Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Pediatr. Author manuscript; available in PMC 2010 December 1.
Published in final edited form as:
PMCID: PMC2784130

Variation in inpatient therapy and diagnostic evaluation of children with Henoch Schönlein purpura

Pamela F. Weiss, MD, Andrew J. Klink, MPH, Kari Hexem, MPH, Jon M. Burnham, MD, MSCE, Mary B. Leonard, MD, MSCE, Ron Keren, MD, MPH, Russell Localio, PhD, and Chris Feudtner, MD, MPH, PhD



To describe variation regarding inpatient therapy and evaluation of children with Henoch Schönlein purpura (HSP) admitted to children’s hospitals across the United States.

Study design

We conducted a retrospective cohort study of children discharged with a diagnosis of HSP between 2000 and 2007 using inpatient administrative data from 36 children’s hospitals. We examined variation among hospitals in the use of medications, diagnostic tests, and intensive care services using multivariate mixed effects logistic regression models.


During the initial HSP hospitalization (N=1,988), corticosteroids were the most common medication (56% of cases), followed by opioids (36%), NSAIDs (35%), and anti-hypertensives (11%). After adjustment for patient characteristics, hospitals varied significantly in their use of corticosteroids, opioids, and NSAIDs; the use of diagnostic abdominal imaging, endoscopy, laboratory testing, and renal biopsy; and the utilization of intensive care services. By contrast, hospitals did not differ significantly regarding administration of anti-hypertensives or performance of skin biopsy.


The significant variation identified may contribute to varying HSP clinical outcomes between hospitals, warrants further investigation, and represents a potentially important opportunity to improve quality of care.

Keywords: opioids, corticosteroids, anti-hypertensives, non-steroidal anti-inflammatory drugs, adolescents, epidemiology

Henoch Schönlein purpura (HSP) is the most common pediatric vasculitis and accounts for half of all vasculitides in the United States. Using the European League Against Rheumatism (EULAR)/Pediatric Rheumatology European Society (PReS) criteria (not yet validated), HSP is diagnosed when palpable purpura is present plus one of the following: diffuse abdominal pain, any biopsy showing predominant IgA deposition, arthritis or arthralgias, and renal involvement (any hematuria or proteinuria).1 Up to 40% of children with HSP require hospitalization2 for management of acute disease manifestations which may include glomerulonephritis, hypertension, severe pain, gastrointestinal bleeding, or arthritis.

Prior epidemiologic studies about the natural course and treatment of HSP are limited by incomplete follow-up over variable time intervals, selection bias, and varying definitions of disease severity and complications. These limitations are evidenced by widely disparate estimates of the short- and long-term morbidities associated with HSP and varying conclusions regarding different therapeutic algorithms. Further, there are no established guidelines for the management of HSP from professional societies. Current inpatient therapy for HSP may include one or a combination of the following: opioids, corticosteroids, non-steroidal anti-inflammatory drugs (NSAIDs), and anti-hypertensive agents. The frequencies at which these medications are used and the variation in practice patterns across hospitals for inpatient care are unknown. Additionally, the frequency at which diagnostic tests such as imaging (abdominal imaging and endoscopy), laboratory studies (complete blood count, blood chemistry, and urinalysis), and renal biopsy are used during the evaluation of children with HSP and the variation in these practices among hospitals are unknown.

We therefore conducted a retrospective cohort study of pediatric patients admitted with acute HSP disease manifestations to 36 children’s hospitals throughout the United States to determine the current patterns of medication and evaluation for this disease, to assess the amount of variation across the hospitals overall, and to differentiate the variation that can be ascribed to differences among patients versus that due to other factors such as admitting hospital.


This study was approved and reviewed by The Children’s Hospital of Philadelphia Committee for the Protection of Human Subjects and institutional review board. This retrospective cohort study utilized the Pediatric Health Information System (PHIS) administrative database to assess between-hospital variation in inpatient therapy and evaluation of hospitalized children. The cohort consisted of patients with a discharge diagnosis International Classification of Diseases-9-Clinical Modification (ICD-9-CM) code of 287.0, indicating “Henoch Schönlein purpura.”

