|Home | About | Journals | Submit | Contact Us | Français|
The wait time to see a physician in US emergency departments (EDs) is increasing and may differentially affect patients with varied insurance status and racial/ethnic backgrounds.
Using a stratified random sampling of 151 999 visits, representing 539 million ED visits from 1997 to 2006, we examined trends in the percentage of patients seen within the triage target time by triage category (emergent, urgent, semiurgent, and nonurgent), payer type, and race/ethnicity.
The percentage of patients seen within the triage target time declined a mean of 0.8% per year, from 80.0% in 1997 to 75.9% in 2006 (P<.001). The percentage of patients seen within the triage target time declined 2.3% per year for emergent patients (59.2% to 48.0%; P<.001) compared with 0.7% per year for semiurgent patients (90.6% to 84.7%; P<.001). In 2006, the adjusted odds of being seen within the triage target time were 30% lower than in 1997 (odds ratio, 0.70; 95% confidence interval, 0.55-0.89). The adjusted odds of being seen within the triage target time were 87% lower (odds ratio, 0.13; 95% confidence interval, 0.11-0.15) for emergent patients compared with semiurgent patients. Patients of each payment type experienced similar decreases in the percentage seen within the triage target over time (P for interaction=.24), as did patients of each racial/ethnic group (P=.05).
The percentage of patients in the ED who are seen by a physician within the time recommended at triage has been steadily declining and is at its lowest point in at least 10 years. Of all patients in the ED, the most emergent are the least likely to be seen within the triage target time. Patients of all racial/ethnic backgrounds and payer types have been similarly affected.
The Institute of Medicine has described the state of American emergency departments (EDs) as a “growing national crisis.”1(pXIII) A notable symptom of this crisis is the increasingly long time that patients wait to see an ED health care provider. In 1997, patients waited a median of 22 minutes after they arrived in the ED to be evaluated by a physician.2 By 2004, patients were waiting a median of 30 minutes, a 36% increase.2 Prolonged ED wait time decreases patient satisfaction,3-5 limits access,6 increases the number of patients who leave before being seen,7 and is associated with clinically significant delays in care for patients with pneumonia, cardiac symptoms, and abdominal pain.8-10 Consequently, in 2008, the National Quality Forum recommended that the median time patients wait to be seen by a health care provider be 1 of 10 new quality metrics for emergency care.11
Wait time in itself is an imperfect measure of the timeliness of emergency care because the appropriateness of a given wait time depends on the patient’s acuity of illness. Less acutely ill patients can safely wait longer. Reports of ED wait time have generally not considered this clinical context, except occasionally to identify the most acutely ill patients for separate analysis.2 Because the average acuity level of ED patients has decreased during the past decade,2,12 reported increases in the aggregate wait time have been difficult to interpret. Instead, the US Government Accountability Office recently reported that, in 2006, 50.4% of emergent visits had wait times that exceeded the time frame recommended at ED triage.13 This approach has the advantage of accounting for triage acuity; however, we have few data regarding time trends for this measure. Given the average decrease in illness acuity of ED patients, it is important to know whether the percentage of patients waiting longer than recommended at triage has changed significantly for patients at each triage level over time. Furthermore, because racial/ethnic disparities exist2,14 and have persisted or even worsened over time2,15 for other measures of timely ED care, such as wait time and length of stay, it is important to investigate whether the same is true for the percentage of patients seen in a timely fashion.
Accordingly, in this study, we used data from the National Hospital Ambulatory and Medical Care Survey (NHAMCS) to examine whether the percentage of ED patients seen within the target triage time frame has changed from 1997 to 2006, and whether these changes affect patients differently according to triage levels. We also investigated whether blacks and Hispanics or the uninsured are disproportionately unlikely to be seen within their triage target times, and whether any such disparities are changing over time.
