Abstract
Background: In the United States, the Emergency Medical Treatment
and Active Labor Act defines broad guidelines regarding interhospital transfer
of patients who have sought care in the emergency department. However, patient
transfers for nonmedical reasons are still considered a common practice. The
purpose of this study was to evaluate the possible risk factors for hospital
transfer in a population of patients unlikely to require transfer to a level-I
center for medical reasons.
Methods: A retrospective case-control national database study was
performed with use of data from the National Trauma Data Bank (version 4.3).
The study group consisted of patients with low Injury Severity Scores (=9)
who were transferred to a level-I trauma center from another hospital. The
controls were patients with low Injury Severity Scores who were treated at any
hospital that was lower than a level-I trauma center and were not transferred.
Hypothesized risk factors for hospital transfer were the age, gender, race,
and insurance status of the patient; the time of day the transfer was
received; and the number and type of comorbidities.
Results: The total sample included 97,393 patients, 21% of whom were
transferred to a level-I trauma center. The odds ratios adjusted for all risk
factors indicated that transfer rates were higher for male patients compared
with female patients (adjusted odds ratio = 1.46), children compared with
seniors (3.54), blacks compared with whites (1.28), evening or night transfers
compared with morning or afternoon transfers (2.25), patients with Medicaid
compared with those with other types of insurance (2.02), and for those with
one or more comorbidities compared with those with no comorbidity (2.79).
Conclusions: These results suggest the need for prospective studies
to further investigate the relationships between hospital transfer and medical
and nonmedical factors.
Level of Evidence: Therapeutic Level III. See
Instructions to Authors for a complete description of levels of evidence.
In the United States, the Emergency Medical Treatment and Active Labor Act
(EMTALA) defines broad guidelines regarding interhospital transfer of patients
after they have sought care in the emergency
department1-3.
Occasionally, however, trauma center physicians question the reasons for the
transfer once the patient has arrived, noting that it was not the injury that
had prompted the transfer but, rather, other factors, such as demographic
characteristics of the patient (age, race, gender, or comorbidities), staffing
issues (time of day), and economic factors (ability to pay or insurance
status).
Few studies have evaluated the factors associated with patient transfer
after injury. Nathens et
al.4 reported on a
series of 2008 patients treated in the central region of the state of
Washington; they found that male gender, younger age, and noncommercial
insurance were associated with patient transfer from a level-III or IV trauma
center to a level-I facility. That series was restricted to patients in an
age-range from sixteen to sixty-five years; it grouped payments made by
Medicare and Medicaid and self-pay into the noncommercial insurance category;
and it included patients with all levels of injuries according to the Injury
Severity Score (ISS).
The purpose of the present study was to use a large national database of
trauma patients to evaluate the association between certain factors, such as
the demographic characteristics of the patients, staffing issues, and economic
factors, and the transfer of patients to a level-I trauma center after they
had sought emergency-department care. It was hypothesized that patients with
lower injury severity, those who were minorities, those who were underinsured
or uninsured, those with preexisting comorbidities including "less
desirable" comorbidities (e.g., human immunodeficiency virus-acquired
immunodeficiency syndrome [HIV-AIDS], hepatitis, alcoholism, or drug abuse),
and those who required treatment outside the normal business hours would be
more likely to be transferred.
Data Source
The data in this study were obtained from the National Trauma Data Bank,
version 4.3, which includes information on >1.21 million trauma patients
from 377 institutions in the United States gathered from 1988 through
20045. The data bank
is managed through the American College of Surgeons and gathers data from 55%
of all level-I trauma centers, 35% of all level-II trauma centers, as well as
many level-III trauma centers throughout the United States. Data entry is
voluntary at participating hospitals and, therefore, may not reflect the
actual number of trauma patients seen at an institution. The National Trauma
Data Bank provides no weighting information that would allow the user to
estimate national incidence rates.
Target Population
The study population was patients initially presenting at a level-II or
lower trauma center with low injury severity. Since the National Trauma Data
Bank provides no mechanism to track a patient across visits to the same or to
different hospitals, all inclusion and exclusion criteria as well as patient
demographic data were based on the information provided by the admitting
hospital. Thus, for the patients who were transferred, the level of the
original or transferring hospital cannot be determined. For this reason, the
cohort of interest was limited to: (1) patients who were not transferred to
and not originally admitted at a level-I trauma center, and (2) patients who
were transferred to a level-I trauma center, under the assumption that
transfer from one level-I facility to another was unlikely.
