycle riders by the severity/treatment levels listed
in Exhibit 3. The distribution of motorcycle riders is different because motorcycle riders
are more likely than passenger vehicle drivers to be injured in police-reported crashes.

Exhibit 4.
Number of Drivers and Motorcycle Riders Contributing to
the CODES Analysis of Effectiveness of Safety Belts and
Motorcycle Helmets, by Severity/Treatment Levels
Severity/Treatment Levels

Passenger
Vehicle
Drivers

Motorcycle
Riders

703,319

2,892

Slightly Injured

81,353

3,128

Transported

78,054

2,378

Inpatient

14,599

1,604

2,345

351

879,670

10,353

Not Injured

Died
Total

Methodology
To evaluate whether safety belts and motorcycle helmets are beneficial in
reducing mortality, morbidity and injury severity, NHTSA used a measure employed in
many previous studies: effectiveness. Effectiveness is defined as the percentage
reduction in injuries or deaths for people wearing safety belts or helmets compared to
people not wearing safety belts or helmets. For example, if the effectiveness of some

CODES Report to Congress

- 12 -

February 1996

device in reducing injuries is 35 percent, then 35 percent of those people who were
injured while not using the device would not have been injured had they used it. In the
CODES study, the group of crash victims (passenger vehicle drivers for belt
effectiveness and all motorcycle riders for helmets) who died was compared with all
other crash-involved victims to estimate effectiveness in reducing mortality (outcome
measure 1). To assess the effectiveness in reducing morbidity (outcome measure
4), the group of crash victims experiencing any injury, i.e., slightly injured, transported,
inpatient, or died, was compared with those not injured. Finally, to assess the benefits
in reducing injury severity, the effectiveness for all outcome measures was analyzed.
To provide the input for computing effectiveness, each CODES grantee was
required to perform a series of regression analyses on its linked data using the
outcome measures listed above. The results of each of these analyses were provided
to NHTSA by the grantees as part of their final project reports.
To estimate the effect of safety belt and motorcycle helmet use on medical
costs, costs were limited to total inpatient charges found on hospital, rehabilitative,
and long-term care patient records. Outpatient and non-medical charges were not
available from the patient medical record nor uniformly available from other sources in
the CODES states. Inpatient charges represent about 60 percent of the total direct
medical expenses. The remaining 40 percent include 25 percent for physician
charges, about 4 percent for emergency room charges, about 1 percent for EMS
charges, and 10 percent for other charges. 2 In addition to the direct medical charges
paid by the people who are injured in motor vehicle crashes, there are external costs
which are paid, sometimes by the public, when someone fails to buckle up or wear a
helmet. 3 Information on these costs are not readily available from the patient unit

CODES Report to Congress

- 13 -

February 1996

record.
To provide the input for comparing average charges, each grantee computed
average inpatient charges for passenger vehicle drivers and for motorcycle riders.
Averages were calculated based on the victims’ belt or helmet use and for various
payers. The results of each of these analyses also were provided to NHTSA by the
grantees as part of their final project reports.
The results presented in this report were statistically combined by NHTSA staff
from the individual state data provided by the grantees. To obtain overall estimates of
effectiveness, state estimates were weighted by the inverse of their standard errors.
For overall estimates of average charges, each state estimate was weighted by the
number of observations which contributed to it. Because these methods of computing
averages are strongly influenced by the number of cases, data from the larger states
(Missouri, New York, Pennsylvania, and Wisconsin) may disproportionately influence
the overall weighted average. However, these methods made it possible to compare
the association between injury level and safety belt/helmet use in each state. For
individual state estimates the reader is referred to the CODES Technical Report which
describes the linkage and analytical processes used in the CODES project.
The reader should note that the overall results presented in this report reflect
only the 7 CODES states and the case selection criteria for the safety belt and
motorcycle helmet analyses. They are not intended to be nationally representative.