PHIS is an administrative database that contains comprehensive inpatient data from pediatric hospitals from all regions of the United States. Participating hospitals are affiliated with the Child Health Corporation of America (Shawnee Mission, KS), a business alliance of children’s hospitals. The data warehouse contains two types of data. Level 1 data contain encrypted patient identifiers, demographics, dates of admission and discharge, and ICD-9-CM primary diagnosis and procedure codes. Level 2 data contain detailed information about the patient encounter including all ICD-9-CM diagnosis and procedure codes and specific financial and utilization data, including pharmacy, supply, laboratory, imaging, and clinical services. Data are de-identified and subjected to rigorous reliability and validity checks prior to inclusion in the database. Data that do not meet an established error threshold are rejected and must be corrected before resubmission.3

Forty-one hospitals contributed to the PHIS database between January 1, 2000 and December 31, 2007. Three hospitals were excluded from this analysis because they did not provide any level 2 entry data during the study interval. Sixty-two hospital-years of data were excluded secondary to data quality issues. A total of 36 hospitals and 233 hospital-years of data remained for the analysis.

The source population for the study was children younger than 18 years of age with discharge dates between 1 January 2000 and 31 December 2007 (N=3,275,947). Inclusion criteria were an ICD-9-CM diagnosis indicating a diagnosis of “HSP” (code 287.0) and discharge from a hospital contributing level 2 data (N=3,522). To ensure that the cohort represented likely incident cases of HSP, subjects with an HSP admission in the six months prior to the study period were excluded. Subjects with a discharge diagnosis of another rheumatic condition such as Wegener’s granulomatosis, systemic lupus erythematosus, juvenile dermatomyositis, or polyarteritis nodosa, were excluded. Subjects with a missing admission diagnosis were excluded (N=286). If subjects had more than one admission during the study period, only the index admission was included. After inclusion and exclusion criteria were satisfied 2,407 hospitalization and 1,988 subjects remained for analysis. Age, sex, race, and Medicaid status were available on all subjects and included in the final analysis model. Race was coded as a dichotomous variable (Caucasian versus non-Caucasian).

The severity level (coded 1-4) is an indicator of the APR-DRG grouping. APR-DRG scores represent illness severity and risk of mortality for the entire hospitalization.4 They are calculated from computer algorithms based upon age, sex, diagnoses, procedures, and discharge status. Only one admission diagnosis is coded per hospitalization. All subjects in the cohort had a discharge diagnosis of HSP, but not all subjects were diagnosed with HSP at the time of admission. We report results only for subjects with complete data, including an admission diagnosis. Because HSP is known to have seasonal variation, discharge month (1-12) was included in the analysis model. Additionally, discharge year (2000-2007) was included in the final model.

Corticosteroid, opioid, NSAID, and anti-hypertensive medication exposure

Medication exposure was determined using pharmacy billing data (Appendix 1 for list of included generic drugs; available at and was defined as receipt of medication at any time during hospitalization.

Diagnostic testing

All subjects’ imaging procedures were determined using their ICD-9-CM procedure codes and billing data and were defined as receipt at any time during hospitalization. Imaging procedures included abdominal imaging (abdominal ultrasound, radiographs, computed tomography, magnetic resonance imaging, and fluoroscopy) and endoscopy. Laboratory testing (complete blood counts, blood chemistry, and urinalysis) was determined using PHIS-specific Clinical Transaction Classification™ (CTC) codes. Receipt of renal and skin biopsy was determined using ICD-9-CM procedure codes and billing data. All tests were defined as receipt at any time during hospitalization.

