PEDIATRICS Vol. 119 No. 6 June 2007, pp. e1319-e1324 (doi:10.1542/peds.2006-2309)
ARTICLE |
Differences in Severity-Adjusted Pediatric Hospitalization Rates Are Associated With Race/Ethnicity
a Department of Pediatrics, George Washington University School of Medicine, Washington, DC; Divisions of
b Emergency Medicine
g Critical Care Medicine, Children's National Medical Center, Washington, DC
c Department of Pediatrics, University of California School of Medicine, Davis, California
d Center for Health Services and Community Research, Washington, DC
e Children's Research Institute, Washington, DC
f Center for Hospital-Based Specialties, Washington, DC
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
OBJECTIVE. Racial/ethnic disparities in health care delivery have been well described, but little is known about such disparities for children who seek emergency care. The objective of this study was to test the hypothesis that severity-adjusted emergency department pediatric admission rates are associated with race/ethnicity.
METHODS. Secondary analyses were conducted of an established database of 16 emergency departments that participated in a national study to validate the Pediatric Risk of Admission II score, which is used to measure severity of illness. Patients were randomly selected by the coordinating center from daily emergency department visit logs. Crude and severity-adjusted admission rates were compared among the 3 most common races/ethnicities: white, black, and Hispanic. Adjusted admission rates were calculated by using the standardized admission ratio, which was calculated by dividing the observed admissions by the predicted admissions, when predicted was calculated from the Pediatric Risk of Admission II score.
RESULTS. After exclusion of 3 sites that recorded race/ethnicity in <10% of patients, there were 13 sites with 8952 patients in the 3 major race/ethnicity groups. Black and Hispanic patients were similar to each other and different from white patients; therefore, these 2 groups were combined for analyses. Both crude (8.2% vs 5.3%) and severity-adjusted (standardized admission ratio: 1.71 vs 1.1) admission rates were higher in white than in nonwhite patients. Standardized admission ratios were close to 1.0 in both race/ethnicity groups in the higher quintiles of illness severity. In contrast, white patients were admitted at 1.5 to 2 times the expected rate in the lowest 2 quintiles of severity.
CONCLUSIONS. There are differences in both crude and adjusted admission rates between white and black/Hispanic patients. The results are more consistent with high rates of discretionary admissions for white patients with low illness severity than with underadmitting severely ill black or Hispanic patients.
Key Words: health services research • health care delivery • disparities • health care quality
Abbreviations: ED—emergency department • PRISA—Pediatric Risk of Admission • PRISA II—Pediatric Risk of Admission, second generation • PEM—pediatric emergency medicine • GEE—generalized estimating equation • SAR—standardized admission ratio • OR—odds ratio
Inequities in health care on the basis of race and/or ethnicity are a serious quality problem facing the American health care system.1,2 In addition to obvious differences in access that are associated with social class and poverty, several authors have raised concerns that racial disparities in health care are not simply an issue of inadequate access. In several adult studies, even when care is available and adjustments are made for insurance status and income, medical care is unequal when racial/ethnic groups are compared.3–6 These disparities occur across a wide range of diseases and health circumstances, ranging from primary care preventive and screening services7,8 to care for end-stage renal disease,9 cancer,10,11 and HIV.12,13 Importantly, some studies have also demonstrated higher mortality rates in minorities associated with these disparities.14–16 In the emergency department (ED) setting, where access is guaranteed, there are significant racial differences in the care of patients with acute myocardial infarction,16–18 and black and Hispanic patients are less likely to receive analgesic medications for bone fractures.19,20 In children, appendicitis is more likely to result in rupture and abscess in nonwhite compared with white patients.21
With the exception of previous work done by our group,21 no studies have examined racial disparities in the treatment of children with emergency conditions. An important barrier to this type of research has been the lack of a method for adjusting for illness severity. We developed and previously described a severity of illness adjustment method called the Pediatric Risk of Admission (PRISA) score.22 PRISA uses the risk for hospital admission as an index of severity and was developed and tested in a single institution.22 PRISA was found accurate in an independent study in a single hospital in Canada23 and was recalibrated in 5 academic medical centers in the United States.24 The second-generation score (PRISA II)25 was redeveloped and validated in a national sample of 16 EDs. We conducted analyses that are reported here to test the hypothesis that there are differences in severity-adjusted rates of pediatric emergency hospitalization associated with race/ethnicity.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The results presented in this article represent secondary analyses of data that were collected for the purpose of developing the PRISA II severity-of-illness score for pediatric emergency patients.25 The details of site and patient selection are described in detail elsewhere.25 In brief, 16 EDs were block randomly selected in a balanced 2 x 3 factorial design to represent the care factors of high or low volume (less than or equal to or more than the median of pediatric visits per year), presence or absence of a pediatric emergency medicine (PEM) subspecialist, and presence or absence of residents delivering ED care. Two sites were selected randomly for each stratum, resulting in 16 study sites. There were 8 sites with high volume and 8 with low volume, 8 sites with a PEM subspecialist and 8 without, and 8 sites with residents and 8 without residents.