In

addition, except as noted, the financial results were based only on data from the linked
inpatient records. These linked records represent a sub-population of all occupants
who were injured and required inpatient care as the result of a motor vehicle crash.
Consequently, financial results generated from this population of linked records vary

CODES Report to Congress

- 14 -

February 1996

from those results generated from a previous study based on all inpatient discharges
with any E-code for a motor vehicle crash from 6 states, only one of which also
participated in the CODES project. 4

Contributing Risk Factors
Assessing whether or not safety belts and motorcycle helmets are effective in
reducing the occurrences of occupant mortality and morbidity, and in reducing the
severity of injury, is not a simple task. While it would be quite easy to compare the
observed injury rates among drivers using safety belts, real differences in injuryreducing effectiveness can be masked by a multitude of factors, (e.g., driver age and
sex) not directly related to safety belt or motorcycle helmet use themselves. To the
extent possible, contributing risk factors need to be identified and included in the
analysis. Exhibit 5 presents the risk factors which were available in all of the CODES
states.

Exhibit 5.
Contributing Risk Factors Available for the CODES
Safety Belt and Motorcycle Helmet Analyses
Analysis
Factor

Safety Belt

Motorcycle
Helmet
Not Used

Type of Crash
Rural/Urban
Age
Male/Female
Posted Speed Limit
Wet/Dry

CODES Report to Congress

- 15 -

February 1996

Night/Day
Intersection
Related
Vehicle Type

Not Used

Seating Position

Not Used

Failure to control for these risk factors might have led to incorrect estimates of
safety belt or helmet effectiveness. Estimates of effectiveness without these factors
incorporated were generated as part of the CODES analysis. These estimates of
effectiveness with and without risk factors were not significantly different in most of the
states.

RESULTS -- BENEFITS OF SAFETY BELT USE
Effectiveness
The results of the analyses for the effectiveness of safety belts in preventing
death and injury are summarized in Exhibit 6.

Exhibit 6.
Safety Belt Effectiveness by Outcome
For Crash-Involved Drivers in the CODES States *
Effectiveness Estimates
Outcome
Measure

Belt Use as
Reported by
Police

Adjusted for
Overreporting of Belt
Use

Died

89%

60%

Died or Inpatient

75%

45%

Died, Inpatient, or
Transported

54%

30%

CODES Report to Congress

- 16 -

February 1996

Any Injury

52%

20%

*Please read the discussion of the safety belt analyses on pages 21-23.

Given involvement in a crash, safety belts are effective both in preventing any
injury and in reducing the likelihood of a fatality. This indicates that, in the 7 CODES
states during the study years, many of those unbelted drivers injured or killed in policereported crashes would not have been injured or killed had they been belted.
However, because of the likelihood of crash involved vehicle occupants to over-report
their safety belt use, it is believed that the study results using as-reported belt use are
inflated. Consequently, estimates adjusted for expected over-reporting were
developed (see the section beginning on page 21 for a discussion on over-reporting).
Safety belts also are effective in reducing the severity of injuries, as evidenced by the
effectiveness rates which increased as the severity of the outcome increases. The
information in Exhibit 6 implies that safety belts cause a downward shift in the severity
of injuries sustained by vehicle drivers injured in a crash; that is, wearing safety belts
results in fewer injuries and less severe injuries on average than would be sustained if
the drivers do not wear safety belts.

Cost of Crash Injuries
Average inpatient charges for passenger vehicle drivers are presented in Exhibit
7. The exhibit also shows the average charges adjusted for the likelihood of being
admitted to an inpatient facility, that is, an average charge for all crash-involved belted
and unbelted drivers.

