OBJECTIVE: Emergency department (ED) crowding prevents the efficient and effective use of health services and compromises quality. Patients who use the ED for nonemergent health concerns may unnecessarily crowd ED services. In this article we describe characteristics of pediatric patients in the United States who use EDs for nonemergent visits.
METHODS: We analyzed data from the 2002–2005 Medical Expenditure Panel Survey. The Medical Expenditure Panel Survey is conducted by the Agency for Healthcare Research and Quality and consists of a nationally representative sample of the civilian noninstitutionalized population of the United States. Our study sample consisted of 5512 person-years of observation. We included only ED visits for children from birth to 17 years of age with a specified International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code. The main dependent variable for our multivariate logistic model was nonemergent ED use, which was constructed by using the New York University ED-classification algorithm. Independent variables were derived from Andersen's Behavioral Model of Health Services Utilization.
RESULTS: We found that from 2002 to 2005, a nationally representative sample of US children from birth to 17 years of age used EDs for various nonemergent or primary care–treatable diagnoses. Overall, children from higher-income families had higher ED expenditures than children from lower-income families. Children with private insurance had higher total ED expenditures than publicly insured or uninsured children, but uninsured children had the highest out-of-pocket expenditures. We found that children from birth to 2 years of age were less likely to use the ED for nonemergent diagnoses (odds ratio [OR]: 0.13; P < .01) compared with older children. Non-Hispanic black children were also less likely to use the ED for nonemergent diagnoses (OR: 0.40; P = .03) than were non-Hispanic white children.
CONCLUSION: Children's sociodemographic characteristics were predictors of nonemergent use of ED services.
WHAT'S KNOWN ON THIS SUBJECT:
ED crowding is a growing concern in health care. Studies have shown that nonemergent ED use may contribute to ED crowding.
WHAT THIS STUDY ADDS:
We describe characteristics of pediatric patients in the United States who use EDs for nonemergent visits.
The increasing number of hospital emergency department (ED) visits over the past decade has led to increased ED crowding and patient wait times, overburden of ED providers, and perhaps poorer overall outcomes.1,–,5 From 1995 to 2006, the annual number of ED visits increased from 96.5 million to 119.2 million.5 The total number of hospital EDs simultaneously decreased by ∼10%, thus increasing the annual number of visits per ED.5 The American College of Emergency Physicians (ACEP) has stated that the increase in visits combined with closures of EDs threatens the safety of patients and will further endanger an already fragile and overburdened system, a concern that is shared by researchers and policy makers.2,6
Studies of ED use among adult patients have revealed that patients with Medicaid made the majority of ED visits and that a large percentage of the visits were potentially preventable.7,–,9 Billings et al7 examined data on ED visits in New York City, New York, and excluded injury, mental health, and alcohol or substance abuse visits. Their results revealed that 4 of 5 visits were for nonemergent conditions, for care that could have otherwise been provided in a primary care setting, or for preventable conditions. A Utah Department of Health analysis of primary care–sensitive ED visits in Utah also revealed that 4 of every 10 ED visits in Utah in 2001 were nonemergent, emergent but treatable by a primary care practitioner, or emergent and needing ED care but for a preventable or avoidable cause (eg, asthma exacerbation).10
The appropriate use of ED services and resources can result in improved care and outcomes: patients with emergent conditions would likely receive more timely care with less crowding in the ED, and patients with nonemergent conditions would experience more continuity of care and preventive services in primary care settings.
Although ED crowding and nonemergent ED use affect all patient populations, few studies have examined pediatric patients and their use of EDs nationally for nonemergent conditions.11,–,13 In addition, sociodemographic factors associated with nonemergent ED use in a nationally representative sample of children have not been well delineated.12,14,15 By using national data from the Medical Expenditure Panel Survey (MEPS) and the New York University ED-classification algorithm (developed by the New York University [NYU] Center for Health and Public Service Research), we describe a profile of nonemergent ED visits by children and adolescents in the United States and the factors associated with nonemergent ED visits.
We posited that children's ED use and nonemergent ED visits are associated with personal, family, and socioeconomic determinants of care. Our conceptual framework was based on Andersen's Behavioral Model of Health Services Utilization.16 Andersen's model is one of the most frequently used social models that capture factors that influence health services use. The model consists of 3 main components including predisposing characteristics, enabling factors, and need.