We conducted a cross-sectional study using data collected for the NHAMCS from 1997 to 2006. The NHAMCS is a 4-stage probability sample of visits to EDs of US general and short-stay hospitals, excluding federal, military, and Department of Veterans Affairs hospitals. Each hospital collects data from a systematic random sample of ED visits during an arbitrarily assigned 4-week period. Data collection, abstraction, and cleaning procedures have been described in detail elsewhere.16 The data set includes weights to allow estimation of national results. Wait time was not included as a variable in 2001 and 2002; consequently, these years are omitted from the study.
Our primary outcome was the percentage of patients with a wait time less than the upper limit of the target time established for each patient at triage, operationalized for each patient as a binary variable set to 0 for wait time within the triage target and 1 for wait time longer than the triage target. This measure places each patient’s wait time within a clinical context. Wait time was defined as the number of minutes between the time the patient arrived at the ED and the time the patient was seen by a physician. Triage assessment is conducted upon arrival to the ED by an ED health care provider, typically using 1 of several scales with 3 to 5 levels.17 Each level of triage corresponds to a time within which the patient should be seen by a licensed independent practitioner. From 1997 to 2004, the NHAMCS data set included a 4-level triage variable: emergent (see in 0-14 minutes), urgent (see in 15-60 minutes), semiurgent (see in 61 minutes to 2 hours), and nonurgent (see in >2-24 hours). In 2005 and 2006, the 0-to 14-minutes category was split into 2 categories: immediate (see in 0 minutes) and emergent (see in 1-14 minutes). To compare trends over time, we recombined these into one 0-to 14-minutes category.
Our independent variables included visit characteristics, patient sociodemographic factors, and hospital characteristics. Visit characteristics included triage categorization, level of pain at presentation, whether the patient was seen by a trainee, season of visit, and day of visit. Patient sociodemographic factors included age, self-reported race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other), sex, method of payment, and mode of arrival at the ED. After confirming that the linearity assumption was met, we treated age as a continuous variable. Hospital characteristics included urban status (location within a metropolitan statistical area or not), ownership (nonprofit, proprietary, and government [nonfederal]), and geographic region (Northeast, Midwest, South, and West).
We used standard descriptive statistics to characterize the sample of patients and hospitals, weighted to estimate national statistics. We compared characteristics of visits with and without missing data using χ2 or t tests as appropriate.
We described trends in ED wait time using median (interquartile range) for the overall sample and for each triage category. To assess the significance of these changes, we used non-parametric testing. We analyzed trends in ED wait time relative to triage target time using χ2 tests of the proportion within the target in each year. We also stratified this analysis by triage level. We calculated the yearly percentage change in wait time and in the proportion seen within the triage target using the geometric growth formula r=(Yt/Y0)(1/t) – 1, where Yt indicates value in the final year; Y0, value in the initial year; t, total years; and r, mean growth rate.
Because of the widespread notion that increases in wait time are a consequence of an increase in ED use by uninsured or black and Hispanic patients,18 we examined whether the distribution of race/ethnicity and of payer type in the sample changed during the study period using χ2 tests. We also examined whether patients of different racial/ethnic groups or payer types had similar distributions of triage levels using χ2 tests. We then determined whether trends in triage acuity over time were significantly different for patients of different racial/ethnic groups and of different payer types in linear regression models using triage level as the outcome variable. We also assessed “excess” use of the ED for nonurgent visits by patients of different racial/ethnic groups or payer types by calculating the difference between the percentage of visits by black and Hispanic or uninsured patients that were nonurgent and the percentage of visits by white or privately insured patients that were nonurgent.
Finally, we estimated a multivariate logistic regression model using all visit, patient, and hospital variables to determine factors associated with being seen within the target triage time. To determine whether patients with different triage acuity levels experienced different trends in rates of being seen within the target time during the study period, we tested the interaction of year and triage level. To determine whether patients of different racial/ethnic backgrounds or payer type experienced different trends in the rates of being seen within the target time, we tested the interaction of year and race/ethnicity or payment type in our multivariate model. Because our initial analysis showed that patients of all races/ethnicities and all payment types experienced similar trends over time in triage acuity, it was not necessary to stratify this analysis by triage level.