CODES Report to Congress

- 17 -

February 1996

Exhibit 7.
Average Inpatient Charge by Safety Belt Use for Inpatient and
All Crash-Involved Passenger Vehicle Drivers in the CODES
States
Safety Belt Use
Group

Inpatient Victims
All Crash-Involved Drivers

Used

Not
Used

Increase for
Not Using Belts

$9,004

$13,937

55%

$110

$562

408%

The difference in average charges (total charges hide the difference between
the belted and unbelted when far more vehicle occupants use safety belts than do not)
between the belted and unbelted groups indicates that unbelted inpatient drivers
experienced higher charges reflecting more severe injuries and longer lengths of stays.
The average inpatient charge is 55 percent higher for crash victims needing inpatient
care who did not use their safety belt than for those who did, a savings of almost
$5,000 for each belted inpatient. When total charges are distributed among all crashinvolved drivers, the average charge, although smaller in magnitude, shows an even
larger percentage increase for unbelted drivers due to the increased chance of
inpatient care. The overall average inpatient charge for all crash-involved victims
increased by 408 percent for unbelted drivers.

CODES Report to Congress

- 18 -

February 1996

Comparison of Charge to Cost
As mentioned earlier, this analysis of the benefits of safety belts was based on
the charge information available from the hospital discharge and other (rehabilitative
and long-term care) inpatient databases. Total charges reported in these databases
are higher than the actual cost to the provider. They include a markup factor
established by each health care provider to ensure the overall viability of the facility by
covering bad debts, cost shifting among the payers, and profit/surplus revenue. The
actual cost of care associated with a specific inpatient stay is very difficult to determine
because actual costs are usually not patient-specific or routinely calculated.
To provide an order-of-magnitude estimate for costs incurred by motor vehicle
crash victims requiring inpatient care, actual costs were estimated separately for each
state using charge-to-cost ratios based on data that are obtained from statewide
Medicare cost reports 5. For the CODES states , the ratios for the CODES data
collection year range from a low of 1.15 to a high of 1.71. Total charges in these states
were $164.4 million. Using the charge-to-cost ratios results in an estimated total cost
of $114.5 million. Estimated savings were then computed and are shown graphically in
Exhibit 8. If all drivers involved in police-reported crashes had been wearing a safety
belt, the savings could be represented as approximately $68 million in reduced
inpatient charges or $47 million in reduced inpatient costs, both 41 percent reductions.

CODES Report to Congress

- 19 -

February 1996

200
Total Charges
Estimated Costs
150

100

50

0
As Reported

Exhibit 8.

All Belted
Safety Belt Use

Savings

Estimated Total Inpatient Charges and Costs for All
CODES States: (left) Given the Existing Mix of Belted and
Unbelted Drivers As Reported; (center) If All Drivers Were
Belted; and (right) Resulting Savings.

Charges by Payer Source
Hospitals and rehabilitative and long-term care facilities seek payment for
charges from several sources. Private health insurance companies, including
Worker’s Compensation, are usually the primary source. The taxpayer is another
source of payment through government programs such as Medicare and Medicaid.
Victims without medical insurance are included in the other category. These "selfpayers" often are unable to pay their bills, and the cost of providing this care is passed
on through higher charges for those who do pay. Exhibit 9 shows the total amount of
charges by each payer source and average inpatient charges by type of payer and
safety belt use.

CODES Report to Congress

- 20 -

February 1996

Exhibit 9.
Average Inpatient Charge and Total Inpatient Charges
by Source of Payment and Safety Belt Use
for Crash-Involved Drivers in the CODES States
Average Inpatient Charge
Source of
Payment

Safety Belt Use
Used

Public 1

Not
Used

Difference

Total
Inpatient
Charges

$13,322

$18,922

$5,600

$26,498,675

Private Insurance 2

$8,581

$14,058

$5,477

$113,156,421

Other 3

$8,180

$10,534

$2,354

$24,788,922

1

Includes all charges to Government Funded Sources including Medicaid, Medicare, etc.
Private Insurance Companies including Worker’s Compensation
3
Usually Self Payment
2

At the time of discharge, private insurance, including Worker’s Compensation,
was the payer for 69 percent of all inpatient charges. Public sources, usually Medicare
and Medicaid, and other government sources , accounted for 16 percent. The balance
(15 percent) was in the other group. Regardless of pay source, the average charge for
an inpatient who was not using a safety belt was higher than the charge for a belted
inpatient. The average charge for unbelted drivers in the private insurance payer group
was 64 percent higher than for those drivers using safety belts. For those not wearing
safety belts in the public payer group the average charge was 42 percent higher than
for the belted public payer group. For the other group, the difference was 29 percent.
Note that the most severely injured people who become medically needy can apply for
Medicaid as a result of their injuries. Therefore, the reader is cautioned not to draw
any unwarranted inferences about higher charges to public payers, a subject that was
not studied in this project.