The coordinating center randomly enrolled 2 patients per day from each ED using a random-number generator to select patients from each site's daily consecutive patient arrival log. Data were collected for 375 consecutive days, but actual record collection ranged from 729 to 753 patients. For sites with >729, patients were randomly eliminated to keep only 729 patients from each site, thereby maintaining a balanced 2 x 3 factorial design.
Masked, photocopied data were abstracted and computerized. Abstracted data included demographic, historic, physiologic, diagnostic, and therapeutic variables. Race/ethnicity were recorded with the following categories, as commonly used by the participating hospitals: white, black, Hispanic, Asian, and other. There were too few patients with Asian or "other" for meaningful analysis; therefore, we compared outcomes among the 3 major categories of white, black, and Hispanic. Preliminary analyses revealed obvious and consistent differences between white patients and the other 2 groups but little difference between black and Hispanic patients. Therefore, apart from initial descriptive data, we compared white patients with the combination of black and Hispanic patients, reasoning that (1) both of these can be considered minorities when compared with white individuals, and (2) the observed differences were in the same direction and of approximately the same magnitude. Socioeconomic data were not available on individual patients; therefore, we used postal zip code–level median household income from the 2000 US census.
The primary outcome was hospital admission, which was defined as admission to an inpatient unit or to an observation area for >12 hours. Mandatory admissions were defined as admissions that received a therapy (eg, intravenous antibiotics) or had a condition (eg, full-thickness burn) that generally required inpatient care. The list of characteristics that defined mandatory admission was validated using a consensus process and has been published previously.22 All classifications of patients regarding admission and mandatory admission status were made by the principal investigator after review of inpatient records. Analyses used the previously developed PRISA II score, which models the risk for medically mandatory hospital admission using historical, physiologic, and therapeutic variables in a multiple logistic regression model. The PRISA II score was developed using generalized estimating equations (GEEs),26 an extension of generalized linear model regression analysis that incorporates the correlated data or effects of patient clustering within institutions.
For the analyses reported here, crude and adjusted admission rates were calculated and compared among the 2 race groups described. Adjusted admission rates were calculated using the standardized admission ratio (SAR; observed admissions divided by predicted mandatory admissions, where number of predicted mandatory admissions was calculated from the PRISA II model). A SAR of 1.0 indicates that the total number of admissions equals the predicted number of mandatory admissions on the basis of illness severity. A SAR >1 indicates admission rates in excess of those predicted by the severity of illness of the group, whereas SARs <1 indicate fewer admissions than predicted by severity. In addition, we used multivariable logistic regression modeling that included demographic variables to determine the additional effect of race on admission probability, expressed as adjusted odds ratios (ORs). Included in this model were the PRISA II score (illness severity), gender, race, and median household income. Household income was determined from the 2000 US census using the patient's zip code. The GEE method was used to adjust for the clustering of patients within institutions. GEEs are an extension of logistic regression analysis that adjust the confidence intervals of the parameter estimates on the basis of the degree of clustering of patients. For example, if most of the Hispanic patients were treated at only a few institutions, then GEE would account for how much of the variance in admission rates was attributable to different admission practices in those institutions as well as how much of the variance is related to differences in the way Hispanic patients were treated compared with white patients. We did not include age as a separate variable because age is incorporated into the PRISA II score, both as a risk factor (age <90 days) and in the determination of reference ranges of physiologic variables.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
A total of 11664 pediatric patients were enrolled, 8668 (74.3%) of whom had 1 of the 3 major race categories recorded (white, black, and Hispanic). Three hospitals recorded race in <10% of patients; these sites were therefore excluded from additional analysis. Within the remaining 13 sites, 525 patients had other race categories (Asian, other, no race recorded) and were also excluded from additional analysis. Thus, the total sample for analysis included 8952 patients from 13 institutions (Fig 1).
