Thurman D, Guerrero J. Trends in Hospitalization Associated With Traumatic Brain Injury. JAMA. 1999;282(10):954-957. doi:10.1001/jama.282.10.954
Author Affiliations: National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, Ga.
Context Traumatic brain injury (TBI) is associated with more than 50,000 deaths
in the United States each year, and recent observations suggest a substantial
decline in TBI-related hospitalizations and deaths.
Objective To analyze long-term trends in TBI-related hospitalization in the United
Design, Setting, and Participants Analysis of existing data from 1980 through 1995 from the National Hospital
Discharge Survey, an annual survey representing the US general population.
The number of participating hospitals ranged from 400 to 494.
Main Outcome Measures Annual rates of TBI-related hospitalization, stratified by age, sex,
severity of injury, and outcome.
Results The annual number of TBI cases identified from the sample during the
study period ranged from 1611 to 3129. Overall rates of hospitalization for
TBI declined an estimated 51%, from 199 to 98 per 100,000 per year. When analyzed
by severity of injury, mild TBIs declined most during this period, from 130
to 51 hospitalizations per 100,000 per year (61% decline; P<.001 compared with intermediate and severe TBI). The decline was
greatest among those aged 5-14 years (−66%) and least among those aged
65 years or older (−9%). The ratio of male to female rates showed little
variation during the study period (ratio, 1.8; 95% confidence interval [CI],
1.6-2.0), as did the in-hospital mortality rate (mean, 5.3 per 100,000; 95%
CONCLUSIONS Changes in hospital practices may be a major factor in the declining
rates of TBI-related hospital admissions. These practices increasingly appear
to exclude persons with less severe TBI from hospital admission and shift
their care to outpatient settings.
In the United States, traumatic brain injury (TBI) is associated with
the death of 51,000 people each year, about one third of all injury deaths.1 Traumatic brain injury is a cause of long-term disability
that annually affects an estimated 70,000 to 90,000 people.1,2
Given these serious consequences and the large number of people affected,
public health efforts to prevent the occurrence and mitigate the consequences
of TBI have received increased attention in recent years.3
Some success in TBI prevention and treatment efforts is suggested by
an analysis of national mortality trends, which indicates a 22% decline in
rates of death associated with TBI from 1979 to 1992.4
To our knowledge, there is no published study describing national trends in
TBI-related hospitalizations over a comparable period, although comparisons
of some studies in selected localities suggest the rate of these hospitalizations
has declined more steeply.5 The purpose of
this study is to describe long-term trends in TBI-associated hospitalization
in the United States; to examine trends in the distribution of these injuries
by age, sex, level of severity, and outcome; and to assess the implications
of these trends with respect to public health and health care delivery.
To estimate annual rates of TBI-associated hospitalization, we analyzed
data collected from 1980 through 1995 by the National Hospital Discharge Survey6 of the Centers for Disease Control and Prevention
National Center for Health Statistics. Each year during this interval, the
number of hospitals participating in the survey ranged from 400 to 494, and
the sample size ranged from 181,000 to 276,000 hospitalizations for all causes.
From 1980 through 1995, the annual number of cases of TBI identified in the
sample ranged from 1611 to 3129. These TBI cases were identified using the
Centers for Disease Control and Prevention case definition,7
ie, all hospital discharge records that contained—in 1 or more of 7
diagnostic data fields—an International Classification
of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)8 diagnostic code in the ranges
800.0-801.9, 803.0-804.9, or 850.0-854.1. The National Hospital Discharge
Survey used a multistage probability sampling design with weights applied
to records to adjust for the probability of selection and nonresponse. These
weighted records were used to generate US population estimates for the total
number of hospitalizations associated with TBI each year. Relative SEs for
annual population estimates, which ranged from 5% to 8%, were calculated using
methods described by the National Center for Health Statistics.6
To classify the severity of TBIs among patients who were hospitalized,
we used ICDMAP-90,9 a computer algorithm that
maps ICD-9-CM diagnostic codes to a 6-level score
approximating the Abbreviated Injury Scale (ICD/AIS).10
We then reclassified the ICD/AIS scores for the head region to a 3-level scale:
ICD/AIS scores of 1 or 2 were defined as mild, 3 as moderate, and 4 through
6 as severe. We could not perform an extensive analysis of TBI incidence by
cause of injury, because such data were incomplete; ICD-9-CM external-cause-of-injury codes (E codes) were present in only 10%
to 47% of cases during the period of this study.