CODES Report to Congress

- 21 -

February 1996

Discussion of Safety Belt Analyses
NHTSA has completed many analyses of the effectiveness of safety belts.
Based on these studies NHTSA believes that the effectiveness of safety belts is in the
range of 40-50 percent for preventing mortality and in the range of 45-55 percent for
preventing morbidity 6. These estimates are not entirely consistent with those produced
from the CODES analyses, and NHTSA believes that the CODES results may be
inflated by over-reporting of belt use on the police crash reports. For most persons
involved in crashes, belt use is self-reported. That is, the police do not observe it.
During the past 10 years while safety belt laws have been put into force, there has
been an increased tendency toward belt over-reporting, i.e., occupants tell the police
officer they were belted when they were not. Over-reporting may be due to the
presence of legal penalties for non-use of safety belts, to discounts offered by some
automobile insurance companies for a signed commitment that the policy holder will
always use his or her safety belt, or to other reasons. NHTSA's estimates of 40-50
percent for preventing mortality and 45-55 percent for preventing morbidity were based
on data obtained prior to efforts to increase belt use through the passage of belt use
laws and insurance incentives which are believed to have contributed to corresponding
increases in observed belt use. Therefore, they are not as likely to be inflated by overreporting.
Belt use rates are higher for drivers in police-reported crashes used in the
CODES analysis than for drivers observed in the general motoring public. Exhibit 10
shows reported belt use, averaged from data for all of the CODES states, for the
different levels of police-reported severity. These data were the most current at the
time of the study: 1990 in Hawaii and Missouri, 1991 in Maine, Pennsylvania, Utah
and Wisconsin, and 1992 in New York. For comparison, the estimated national use

CODES Report to Congress

- 22 -

February 1996

100

80

60

40

20

0
None

PossibleNonincapacitaing Severe

Fatal

Police Reported Injury Severity
Exhibit 10. Police-Reported Belt Use, by Injury Severity for CrashInvolved Drivers in the CODES States. The horizontal line
at 51 percent is the average from NHTSA's '19 Cities
Survey.'

rate for 1991 based on observational data from NHTSA's 19 Cities Survey was 51
percent.

It is unlikely that crash-involved drivers would have higher safety belt use

rates than the general motoring public, since behavior that leads to increased risk of
crash involvement is associated with risks such as driving after drinking or not wearing
safety belts. The effect of higher reported safety belt use, especially among the less
seriously injured and uninjured vehicle drivers, is to make safety belts appear more
effective than they actually are.
Comparison of the observed and reported belt use rates for the CODES states
suggests that 35 percent of the belted who are uninjured or slightly injured may have

CODES Report to Congress

- 23 -

February 1996

misreported their belt use at the time of the crash. When this over-reporting rate is
factored into the analyses, the effectiveness rates are closer to 60 percent for
preventing mortality and 25 - 45 percent for preventing morbidity, estimates much
closer to NHTSA's current estimates of belt effectiveness.
For the average inpatient charges reported in Exhibit 7, it is possible to assume
that an injury severe enough to require inpatient care may be more likely to provide
tangible evidence of belt use or non-use, thus discouraging the tendency to overreport. Therefore, the 55 percent figure would stay approximately the same.
For the average charges for the group of all crash-involved drivers, the situation
is much more complicated, because all drivers were included. Assuming a 10 percent
over-reporting by the inpatient drivers, and a 35 percent over-reporting by non-inpatient
drivers, then the percent cost benefit of wearing a belt for crash-involved drivers
approaches the same for drivers receiving inpatient care, or 60-70 percent.

