Emergency Department Visits for Nonurgent Conditions: Systematic Literature Review
Lori Uscher-Pines, PhD, MSc; Jesse Pines, MD, MBA; Arthur Kellermann, MD, MPH; Emily Gillen, MA; and Ateev Mehrotra, MD, MS
Nonurgent emergency department (ED) visits are typically defined as visits for conditions for which a delay of several hours would not increase the likelihood of an adverse outcome.1,2 Most studies find that at least 30% of all ED visits in the United States are nonurgent, although select studies such as those using National Hospital Ambulatory Medical Survey data report lower percentages (<10%).3-8 Visiting the ED instead of another care site (eg, physician’s office, retail clinic, urgent care) for a nonurgent condition may lead to excessive healthcare spending and unnecessary testing and treatment, and represent a missed opportunity to promote longitudinal relationships with primary care physicians (PCPs).4-6,9-12 A recent study projected $4.4 billion in annual savings if nonurgent ED visits were cared for in retail clinics or urgent care centers during the hours these facilities are open.13 With increasing demand and a shortage of PCPs, nonurgent ED use will likely increase in the near future. Recent predictions suggest that implementation of the Affordable Care Act and resulting expansions of insurance coverage will contribute to even higher levels of ED usage.14,15
There is widespread interest in interventions to discourage nonurgent ED visits. A 2006 survey found that 30% of emergency physicians work in hospitals that have implemented practices to discourage nonurgent visits.16 Interventions by health systems and payers have included patient education on what is appropriate ED use, financial disincentives such as higher copayments for ED visits, and encouragement of PCPs to provide care on evenings and weekends.17-19 Despite these efforts, nonurgent ED visits have continued to rise.20 One explanation could be that prior interventions have not adequately addressed the underlying issues that lead patients to visit EDs for nonurgent conditions.7 Moreover, policies to deter ED use can have negative, unintended consequences. For example, enrollees in high-deductible health plans, who bear a higher share of the costs of an ED visit, are less likely to seek care for a true emergency.21 Nonurgent ED use has been discussed in the peer-reviewed literature for the last 3 decades.12 However, no systematic review of nonurgent ED use in the United States has been published to date.
We conducted a systematic review of the literature and developed a conceptual framework to understand why individuals visit the ED for nonurgent conditions. Our goal was to highlight gaps in knowledge, inform future research on this topic, and empirically inform future interventions that attempt to decrease the number of nonurgent ED visits.
METHODS Study Design
We conducted a systematic review of the peer-reviewed and grey literature to identify factors associated with nonurgent ED use by adults in the United States. Studies outside the United States were excluded because they may not generalize to the unique features of the US healthcare system.22 A health sciences research librarian worked with the study team to develop our search strategy. We searched multiple databases including Cumulative Index to Nursing and Allied Health (CINAHL), OAIster, ISI Web of Science, New York Academy of Medicine Grey Literature database, PsycINFO, and PubMed. Searches used the following free text and medical subject headings terms: (emergency service, hospital OR emergency room, OR emergency department) AND (nonurgent OR nonurgent OR unnecessary OR inappropriate). We also used the “related citations” function in PubMed to identify any articles determined to be similar to articles selected for inclusion, and we hand-searched the reference lists of all included articles. The search for abstracts was conducted in January 2011.
Two reviewers (LU-P and EG) independently examined each abstract returned by the PubMed search, and 1 reviewer (LU-P) reviewed the abstracts returned by the other search engines (fewer than 10% of the total abstracts reviewed). If either or both reviewers determined that an abstract met inclusion criteria, it underwent a more thorough full-text review. One reviewer (LU-P) evaluated the full-text articles on whether they met inclusion criteria and extracted data on all included articles. To meet inclusion criteria, articles had to be published after January 1990, be written in English, and present some quantitative data (including descriptive data) on nonurgent ED use. We excluded dissertations, articles without abstracts, and articles exclusively focused on pediatric or non-US populations. Articles that presented qualitative data only or reviewed existing literature were not formally included in the review, but were used to inform the creation of a conceptual framework.5,6,22-31
To facilitate data extraction, we created a standardized data form to collect information from included articles. The information gathered, as available, included study population, sample size, setting, design, comparison group, response rate, definition of a nonurgent visit, independent and dependent variables, key findings, and use of a conceptual framework. A variety of terms were used to describe nonurgent visits including inappropriate visits,32avoidable visits,16nonemergency visits,33 and minor illness visits.34 In this article we chose the most prevalent term: nonurgent visits. The research team elected not to rate the quality of articles because all the studies were observational in nature and the majority did not use multivariate statistics.