We used commercially available statistical software to calculate crude
hospitalization incidence rates for each year, using as denominators the US
Census estimates for the national civilian population at midyears during this
interval. Generalized variance estimation functions that were provided by
the National Center for Health Statistics were used to generate confidence
intervals (CIs) for rates. Estimated crude hospitalization incidence rates
by outcome status, injury severity, sex, and age by year, as well as average
rates for the entire time interval, are presented. We compared 2-year aggregate
data from 1980-1981 and 1994-1995 to measure changes between end points of
the entire time interval, using z scores to test
for differences in rates. To test for differences in percent change of hospitalization
rates between categories of TBI severity, we calculated z scores using variance approximation methods.11
To compare estimated rates of TBI-associated hospitalization with TBI-associated
emergency department (ED) visits, we used data collected by the National Hospital
Ambulatory Medical Care Survey from 1992 (the first year of the survey) to
1995.12 Because the variance estimates associated
with individual years of the survey were large, we calculated the average
annual rate of TBI-associated ED visits for the interval 1992 to 1995 using
methods described elsewhere.13 As with the
analysis of National Hospital Discharge Survey data, TBI cases were identified
among National Hospital Ambulatory Medical Care Survey records by searching
all 3 diagnostic fields for ICD-9-CM codes consistent
with the Centers for Disease Control and Prevention case definition. Traumatic
brain injury–associated ED visits in 1995 that included a computed tomography
(CT) scan evaluation of any body region were identified by the presence of
a CT procedure code in the case record. Our analyses of National Hospital
Ambulatory Medical Care Survey data included only those TBI-associated ED
visits that did not lead to hospital admission or death.
From 1980 to 1995, the estimated annual incidence rate of hospitalization
associated with TBI declined 51%, from 199 to 98 per 100,000 people (Table 1 and Figure 1). The rates for males, which peaked at 260 per 100,000
in 1981, were consistently higher than the rates for females over the entire
time interval. The average ratio of male-to-female rates was 1.8 (95% CI,
1.6-2.0), showing little variation during this time. The in-hospital mortality
rate also showed little change during this period (Figure 1), with a mean of 5.3 (95% CI, 3.6-7.1) deaths per 100,000.
External cause of injury was identified in 27% of TBI-related hospitalizations
during this interval; 13% of injuries were attributed to motor vehicle crashes,
7% to falls, and 7% to all other known causes.
The decline in rates of TBI-related hospitalization varied among different
age groups (Table 1). The decline
was greatest among persons 5 to 14 years old and least among persons 65 years
old or older.
The incidence rate of hospitalization for mild TBI declined significantly
(P<.001) compared with the rates of hospitalization
for moderate and severe TBI (Table 1
and Figure 2). From 1980 to 1995,
TBIs classified as mild declined 61%, from 130 to 51 hospitalizations per
100,000 people, TBIs classified as moderate declined 19%, from 26 to 21 hospitalizations
per 100,000, and TBIs classified as severe increased 90%, from 10 to 19 hospitalizations
per 100,000. In 1995, the total number of hospitalizations for TBI was estimated
at approximately 260,000.
For 1992 to 1995, the estimated annual rate of TBI-related visits to
hospital EDs was 394 (95% CI, 357-430) per 100,000 people. Among all TBI-associated
ED visits in 1995, the proportion that included a CT scan evaluation was 26%.
Among ED visits in 1995 that were associated with TBI without diagnosis of
injury to other body regions, 27% included a CT scan evaluation.