RESULTS -- BENEFITS OF MOTORCYCLE HELMET USE
Effectiveness
In the analysis for helmet effectiveness, NHTSA used data from six of the 7
states for which data were available. Utah was excluded because there was no place
on their police crash report to identify an unhelmeted motorcyclist -- all records were
coded either "helmet worn" or "unknown." Also, all risk factors except helmet use were
excluded to avoid reducing the sample size excessively due to missing data. In the
CODES Report to Congress

- 24 -

February 1996

safety-belt analyses, dropping all the risk factors except safety-belt use had no
significant effect on the results. In the helmet analyses, if only cases with no missing
data for all risk factors were used, the results would have been extremely unstable due
to the small number of cases. The combined results for motorcycle helmet
effectiveness are summarized in Exhibit 11.

Exhibit 11.
Effectiveness of Motorcycle Helmets by Outcome
Measure
for Crash-Involved Motorcycle Riders in CODES States*
Outcome Measure

Effectiveness

Died

35%

Died or Inpatient

26%

Died, Inpatient, or Transported

26%

Any Injury
*Excludes Utah.

9%

Given involvement in a crash, CODES results show that motorcycle helmets are
35 percent effective in preventing fatality, but only 9 percent effective in preventing any
injury. Effectiveness of motorcycle helmets was higher for more severe injuries, the
same pattern that was found for safety belts. However, motorcycle helmets were not
designed to prevent injuries other than head injuries.

Cost of Crash Injuries to Motorcycle Riders
An analysis similar to the one described previously for safety belts was done to
determine the benefits of motorcycle helmets with respect to charges incurred by
CODES Report to Congress

- 25 -

February 1996

those injured. Again, the analysis has been restricted to the charge information
available from the inpatient databases. A summary of these results is presented in
Exhibit 12.

Exhibit 12.
Average Inpatient Charge by Motorcycle Helmet Use
For Inpatient Victims and All Crash-Involved
Motorcycle Riders in CODES States*
Group

Motorcycle Helmet
Use
Used

Inpatient Victims
All Crash Involved
Riders
*Excludes Utah.

Not Used

Increase
for Not
Wearing
Helmet

$14,377

$15,578

8%

$2,064

$2,808

36%

The average inpatient charge for motorcycle crash victims receiving inpatient
care was $14,377 for those who used helmets, and $15,578 for those who did not, an 8
percent increase ($1,201) in charges for those electing to not wear a helmet. When
adjusted for all crash-involved motorcycle riders, the difference in the average inpatient
charge, $744, is 36 percent higher for those not wearing a helmet. These differences
are not as dramatic as those seen for safety belts. This is likely to be an effect of both
the smaller sample sizes involved and the likelihood that motorcycle riders will be
injured in a crash, regardless of whether or not they are wearing a helmet, in large part
because there is little or no protective vehicle structure.

Comparison of Charge to Cost

CODES Report to Congress

- 26 -

February 1996

The same charge-to-cost ratios by state as used in the safety belt analysis were
applied to the data from the 6 states which contributed to the motorcycle helmet
analysis. If all motorcycle riders involved in police-reported motorcycle crashes had
been wearing a motorcycle helmet, about 3 percent of total inpatient charges or 7
percent of actual costs would have been saved. This result is shown graphically in
Exhibit 13.

30
Total Charges
Estimated Costs

25
20
15
10
5
0

As Reported
All Helmeted
Motorcycle Helmet Use

Savings

Exhibit 13. Estimated Total Inpatient Charges and Costs for 6 CODES
States (Excluding Utah): (left) Given the Existing Mix of
Helmeted and Unhelmeted Riders As Reported; (center) If
All Riders Were Helmeted; and (right) Resulting Savings.

Charges by Payer Source
Exhibit 14 presents the average inpatient charge by type of payer and
motorcycle helmet use for inpatient motorcycle crash victims.