|
The 3 major race/ethnicity groups were compared with respect to patient characteristics and the most common diagnoses (minor injuries, otitis media, and fever), as shown in Table 1. Black children were more likely to arrive by emergency medical services. Hispanic patients had higher baseline severity (ie, PRISA II scores) compared with the other groups. White children were more likely to present with minor injuries and less likely to have otitis media than black and Hispanic children.
|
Crude and adjusted admission rates by race are presented in Table 2. White patients had higher overall rates of admission and higher rates of admissions to ICUs. The difference in overall rates of admission persisted even after adjustment for illness severity using the SAR (Table 2 and Fig 2). Figure 3 depicts the SAR for the race/ethnicity groups for each quintile of predicted admission risk. It is apparent that as the severity of illness quintile rises, admissions are increasingly equivalent and close to predicted. However, in the 2 lowest risk quintiles, white patients are admitted at 1.5 to 2 times the rate predicted by the need for therapies or their physiologic status, which is not true for black and Hispanic patients, for whom the SAR is close to 1 in all quintiles. Table 3 depicts the GEE model (multiple logistic regression model), which incorporates demographic variables, including race, gender, median household income, and severity-of-illness score (PRISA II). The adjusted OR for admission for black and Hispanic children is approximately half (0.54) that of the reference group of white children.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
These results are consistent with a growing body of literature demonstrating disparities in health care that are associated with race. Both crude and severity-adjusted admission rates were lower for black and Hispanic children when compared with white children. In multivariate analyses, the probability of hospital admission for black and Hispanic children was lower even after controlling for illness severity and sociodemographic characteristics. Therefore, in this patient sample, the effect of race/ethnicity on admission probability is independent of income measured at the zip code level.
Three important points about these results are worth noting. First, more health care is not always better. These data are more consistent with a practice of overadmitting white patients who are less severely ill than underadmitting black and Hispanic patients who are more severely ill. As depicted in Fig 3, white patients who are not severely ill are admitted in excess of predicted. It is in such less seriously ill patients that there is greater physician discretion in admission decision-making. In contrast, for both white and nonwhite patients, admission rates are very close to predicted in the more severely ill quintiles. These results are consistent with a national study of myringotomy in preschool children. White patients were more likely to receive a myringotomy,27 a surgical procedure for which there are clear medical indications in only 41% of patients.28 The second point is the critical importance of accounting for illness severity when studying health care disparities. Without the ability to measure severity quintiles, as displayed in Fig 3, we might have erroneously concluded that black and Hispanic patients were being denied essential hospital admissions. Instead, our results suggest that white patients are overadmitted when not severely ill. Third, the magnitude of the effects of race/ethnicity that were observed in this study (adjusted OR: 0.54; Table 3) is consistent with previous literature, which demonstrated adjusted ORs of 0.47 to 0.60.3,8,9,12
The appropriate disposition of patients to either inpatient care or outpatient care is one of the most important decisions made by ED physicians. Admitting patients unnecessarily creates iatrogenic risk29,30 and unnecessarily increases health care costs,29 whereas failing to admit patients who require hospitalization may allow progression of disease without adequate therapy. Therefore, we chose to study the association of hospital admission with race/ethnicity in this large national database of ED visits. One significant advantage of this study is the ability to control for illness severity, including physiologic variables. The PRISA II score was developed and validated using this patient sample.25 Therefore, we were able to assess the independent contribution of race/ethnicity to admission probability after controlling for physiologic and other indicators of illness severity.