We used the household component files and the ED event files from the 2002–2005 MEPS, a national survey on the use and expenditures of medical care conducted by the Agency for Healthcare Research and Quality.17,–,21 The MEPS has been conducted annually since 1996 and consists of a nationally representative sample of the noninstitutionalized civilian population of the United States. Each family that was surveyed by the MEPS participated in 5 rounds of data collection. A core questionnaire was administered by using computer-assisted interviewing to obtain information about each family member's health status, use of medical services, health care expenditures, income, and insurance coverage. Different versions of the health-status items were used for adults and children. In addition, data were collected each round for specific medical events and compiled in the event files. The events included office-based medical provider visits, outpatient-department visits, ED visits, hospital-inpatient stays, dental visits, prescribed medicines, home health care, and other medical expenses. The ED event files provided detailed information on the household-reported use of ED services. Each record in the file represented 1 ED visit for members in the household, with each being a unique event. The information was collected primarily on the basis of household reports and data from medical providers that were used for verification when appropriate.
The pediatric condition associated with each medical-care event (in this case, ED visit) was reported by the primary household respondent every year during data collection. This information was compiled in the MEPS by using the collapsed International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (3-digit codes) for privacy and protection from deductive disclosure. The ICD-9-CM codes were verified by professional coders with a MEPS-reported error rate of <3%. A total of 27976 ED visits were reported from 2002 to 2005, and 7104 of these visits were made by children from birth to 17 years of age.
We categorized nonemergent ED visits at the visit level by using the NYU ED-classification algorithm (developed by the NYU Center for Health and Public Service Research, in collaboration with the Robert Wood Johnson Foundation and Commonwealth Fund).7 The NYU ED-classification algorithm was developed by a panel of ED and primary care physicians on the basis of actual examinations of ∼6000 complete ED records. The NYU ED-classification algorithm computed the probability of an ED visit being emergent or nonemergent by using ICD-9-CM codes, where nonemergent cases were defined as those in which immediate medical care would not be required within 12 hours on the basis of the patient's complaint, presenting symptoms, vital signs, medical history, and age. Emergent visits were further subdivided into 2 subcategories: the primary care–treatable subcategory, for which treatment would be required within 12 hours but could have been safely completed in a primary or urgent care setting, and the ED care–needed subcategory, for which the resources of an ED would have been necessary within 12 hours. After processing the MEPS ED-visit data by using the NYU ED-classification algorithm, we aggregated all ED visits each year to person-level data, resulting in a final analytical data set representing 5512 person-years. For children with >1 ED visit, the NYU ED-classification algorithm probabilities were averaged across visits.
Although the NYU algorithm is a validated approach for classifying ED visits, it excluded some diagnostic categories: mental health–related, injury, alcohol and substance abuse–related, and other unclassifiable conditions with variable severity. To complement this gap, our 3-physician panel created 6 diagnostic categories for pediatric ED visits on the basis of a literature review and clinical guidelines.11,13,22,–,24 These categories consist of common ED visits for children and adolescents that are considered as nonemergent or more appropriate in primary care settings: asthma, influenza or other viral symptomology, otitis media, allergic symptoms (including skin), minor muscular/skeletal or sports injury, and preventive, immunization, or well-child care (see the Appendix).
We used descriptive statistics, including univariate and bivariate approaches, to summarize the characteristics of children with reported ED visits. We used multivariate logistic regression to test the hypothesis that patient-level characteristics (from Andersen's behavioral model) are associated with nonemergent ED use. Nonemergent ED use for an individual child was defined as having a >50% average probability of a visit being classified as nonemergent or primary care treatable by the NYU algorithm within that year. We conducted all analyses with the Stata 10 statistical package (Stata Corp, College Station, TX, 2008).
We estimated multivariate logistic regression by using the following independent variables: indicator variables for child's gender, categorized as male or female; race/ethnicity, categorized as non-Hispanic white, non-Hispanic black, Hispanic, or other; family income, categorized as poor (<100% federal poverty level [FPL]), near poor (100%–124% FPL), low income (125%–199% FPL), middle income (200%–399% FPL), or high income (≥400% FPL); and insurance coverage, categorized as uninsured, private insurance, or public insurance. The term uninsured was defined as being without insurance for the entire year; private insurance was defined as any private coverage for the entire year (eg, employer, union, group/school, or self-purchased); public insurance was defined as having coverage under TRICARE, Medicare, Medicaid, a State Children's Health Insurance Program, or other public hospital programs.