CODES Report to Congress

- 27 -

February 1996

Exhibit 14.
Average Inpatient Charge and Total Inpatient Charges
by Source of Payment and Motorcycle Helmet Use
for Crash-Involved Motorcycle Riders in CODES States*
Average Inpatient Charge
Source of
Payment

Motorcycle Helmet
Use
Used

Difference

Total
Inpatient
Charges

Not
Used

Public 1

$23,793

$24,925

$1,132

$5,364,759

Private Insurance 2

$13,617

$15,687

$2,070

$14,764,706

Other 3

$10,565

$8,913

($1,653)

$3,403,183

*Excludes Utah.
1
Includes All Charges to Government Funded Sources including Medicaid, Medicare, etc.
2
Private Insurance Companies and Worker’s Compensation
3
Usually Self Payment

The motorcycle helmet results are similar to those shown for the safety belt
analysis except for the other category. At the time of discharge, the expected pay
source for a majority of charges was a private insurance company. Billings to private
insurance companies covered 63 percent of all charges, and the average charge for an
unhelmeted motorcyclist was almost 15 percent higher than the charge for a helmeted
rider in this group. On the other hand, the average charge for an unhelmeted
motorcycle occupant in the other payer group was lower than for helmeted riders.
Public sector sources covered about 23 percent of the inpatient charges for
motorcycle crash victims in the 6 states in this analysis. The average inpatient charge
for those injured motorcyclists who used a public payer source was more than 5
percent higher for motorcyclists who did not wear a helmet than for those who did.

CODES Report to Congress

- 28 -

February 1996

Prevention of Brain Injury
Helmets were not designed to protect the rider from most types of injuries which
could affect a motorcycle rider. Their main function is to reduce injuries to the head
and, especially, the brain. Brain injury is more likely to result in expensive and longlasting treatment, sometimes resulting in lifelong disability, whereas other head injuries,
concussions and skull fractures (without damage to the brain itself), are more likely to
result in full recovery. To examine whether motorcycle helmets would be more
effective in reducing the injuries they were designed to prevent, NHTSA performed a
separate analysis restricting the outcome measure to whether or not the motorcyclist
received inpatient care for a brain injury. One state, Wisconsin, had subdivided its
inpatients with head injuries into brain injury, concussion, and simple skull fracture
groups. The inpatient files from 5 other states were added to Wisconsin's data
following the definitions used by Wisconsin. Again, Utah was not used since the crash
report did not include a code for not wearing a helmet. The model was revised to
include only helmet use as a risk factor to maximize the number of cases which could
be included in the analysis.
The resulting analysis of effectiveness revealed that motorcycle helmets are 67
percent effective in preventing brain injuries. Thus, if all motorcyclists had been
wearing helmets, 67 percent of those unhelmeted motorcyclists who received inpatient
care for a brain injury would not have sustained the brain injury. In other words,
unhelmeted motorcyclists were over three times as likely to suffer a brain injury as
were helmeted motorcyclists.
Average charges for inpatient motorcycle riders by brain injury status and
helmet use are shown in Exhibit 15.

CODES Report to Congress

- 29 -

February 1996

Exhibit 15.
Average Inpatient Charge
by Motorcycle Helmet Use and Brain Injury Status
For Inpatient Motorcycle Riders in CODES States*
Motorcycle Helmet Use
Group/
Payer
Source

Used

Not Used

Brain
Injured

Not
Brain
Injured

Brain
Injured

Not
Brain
Injured

$26,985

$12,736

$26,805

$11,730

Public

$33,764

$22,066

$46,347

$11,596

Private

$29,610

$11,834

$24,461

$12,807

$16,664
*Excludes Utah.