Data analysis

We used data on patient demographics, hospitalization severity, and diagnosis to assess how much of observed variation in utilization of medications could be attributed to differences in patient case-mix across hospitals and how much might be attributed to unmeasured variation. Patient demographic variables (age, sex, race, and Medicaid status) were examined using medians with ranges or percentages. We used two models to assess (1) the effect of standardization on patient-level factors on the hospital-level rates of therapy, and (2) the degree to which observed variation in use of therapy across hospitals varied beyond what would be expected after standardization. To standardize hospital-level rates of therapy, we used logistic regression to estimate population-averaged rates of therapy on the probability scale that would be expected (E) at each hospital based on their patients’ characteristics. We then compared these expected rates to those observed (O) for each hospital. To evaluate variation beyond expected, we used a mixed effects logistic regression model with prespecified individual level covariates (demographics (age, sex, race, Medicaid status), admission diagnosis of HSP, medication exposures (opioids, corticosteroids, NSAIDs, and anti-hypertensive drugs) and discharge month and year) as fixed effe Ects and admission hospital as a random effect (random intercept). The model thus takes the form logit(E(pij))=Xßij + bi, where ßij are the patient level covariates for the jth patient in the ith hospital, and bi represents the random effect for the ith hospital. The random effect represents the degree to which a hospital’s medication, diagnostic testing, and use of the intensive care unit departs from what would be expected on average for a hospital with the same set of patients. This method avoids issues of multiple comparisons, because each hospital’s random effect represents a weighted average of a hospital’s individual experience and that of the entire group of hospitals. Only hospitals with more than 25 HSP cases were included in the final analysis. All analyses were performed using the functions “xtlogit” and “xtmelogit” in Stata 10.1 (StataCorp, College Station, TX, USA).


During the eight-year study period there were 2,407 hospital admissions for HSP. The median age for the cohort was six years (range: 2 months-17 years) and 90% of children were between the ages of 2 and 15 years; 59% of subjects were male. The HSP admissions (83%; n=1,988) were for an initial episode and 17% (N=419) were for a readmission. The median length of stay was 3 days for the initial hospitalization (range: 2-94 days) and 4 days for readmissions (range: 2-66 days). The distributions of sex, age, and race, as shown in Table I are consistent with observations in previously published studies.2, 6-8 Forty-three percent of initial and 39% of readmissions had an admission diagnosis of HSP. The admission and various discharge diagnoses, in addition to HSP, assigned to subjects are listed in Table I.

Table 1
Subject demographics and HSP characteristics

Characteristics of initial hospitalizations and readmission were different to a statistically significant degree in regards to age, Medicaid status, length of stay, hospital severity score, admission and discharge diagnoses (Table I). Due to the complexity of understanding and studying readmissions, our subsequent analyses are restricted to only initial hospitalizations (N=1,988), a restriction that makes interpretation of the analyses more straightforward.

Hospital-level variation in medication use during initial hospitalization

Corticosteroids were administered in 56% of initial hospitalizations. Variation in corticosteroid use across hospitals was substantial across hospitals before and after adjustment for subject characteristics (p<0.001). Unadjusted corticosteroid use varied from 32% to 73% (Table II). Proportions adjusted for subject characteristics varied from 31% to 74% (Table II). After adjustment, nine of 35 hospitals (26%) had significantly different corticosteroid use from the average of all PHIS hospitals (Figure, A); three hospitals had significantly more corticosteroid use, and six hospitals had significantly less corticosteroid use. Intermediate-acting corticosteroids (i.e., methylprednisolone, prednisolone, and prednisone) accounted for 97% of corticosteroid use.

Variation in medication use among hospitals in unadjusted and adjusted models
Table 2
Treatment and evaluation of children with HSP during initial hospitalization

Opioid analgesics were administered during hospitalization in 36% of initial admissions. Unadjusted proportions of opioid use ranged widely, from 2% to 60%. Significant between-hospital variation remained after adjustment for subject characteristics (p<0.001) (Table II). After adjustment, six of 35 hospitals (17%) had significantly different opioid use from the average of all the PHIS hospitals (Figure, B); four hospitals had significantly more use and two hospitals had significantly less use. When the opioid category was subdivided into morphine sulfate, fentanyl, and all other opioids, significant variation remained among all categories (Table II).

NSAIDs were administered during hospitalization in 35% of initial admissions. Unadjusted proportions of NSAIDs use at each hospital ranged from 17% to 59%. Variation in NSAIDs use across hospitals remained significant after adjustment for subject characteristics (p<0.001). Proportions adjusted for subject characteristics ranged from 17% to 54% (Table II). After adjustment, six of 35 hospitals (17%) had significantly different NSAIDs use from the aver Dage of all PHIS-participating hospitals (Figure, C); three hospitals had significantly more and three hospitals had significantly less NSAIDs use than expected by chance.

