ORIGINAL CONTRIBUTION
The Effect of Emergency Department Crowding on Reassessment of Children With Critically Abnormal Vital Signs Holly E. Depinet, MD, MPH, Srikant B. Iyer, MD, MPH, Richard Hornung, PhD, Nathan L. Timm, MD, and Terri L. Byczkowski, PhD, MBA
Abstract Objectives: The objective was to determine whether several measures of emergency department (ED) crowding are associated with an important indicator of quality and safety: time to reevaluation of children with documented critically abnormal triage vital signs. Methods: This was a retrospective cross-sectional study of all patients with critically abnormal vital signs measured in triage over a 2.5-year period (September 1, 2006, to May 1, 2009). Cox proportional hazard analysis was used to determine rate ratios for time to critically abnormal vital sign reassessment, when controlled for potential confounders. Results: In this 2.5-year sample, 9,976 patients with critically abnormal vital signs in triage (representing 3.9% of 253,408 visits) were placed in regular ED rooms with electronic alerts prompting vital sign reassessment after 1 hour. Overall, the mean time to reassessment was 84 minutes. The rate of vital sign reassessment was reduced by 31% for each additional 10 patients waiting for admission (adjusted odds ratio [OR] = 0.98; 95% confidence interval [CI] = 0.98 to 0.99), by 10% for every 10 patients in the lobby (adjusted OR = 0.94; 95% CI = 0.93 to 0.96), and by 6% for every additional 10 patients in the overall ED census (adjusted OR = 0.97; 95% CI = 0.97 to 0.98). Conclusions: Emergency department crowding was associated with delay in the reassessment of critically abnormal vital signs in children; further work is needed to develop systems to mitigate these delays. ACADEMIC EMERGENCY MEDICINE 2014;21:1116–1120 © 2014 by the Society for Academic Emergency Medicine
T
he 1999 Institute of Medicine report To Err is Human identified the emergency department (ED) as the area of the hospital with the highest risk of adverse events.1 The simultaneous care of multiple patients with varying acuity levels, frequent interruptions, time constraints, high volume, and the need to institute critical diagnostic and therapeutic regimens with limited information contribute to an increased risk for medical error in the emergency setting. Exacerbating these risks, ED crowding has been increasingly recognized as a problem in our health care system and has been associated with a negative effect on quality of care measures in the adult and pediatric ED settings.2–4
An important system for preserving patient safety in the ED involves identifying and reassessing patients with critically abnormal vital signs.5 This is especially important in pediatric patients, for whom vital sign assessment (and response to abnormalities) poses significant challenges: vital signs can be difficult to assess in an uncooperative child; vital sign norms vary by age, making it more difficult for clinicians to immediately recognize abnormalities; and children have significant abilities to compensate physiologically for illness and appear well early in their course of illness, making vital sign abnormalities an important early warning sign of subsequent deterioration.
From the Division of Emergency Medicine, Department of Clinical Pediatrics, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH. Received February 28, 2014; revision received May 27, 2014; accepted June 9, 2014. Presented at the Pediatric Academic Societies Annual Meeting, Vancouver, British Columbia, May 2014. The authors have no relevant financial information or potential conflicts of interest to disclose. Supervising Editor: Bema Bonsu, MD. Address for correspondence and reprints: Holly E. Depinet, MD, MPH; e-mail:
[email protected].
