The correspondents raise concerns that our study suffers from selection bias; lacks sufficient statistical power to prove the safety of the deferred care criteria; fails to follow standard methods for decision rule development; fails to address the root cause of emergency department overcrowding; and proposes an unwise, unsafe, and unfair solution to the current access crisis in emergency care. This letter addresses each of these concerns.

Our study's goal was to test the hypothesis that a population of ambulatory patients exists who seek emergency department care for non-life-threatening conditions and can safely await primary care evaluation on the following day. Had we set out to estimate the prevalence of such patients in the general population of emergency department patients, our sampling method would have represented a serious threat to the study's external validity. In fact, the study's design and purpose render the issue of external validity moot but preserve internal validity. Our findings apply to exactly the group of patients we chose to study and support safe next-day care in this group. Our discussion challenges others to replicate this finding in other settings as a prerequisite to wider implementation.

Our study had adequate power to exclude clinically important differences in our primary outcome, change in health status. The 95% CI for change in health status excluded a clinically important difference between groups. For our secondary outcomes, much was made of the fact that the 95% CI included the possibility that the deferred care group may have had 1 extra bed day because of illness. We note that the 95% CI also included the possibility that the deferred care group had 0.8 less bed day because of illness.

The excellent work of McGinn and colleagues in developing decision rules, which Dr. Pitts cited, represents one but not the only approach. A large body of literature supports the RAND-UCLA Appropriateness Method we chose (1). Both methods require large validation studies in diverse populations, as emphasized in our Discussion.

We thank Dr. Kellermann for citing previous work confirming that a minority of patients presenting to emergency departments consider their conditions emergent or urgent (2). Many studies show that emergency department overcrowding has many causes, including internal patient flow issues and patient volume (3). We have addressed a single facet of the problem.

Dr. Kellermann's contention that our intervention was designed to benefit hospitals and third-party payers rather than the study participants reflects a narrow view of emergency department overcrowding. Previous work has amply shown that prolonged waiting times in the emergency department are also unsafe (4). Readers concerned about possible legal consequences of deferring care should consider that long waiting times can also violate the patient antidumping statute (5). Finally, we dispute the claim that deferring primary care to a clinic setting is less fair than a "treat and refer every patient" approach. The differences in quality, outcomes, and costs between the two approaches are worth additional study.