Several factors could account for the major decline in the estimated
national rate of hospitalizations for TBI during the 16 years of the study
period. First, during almost the same interval (1979-1992), the TBI-associated
death rate decreased from 24.6 per 100,000 people to 19.3 per 100,000, a development
that has been attributed in part to successes in injury prevention, especially
those associated with motor vehicles.4 In our
study, because external cause of injury was not consistently available in
the National Hospital Discharge Survey data, we cannot directly assess any
effects of injury prevention on trends in TBI-related hospitalizations during
the study interval. Nevertheless, if success in injury prevention were the
only factor contributing to the decrease in TBI-related hospitalization, then
we would expect the decreases in rates of TBI-related hospitalization and
TBI-related death to be similar in proportion; we might also expect a similar
decrease in TBI-related hospitalization rates among all levels of injury severity.
However, in contrast to the 22% decline in the TBI-associated death rate between
1979 and 1992, our data demonstrate a disproportionately greater decline in
TBI-related hospitalization rates—51% during the interval of this study—with
the decline found mainly among injuries of mild severity. Second, in the United
States from 1980 to 1994, the rate of hospitalization for all causes declined
29%, suggesting a change in hospital practices for admissions.14
In a study of hospital service utilization and expenditures during this period,
a reduction in inpatient services was associated with an increase in outpatient
services, a shift attributed to the growing influence of prospective payment
system reimbursement and managed care.15 Thus,
hospital admission policies may now exclude a larger proportion of patients
with mild TBI.
Hospital admission practices for TBI also may be affected by improved
diagnostic technology. More frequent ED-based CT scanning may shift treatment
of milder TBIs from hospital inpatient services to outpatient services. However,
this effect may be limited, since only 26% of ED visits for TBI in 1995 included
CT scan evaluations. A shift in TBI care toward the ED cannot be measured
directly, because we are unable to compare rates of TBI-related ED visits
with rates of TBI-related hospital admissions throughout the study period.
However, the estimated rate of TBI-related ED visits in 1992 through 1995—approximately
1 million cases each year in the United States—indicates that about
80% of persons now evaluated for TBI are not admitted to hospitals.
The increased rate of TBIs classified as severe runs counter to all
other trends found in these data. Three factors might account for this phenomenon.
The increased number of cases classified as severe may be influenced by changes
in ICD-9-CM coding practices during this period that
decreased the assignment of nonspecific codes and increased the assignment
of codes with specific severity information. Also, increased reliance on CT
scanning and improved imaging during this study period may have increased
the detection of trauma-related intracranial lesions, again leading to the
assignment of more specific ICD-9-CM codes that indicate
greater severity. In addition, improvements in trauma care during the interval
of this study may have increased the proportion of critically injured persons
who survive long enough to be admitted to hospitals, thereby increasing the
number of cases classified as severe.
Our analysis has several limitations. The accuracy of our estimate of
trends in the TBI severity distribution depends on correct ICD-9-CM diagnostic coding that is consistent over time among all hospitals
in the sample. Limitations in the accuracy of diagnostic coding performed
in hospitals have been noted since the beginning of this study period.16 Furthermore, coincident with the increased use of
prospective payment systems, coding practices appear to have evolved toward
increased use of codes denoting more specific and severe diagnostic categories.17- 19 However, changes
in coding practices documented during this period do not appear sufficient
to account for the very large decline in rates of hospitalizations for TBIs
classified as mild.
The accuracy of our estimate of trends in the TBI severity distribution
also depends on the accurate classification of these injuries by ICDMAP-90
software. Misclassification of these data would affect our estimated relative
proportions of TBI by level of severity more than the trends for each level
of severity, because this classification algorithm is applied consistently
to the data across all years examined.
Despite a substantial overall reduction of TBI-related hospitalization
rates between 1980 and 1995, TBI continues to represent a major public health
problem that calls for the application of more effective methods of injury
prevention. The shift toward treating persons with less severe TBI as outpatients
raises additional concerns. The proportion of those with mild TBI who experience
effects later is unknown, although cognitive and emotional sequelae resulting
in extended or long-term disability have been reported in a minority of cases.20- 24
Our data provide no information about the number of persons seen in EDs or
other outpatient facilities with such sequelae who may need follow-up care
or evaluation. Assessing these needs and ensuring the availability of appropriate
services is an important function of public health; additional studies are
needed to address these issues.