The inclusion criteria for this study were all patient records in which (1)
the trauma center level was coded; (2) the hospital transfer code (either yes
or no) was indicated and the transfer was to a level-I facility; and (3) the
injuries did not suggest the need for a transfer, with such injuries defined
in the present study as (a) an ISS of =9, which is the National Trauma Data
Bank definition of a "minor" injury, and (b) those with no level-I
emergency-department disposition suggesting serious injury or special needs,
which included intensive care unit, burn unit, operating room, or death (dead
on arrival or while being treated in the emergency department). Patients who
were transferred were all direct hospital-to-hospital transfers.
The exclusion criteria were the absence of valid codes related to the
following factors that were identified as being of interest in the evaluation
of hospital transfer status: basic demographic characteristics (age, gender,
and race), time of day the hospital transfer was received, and health
insurance status of the patient.
Study Design
This was a case-control study. The study group consisted of patients who
had been transferred to a level-I trauma center from another hospital, and the
controls were patients who had been treated at a hospital that was lower than
a level-I trauma center and had not been transferred (i.e., they were treated
where they presented).
Hypothesized Risk Factors
The age, gender, race, comorbidity status, and insurance status of the
patient and the time of day the transfer was received were evaluated as
possible risk factors for hospital transfer. Patient age was used to create
three subgroups: children (zero to seventeen years), adults (eighteen to
sixty-four years), and seniors (sixty-five years or older). Race was coded as
white, black, Hispanic, or other. Insurance status was categorized according
to method of payment: (1) insurance (i.e., commercial, no-fault, Workers'
Compensation, or Medicare); (2) underinsured (Medicaid); or (3) no insurance
(self-pay). Although some insurance companies in some states reimburse at
rates similar or identical to Medicaid rates, the National Trauma Data Bank
provides limited specific insurance company information; therefore, only
Medicaid status was classified as underinsured. Self-pay status was grouped
with no insurance. The time of day that the transfer was received was
categorized into four levels: morning (6 am to noon), afternoon
(noon to 6 pm), evening (6 pm to midnight), or night
(midnight to 6 am).
Although Injury Severity Scores (ISS) of =9 are categorized as minor
injuries in the National Trauma Data Bank, higher scores in this range could
reflect an isolated severe injury that might be considered a valid reason for
transfer. For example, an isolated femoral fracture would result in an ISS of
9. Therefore, the ISS codes were subdivided into two groups, 0 to 3 and 4 to
9, where 0 to 3 was considered a very minor injury. This variable was included
as another primary risk factor for hospital transfer.
The National Trauma Data Bank list of patient comorbidities was used to
evaluate the comorbidity status of patients by implementing the Deyo-Charlson
comorbidity
index6,7.
This index was then categorized into two levels: 0 and =1 reported
comorbidities. In addition, exploratory analyses of individual preexisting
comorbidities coded in the data bank were performed to assess their
relationship to hospital transfer status.
Analysis Plan
Logistic regression techniques were used to determine the odds ratios and
adjusted odds ratios associated with each risk factor alone and adjusted for
all other risk factors. All analyses were conducted with use of SAS software
(version 9.1; SAS Institute, Cary, North Carolina) on the Windows XP
Professional operating system.
Inclusion-Exclusion Summary
Of the 1,218,510 patient records in version 4.3 of the National Trauma Data
Bank, 136,685 or 11.2% met the study inclusion criteria, as they had (1) a
hospital transfer code, (2) a hospital-designated level code, and (3) no
reason for hospital transfer, which was defined in this study as (a) an ISS
score of =9, and (b) no emergency-department disposition code typically
considered as a reasonable cause for transfer, as summarized in the Materials
and Methods section.
Exclusion criteria focused on missing data for primary risk factors. Of the
136,685 eligible patients, 39,292 (28.75%) had some missing data for the
following categories: age (7.37%), gender (0.06%), race (8.46%), insurance
status (11.28%), and/or time of day the transfer was received (4.88%), leaving
97,393 patients in the sample. The percentages of patients with missing data
for each criterion sum to more than the total percentage of patients with
missing data as some patients had missing data for more than one criterion.