All analyses were conducted using SAS statistical software, version 9.1.2 (SAS Institute Inc, Cary, North Carolina). We adjusted all standard errors to account for the complex sampling design of the study and weighted all results to reflect national estimates, using the SAS procedures surveyfreq and surveylogistic. It was not possible to statistically adjust for potential clustering of hospitals over time because hospital identity is masked. In addition, because of software limitations, median wait time results could not be weighted. All statistical tests were 2-tailed, and we used a P value of .05 to determine statistical significance.
The full data set comprised 239 405 patient visits weighted to represent 865 million visits during the 8 years of the study. A total of 61 568 visits (25.7%) were missing wait time data, and 25 838 of the remaining visits (10.8%) were missing triage assignment data. Therefore, 151 999 patient visits (63.5%), which, when weighted, represented 539 million visits, were included in the study. Patients with missing data were less likely to report having pain, to arrive by ambulance, to be admitted to the hospital, to be cared for in the South, and to be from later years in the study (Table 1).
Changes in selected characteristics of the study sample over time are shown in Table 2. The weighted number of visits by emergent patients did not change during the study period (mean annual change, −0.02%), whereas the overall weighted visit total increased by an average 3.9% each year, from 74 million in 1997 to 103 million in 2006. Consequently, the percentage of patients with triage assessments who were classified as emergent decreased from 26.9% to 18.3% during the decade of the study (P<.001). The number of patients seen each year by triage category is shown in Figure 1. The percentage of uninsured patients remained relatively constant across all years of the study (between 16% and 17% of visits), whereas the proportion of white patients declined from 68.2% to 60.3%.
Patients of each payment type experienced similar decreases in triage urgency over time (P for interaction=.18), as did patients of all racial/ethnic groups (P for interaction=.73). In 2006, 17.0% of uninsured ED patients were categorized as nonurgent, compared with 13.9% of privately insured ED patients (absolute difference, 3.1%). This 3.1% excess use of the ED by uninsured patients relative to privately insured patients for nonurgent visits accounted for approximately 567 000 visits, or 0.5% of total visits in 2006. Similarly, excess use of the ED by black and Hispanic patients relative to white patients for nonurgent visits (difference of 3.5%) accounted for approximately 1.3 million visits, or 1.1% of all ED visits, in 2006.
Between 1997 and 2006, median wait time for all patients increased 4.6% per year, from 22 minutes (interquartile range, 10-47) to 33 minutes (interquartile range, 15-69) (P<.001). Median wait time increased by 4.6% per year for emergent patients, 2.8% per year for urgent patients, 3.9% per year for semiurgent patients, and 1.6% per year for nonurgent patients (Table 3 and Figure 2). These increases in median wait time were statistically significant for every level of urgency (Table 3).
The percentage of patients seen by a physician within the time recommended at triage declined 0.8% a year, from 80.0% in 2000 to 75.9% in 2006 (P<.001) despite the decrease in the percentage of emergent patients during this time period. From 1997 to 2006, the percentage of emergent patients seen within the triage target time declined 2.3% per year, from 59.2% to 48.0% (P<.001); for urgent patients, it declined 1.1% per year, from 84.0% to 76.3% (P<.001); and for semiurgent patients, 0.7% per year, from 90.6% to 84.7% (P<.001). All nonurgent patients were seen within the triage target time in every year (Table 3 and Figure 3).
In bivariate analyses (Table 4), higher acuity of illness was strongly associated with waiting longer than the triage target time. Overall, 56.6% of emergent patients were seen within the triage target time compared with 100% of nonurgent patients (P<.001). Patients ultimately admitted to the hospital were less likely to be seen within the triage target time (72.6% vs 80.4%; P<.001), as were patients seen by trainees (74.0% vs 79.8%; P<.001) and those visiting urban hospitals (77.8% vs 85.4%; P<.001). Hispanic and black patients were significantly less likely than white patients to be seen within the triage target time (76.0% for Hispanics and 77.7% for blacks vs 80.4% for whites; P<.001), even though they were less likely than white patients to be triaged as emergent. Uninsured, Medicaid, and privately insured patients were seen within the triage target time in similar proportions. Significant differences were also seen in several other independent variables (Table 4).