RESULTS Identification of Relevant Articles
The initial search strategy generated 1983 abstracts. An additional 7 abstracts were obtained by hand-searching the reference lists of full-text articles and using the related citations feature in PubMed. From this list, the reviewers identified 63 articles for full-text review, of which 26 satisfied criteria for inclusion (Figure 1). The primary reasons for exclusion included lack of quantitative data and an exclusive focus on non-US patients.
Overview of Articles and Definition of Nonurgent Condition
Six studies (23%) described only visits for nonurgent connditions (Table 13,9,33,35-37). Of those, 4 articles (16%) described nonurgent visits to the ED and 2 articles (8%) compared nonurgent ED visits with PCP visits for similar conditions.33,37 The other 20 articles (77%) compared nonurgent ED visits with other types of ED visits (Table 21,2,12,16,32,34,38-51), including urgent visits, urgent and emergent visits,1,47 and all ED visits.16,34
No 2 studies used the same exact definition of nonurgent visits. A total of 11 articles (42%) identified nonurgent visits through retrospective review of medical records, 11 (42%) identified nonurgent visits prospectively at triage, and 3 articles (12%) used retrospective patient self-report. (See eAppendix at www.ajmc.com for additional detail on definitions.) Across the relevant articles, the average fraction of all ED visits that were judged to be nonurgent (whether prospectively at triage or retrospectively following ED evaluation) was 37% (range 8%-62%). Four articles (15%) presented a conceptual framework to guide the study design and interpretation of results. Three articles used the Anderson model of healthcare utilization,12,33,35 and 1 article used Mechanic’s model of illness behavior.47
In the reminder of this article, we summarize findings from the subset of articles (n = 16) that included a comparison group of either urgent ED patients or all ED patients and examined whether differences among these groups were statistically significant. We also include illustrative examples from the remaining studies (n = 10) regarding self-reported reasons for nonurgent ED use and barriers to use of alternative locations.
Factors Associated With Nonurgent Emergency Department Use
We summarize our findings on sociodemographic factors and other factors associated with nonurgent ED use in Table 3 and Table 4, respectively. These factors are discussed below.
Age. Among the 9 articles that examined age, 6 found that younger adults were more likely to have nonurgent visits compared with older adults.32,41,46,48,50,51 Effect sizes were generally large (odds ratio [OR] >2). Three articles found no association between nonurgent ED use and age.12,34,45
Race. Among the 9 articles that examined race, 4 articles found that blacks were more likely than whites to have a nonurgent visit.12,42,46,51 However, 5 articles reported no association.16,34,45,48,50 One study pointed out that blacks had higher rates of nonurgent ED visits despite the fact that they were less likely to utilize healthcare in general.12
Sex. Findings were inconsistent across the 10 articles that examined gender. Four articles found that women were more likely than men to have a nonurgent visit,32,46,48,50 and 2 articles concluded the opposite (ie, men were more likely than women to have a nonurgent visit).34,41 Four articles found no association.12,16,45,51
Income. Among the 4 articles that assessed income,12,16,34,50 2 reported that persons with low incomes were more likely to have nonurgent ED visits.12,50 Effect sizes were generally moderate (OR <2).
Insurance. Among the 13 articles that examined the uninsured, 2 found that uninsured patients were less likely to use the ED for nonurgent visits,12,49 2 found that the uninsured were more likely to use the ED for nonurgent visits,32,34 and 5 identified no association.1,16,43,45,48 One study found that the uninsured were more likely than HMO patients but less likely than Medicaid patients to have a nonurgent ED visit.44 Articles that looked at Medicaid patients found that either Medicaid was predictive of nonurgent ED use12,32,44,46,51 or there was no association.16,34,43,49 Effect sizes were generally moderate (OR <2).