A P value of <.05 was chosen as the criterion for statistical significance in all of the analyses. We weighted all analyses by using weights that reflected the sample design of the MEPS as well as survey nonresponses, and we adjusted all SEs for clustering at the person level by using the Huber-White sandwich estimator.25 Our study was exempt from institutional review board approval because it used public data with no identifiable information.
Finally, we conducted a sensitivity analysis of our multivariate models to include an indicator variable for patients who reported that they had a regular source of care.
Our study sample consisted of children who belonged to various age groups: 22.1% were from birth to 2 years of age; 19.3% were aged 3 to 5 years; 28.9% were aged 6 to 11 years; and 29.7% were aged 12 to 17 years. Boys made up ∼55% of the study sample. The majority of children were non-Hispanic white (61%), privately insured (60%), and from middle- to high-income families (31% and 25%, respectively). Descriptive statistics for the sample are listed in Table 1.
Table 2 summarizes children with ED visits in the 6 diagnosis categories that were constructed by our 3-physician panel. These diagnosis categories were generally considered to be nonemergent or primary care treatable on the basis of our literature review. Among these categories, ED visits for well-child care and/or immunizations were the least prevalent. Children most frequently visited the ED for influenza or other viral illnesses. For influenza or other viral illnesses, a large percentage of children seen in the ED were from birth to 2 years of age (44%). For asthma, a large percentage of children seen in the ED were poor (<100% FPL) (44%). For otitis media, ED visits among younger children (aged 0–5 years) were more prevalent than with older children (aged 6–17 years). For allergy-related symptoms, non-Hispanic white children were the most prevalent racial/ethnic group (46%). For minor injury, ED visits were much more prevalent among older children (aged 6–17 years, 86%) than young children (aged 0–5 years, 13%). Overall, 27% of the ED visits in our data were for 1 of these 6 (likely nonemergent) categories.
Children from higher-income families, as expected, had higher ED expenditures than children from lower-income families. Children with private insurance had higher total ED expenditures than children with public insurance or those who were uninsured; however, uninsured children had the highest out-of-pocket ED expenditures.
Table 3 illustrates the likelihood of nonemergent ED use compared with emergent visits on the basis of the NYU ED-classification algorithm and adjusted for other covariates. We found that children from birth to 2 years of age were less likely to use the ED for nonemergent diagnoses (odds ratio [OR]: 0.13; P < .01) compared with older children. Non-Hispanic black children were also less likely to use the ED for nonemergent diagnoses (OR: 0.40; P = .03) than were non-Hispanic white children. It is interesting to note that we found that, all else being equal, girls were more likely than boys to use the ED for nonemergent diagnoses (OR: 1.93; P = .02).
Over the past decade, the annual number of ED visits has increased by ∼20%.5 Several national studies of adult patients and regional studies of pediatric patients have suggested that a significant proportion of ED visits were nonemergent or primary care treatable and preventable.6,–,8,13,–,15 In this study, which used data from a nationally representative pediatric sample, we reached similar conclusions. In addition, we found that nonemergent ED visits among children were associated with several sociodemographic factors.
A review of the literature showed that ED crowding can be attributed, at least in part, to inappropriate or nonemergent ED use.1,–,8,10 ED crowding can lead to increased mortality and morbidity, service delays, patient dissatisfaction, and financial burdens.1 One potential solution to alleviate ED crowding would be to identify and reduce inappropriate or nonemergent ED use. Several studies with adult patients have identified age, education, insurance, and other barriers to primary care as factors that are associated with inappropriate ED use.1,6,8,10,26,–,29 Specifically, patients identified access barriers in the primary care clinic as the major reason for choosing the ED instead of the clinic.27 They reported a cumbersome scheduling system, long waiting times for appointments, and no availability of walk-in care.27 In a study by Billings et al, New York City residents chose to seek care in an ED (despite most patients knowing their conditions were not an emergency) because of factors such as the convenience and level of service offered by EDs, as opposed to the longer wait times and inconvenient hours when compared with alternative care settings.28,30
Our findings have shown that nonemergent ED use with children is associated with several sociodemographic factors, as conceptualized by Andersen's Behavioral Model of Health Services Utilization. Although the MEPS data were not suited to study causative mechanisms for nonemergent ED use, evidence from the literature suggests that predisposing factors such as age, gender, and race; enabling factors such as health literacy, knowledge or education, insurance coverage, and the availability and ease of primary care services; and the perceived need for urgency can all play a role.