Anti-hypertensive drugs were administered during hospitalization in 11% of initial admissions. Prior to adjustment for subject characteristics anti-hypertensive use ranged from 0% to 25%. Proportions adjusted for subject characteristics ranged from 0% to 25% (Table II). Unlike opioid, corticosteroid, and NSAIDs use, variation in anti-hypertensive use between hospitals was not significant after adjustment for subject characteristics.

Hospital-level variation in diagnostic tests during initial hospitalization

The unadjusted and adjusted means and distributions for each of the following diagnostic tests performed at each hospital during the initial admission are presented in Table II: abdominal imaging, endoscopy, complete blood count, blood chemistry, urinalysis, renal biopsy, and skin biopsy. After adjustment for subject characteristics, significant variation between hospitals was found in the use of imaging, endoscopy, complete blood count, blood chemistries (blood urea nitrogen and creatinine), and urinalysis, and renal biopsy. Use of skin biopsy did not vary significantly between hospitals after model adjustment.

Hospital variation in intensive care unit level of care

The unadjusted and adjusted mean and distribution for use of the intensive care unit level of care during the initial admission are presented in Table II. After adjustment significant variation between hospitals remained.

Sensitivity Analyses

Conceivably some of the between-hospital variation in medication use is driven by cases that are either more severe or with underlying renal abnormalities. The hospitalization severity level (coded 1-4) is an indicator of illness severity and risk of mortality for the entire hospitalization. When the hospitalization severity level was added to the analysis significant variation in the use of corticosteroids (range: 0.30, 0.75), opioids (range: 0.03, 0.55), and NSAIDs (range: 0.16, 0.52) remained (p-value for all three analyses <0.001).

Further, when a dichotomous variable indicating admission to the ICU at any point during hospitalization was added to the model (as a surrogate marker of severity), none of the adjusted rates of medication use were significantly altered. When the analysis was restricted to only those subjects with a discharge diagnosis indicating a renal or urinary abnormality (N=412) substantial variation remained in the use of corticosteroids (range: 0.43, 0.99), opioids (range: 0, 0.89), and NSAIDs (0, 0.62), albeit statistically insignificant for the use of corticosteroids and NSAIDs.

We performed additional regression analyses to further explore potential hospital-level factors that may have had an effect on the observed variations in medication use. When average annual production, as represented by the number of inpatients recorded for each hospital, was included in the regression model as a categorical variable substantial variation in the use of corticosteroids, opioids, and NSAIDs remained (p-value for all three analyses <0.001). Further, significant variation remained after a categorical variable for HSP case volume was added to the regression model (p-value for all three analyses <0.001).

Medical evaluation and treatment of HSP could conceivably differ based upon the presence of pediatric subspecialists, including nephrologists, rheumatologists, and dermatologists. When a dichotomous variable indicating the presence of each of these three subspecialties at the participating hospitals was added to the regression model, substantial variation in the use of corticosteroids, opioids, and NSAIDs remained (p-value for all three analyses <0.001).

Some of the variation in medication usage may be secondary to the range of unusual cases seen at the different hospitals. The clinical features of HSP are more atypical at the extremes of age with milder disease in children less than 2 years and more severe disease in adolescents and adults. Therefore, we restricted our analysis to include those children between the ages of 4 and 9 (25-75% for age of this cohort). After restriction, significant variation remained in the use of corticosteroids, opioids, and NSAIDs (p-value for all three analyses <0.001).

The variation in NSAID and opiate use might also be explained by whether subjects with an admission or discharge diagnosis codes indicating joint symptoms or abdominal pain. Variation in the use of opioids and NSAIDs among those with a diagnosis indicating joint or abdominal symptoms was nearly identical to those without joint or abdominal diagnoses.


In this large, multi-center cohort study of children hospitalized with new-onset HSP, corticosteroids were the most common medication administered during the initial hospitalization, followed by opioids, NSAIDs, and anti-hypertensive medications. Large variations were seen between hospitals. After controlling for patient-level factors and the random effects of each hospital, significant differences between hospitals remained.