1116 1116
ISSN 1069-6563 1116 PII ISSN 1069-6563583
© 2014 by the Society for Academic Emergency Medicine doi: 10.1111/acem.12478
ACADEMIC EMERGENCY MEDICINE • October 2014, Vol. 21, No. 10 • www.aemj.org
Despite the importance of vital signs in the assessment of ill and injured children, the reliability with which vital signs are monitored in the ED has not been well described in the literature. Although the effects of crowding have been well documented in adult settings, less information is available around the effects in pediatric settings. A recent study in our pediatric ED demonstrated that overcrowding was associated with a delay in the provision of antibiotics to febrile young infants.6 Because crowding has been shown to be associated with delays in other time-sensitive processes, we hypothesized that ED crowding would be associated with delays in the reassessment of children with critically abnormal vital signs. METHODS Study Design This was a retrospective cross-sectional study examining the association between measures of ED crowding and time to reassessment of critically abnormal vital signs, using data from an electronic health record database from September 1, 2006, to May 1, 2009. This was a time frame during which time a computerized clinical decision support tool and triage protocols to support the timely reassessment of children with critically abnormal vital signs were well established and without changes for more than 1 year. This study was approved by our institutional review board. Study Setting and Population The Cincinnati Children’s Hospital Medical Center (CCHMC) is an urban, tertiary-care hospital with a pediatric ED that has 86,000 visits/year and 45 beds (plus four trauma/resuscitation beds). It is a Level I trauma center, is staffed with pediatric emergency medicine faculty around the clock, and has 24-hour availability of subspecialist consultation and diagnostic imaging capabilities. In our ED, an electronic health record–based best practice alert was developed to help clinicians identify and respond to children with critically abnormal vital signs in a timely manner; this was applied to all children as they were being triaged (excluding those triaged to the shock-trauma suite for emergent resuscitation). This
1117
alert identified age-adjusted critically abnormal vital signs in real time and generated a nursing order to prompt a reassessment of the abnormal vital sign within 1 hour. We included all patients 0 to 21 years of age seen in the CCHMC ED, in whom a critically abnormal vital sign alert was triggered in triage. During this time period, our practice was that all patients had their first set of vital signs documented in triage (i.e., there was no in-room triage). Vital signs were flagged as critically abnormal if they were outside the 95th percentile for age, based on norms developed on our CCHMC ED population using a 3-year baseline (Table 1). We excluded children who were taken from triage to the shock-trauma suite for immediate care, since they all underwent immediate, continuous vital signs monitoring. Study Protocol All patient-specific data were extracted from EmSTAT, a proprietary electronic health record database that was in place for the duration of the study. EmSTAT contained information about patient-specific variables including vital signs, nursing documentation, order entry and completion, limited computerized clinical decision support, triage acuity (based on the Emergency Severity Index, a widely accepted tool),7 and demographic information. Data about nurse staffing ratios were obtained from analysis of the daily nursing schedule (hours per expected patient visits). The primary outcome variable was time to vital sign reassessment in minutes. Validated measures of ED crowding were used, as outlined below; together they measured all three aspects of ED crowding and flow: input, throughput, and output. Although there are many potential measures of ED crowding reported in the literature, and composite indices have been developed to predict crowding condition in real time and have also been used in research studies, we specifically choose these measures because they have been used in the pediatric ED crowding literature and are useful in demonstrating the effect of several aspects of crowding at a very granular level.6 Each indicator was measured at the top of the hour for each patient. Measures of ED input: number of patients waiting to go to ED rooms–total number of patients who
•
Table 1 Critically Abnormal Vital Sign Parameters Vital sign Heart rate (beats/min) High Low sBP (mm Hg) High Low dBP (mm Hg) High Low High respiratory rate (breaths/min)
0–5 months
6–12 months
1–3 yr
4–6 yr
7–12 yr
>12 yr
>185 185 175 160 150 No alarm
>140 No alarm
>140 140 140 140 150 160 90 70
>90 70
>90 65
>90 55
>95 45
>100 40
1118
• •
•
Depinet et al. • ED CROWDING AND CHILDREN WITH CRITICALLY ABNORMAL VITAL SIGNS
have not been placed in their ED rooms (are in triage, the lobby, waiting to be triaged, etc.). Measures of ED throughput: hourly ED census— total number of all patients in all parts of the ED (i.e., triage, lobby, room, boarding); cumulative hours in the ED—cumulative number of hours all of the patients together have been in the ED. Measures of ED output: number boarding—total number of patients with disposition status of “admitted” who are in the ED waiting to go up to their rooms; cumulative hours—”admitted” patients have been waiting to move to their rooms (from time admission disposition was set). Covariates: we controlled for the following covariates: age, race, shift (day, evening, night), triage acuity, nurse staffing ratios, and disposition of the patient (admit vs. discharge).
