 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Center for Clinical Effectiveness, Department of Surgery, Duke University Medical Center, Durham, North Carolina
Pharmaceutical Policy and Evaluative Sciences, School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
Division of Pediatric Emergency Services, Department of Surgery and Pediatrics, Duke University Medical Center, Durham, North Carolina
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Methods. Secondary analysis of the household component of the 1997 Medical Expenditure Panel Survey on 10 193 children younger than 18 years. The main outcome measures were annual overall ED utilization and ED utilization for nonurgent problems.
Results. During 1997, 10.8% of children were uninsured for the entire year. A total of 17.5% of children were publicly insured the entire year, whereas 55.3% of children held private insurance the entire year. There were also 16.5% of children who were insured only part of the year. Without adjusting for covariates, publicly insured children were more likely to have an ED visit during the year than both privately insured children (unadjusted odds ratio [OR]: 1.26; 95% confidence interval [CI]: 1.03–1.55) and uninsured children (unadjusted OR: 1.46; 95% CI: 1.1–1.95). The difference between publicly insured and privately insured children (adjusted OR: 0.90; 95% CI: 0.70–1.16) and between publicly insured and uninsured children (adjusted OR: 1.12; 95% CI: 0.84–1.49) became insignificant after controlling for covariates. With or without adjustments for covariates, there was no significant difference in the likelihood of having an ED visit between privately insured and uninsured children. Similar to the utilization pattern of overall ED visits, publicly insured children were more likely to have a nonurgent ED visit than both privately insured (unadjusted OR: 1.86; 95% CI: 1.36–2.53) and uninsured children (unadjusted OR: 1.81; 95% CI: 1.15–2.84). Both differences disappeared after controlling for covariates. There was no significant difference in the likelihood of nonurgent ED visits between privately insured and uninsured children with or without adjustments for covariates.
Conclusions. Health insurance status was not associated with children’s overall ED use or children’s ED use for nonurgent problems at the national level. Our findings suggest that policy efforts in an attempt to relieve ED overcrowding conditions should look for measures beyond solely making changes in health insurance coverage for children.
Key Words: emergency • nonurgent • insurance • children • utilization
Abbreviations: ED, emergency department • HMO, health maintenance organization • MEPS, Medical Expenditure Panel Survey • OR, odds ratio • CI, confidence interval
Health insurance status has a profound impact on children’s access to primary care. Lack of health insurance is associated with fewer physician visits, inadequate preventive services, and a lack of a usual source of medical care.1,2 The implementation of public insurance programs such as Medicaid has improved, to some extent, children’s access to a usual source of routine or sick care, but poor children with Medicaid coverage were still less likely to receive routine care in physicians’ offices and were more likely to lack continuity of care than nonpoor children.3 The lack of primary care among uninsured and Medicaid-covered children may result in their overdependence on emergency departments (EDs). Studies demonstrated that uninsured children were more likely to identify the ED as their usual site of care as compared with insured children.1 Of insured children, those with Medicaid coverage tended to have higher rates of nonurgent ED visits than those with private insurance.4
ED overcrowding is commonly observed and has become a national problem in recent years.5 Each year, millions of children in the United States use the ED for medical care, and the pediatric population accounted for nearly 1 in every 4 ED visits.6 Studies indicate that between one third and one half of these visits are for nonurgent problems.6 In some places, rates of nonurgent ED visits reached 70%.7 This type of medical service utilization may contribute significantly to the overcrowded conditions in EDs. In 1997, Congress enacted the State Children’s Health Insurance Program (SCHIP) to authorize >$20 billion over 5 years to extend health care coverage to uninsured low-income children. At the same time, the Medicaid program was undergoing change. Some state Medicaid programs began to make the transition from a fee-for-service model to managed care and tried to diminish the gap with private insurance.8 It was hoped that such health care reforms could provide uninsured or publicly insured children better access to primary care and consequently reduce their dependence on emergency care.9,10 To date, there are few data to suggest that such changes have led to the expected results.