There are several limitations to this study. First, race/ethnicity was recorded on medical charts by various personnel, including physicians, nurses, and clerical staff. Race is a social construct rather than a biological trait; therefore, the accuracy of assignment is not really at issue. The results of this study suggest that however race is defined in EDs, it is associated with differential probability of hospitalization. Second, we did not collect insurance information or household income on individual patients, and there are well-described limitations of using postal code median household income as a proxy for individual income.31 Finally, there are many immeasurable factors in the complex decision to hospitalize a child, including parental preferences and community and family resources. Given our study design, it is impossible to determine the reasons for the observed differences in treatment. Language barriers, for example, might be associated with admission decisions for Hispanic patients but would be unlikely to affect black patients. Despite these limitations, the results reported here may be the first to identify a racial/ethnic differential in pediatric ED disposition independent of physiology and medical factors. Additional studies are required to explore possible explanations for the relationship between ED physician decision-making and patient race/ethnicity.
|
ACKNOWLEDGMENTS |
|---|
This study was supported by Agency for Healthcare Quality and Research grant RO1 HS10238-02. The agency acted solely as the funding body for this investigator-initiated research and did not have input into the study design, conduct of the study, data analysis, or generation of the manuscript.
This article is dedicated to the memory of Dr Patel.
|
FOOTNOTES |
|---|
Accepted Nov 28, 2006.
Address correspondence to James M. Chamberlain, MD, Division of Emergency Medicine, Children's National Medical Center, 111 Michigan Ave, NW, Washington, DC 20010. E-mail: jchamber{at}cnmc.org
The authors have indicated they have no financial relationships relevant to this article to disclose.
Dr Joseph's current affiliation is Center for Health Services and Community Research, Washington, DC.
Drs Chamberlain and Patel had full access to all of the data in the study. Dr Chamberlain takes responsibility for the integrity of the data and the accuracy of the data analysis.
Deceased. 
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
1. Institute of Medicine. Crossing the Quality Chasm. Washington, DC: National Academy Press; 2001
2. Smedley BD, Stith AY, Nelson AR, eds. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, DC: National Academies Press; 2004
3. Ayanian JZ, Udvarhelyi IS, Gatsonis CA, Pashos CL, Epstein AM. Racial differences in the use of revascularization procedures after coronary angiography.
JAMA. 1993;269
:2642
–2646
4. Hannan EL, van Ryn M, Burke J, et al. Access to coronary artery bypass surgery by race/ethnicity and gender among patients who are appropriate for surgery. Med Care. 1999;37 :68 –77[CrossRef][Web of Science][Medline]
5. Ramsey DJ, Goff DC, Wear ML, Labarthe DR, Nichaman MZ. Sex and ethnic difference in use of myocardial revascularization procedures in Mexican Americans and non-Hispanic whites: the Corpus Christi Heart Project. J Clin Epidemiol. 1997;50 :603 –609[CrossRef][Web of Science][Medline]
6. Canto JG, Allison JJ, Kiefe CI, et al. Relation of race and sex to the use of reperfusion therapy in Medicare beneficiaries with acute myocardial infarction.
N Engl J Med. 2000;342
:1094
–1100
7. Collins SC, Hall A, Neuhaus C. US. Minority Health: A Chartbook. New York, NY: Commonwealth Fund; 1999
8. Schulman KA, Berlin JA, Harless W, Kerner JE, et al. The effect of race and sex on physicians' recommendations for cardiac catheterization.
N Engl J Med. 1999;340
:618
–626
9. Kasiske BL, London W, Ellison MD. Race and socioeconomic factors influencing early placement on the kidney transplant waiting list. J Am Soc Nephrol. 1998;9 :2142 –2147[Abstract]
10. Imperato PJ, Nenner RP, Will TO. Radical prostatectomy: lower rates among African American men. J Natl Med Assoc. 1996;88 :589 –594[Medline]
11. Bernabei R, Gambassi G, Lapane K, et al. Management of pain in elderly patients with cancer. SAGE Study Group. Systematic Assessment of Geriatric Drug Use via Epidemiology [published correction appears in JAMA. 1999;281:136].
JAMA. 1998;279
:1877
–1882
12. Moore RD, Stanton D, Gopalan R, Chaisson RE. Racial differences in the use of drug therapy for HIV disease in an urban community.
N Engl J Med. 1994;330
:763
–768
13. Shapiro MF, Morton SC, McCaffrey DF, et al. Variations in the care of HIV infected adults in the United States: results from the HIV Cost and Services Utilization Study.