The relative differences in the retained sample (97,393 patients) and those
who were lost because of missing values (39,292 patients) are summarized in
Figure 1. Overall, the
distributions of gender and the time of day that the transfers were received
were very similar in the lost and retained samples. For race, whites were
somewhat overrepresented in the retained sample, and Hispanics were
underrepresented. Similarly, for age-groups, those who were sixty-five years
or older were somewhat overrepresented and those in the eighteen to
sixty-five-year-old group were underrepresented in the retained sample. For
ISS, patients with a very minor injury (an ISS of 0 to 3) were somewhat
overrepresented in the retained sample. For insurance status, insured patients
were slightly underrepresented and Medicaid and noninsured patients were
overrepresented in the retained sample.
Since losses due to missing values can result in biased results, logistic
regression analyses to determine the association between hospital transfer and
the hypothesized risk factors were done with all patients who met the
inclusion criteria, with missing values for each factor included as a valid
level (136,685 patients), and were compared with the model with use of only
patients with no missing data (97,393 patients). Of the twenty-three adjusted
odds ratios comparing various factor levels (e.g., male patients compared with
female patients and blacks compared with whites), the reduced sample had
thirteen stronger and nine weaker adjusted odds ratios compared with the full
sample of all eligible patients, but with no differences in interpretation.
The average difference (and standard deviation) in adjusted odds ratios was
0.010 ± 0.15 (range, —0.40 to 0.35). The average absolute
difference in adjusted odds ratio was 0.12 ± 0.11 (range, 0.0003 to
0.040). Thus, it was decided that biasing effects associated with the reduced
sample size were minimal and, therefore, the results from the sample with
complete data (97,393 patients) are reported.
Sample Description
In the final cohort of 97,393 patients, the majority (56.72%) was male. The
average age (and standard deviation) was 44.5 ± 23.7 years (range, less
than one year to eighty-nine years), with 15.9% who were less than eighteen
years, 56.1% who were between eighteen and sixty-four years, and 28.1% who
were sixty-five years or older. With respect to race, 75.25% of the patients
were white, 12.05% were black, 8.53% were Hispanic, and 4.16% were other
races. The average ISS (and standard deviation) was 5.1 ± 3.1 (range, 0
to 9). According to the Deyo-Charlson comorbidity index, 6.11% of the patients
had at least one preexisting comorbidity. With respect to insurance status,
72.5% of the patients were categorized as insured; 10.8%, as having Medicaid;
and 16.7%, as uninsured (self-pay).
In total, 20,647 patients (21.2%) were transferred to a level-I trauma
center, with the remaining 76,746 (78.8%) treated at the lower-level trauma
center where they initially presented. For those who were transferred, 14.3%
were received at the level-I trauma center at night (midnight to 6
am); 20.1%, in the morning (6 am to noon); 33.7%, in the
afternoon (noon to 6 pm); and 32.0%, in the evening (6
pm to midnight).
Risk Factors for Hospital Transfer
One set of logistic regressions was done for the primary risk factors in
the design: patient age, gender, ISS group, race, insurance status, time that
the transfer was received, and overall comorbidity status. A second set of
logistic regressions was done to explore the association of individual
comorbidities coded as preexistent conditions in the National Trauma Data
Bank.
Primary Risk Factors
Table I summarizes the odds
ratios and adjusted odds ratios for the primary risk factors in this study.
Odds ratios and adjusted odds ratios were consistent with respect to their
significance and direction for gender, ISS, age, number of comorbidities, and
time that transfer was received. Analysis of the adjusted odds ratios
demonstrated that (1) male patients were 46% more likely to be transferred
than female patients (adjusted odds ratio = 1.46; 99% confidence interval,
1.39 to 1.53); (2) patients with ISS scores of 4 to 9 were 57% more likely to
be transferred than patients with scores of 0 to 3 (adjusted odds ratio =
1.57); (3) children were 19% more likely to be transferred than adults
(adjusted odds ratio = 1.19), seniors were 67% less likely to be transferred
than adults (adjusted odds ratio = 0.33), and children were 254% more likely
to be transferred than seniors (adjusted odds ratio = 3.54); (4) those with
one or more comorbidities were 179% more likely to be transferred than those
with none of the comorbidities included in the Deyo-Charlson index (adjusted
odds ratio = 2.79); and (5) transfers were 125% more likely in the evening or
at night than in the morning or afternoon (adjusted odds ratio = 2.25).
Adjustment for the other risk factors had variable effects on the
association between the factors of race and insurance and the occurrence of a
hospital transfer. The adjusted effects of race remained significant (p <
0.01), but they were weaker when blacks were compared with whites and
Hispanics; and differences between Hispanics and whites were only significant
(p < 0.01) when looking at adjusted values. Blacks were 28% more likely to
be transferred than whites (adjusted odds ratio = 1.28) and were 81% more
likely to be transferred than Hispanics (adjusted odds ratio = 1.81).