Social Support. The only social support measure reported in the literature was marital status. Among the 4 articles that looked at the relationship between nonurgent ED use and marital status, no article identified an association.16,34,45,48
Health Status. Among the 4 articles that examined health status, 2 found that persons with poor health were more likely to have nonurgent visits,12,50 and 2 identified no association.16,48
Previous Healthcare Experiences. Previous healthcare experiences refer to an individual’s utilization history both within and outside of the ED. Two articles examined previous healthcare experiences. One article found that a recent hospitalization was associated with lower odds of having a nonurgent visit, more frequent ED visits were associated with higher odds of having a nonurgent visit, and the number of primary care visits had no association with having a nonurgent visit.45 In contrast, another article found that the average number of physician visits in an outpatient setting other than the ED was higher for persons with nonurgent ED visits.12
Culture/Community Norms and Personality. Culture/ community norms refers to the practices of others within one’s community (eg, the propensity of neighbors to use the ED). Personality factors are those related to an individual’s emotional, attitudinal, and behavioral response patterns. Examples of relevant traits include decision-making style and risk aversion. No article that compared nonurgent with urgent patients assessed culture or community norms or personality factors; however, 1 study of nonurgent patients found that personality factors such as coping mechanisms were not associated with going to the ED versus PCP for a nonurgent condition.37
Perceived Severity. Perceived severity refers to the patient’s perception of the urgency of his/her illness, which is a function of both personal beliefs and knowledge about what an emergency is. No article that compared nonurgent with urgent patients explored perceived severity; however, 4 articles that focused only on nonurgent ED visits described patients’ perceptions of the urgency of their conditions. In these cases, the vast majority of patients (>80%) felt that their condition was urgent/could not wait for treatment.3,9,36,38
Convenience. Convenience refers to the ease with which a patient can seek care, including travel, timing, and location. Among the 3 articles that discussed convenience,16,34,50 all found that convenience factors played a role in driving nonurgent ED use. For example, 1 study reported that the leading reason why the nonurgent group used the ED was “ease of use.”34 A descriptive study of nonurgent ED users found that 60% of nonurgent ED patients felt that the ED was more convenient than their PCP.9
Cost. Cost refers to the financial burden incurred by the patient. While no article that compared nonurgent with urgent patients assessed cost, 1 study of just nonurgent ED patients found that 42% chose the ED because of payment flexibility (ie, no requirement to pay at the time of care).3
Access. Access refers to the ability of the patient to obtain timely care outside the ED. Four articles found an association between poor access (eg, difficulty in obtaining healthcare, not having a regular physician) and nonurgent ED use.1,16,48,50 Only 1 article (which focused exclusively on a population of homeless adults) identified no association between poor access and likelihood of having a nonurgent visit.45 Furthermore, a Harris Interactive survey reported that ED physicians felt that waiting times for appointments with PCPs and limited access to physicians on weekends were the leading reasons for nonurgent ED use.16 In a descriptive study of nonurgent ED patients, authors reported that the most significant barrier to getting care outside the ED was inability to get an appointment at a clinic.35
Referral/Advice. Referral/advice refers to being counseled to go to the ED by a provider. Two articles (1 with a comparison group and 1 on only nonurgent ED users) suggested that healthcare provider referral may be a substantial driving force in nonurgent attendance.9,34 One article found that about half of the nonurgent patients who presented during business hours were advised to go there by a PCP.9 Beliefs and Knowledge About Alternatives. A total of 3 articles (2 with comparison groups and 1 on only nonurgent ED users) directly addressed beliefs about alternatives. One article reported that 76% of nonurgent ED users chose the ED because they felt they would receive better care there.3 A Harris Interactive survey reported that nonurgent ED users were more likely to think that other places were more expensive than the ED.16 Finally, another article found that persons who were not satisfied with their regular source of care were more likely to make a nonurgent visit to an ED.50
Due to the heterogeneity and limitations of the articles, it is challenging to summarize what drives the decision to seek ED care for nonurgent conditions. The limited evidence suggests that younger age, greater convenience of the ED compared with other ambulatory care alternatives, referral to the ED by a healthcare provider, and negative perceptions of non-ED care sites all play a role in decisions to seek care in the ED for nonurgent problems. Other factors appear unrelated to nonurgent ED use, or more commonly, the results are inconclusive due to inconsistencies across studies or because the factors have rarely been studied. Because of the weak evidence base, we argue that all of the factors assessed in the literature are candidates for future research.
We believe a key limitation of these prior studies is the lack of a robust theoretical framework on what drives nonurgent ED use. To potentially guide future work, we created a theoretical model of the decision-making process and factors that may influence a patient’s decision to visit the ED for a nonurgent condition. We based the model on review of included studies, as well as qualitative studies and commentaries. 6,7,22,24,26,27,30,31,52 Qualitative studies that used patient interviews and focus groups were important to include because they generated hypotheses regarding reasons for use that can be probed in future empirical work.
The model depicted in Figure 2 suggests that a patient arrives at a decision to seek care in an ED by consciously or unconsciously weighing several considerations. First, the patient experiences acute symptoms—either a new problem or a flare-up of a chronic condition that is not immediately debilitating or clearly emergent (eg, chest pain, signs of stroke). The patient then considers various options including going to the ED, going to another location, or not seeking care.
In our model, the decision to go the ED is influenced by an array of causal pathway factors and associated factors. While all of the factors depicted in the model likely influence nonurgent ED use, the causal pathway factors act as independent predictors. In contrast, we believe associated factors influence ED use via one of the causal pathway factors. For example, while certain models suggest that gender may be associated with nonurgent use, there is no a priori explanation as to why gender would be influential. We believe that gender, an associated factor, could possibly impact the decision to seek care in the ED for a nonurgent condition by affecting the perceived severity of the condition and beliefs and knowledge about alternatives (both causal pathway factors). In our review, the distinction between causal pathway and associated factors is also important, as almost all interventions to decrease nonurgent ED use focus on causal pathway factors.