In multivariate logistic regression analysis adjusting for all visit, patient, and hospital characteristics (Table 5), level of triage urgency remained strongly associated with the likelihood of being seen within the triage target time. The odds of being seen within the target time were 87% lower for emergent patients compared with semiurgent patients (odds ratio [OR], 0.13; 95% confidence interval [CI], 0.11-0.15). For each year until 2005, the adjusted odds of being seen within the triage target time were not significantly different from 1997, despite a gradual increase in the median wait time. However, in 2005 and 2006, the adjusted odds of being seen within the triage target time were significantly lower than in any year from 1997 through 2004. In 2006, the adjusted odds of being seen within the triage target time were 30% lower than in 1997 (OR, 0.70; 95% CI, 0.55-0.89) and 24% lower than in 2004 (0.76; 0.65-0.91).
Although in multivariate analyses the odds of being seen within the triage target time were significantly lower for Hispanic (OR, 0.77; 95% CI, 0.69-0.85) and African American (0.80; 0.72-0.88) patients, the ORs for other covariates were lower (Table 5). Insurance status was not significantly associated with being seen within the triage target time in multivariate analyses. The interaction of year and triage category was not significant (P=.29), suggesting that patients of all triage levels were increasingly less likely to be seen within the triage target time during the study period. A similarly nonsignificant interaction term was found for payment type (P=.24). The interaction of race/ethnicity and time was of borderline significance (P=.052). The C statistic for the full model was 0.76.
In this examination of 152 000 visits to EDs in US hospitals, we find that the odds of being seen by a physician within the time recommended at triage declined by 30% from 1997 to 2006. In 1997, 1 in 5 patients waited longer to be evaluated by an ED physician than recommended at triage. By 2006, this proportion had risen to 1 in 4 patients. Most distressingly, in 2006, more than half of patients assessed at triage as needing emergent care waited longer to be seen by a physician than recommended. Although African American, Hispanic, and uninsured patients were less likely than white or privately insured patients to be seen on time in any given year, all patients suffered similarly from a decreased likelihood of being seen on time during the study period.
Emergency departments are increasingly overcrowded, thereby straining resources.1 Triage assessment is intended to mitigate this strain by ensuring that the most acutely ill patients are prioritized for assessment, regardless of the competing demands on ED physicians’ time. Considering wait time within the clinical context of triage assessment therefore allows for a more nuanced understanding of the timeliness of ED care than wait time in aggregate. Unfortunately, we found that the wait times of emergent patients increased over time in parallel with those of less urgent patients, despite the theoretically mitigating effects of triage assessment. Furthermore, these patients experienced the largest percentage increases in wait time and were consistently the least likely to be seen on time.