Hispanics were 30% less likely to be transferred than whites (adjusted odds
ratio = 0.70). For insurance status, adjustments weakened the odds ratios.
Patients with Medicaid were 102% more likely to be transferred than insured
patients (adjusted odds ratio = 2.02), and there was no significant difference
in hospital transfer rates for patients with insurance compared with those
without (adjusted odds ratio = 1.01).
Given the higher rate of transfer for patients with an ISS of 4 to 9, a
subgroup analysis was done for patients with an ISS in the 0 to 3 range.
Table II summarizes these
results. In general, the pattern of results for the risk factors for hospital
transfer was remarkably similar from Table
I to Table II.
There were, however, two differences. For the subgroup of patients with an ISS
between 0 and 3, transfer rates were no longer greater for blacks compared
with whites (adjusted odds ratio = 1.07; 99% confidence interval, 0.95 to
1.21), and transfers received in the afternoon were no longer more likely than
those received in the morning (adjusted odds ratio = 1.10; 99% confidence
interval, 0.95 to 1.28).
Preexistent Specific Comorbidities
A total of forty-one distinct preexisting comorbidities, listed in
Tables III and in the Appendix,
were coded in the National Trauma Data Bank for the cohort. The low prevalence
for many of the comorbidities did not allow for similar subgroup analysis of
patients with very minor injuries defined as ISS scores ranging from 0 to
3.
Table III summarizes both
the unadjusted and adjusted odds ratios for the set of five comorbidities
identified a priori as possible risk factors for hospital transfer. The odds
ratios and adjusted odds ratios for four of these comorbidities were
associated with increased transfer rates and, therefore, only adjusted odds
ratios are summarized: (1) a history of psychiatric disorders (adjusted odds
ratios = 4.12); (2) chronic alcohol abuse (adjusted odds ratios = 3.45); (3)
chronic drug abuse (adjusted odds ratios = 3.37); and (4) history of cirrhosis
(adjusted odds ratios = 1.85). With regard to HIV-AIDS status, the fifth
comorbidity, no instances had been recorded in the transferred group and,
therefore, no odds ratio could be calculated.
The results for the remaining thirty-six coded preexisting comorbidities
that were not initially identified as risk factors for hospital transfer are
summarized in the Appendix. Eleven comorbidities were considered significant
(p < 0.01) risk factors alone and when adjusted for other patient factors;
one comorbid status (Alzheimer disease) demonstrated a significant (p <
0.01) protective effect for transfer when considered alone but a significant
(p < 0.01) risk factor for transfer when adjusted for demographics; five
comorbidities (rheumatoid arthritis, coronary artery disease, congestive heart
disease, chronic obstructive pulmonary disease, and myocardial infarction)
were significant risk factors only when looking at adjusted odds ratios; one
comorbid status (transplants) was a significant transfer risk factor only when
looking at unadjusted odds ratios; and the remaining eighteen comorbidities
demonstrated no significant effects.
In this study cohort, 21% of the patients with a low injury severity (an
ISS of 0 to 9) were transferred to a level-I trauma center and were compared
with a control group of patients also with a low injury severity who remained
at a lower-level trauma center. In the analysis of the subgroup of patients
who had a very minor injury (an ISS of 0 to 3), 18% of the patients were
transferred to a level-I trauma center. Although there are few other studies
with which to compare these results, the transfer of 18% to 21% of patients
with low injury severity seems high. Nathens et al. reported a transfer rate
of 12% for patients in Washington State who were initially treated at a
level-III or IV trauma center and then transferred to a level-I
facility4.
Although the rate reported by Nathens et al. may reflect a regional rather
than a national rate, the number of transfers in the current study may be
inflated as the National Trauma Data Bank does not allow patient linkage
between hospitals and, therefore, it is not possible to limit the cohort to
patients transferred from one National Trauma Data Bank hospital to another.
In other words, it cannot be determined what percentage of patients
transferred to a level-I hospital in the data bank was transferred from a
hospital that was not in the data bank. For example, if hospitals in the
National Trauma Data Bank transferred patients more often to hospitals that
were not in the data bank than they transferred patients to hospitals that
were in the data bank, the transfer rate in this study would be
underestimated. However, the majority (55%) of level-I trauma centers in the
country are part of the National Trauma Data Bank; therefore, the opposite
pattern, in which non-data-bank hospitals were more likely than not to
transfer a patient to a level-I hospital that was in the data bank, is more
probable, and the 21% rate of transfers reported in this study would be an
overestimation of the true rate.