Numerous studies have been conducted to examine the relationship between health insurance status and children’s ED utilization.4,11–16 The studies can be divided into 2 groups, 1 group of which examined overall ED utilization. Using national survey data sets, researchers in these studies found that publicly insured children were more likely to visit the ED than privately insured or uninsured children. However, these studies were merely descriptive, and potential confounding factors were not considered.11,12,15 Another group focused on ED utilization for nonurgent problems. Most of the studies in this group used data from 1 or 2 local hospitals, and their results cannot be generalized nationally.4,13,16 One study did use a national survey data set but examined only the effects of a health maintenance organization (HMO) on nonurgent ED visits. Results indicated that children with private, non-HMO insurance were more likely to visit the ED for nonurgent problems than children with HMO insurance.14 Whether the nonurgent ED visits differed between privately insured and publicly insured or uninsured children was not examined.
On the basis of our literature search, little is known about the effects of insurance status on children’s overall ED utilization or ED utilization for nonurgent problems at the national level in the United States. The objective of this study was to use the 1997 Medical Expenditure Panel Survey (MEPS), a national survey data set, to investigate the relationship between health insurance status and children’s overall ED utilization and the relationship between health insurance status and children’s ED utilization for nonurgent problems. Our null hypothesis was that publicly insured or uninsured children were more likely to visit the ED and more likely to have a nonurgent ED visit in a year than privately insured children.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Health Insurance Coverage
Children’s health insurance coverage for 1997 was grouped into 5 categories: uninsured all year, privately insured all year, publicly insured all year, insured only part of the year, and insured all year but switched between private and public insurance. Because <1% of children belong to the last category, this study investigated only the first 4 categories of health insurance.
Overall ED Utilization
Overall ED utilization was defined as a dichotomous variable. It measured whether a child had any ED visits in 1997.
ED Utilization for Nonurgent Problems
ED utilization for nonurgent problems was also defined as a dichotomous variable. It measured whether a child had any nonurgent ED visits in 1997. Because MEPS does not provide direct information about whether an ED visit is for urgent or nonurgent problems, we used a variety of household-reported data sources to classify the visits. Households reported the type of care that a child received during the ED visit; categories include 1) diagnosis or treatment, 2) emergency, 3) psychotherapy or mental health counseling, 4) follow-up or postoperative visit, 5) immunization or shots, 6) pre/postmaternity care, and 7) other. When the type of care that a child received in the ED was designated as "emergency," the visit was considered urgent. Other visits that were not designated as "emergency" but were also considered urgent include 1) visits that led to admission and 2) visits in which children received tests or procedures that were not immediately available in primary care offices and were needed for the diagnosis and treatment of some potential urgent problems. These tests or procedures include radiography, computed tomographic scan or magnetic resonance imaging, electrocardiogram, electroencephalogram, and any surgical procedures. The remaining visits were classified as nonurgent. On the basis of our criteria, 38.6% of the ED visits was classified as nonurgent. Because our focus was on nonurgent ED visits, we tried to identify ED visits that have a relatively high certainty of being nonurgent. We therefore used somewhat more relaxed criteria to classify urgent visits.
We tested the validity of our criteria for nonurgent ED visits by using International Classification of Diseases, Ninth Revision codes. Although International Classification of Diseases, Ninth Revision codes provide little information about severity or associated complications of a condition, specific conditions such as otitis media, nasopharyngitis, and viral infection that, in most cases, could be treated in primary care offices and are unlikely to be life threatening.17 These conditions are likely to be nonurgent. On the basis of this rationale, if our criteria were valid, then we would expect that ED visits for each of the above conditions were more likely to be classified as nonurgent than urgent. Data analysis was consistent with our expectation. For otitis media, 77.3% of the ED visits were classified as nonurgent versus 22.7% of the ED visits as urgent (P < .0001). For nasopharyngitis, 81.3% of the ED visits were classified as nonurgent versus 19.7% of the ED visits as urgent (P < .0001). For viral infection, 72.1% of the ED visits were classified as nonurgent versus 27.9% of the ED visits as urgent (P < .0001).