The causes of increasing wait time in the ED are likely many. Per capita ED use has increased substantially, from 34.2 visits per 100 persons in 1996 to 40.5 visits per 100 persons in 2006; much of the increase has been among the population of less acutely ill patients.12 Debate exists as to the reasons for this growth; however, our study confirms that excess use of the ED for nonurgent care by uninsured patients is not a major factor,18 accounting for only 0.5% of ED visits in 2006. Similarly, excess use by black and Hispanic patients for nonurgent care relative to white patients accounted for only 1.1% of all ED visits in 2006. Rather, it appears that decreased access to primary health care for all patients and an aging population are more important contributors to the per capita increase in ED utilization.18 The nation’s population has also grown in the past decade. Together, these trends have resulted in a 32% increase in ED volume from 1996 to 2006.12 At the same time, the number of EDs has fallen from 4019 to 3833, leading to a dramatically increased volume per ED.12
Prolonged wait time, however, is not only a function of increased volume. Many EDs still rely on older, less accurate 3-level triage systems, rather than the more predictive 5-level systems.17,19,20 This hampers the ability of ED health care providers to accurately prioritize the acutely ill. Physician bias may play a role: patients with “undesirable” chief complaints, such as back pain and constipation, wait longer to be seen than those with “desirable” complaints, such as fractures and palpitations.21 Inefficient ED processes of triage, registration, evaluation, treatment, laboratory testing, radiologic evaluation, staffing, and so forth contribute to ED crowding and prevent physicians from being able to see new patients.22 The Institute of Medicine has suggested that hospitals might be prioritizing elective admission patients over ED admissions because of higher reimbursement rates for elective admissions, leading to longer wait times for ED patients.1 Some, however, have found that ED admissions are more profitable than elective admissions.23 Finally, high hospital occupancy rates reduce the number of beds available for admitted ED patients and contribute to overcrowding and delays in care.24-28 Recent research suggests that such output factors may in fact be the most important drivers of ED crowding.13
The multifactorial nature of prolonged ED wait time lends itself to numerous avenues for improvement. Increasing patients’ access to alternate sites of care may divert some nonurgent visits, although this might not be very effective at reducing wait times.29 Interventions to improve triage, registration, staffing, laboratory testing, radiologic testing, and numerous other processes have shown promise.5,30-46 Redesign of the physical plant, while costly, can increase efficiency.32 Outside the ED, efforts have been made to even patient flow by discharging in-patients earlier in the day or reorganizing operating room scheduling.44,47 In all of these efforts, nonmedical industry analytics such as queuing theory or Toyota’s lean manufacturing approaches to workflow and process design are gaining traction and may be useful.1,48-50 To date, however, large-scale, high-quality studies comparing or carefully assessing the impact of improvement activities are lacking. Comparative research into the most effective methods of reducing ED crowding, decreasing ED length of stay, and limiting ED wait times is urgently needed to help EDs prioritize their quality improvement activities and maximize their impact.
Our study has several limitations. Owing to the fairly recent widespread adoption of triage assessment in the United States,19,20 we were less able to determine patients’ wait time relative to triage target times for earlier years of the study. However, because an institutional lack of triage assessment affects patients of every acuity level, this problem was unlikely to bias the results within each triage level appreciably. Confirming this, the mean wait time of patients with no triage assignment was similar to the overall mean wait time. Triage assessment reliability has been reported to be fair to excellent, depending on the triage method used,17,51-54 and may have varied among hospitals. Furthermore, in earlier years of the study, less predictive 3-level triage systems were standard, in contrast to the more predictive 4- or 5-level systems in use more recently.19,20 Consequently, triage assessments in more recent years were more likely to be accurate. In addition, beginning in 2005, NHAMCS data regarding ED wait times were manually checked by the Centers for Disease Control and Prevention for consistency and corrected when appropriate (L. F. McCaig, MPH, and S. M. Schappert, MA, written communication, June 17, 2009). Data may also have been more accurately collected in recent years with the advent of electronic records. We therefore have the most confidence in the most recent data, during which the proportion of patients seen within the triage target time was also the smallest. Finally, data are collected from medical records by each hospital and may not reflect actual practice, particularly for the most emergent patients, who might be examined first and documented later. Wait time performance for emergent patients may therefore be better than the data suggest. It is unlikely, however, that such documentation lags would have worsened systematically over time; consequently, this inaccuracy is unlikely to have biased the trend results substantially.
In summary, this study of the time patients wait to be seen in the ED demonstrates that an increasing proportion of patients are waiting longer than recommended by triage assessment. The increasing lack of timely care has affected all patients, including emergently ill patients, who are the least likely to be seen on time. Interventions for reducing wait times are urgently needed.
Funding/Support: This study was supported by Clinical and Translational Science Award grants UL1 RR024139 and KL2 RR024138 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH roadmap for Medical Research. Dr Horwitz is supported by Yale–New Haven Hospital and by the NCRR.
Financial Disclosure: None reported.
Disclaimer: The contents of this study are solely the responsibility of the authors and do not necessarily represent the official view of the NCRR or the NIH.
Role of the Sponsor: No sponsor had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, and approval of the manuscript.