Risk Factors for Hospital Transfer
Of greatest interest and concern in this study were the results
demonstrating that nonclinical factors that should not be relevant for
transfer, such as the gender, age, race, and insurance status of the patient
and the time of day, were all significant risk factors for hospital transfer,
even when controlling for injury severity. The patterns of these results were
somewhat consistent with stereotypes and/or expectations, with male patients
more likely to be transferred than female patients, with children more likely
to be transferred than adults, with blacks more likely to be transferred than
whites, with patients with Medicaid more likely to be transferred than
patients with insurance, and with transfers received in the evening and night
more likely than transfers received in the morning and afternoon. It should
also be noted that results from the subgroup analyses of patients with very
minor injury severity (an ISS of 0 to 3) were remarkably similar with the
exception that transfer rates were no longer significantly greater for blacks
compared with whites on the basis of adjusted odds ratios.
Some patterns were unexpected, with seniors less likely to be transferred
than adults, with patients with undetermined insurance status no more likely
to be transferred than patients with insurance, and with Hispanics less likely
to be transferred than whites. These unexpected results, especially the
finding that uninsured patients were not more likely to be transferred than
insured patients, may reflect multicollinearity among the risk factors since
the unadjusted odds ratios typically resulted in more expected relationship
patterns. On the other hand, this result may reflect a weakness in the coding
system, which, for example, does not allow the user to distinguish between
uninsured patients who can afford to pay their health care costs compared with
those who have no ability to pay.
The associations found in this study are similar to those reported by
Nathens et al.4. In
a smaller series of 2008 patients treated in the Washington State central
region, they reported that male gender, younger age, and noncommercial
insurance were risk factors for transfer from a level-III or IV trauma center
to a level-I facility. Race, however, was not a risk factor for transfer.
Interestingly, they found that insurance status was a risk factor for hospital
transfer only for patients with lower levels of injuries. There seems to be no
obvious explanation for the gender and age bias found in either study.
Perhaps, as hypothesized by Nathens et al., the triaging physician may have
had additional information not available from the existing databases on which
to base the decision to transfer.
The notion that hospital transfer may occasionally reflect nonmedical
economic realities seems relevant, given that patients with Medicaid were more
likely to be transferred than patients with other insurance, and evening and
night transfers were more likely when staffing is more problematic. Although
it is convenient to hypothesize that hospitals would want to transfer patients
during the evening or night secondary to staffing issues, transfer arrival
during the evening or night may only reflect the length of time for the
transfer to be initiated and completed. Unfortunately, since the National
Trauma Data Bank does not have information on the time the transfer was
initiated, this rival explanation for the time a transfer was received could
not be tested. No other explanation for the higher transfer rate for the
Medicaid patients seems appropriate since, after adjusting for other risk
factors, patients with no insurance or categorized as self-pay were not
transferred at higher rates than insured patients.
In general, patients with any medical comorbidity were more likely to be
transferred than patients with no comorbidities. Overall, the specific
comorbidities hypothesized to affect hospital transfer had the expected
result: patients with a history of a psychiatric disorder, chronic drug abuse,
chronic alcohol abuse, and cirrhosis were all significantly more likely to be
transferred (p < 0.01). The extremely low rate of HIV-AIDS in the National
Trauma Data Bank may be due to a number of factors, including patient
reluctance to volunteer this information and the fact that some hospitals may
not disseminate this information. In addition, hospital transfers were
significantly higher for eleven additional comorbidities identified as
preexisting states. This may reflect the desire of hospitals to treat less
medically complex patients and transfer the patients who may require more
intense and costly medical services. Alternatively, it may reflect the ability
of the hospitals to recognize their limitation due to a lack of services that
are more readily available at a level-I trauma center.
Strengths and Limitations
The strengths of this study include the large number of patients and the
cohort definition used to limit trauma or illness severity as defined by the
ISS (0 to 9 indicating a "minor" injury and 0 to 3 indicating a
"very minor" injury) to eliminate the severity of trauma as a
potential reason for transfer and removing patients requiring specialized
care, such as those being sent to intensive care or burn units. In addition,
this case-control study defined cases and controls in a unique fashion. The
cohort included patients with low injury severity who did not require
specialized care, with the study patients, or cases, defined as those who were
transferred to a level-I trauma center and the controls defined as those who
were not transferred and were treated at the level-II or lower trauma
center.