$9,585

$10,238

$8,593

All Inpatient Victims

Other

Regardless of helmet use and payer source, the average charge for inpatient
care for a motorcyclist who sustained a brain injury is more than twice the average
charge for motorcyclists receiving inpatient care for other injuries. Inpatient charges for
unhelmeted motorcyclists receiving care for a brain injury ($26,805) are 2¼ times
greater than the average charge for care for unhelmeted inpatient motorcyclists not
sustaining a brain injury ($11,730). Therefore, if all motorcyclists wore helmets,
approximately $15,000 in inpatient charges would be saved during the first 12 months
for every motorcycle rider who, due to wearing the helmet, did not sustain a brain
injury. Additional savings would accrue from avoiding the continual costs for care over
a lifetime.

Discussion of Motorcycle Helmet Analysis
Regarding the effectiveness of motorcycle helmets in reducing fatalities and
CODES Report to Congress

- 30 -

February 1996

injuries, the results are also consistent with previous analyses NHTSA has conducted.
The 34 percent figure in Exhibit 11 is very close to a 1989 NHTSA analysis

7

which

estimated that motorcycle helmets were 29 percent effective against fatal injury. The
minimal effectiveness of helmets when lesser injuries are added to the analysis should
not be surprising. Helmets prevent head injury, not all injuries.
This makes the analysis of brain injury all the more important, because it shows
that helmets are effective in reducing the types of injury they were designed to reduce.
They were 67 percent effective in the 6 selected states, which is more than twice the
fatality effectiveness. Helmets also reduce the cost where it counts. In these 6 states,
cases with brain injury were more than twice as costly as non-brain injury during the
first 12 months.
With motorcycle helmets, the over-reporting problem does not exist, because it
is easier to see "helmet use" than "belt use." There is no substantial group of
motorcyclists claiming they were wearing helmets when they were not. No
adjustments need be contemplated as in the safety-belt analysis. However, there is a
problem with missing data on motorcycle helmet use. New York, with a helmet-use
law, showed "unknown helmet use" on 38 percent of its motorcyclist records.
Wisconsin, without a helmet-use law, showed only 9 percent. In general, states with
laws are more likely to have missing data. Police may be reluctant to give a ticket for
not wearing a helmet to a motorcyclist who has just suffered a crash.

CODES Report to Congress

- 31 -

February 1996

CONCLUSIONS
Study Results
The CODES study results confirm earlier NHTSA analyses that safety belts are
highly effective in preventing injury and fatality in motor vehicle traffic crashes, and
that they cause a downward shift in the severity of injuries. The earlier NHTSA studies
estimated that safety belts are 40 - 50 percent effective in preventing mortality and 45 55 percent effective in preventing morbidity.
The CODES results also show that the average inpatient charge for a driver
admitted to an inpatient facility as a result of a motor vehicle injury is 55 percent higher
if that person was unbelted. For the 7 states contributing data for this study, the
average charge for a belted crash-involved driver receiving inpatient care was $9,004,
while the average charge for an unbelted driver was $13,937. If, in these states, all
unbelted passenger vehicle drivers had been wearing safety belts, it is estimated that
the reduction in inpatient charges would have been approximately $68 million, or an
estimated $47 million in actual inpatient costs. Private insurance accounted for 69
percent of the inpatient charges compared to 16 percent for public and 15 percent for
other sources. In all cases, the average inpatient charge was greater for drivers who
were unbelted.
The study results confirmed NHTSA's estimate of motorcycle helmet
effectiveness (29 percent) in preventing fatalities. CODES results showed that
helmets were 35 percent effective in preventing a fatality. The average inpatient
charge for motorcycle crash victims was $14,377 for those who used helmets, and
$15,578 for those who did not, an 8 percent increase for those electing not to wear a
CODES Report to Congress

- 32 -

February 1996

helmet. Seven percent in actual inpatient costs, and more in lifetime care, would be
saved if unhelmeted motorcyclists wore helmets. Private insurance sources
accounted for 63 percent of inpatient charges compared to 23 percent for public and 14
percent for other sources. For both the private and public sources, average inpatient
charges for motorcycle crash victims were greater for the unhelmeted.
Helmets cannot protect the rider from most types of injuries. However, further
analysis revealed that motorcycle helmets are 67 percent effective in preventing brain
injuries. In other words, unhelmeted motorcyclists were over three times as likely to
suffer a brain injury as were helmeted motorcyclists.
Examination of the average inpatient charges revealed that the average charge
for inpatient care for a motorcyclist who sustained a brain injury is more than twice the
average charge for motorcyclists receiving inpatient care for other injuries. On
average, approximately $15,000 in inpatient costs would be saved during the first 12
months for every injured motorcycle rider who did not also sustain a brain injury.
Therefore, if all motorcycle riders wore helmets, fewer victims would incur the high cost
of inpatient care associated with brain injury.

