Emergency Department Crowding and Outcomes After Emergency Department Discharge

Abstract

Study objective: We assess whether a panel of emergency department (ED) crowding measures, including 2 reported by the Centers for Medicare & Medicaid Services (CMS), is associated with inpatient admission and death within 7 days of ED discharge.

Methods: We conducted a retrospective cohort study of ED discharges, using data from an integrated health system for 2008 to 2010. We assessed patient transit-level (n=3) and ED system-level (n=6) measures of crowding, using multivariable logistic regression models. The outcome measures were inpatient admission or death within 7 days of ED discharge. We defined a clinically important association by assessing the relative risk ratio and 95% confidence interval (CI) difference and also compared risks at the 99th percentile and median value of each measure.

Results: The study cohort contained a total of 625,096 visits to 12 EDs. There were 16,957 (2.7%) admissions and 328 (0.05%) deaths within 7 days. Only 2 measures, both of which were patient transit measures, were associated with the outcome. Compared with a median evaluation time of 2.2 hours, the evaluation time of 10.8 hours (99th percentile) was associated with a relative risk of 3.9 (95% CI 3.7 to 4.1) of an admission. Compared with a median ED length of stay (a CMS measure) of 2.8 hours, the 99th percentile ED length of stay of 11.6 hours was associated with a relative risk of 3.5 (95% CI 3.3 to 3.7) of admission. No system measure of ED crowding was associated with outcomes.

Conclusion: Our findings suggest that ED length of stay is a proxy for unmeasured differences in case mix and challenge the validity of the CMS metric as a safety measure for discharged patients.

Figure 1

Measures of ED crowding.

Figure 2

Study cohort. KPSC, Kaiser Permanente South California.

Figure 3

A-I, Logistic regression assessing the association between crowding measures and 7-day admission after ED discharge. Relative risks of the odds of admission was calculated according to the odds ratio divided by the median odds ratio. Each graph is a representation of a model that adjusts for age, sex, race or ethnicity, ambulance versus nonambulance arrival, triage blood pressure and pulse, ESI triage score, preexisting comorbidities, primary discharge diagnosis, shift, weekend versus weekday, month, medical center, year, and interaction between medical center and year. The dotted line depicts the 95% CI.

Figure 3

A-I, Logistic regression assessing the association between crowding measures and 7-day admission after ED discharge. Relative risks of the odds of admission was calculated according to the odds ratio divided by the median odds ratio. Each graph is a representation of a model that adjusts for age, sex, race or ethnicity, ambulance versus nonambulance arrival, triage blood pressure and pulse, ESI triage score, preexisting comorbidities, primary discharge diagnosis, shift, weekend versus weekday, month, medical center, year, and interaction between medical center and year. The dotted line depicts the 95% CI.