The limitations of this investigation include the retrospective database
analysis design. Similar to most database projects, users cannot independently
verify the accuracy of the data, its standardization, or its input.
Retrospective studies often do not capture some important data elements
specific to the research question. For example, since no information was
gathered about the types of services provided by individual hospitals, we do
not know whether the transfers were related to a lack of facilities or
services. Similarly, the National Trauma Data Bank does not provide
information regarding patient requests for transfer, or explicit information
about a clinician's judgment regarding the patient's clinical stability.
A number of additional database-related problems are specific to the
National Trauma Data Bank. The data bank comprises data from multiple
institutions gathered over several years, with year-to-year variations in
hospitals participating in the voluntary program. Thus, trends across time are
not readily interpretable. Second, the National Trauma Data Bank only includes
information on direct hospital-to-hospital transfers. Therefore, there is no
information on patients who were seen at one hospital, discharged, and told to
seek treatment at a level-I trauma center. Third, some of the variables of
particular interest were notable for the lack of complete data. For example,
lack of complete patient demographic information resulted in the exclusion of
29% of the patients who met the preliminary inclusion criteria. The bias
introduced in these results due to this loss of sample size is not readily
estimated. Finally, one particularly frustrating limitation of this study was
the absence of the date of admission data in the National Trauma Data Bank.
With this information, the day of the week and holiday dates could have been
identified. It is a common belief among trauma center personnel that patient
transfers occur more frequently before weekends or holidays. It would have
been helpful to provide evidence to support or refute this notion.
An Injury Severity Score of =9 was used as a determinant of a low level
of injury severity; however, this scoring system may not truly reflect the
actual level of injury since the score is calculated as the sum of the squares
of the highest Abbreviated Injury Score grades in the three most severely
injured body regions. For example, a patient with multiple injuries in the
same body region (e.g., bilateral femoral fractures) would have the same ISS
as another patient with only one injury in that body region (e.g., unilateral
femoral fracture). The ISS recognizes only the three most severely injured
regions, potentially underestimating the true severity of a multiply injured
patient. Additionally, this score may not characterize important features of
injuries that may require transfer to a level-I center, such as soft-tissue
injury associated with a fracture. Emergency-department disposition codes were
used to identify appropriate reasons for transfer, but other factors not
considered in the present study, such as the lack of specialized personnel,
may have been valid reasons for transfer. However, given the large sample size
in this study, these other factors would have needed to affect thousands of
patients to have made a significant change in the observed pattern of results.
Furthermore, the stability of these findings is strengthened by the similar
pattern of results for the subgroup analyses of patients with very minor
injury.
Despite the many limitations associated with database studies in general
and with the National Trauma Data Bank in particular, as summarized above,
these data do support the notion that hospital transfers of injured patients
occur for reasons other than medical necessity. These results suggest the need
for prospective studies to further investigate the relationships between
hospital transfer and both medical and nonmedical factors.
A table showing odds ratios for all remaining coded preexisting
comorbidities is available with the electronic versions of this article, on
our web site at
(go to
the article citation and click on "Supplementary Material") and on
our quarterly CD-ROM (call our subscription department, at 781-449-9780, to
order the CD-ROM). ?
Consolidated Omnibus Budget Reconciliation Act of 1985
(COBRA). PL 99-272, Title IX, Section 9121, 100 Stat 167
(1986).
1986
Omnibus Budget Reconciliation Act of 1989
(OBRA). PL 101-239, Section 6211 (h)(2), 102 Stat 2106, 42 USC
1395dd.
Bitterman RA. Providing emergency
care under federal law: EMTALA. Dallas: American College of Emergency
Physicians; 2001.
2001
Nathens AB, Maier RV, Copass MK,s
Jurkovich GJ. Payer status: the unspoken triage criterion. J
Trauma. 2001;50:
776-83.50776
2001
[PubMed][CrossRef]
Charlson ME, Pompei P, Ales KL,
MacKenzie CR. A new method of classifying prognostic comorbidity in
longitudinal studies: development and validation. J Chronic
Dis. 1987;40:
373-83.40373
1987
[CrossRef]
Deyo RA, Cherkin DC, Ciol MA. Adapting a
clinical comorbidity index for use with ICD-9-CM administrative databases.
J Clin Epidemiol. 1992;45:
613-19.45613
1992
[PubMed][CrossRef]