Significance of CODES
The linked data provide unique insights into the financial outcome of highway
crashes. Police crash reports provide information about the crash environment and
driver/occupants; EMS reports and hospital discharge data add medical information
about injury type and severity; and hospital discharge and insurance claims data reveal
the financial consequences. Taken together, these linked data generated greater

CODES Report to Congress

- 33 -

February 1996

value than when considered alone.
Hundreds of thousands of police-reported crashes were included in the
statewide linked data. This large volume of vehicle-related information increases
the available statistical power to discriminate among specific vehicle attributes while
controlling for nonvehicle-related factors making it possible to generate cost benefit
analyses of vehicle safety performance standards.
An important concern of the public health community relates to the availability
of medical services and their impact on outcome. The availability of linked injury and
crash information supports collaboration between the nonmedical and medical
communities. These data can be used to demonstrate the effectiveness of the
emergency response by police, EMS, and the acute care system, and to predict the
need for an aggressive medical response when specific crash, vehicle, and behavioral
characteristics exist.
NHTSA often examines state data to evaluate the benefits of specific traffic
safety countermeasures. CODES linked data allow the agency to examine not only a
more accurate description of injury consequences, but also the public health cost
savings associated with highway safety initiatives. Since a high percentage of these
costs are funded by citizens through increased taxes to cover the expenses of
uninsured and underinsured crash victims, documentation of the costs is important to
motivate public and legislative support for stricter laws and enforcement actions.
CODES provides documentation, generated from a state's own linked data, that is
more credible among local decision makers who may be tempted to repeal the safety
mandates, such as helmet legislation. CODES information has the capability to

CODES Report to Congress

- 34 -

February 1996

demonstrate the increased costs associated with head injuries for unhelmeted riders,
to identify the health care costs for specific vehicles, crashes, and behaviors (e.g.,
alcohol involvement, unsafe driving actions), to generate community-based information
to support community-based traffic safety programs, and to target specific populations
at risk at the local, regional, or state levels. All of this information identifies and
supports outcome-based injury control activities that have the most potential for
reducing health care costs.

CODES Report to Congress

- 35 -

February 1996

FOOTNOTES
1. Sensitivity Index Demonstration Project--Final Report. Johnson, S. W. And
VanSciver, E. Washington, DC: National Association of Governors’ Highway Safety
Representatives, 1991 .
2. Dorothy P. Rice, Ellen J. MacKenzie, and Associates. Cost of Injury in the United
States. A Report to Congress 1989. The Johns Hopkins University, 1989.
3. Testimony of Ted Miller before House Subcommittee on Surface Transportation,
March 3, 1994.
4. Harris, Joan A. Source of Payment for the Medical Cost of Motor Vehicle Injuries in
the United States. DOT HS 807 800 National Technical Information Service,
Springfield, Virginia, 1992.
5. Almanac of Hospital Financial and Operating Indicators. Cleverley, William O.
Center for Health Care Industry Studies, Ohio State University (published annually).
6. Final Regulatory Analysis, Amendment to FMVSS No.208, Passenger Car Front
Seat Occupant Protection. DOT, NHTSA, Office of Plans and Policy, July, 1984.
7. The Effectiveness of Motorcycle Helmets in Preventing Fatalities. Wilson, D. DOT
HS 807 416, NHTSA Technical Report, March, 1989.

CODES Report to Congress

- 36 -

February 1996

DOT HS 808 347
NRD-31
2/96