Data Analysis Descriptive statistics were developed for the primary outcomes, crowding measures, and covariates. Frequency distributions were developed for categorical variables, and means and standard deviations (SDs) were calculated for continuous variables. The relationship between time to vital sign reassessment (measured continuously in minutes) and the five measures of crowding were examined using Cox proportional hazards models. Cox proportional hazard analysis has the advantage of being able to handle censored data (patients who eloped or were discharged before vital signs were reassessed were treated as “lost to follow up,” or censored, at their last recorded time in the ED system), while also controlling for covariates. Separate models were developed for each crowding measure due to the high degree of correlation among them. Covariates included in the model were patient age, race, acuity, disposition (admit vs. discharge), shift, and nurse staffing ratio. All analysis was completed using SAS 9.3. RESULTS Figure 1 describes patients who met inclusion criteria or exclusion criteria or had censored data. Of 16,890 patients in the study period, 893 were excluded because they were taken immediately to the shock-trauma suite for resuscitation, eight were excluded because they were immediately transferred to another site, and 5,850 were excluded because their critically abnormal vital signs were not present during triage (i.e., they were noted during the remainder of the ED visit). Overall, we included 10,139 patient encounters with a critically abnormal vital sign documented in triage; 163 were lost to follow-up and thus had censored data (i.e., because they left the department before vital signs were reassessed). Table 2 summarizes the demographic characteristics of the sample and other patient variables. It also contains descriptive statistics for the crowding measures and the primary outcome measure of vital sign reassessment. Of the 9,976 encounters with complete data, the mean time to vital sign reassessment was 84 minutes.
253,408 Patient encounters screened
16,890 Encounters with a critically abnormal vital sign during an ED visit 893 Sent directly to shock trauma suite (excluded)
5,850 Abnormal vital sign was after triage (excluded)
8 Immediately transferred to other site (excluded)
163 Encounters partially missing data (excluded) 9,976 Study population
Figure 1. Study population and inclusion/exclusion criteria.
Table 3 shows the results of the multivariable Cox proportional hazards models examining time to vital sign reassessment; each crowding variable showed a statistically significantly decreased rate ratio for the time to vital signs reassessment. The estimated rate ratios ranged from 0.69 (per 10 patients) for number of patients awaiting admission to 0.98 (per 4 hours) for hourly boarding time. The rate ratios can be interpreted as a reduction in the instantaneous rate of vital sign reassessment for the stated increment, for example, the rate of reassessment is reduced by 31% for each additional 10 patients waiting for admission, by 10% for every 10 patients in the lobby, by 7% for every additional 100 hours spent in the ED by all current ED patients, by 6% for every additional 10 patients in the overall ED census, and by 2% for every additional 4 hours of boarding time. We additionally analyzed each of our covariates (age, sex, acuity, shift), for each of the five predictor models; we found that although each one was statistically significant (likely due to large sample size), the effect sizes were not clinically significant. Nursing staffing ratios were the only covariate that showed both clinically and
ACADEMIC EMERGENCY MEDICINE • October 2014, Vol. 21, No. 10 • www.aemj.org
Table 2 Descriptive Statistics for Visit, Patient-related Variables, and Crowding Measures (n = 9,976) Outcome measures Time to vital sign reassessment (min) Crowding measures Input Number of patients waiting for an ED room Throughput Hourly ED census Cumulative hours spent in the ED Output Number boarding Cumulative hours boarding Patient-related variables Patient age (yr) Race Black White Other Sex, female Acuity, emergent Admitted Shift Day Evening Night Nurse staffing ratio*
84.7 (57.1) 24.7 (12.3) 53.8 (23.9) 167.8 (206.7) 4.4 (3.1) 8.3 (13.5) 4.2 (4.9) 37.0 52.8 10.2 45.2 28.9 27.3 32.2 49.6 18.2 1.4 (0.2)
Data are reported as mean (SD) or percent. *Our expected staffing ratio was 1.4 nursing hours per patient visit; lower numbers indicate lesser staffing and higher numbers indicate increased staffing.
Table 3 Cox Proportional Hazard Model Results* Crowding Variable Input Number waiting for an ED room Throughput Hourly ED census Cumulative hours spent in the ED† Output Number of patients waiting for admission Cumulative hourly boarding time
Rate Ratio
95% CI
0.90 (per 10 patients)
0.88–0.92
0.94 (per 10 patients) 0.93 (per 100 hours)
0.93–0.95
0.69 (per 10 patients) 0.98 (per 4 hours)
0.90–0.96 0.66–0.75 0.97–0.99
*Adjusted for patient age, race, acuity, admit status, shift, and nurse staffing ratio. †Due to highly skewed data, these results are based on restricting hours to