Statistical Analysis
MEPS is a stratified, multistage, complex design survey. Two issues need to be considered when analyzing such complex survey data. First, each sample involved in the survey was not equally selected, and sampling weight is needed for the unbiased national estimates. Second, the complex sample design including stratification and clustering of samples within 1 primary sampling unit will have impacts on the estimation of sample variance. Given these considerations, all estimates presented in the text and tables have been weighted to reflect national estimates. The standard errors used in computing test statistics were calculated by using survey modules of Stata 7.0 that can account for complex survey design (Stata Corp, College Station, TX). The threshold for statistical significance was set at P = .05.
Our analysis proceeds in 3 steps: descriptive analysis, simple logistic regression analysis, and multivariate logistic regression analysis. In the descriptive analysis, for each of the 4 insurance groups, the percentage of children with any ED visits in the year and the percentage of children with any nonurgent ED visits in the year were calculated. The results were weighted to be nationally representative, and the estimation of standard errors has accounted for the complex design of the survey.
After the descriptive analysis, simple logistic models were estimated to examine whether the likelihood to have an ED visit in the year and the likelihood to have a nonurgent ED visit in the year were significantly different among the different insurance groups. Both models were specified as a function of insurance status only. In the first model, the dependent variable was whether an ED visit occurred in a year. In the second model, the dependent variable was whether a nonurgent ED visit occurred in a year.
The final analysis was to examine the independent effect of insurance status by controlling for potential confounding factors in a multiple logistic regression model. Potential confounding factors included in the multivariate models were selected on the basis of a review of theoretical18 and empirical literatures16,17,19 addressing access to health care, especially access to emergency care. These covariates included age, gender, ethnicity, residential area, family income, number of parents, family size, and health status. Ethnicity was classified as whether a child was black or Hispanic. Residential area was classified as whether a child lived in metropolitan area. Family income was determined by the household income as a percentage of the federal poverty guideline. It was grouped into the following 5 categories: <100% poverty line, 100% to 124% poverty line, 125% to 199% poverty line, 200% to 399% poverty line, and 
400% poverty line. Number of parents was dichotomized into those with both parents and those with single or no parent. Health status was determined by the self-report overall health. MEPS contains 3 rounds of data for most variables in 1997. Information from the first-round survey of 1997 was used for the confounding factors.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Table 1 shows children’s demographic characteristics by insurance status. The demographic characteristics were different among different insurance groups, especially between publicly insured and privately insured children (Table 1). As compared with privately insured children, publicly insured children tended to be black or Hispanic and were much more likely to have family income below the poverty line. Children who were uninsured all year or insured part of the year were also more likely to be black or Hispanic and more likely to have family income below the poverty line than privately insured children. However, the difference was smaller than that with publicly insured children (Table 1).
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
We found no significant difference in children’s overall ED use among the different insurance groups. Contrary to our findings, by using national survey data sets, several studies found that publicly insured children were more likely to use the ED than privately insured and uninsured children.11,12,15 However, these studies were merely descriptive and did not control for potential confounding factors. Without adjusting for covariates, we also found that publicly insured children were more likely to have an ED visit than privately insured and uninsured children, but after covariates were controlled for, the difference became insignificant (Tables 2 and 3). This suggests that other factors related to health insurance status may account for the difference observed in previous studies and in our simple analysis. Indeed, we found that factors such as family income and ethnicity were significantly associated with overall ED use in our multivariate analysis (Table 4).
Not all children’s ED visits are for emergency care. In fact, a significant number of children’s ED visits are for nonurgent problems.7 The relationship between insurance status, especially Medicaid, and children’s ED use for nonurgent problems has been studied before, but the results have been controversial. Several studies demonstrated that children who are covered by Medicaid were more likely to use the ED for illnesses with lower severity scores.4,13 Other studies found that Medicaid coverage was not a predictor for pediatric nonurgent ED visits.16 Because all of these studies were based on limited observations from particular EDs, the varying samples and different methods may cause such disparities. Our study results were obtained from a set of logistic regression analyses of the 1997 MEPS, a nationally representative database. Without adjusting for covariates, we found that publicly insured children were more likely to have a nonurgent ED visit than privately insured and uninsured children, but the difference became insignificant in multivariate analysis. Just like the overall ED use, factors related to insurance status, such as family income and ethnicity, may account for the difference observed in the simple analysis.