JAMA. 1999;281
:2305
–2375
14. Bach PB, Cramer LD, Warren JL, Begg CB. Racial differences in the treatment of early-stage lung cancer.
N Engl J Med. 1999;341
:1198
–1205
15. Bennett CL, Horner RD, Weinstein RA, et al. Racial differences in care among hospitalized patients with Pneumocystis carinii pneumonia in Chicago, New York, Los Angeles, Miami, and Raleigh-Durham.
Arch Intern Med. 1995;155
:1586
–1592
16. Peterson ED, Shaw LK, DeLong ER, Pryor DB, Califf RM, Mark DB. Racial variation in the use of coronary-revascularization procedures: are the differences real? Do they matter?
N Engl J Med. 1997;336
:480
–486
17. Allison JJ, Kiefe CI, Centor RM, Box JB, Farmer RM. Racial differences in the medical treatment of elderly medicare patients with acute myocardial infarction. J Gen Intern Med. 1996;11 :736 –743[Web of Science][Medline]
18. Weitzman S, Cooper L, Chambless L, et al. Gender, racial, and geographic differences in the performance of cardiac diagnostic and therapeutic procedures for hospitalized acute myocardial infarction in four states. Am J Cardiol. 1997;79 :722 –726[CrossRef][Web of Science][Medline]
19. Todd KH, Deaton C, D'Adamo AP, Goe L. Ethnicity and analgesic practice. Ann Emerg Med. 2000;35 :11 –16[CrossRef][Web of Science][Medline]
20. Todd KH, Samaroo N, Hoffman JR. Ethnicity as a risk factor for inadequate emergency department analgesia.
JAMA. 1993;269
:1537
–1539
21. Guagliardo MF, Teach SJ, Huang ZJ, Chamberlain JM, Joseph JG. Racial and ethnic disparities in pediatric appendicitis rupture rate. Acad Emerg Med. 2003;10 :1218 –1227[CrossRef][Web of Science][Medline]
22. Chamberlain JM, Patel KM, Ruttiman UE, Pollack MM. Pediatric Risk of Admission (PRISA): a measure of severity of illness for assessing the risk of hospitalization from the emergency department. Ann Emerg Med. 1998;32 :161 –169[CrossRef][Web of Science][Medline]
23. Gravel J, Gouin S, Amre D, Bergeron S, Lacroix J. Evaluation of the pediatric risk of admission score in a pediatric emergency department. Ann Emerg Med. 2003;41 :630 –638[CrossRef][Web of Science][Medline]
24. Chamberlain JM, Patel KM, Pollack MM, et al. Recalibration of the Pediatric Risk of Admission (PRISA) score using a multi-institutional sample. Ann Emerg Med. 2004;43 :461 –468[CrossRef][Web of Science][Medline]
25. Chamberlain JM, Patel KM, Pollack MM. The Pediatric Risk of Hospital Admission (PRISA II) score: a second-generation severity-of-illness score for pediatric emergency patients.
Pediatrics. 2005;115
:388
–395
26. SAS Institute. The GENMOD Procedure, Generalized Estimating Equations. SAS/STAT Software: Changes and Enhancements Through Release 6.12. Cary, NC: SAS Institute; 1997:247–362
27. Kogan MD, Overpeck MD, Hoffman HJ, Casselbrant ML. Factors associated with tympanostomy tube insertion among preschool-aged children in the United States.
Am J Public Health. 2000;90
:245
–250
28. Kleinman LC, Kosecoff J, Dubois RW, Brook RH. The medical appropriateness of tympanostomy tubes proposed for children younger than 16 years in the United States.
JAMA. 1994;271
:1250
–1255
29. DeAngelis C, Joffe A, Wilson M, Willis E. Iatrogenic risks and financial costs of hospitalizing febrile infants.
Am J Dis Child. 1983;137
:1146
–1149
30. Steel K, Gertman PM, Crescenzi J, Anderson J. Iatrogenic illness on a general medical service at a university hospital. N Engl J Med. 1981;304 :638 –642[Abstract]
31. Krieger N, Chen JT, Waterman PD, Rehkopf DH, Subramanian SV. Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area-based socioeconomic measures—the Public Health Disparities Geocoding Project.
Am J Public Health. 2003;93
:1655
–1671
PEDIATRICS (ISSN 1098-4275). ©2007 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
