ACADEMIC EMERGENCY MEDICINE • March 2001, Volume 8, Number 3
265
gatekeeping than did other EDs in the Philadel-
phia area. However, in selecting EDs no advance
determination was made of the strength of asso-
ciation between ethnicity and gatekeeping, and
there is no reason to expect that institutions with
a greater frequency of gatekeeping or a greater
prevalence of Medicaid patients would have a
stronger association between gatekeeping and eth-
nicity than would other institutions.
L
IMITATIONS AND FUTURE QUESTIONS
The study results must be interpreted with consid-
eration of several potential limitations of the study
methodology. First, all of the data used for this
study were collected retrospectively. Therefore,
there were limitations in the variables that could
be measured and there were some missing and am-
biguous data.
It is also possible that the association observed
in Philadelphia does not exist elsewhere. However,
there is no characteristic of the Philadelphia-area
health care delivery system that leads us to expect
that the findings of this study would be unique to
Philadelphia. Managed care penetration at the
time of the study was less than in many other
regions of the United States but characteristics of
the patient population, providers, EDs, and the
managed care gatekeeping process were similar.
Nevertheless, the results would need to be tested
in other communities to validate this assumption.
To the extent that managed care gatekeeping of
ED visits persists, a carefully designed prospective
study with complete and accurate data collection
on a wide array of potential confounding variables
will be needed to verify the findings reported here
and to assess their generalizability to other com-
munities. Additional qualitative studies observing
the participants in the gatekeeping process, to as-
sess their potential contributions to the gatekeep-
ing decision, might help to identify a mechanism
for the observed racial difference in access to emer-
gency medical care.
Another limitation of the study is the possibility
that the white patients included in this study were
systematically different from the African American
patients in ways other than their ethnicity, and
that these differences account for the observed as-
sociation between ethnicity and gatekeeping. In
other words, it is possible that incomplete control
of confounding is responsible for the association
between ethnicity and gatekeeping. To the extent
feasible with the administrative data used for this
study, we attempted to adjust for such confound-
ing. For instance, one might speculate that socio-
economic status rather than ethnicity was the
causal variable. However, membership in Medicaid
vs commercial MCO can be used as a proxy for
socioeconomic status. Our analyses showed that
that the ethnicity–gatekeeping association was
present in both Medicaid and commercially in-
sured patients, making this explanation less likely.
One might also speculate that acuity rather
than ethnicity was the causal variable. Although
we used triage score to adjust for acuity, triage
19
score is limited in its ability to measure severity.
For confounding by severity to explain the ob-
served association between ethnicity and gate-
keeping, two preconditions would have to exist: 1)
African Americans would have to be less severely
CONCLUSIONS
ill than whites; and 2) gatekeepers, functioning by African Americans were approximately 1.5 times
telephone, would have to ascertain severity more more likely than whites to be denied authorization
reliably than was captured by the triage score. One for ED visits by their managed care gatekeepers.
of these preconditions for confounding is probably Assuming that bias or uncontrolled confounding
true; triage scores for African Americans were less does not fully explain these findings, they raise im-
severe than for whites. However, to the best of our portant questions about the equitable application
knowledge, the ability of clinicians to make triage of gatekeeping across racial groups and, therefore,
decisions by telephone that are more accurate than the appropriateness of using gatekeeping to reduce
decisions made by triage nurses, who had the op- ED utilization.
portunity to see and examine the patients in per-
son, has not been studied.
References
Another issue with these findings is their
generalizability. Both the pilot data and the pres- 1. McCaig LF. National Hospital Ambulatory Medical Care
Survey: 1998 emergency department summary. Advance data
ent study were from one community, and the in-
from Vital and Health Statistics; no. 313. Hyattsville, MD: Na-
tional Center for Health Statistics, 2000.
stitutions were not randomly selected within that
community. We sought EDs with sufficient varia- 2. Tyrance PH Jr., Himmelstein DU, Woolhandler S. US emer-
bility in the predictor variable (ethnicity) and the
outcome variable (gatekeeping decision) to allow
analysis of their potential association. We also United States. MMWR. 1994; 43:581–6.
sought EDs with a substantial Medicaid popula-
tion. In theory, it is possible that the selected EDs
gency department costs: no emergency. Am J Public Health.
1
3
996; 86:1527–31.
. Centers for Disease Control. Medical-care spending—
4
. Stussman BJ. National Hospital Ambulatory Medical Care
Survey: 1994 emergency department summary. Advance data
from Vital and Health Statistics; no. 275, Hyattsville, MD: Na-
had a stronger association between ethnicity and tional Center for Health Statistics, 1996.