It is interesting to find that black children and Hispanic children were less likely to have an ED visit and less likely to have a nonurgent ED visit than other children in multivariate analysis. This is consistent with a recent study in which a higher proportion of white children had an ED visit when compared with Hispanic or black children and children of other racial/ethnic groups.12 However, in contrast to our results and results from the recent study, most of the studies in this area found that blacks are more likely than whites to use the ED6,20 or to use the ED for nonurgent problem17 or that there was no significant difference in ED use among blacks, Hispanics, and whites.19 Although ethnicity is not a focus for this study, it will be important to investigate in the future why our studies are different from some of the previous studies.
This study is subject to a few limitations. First, our definition of a nonurgent ED visit was based on survey respondents’ reports about the various characteristics of an ED visit. Because information provided by household respondents is subject to recall and reporting bias and because respondents’ perception of the seriousness of the conditions were not addressed directly, such definitions may not accurately reflect conditions that would cause a "prudent layperson" to seek emergency care. Also, because we wanted to identify ED visits that have a relatively high certainty of being nonurgent, we used somewhat more relaxed criteria to determine urgent ED visits. This may have led us to misclassify some nonurgent ED visits as urgent. However, the validity of our classifying criteria has been supported by the analysis showing that the ED visits were much more likely to be classified as nonurgent than urgent for some specific conditions that in most cases were nonurgent. Moreover, there are no standard or widely accepted criteria to define urgent or nonurgent ED visits.21 Second, to understand the independent effects of insurance status, we tried to adjust for a range of potential confounding factors in our multivariate analysis. There are still some variables that we may have failed to control for. For example, we did not include mother’s education in the multiple logistic regression analysis. This factor has been found by previous studies to be associated with children’s use of the ED for routine care.22 We chose not to include it in the multiple regression models because nearly 10% of children did not have this information. The study sample size would be reduced substantially if this variable was included in the model.
ED overcrowding has become a national problem, leading to various negative impacts. The causes of overcrowding are complex and often interwoven. Reports from the US General Accounting Office indicated that growth in ED visits has been particularly pronounced among Medicaid and Medicare recipients and uninsured patients.23 Some policy makers believe that the number of ED visits will decline if uninsured Americans become eligible for universal insurance coverage or problems with public insurance are resolved.10 However, our study focusing on the national pediatric population indicated that uninsured children were not significantly different from publicly insured or privately insured children in their overall ED utilization or ED uses for nonurgent problems. There was also no significant difference in overall ED uses or ED uses for nonurgent problems between privately insured and publicly insured children. A strong policy implication from these findings is that reducing children’s ED visits and relieving ED overcrowded conditions will require more than just health insurance coverage. Other determinants of pediatric ED use must be taken into consideration.
|
FOOTNOTES |
|---|
Reprint requests to (X.L.) DUMC 3645, Duke University Medical Center, Durham, NC 27710. E-mail: luo00003{at}mc.duke.edu
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  M. A. Allison, L. A. Crane, B. L. Beaty, A. J. Davidson, P. Melinkovich, and A. Kempe School-Based Health Centers: Improving Access and Quality of Care for Low-Income Adolescents Pediatrics, October 1, 2007; 120(4): e887 - e894. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Pileggi, G. Raffaele, and I. F. Angelillo Paediatric utilization of an emergency department in Italy Eur J Public Health, October 1, 2006; 16(5): 565 - 569. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. D. Mistry, R. G. Hoffmann, J. S. Yauck, and D. C. Brousseau Association Between Parental and Childhood Emergency Department Utilization Pediatrics, February 1, 2005; 115(2): e147 - e151. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. J. Chung and M. A. Schuster Access And Quality In Child Health Services: Voltage Drops Health Aff., September 1, 2004; 23(5): 77 - 87. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Committee on Pediatric Emergency Medicine Overcrowding Crisis in Our Nation's Emergency Departments: Is Our Safety Net Unraveling? Pediatrics, September 1, 2004; 114(3): 878 - 888. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2003 American Academy of Pediatrics. All rights reserved. |
||
|
||