Our study had several limitations. First, the NYU ED-classification algorithm excluded injuries, visits related to alcohol and other substance use, and mental health visits. Thus, we were unable to determine, among ED-visit data that are representative of US children nationally, what proportion were nonemergent. Second, ICD-9-CM code-based classification of emergent versus nonemergent visits is crude. For the same ICD-9-CM code (eg, ED visit for asthma), the condition can be emergent or nonemergent depending on the disease severity. The existing literature has confirmed that there is a lack of an accepted, feasible, and valid method for determining ED-visit urgency. Clearly, comprehensive ED medical chart review is the gold standard for determining urgency; however, at the national level, this is not feasible and is extremely costly.31 By using the NYU ED algorithm on ICD-9-CM codes as a crude approach for visits classified in our data as nonemergent, 0.5% resulted in direct hospital admission, and 11% had additional expenditures (eg, further workup or observation). Third, as with any secondary data analysis, our study was subject to omitted-variable bias in that the MEPS or Andersen's behavioral model may not have captured some confounding factors. Finally, because our study sample included only children who visited an ED, we cannot comment on the sociodemographic differences between those patients who used ED services versus those who did not.
There are several important implications for nonemergent or inappropriate ED use. Some of the negative consequences of nonemergent ED use include longer waiting times, a more stressful environment for ED physicians, decreased quality, and delays in care. In addition, the financial impact on the sustainability of an ED to provide care in an urban setting is enormous. Our findings can help ED providers, working with community pediatricians and family physicians, to identify and educate parents about nonemergent ED use. Potential ideas to help resolve some of these problems may include collaborations between EDs and local physicians to create an after-hours telephone line to which patients can call and ask if an ED visit is warranted, the distribution of handouts to address similar topics, and the joint creation of an after-hours clinic.
Dr Chen is funded by a National Institutes of Health K23 award.
- Accepted September 2, 2009.
- Address correspondence to Alex Y. Chen, MD, MS, Childrens Hospital Los Angeles, 4650 Sunset Blvd, Mail Stop 30, Los Angeles, CA 90027. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- ED =
- emergency department •
- ACEP =
- American College of Emergency Physicians •
- MEPS =
- Medical Expenditure Panel Survey •
- NYU =
- New York University •
- ICD-9-CM =
- International Classification of Diseases, Ninth Revision, Clinical Modification •
- FPL =
- federal poverty level •
- OR =
- odds ratio
- Trzeciak S,
- Rivers EP
- Schappert SM,
- Burt CW
- Pitts SR,
- Niska RW,
- Xu J,
- et al
- Billings J,
- Parikh N,
- Mijanovich T
- Weinick R,
- Billings J,
- Thorpe J
- 10.↵Utah Department of Health, Health Data Committee, Center for Health Data, Office of Health Care Statistics. Primary Care Sensitive Emergency Department Visits in Utah 2001. Salt Lake City, UT: Utah Department of Health, Health Data Committee; 2004
- Freid VM,
- Makuc DM,
- Rooks RN
- Luo X,
- Liu G,
- Frush K,
- Hey LA
- 18.↵Agency for Healthcare Research and Quality, Center for Financing Access and Cost Trends. MEPS HC-070: 2002 full year consolidated data file. Available at: www.meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h70/h70doc.pdf. Accessed May 20, 2008
- 19.↵Agency for Healthcare Research and Quality, Center for Financing Access and Cost Trends. MEPS HC-079: 2003 full year consolidated data file. Available at: www.meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h79/h79doc.pdf. Accessed May 20, 2008
- 20.↵Agency for Healthcare Research and Quality, Center for Financing Access and Cost Trends. MEPS HC-089: 2004 full year consolidated data file. Available at: www.meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h89/h89doc.pdf. Accessed May 20, 2008
- 21.↵Agency for Healthcare Research and Quality, Center for Financing Access and Cost Trends. MEPS HC-097: 2005 full year consolidated data file. Available at: www.meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h97/h97doc.pdf. Accessed May 20, 2008
- Sharma V,
- Simon SD,
- Bakewell JM,
- Ellerbeck EF,
- Fox MH,
- Wallace DD
- Huber PJ
- Billings J,
- Parikh N,
- Mijanovich T
- Copyright © 2010 by the American Academy of Pediatrics