Although our model does not directly address healthcare supply because we focused on the perspective of the individual patient, one could imagine that the availability (or lack thereof) of options, including a limited supply of providers or an extended wait to be seen, could raise or lower the threshold for seeking care. In addition, while features of the healthcare system such as overall access to care or societal context are not the focus of our framework, they play a role in an individual’s decision making by influencing their knowledge, beliefs, and attitudes about alternative locations for care.
The literature we reviewed on nonurgent ED use has several key limitations. First, descriptive studies of just nonurgent ED visits are hard to interpret. For example, although the selfperceived severity of their problem was high among patients who visited the ED for what others judged to be nonurgent, we do not know whether perceived severity is similar among those who go to other care sites. Second, the comparison of urgent with nonurgent ED visits used in the vast majority of studies might be flawed. Urgent problems (eg, chest pain) are qualitatively different than nonurgent problems (eg, sore throat). The more relevant question is: why does the patient with a self-recognized nonurgent problem choose the ED rather than seek care at an alternative location or simply stay home? Only 2 studies compared nonurgent ED visits with nonurgent PCP visits.33,37 However, we cannot draw conclusions based on these papers because they did not evaluate similar independent variables. Ideally, future studies would also include patients who became ill with a time-limited condition but chose not to seek care. Third, studies disproportionately focused on associated factors (eg, age, sex) that are easy to measure and classify but do not provide a causal mechanism for driving nonurgent ED use and are difficult or impossible to modify. We hope that our theoretical model can guide future work to assess the frequency and relative importance of different causal factors.33,37 Fourth, there are problems in clarifying the relationship between predictors of nonurgent ED use and the definition of nonurgent use itself. For example, based on current research it is unclear whether older adults are in fact less likely to go to the ED for minor conditions or whether their visits are more likely to be deemed “urgent” because they are frail or have multiple comorbid conditions. Lastly, health services research often makes broad generalizations about populations. Because nonurgent ED users are likely a diverse group, the better approach might be to try to break up nonurgent ED users into different strata.34 For example, some individuals may be using the ED due to habit, preference, or lack of education regarding alternatives. The ideal intervention might vary by the different strata. Prior to applying them, the precise issues or challenges need be identified so that the correct intervention(s) is applied to encourage or enable desired behavior by patients.
It is widely presumed that redirecting nonurgent visits to alternate settings is a desirable policy goal, if for no other reasons than to reduce healthcare spending and enable EDs to focus their efforts on more acutely ill and injured patients. However, efforts to deter nonurgent ED use could produce unintended consequences. Imposition of steep copayments and deductibles to discourage ED use might deter some patients from timely care-seeking for serious or even life-threatening problems. Even steering patients to alternate settings from the ED triage desk is not without risk. Some studies have shown that as many as 3% to 5% of patients triaged as nonurgent require immediate hospitalization after further evaluation in the ED.1 Another unintended consequence to consider is increased utilization; efforts to encourage alternatives to the ED (eg, retail clinics) might induce patients who previously would have stayed at home to seek care. Likewise, it is only acceptable to discourage nonurgent use in communities where patients have real alternatives, such as accessible PCPs. High rates of nonurgent ED visits can in fact be an indicator of poor primary care access, as suggested by the ED Use Profiling Algorithm that classifies ED visits by whether they could be treated elsewhere or, although emergent, could have been prevented by earlier access to primary care.53
The major limitation of this review is that the validity of findings is limited by the quality of included articles. Few studied used multivariate statistics, so we are unsure whether the identified factors are associated with nonurgent ED use controlling for other factors. Also, the diverse (and controversial) criteria used to define nonurgent visits limit the comparability of findings. As described above, no 2 studies used the same exact definition of nonurgent visits, identifying nonurgent visits prospectively at triage (eg, based on symptoms) and/or retrospectively (eg, based on ultimate diagnosis). While nonurgent visits seem to represent a significant fraction of all visits, prudent layperson standards that now broadly apply to all health plans require insurers to cover emergency services if a prudent layperson believed he or she was experiencing a medical emergency (regardless of the final diagnosis).54 The standard, advocated by the American College of Emergency Physicians for more than 2 decades, conflicts in principle with the 11 articles that defined urgency based on retrospective review of medical records.
Despite the significant policy interest in deterring nonurgent ED use, our literature review highlights both the limited understanding of what drives nonurgent ED use and flaws in most of the published studies. If health plans, policy makers, and providers want to reduce use of the ED for nonurgent problems, they must ensure that their interventions are evidence- based and tailored to address the needs and concerns of the populations they are designed to serve.