With the exception of anti-hypertensives, substantial variation exists in the frequency of medication use that does not appear to be related to the individual characteristics of the affected children. Use of anti-hypertensives is likely predicated on standard definitions of hypertension and generally accepted thresholds of blood pressure beyond which therapy should be initiated. Conversely, the decision as to whether or not children receive corticosteroids, opioids, and NSAIDs might not be as straightforward. Use of these medications does not appear to be affected by the patient characteristics but varies significantly and substantially across admitting hospitals. Surprisingly, routine laboratory testing that should be a part of the initial diagnostic evaluation of all children with new-onset HSP such as a complete blood count, blood chemistry, and urinalysis, also differed significantly between hospitals. This is problematic and indicates the need to establish clinical guidelines. Even though it is debatable whether a complete blood count and chemistry need to be performed on every child with HSP, a urinalysis should most certainly be checked and followed. Renal involvement is not always detectable at initial presentation and is the major long-term feared complication of HSP.

Why does the routine care of children with HSP differ so markedly between hospitals? Our results demonstrate this variation does not reflect differences in the subjects admitted to the various hospitals. Although the PHIS hospitals differ in case volume, geographic location, and number of subspecialists, these attributes do not explain observed variation in our study. All the PHIS hospitals are tertiary care centers and teaching hospitals located in metropolitan areas. Further, our sensitivity analyses demonstrated that including the hospitals’ annual production, HSP case volume, and the presence of pediatric subspecialists in the regression models did not alter the significance of variation. These differences likely represent local and regional variation in the therapy and evaluation of a common childhood vasculitis for which formal management guidelines do not exist. This wide variability in care should raise concern over whether clinical outcomes of children treated at the different centers are similarly disparate and related to the medications and evaluations performed.

Limitations in the PHIS database raise cautions against over-interpretation of our results. First, PHIS does not provide access to outpatient records so we cannot be certain of any diagnostic evaluation or therapy initiated prior to hospitalization. Second, an administrative database might miscode patients who were diagnosed by the treating clinicians as having HSP; such administrative misclassification of diagnosis would likely be random and non-differential. Third, the PHIS data might not include all information on differences in patients’ indications for medication and evaluation across hospitals and does not include information about the threshold for admission across hospitals. Any patient-level factor or factors that might account for the remaining variation in proportions of medication and resource utilization after case-mix adjustment would have to Dbe obvious and substantial; the existence of such factors is possible but unlikely.

The significant between-hospital variations should rais e concern over the quality of care for children with HSP and highlight the importance of elucidating the impact of corticosteroids, pain control, anti-hypertensives, and resource utilization on clinical outcomes of these children. Certainly, this wide variation in care across hospitals argues for the need for a national quality improvement effort. The impact of treatment variations and implementation of standardized care guidelines has been studied in bronchiolitis,9 pediatric asthma,10 post-surgical care,11 and pediatric appendicitis;12 all of these studies conclude that evidence-based guidelines can beneficially influence utilization and clinical outcomes. Future studies should 1) address whether diversity of treatment and evaluation of children hospitalized with HSP is associated with disparate costs and clinical outcomes, and 2) evaluate the impact of standardized care guidelines for the management of HSP on the utilization and outcomes of care.


P.F.W. is supported by a National Institutes of Health clinical pharmacoepidemiology training grant.


Henoch Schönlein purpura
Pediatric Health Information System
International Classification of Diseases-9-Clinical Modification

APPENDIX 1: Generic drugs included in analyses


Methylprednisolone, Prednisolone, Prednisone, Adrenal combination corticosteroids, Dexamethasone, Triamcinolone

Opioid medications

Alfentanil HCl, Butorphanol tartrate, Codeine, Fentanyl, Hydromorphone HCl, Meperidine HCl, Methadone HCl, Morphine sulfate, Nalbuphine HCl, Narcotic analgesic combinations, Nonnarcotic analgesic and barbiturate combinations, Oxycodone HCl, Remifentanil HCl, Tramadol HCl

Non-steroidal anti-inflammatory medications

Aspirin, Aspirin and other salicylate combinations, Celecoxib, Ibuprofen, Indomethacin, Ketorolac tromethamine, Nabumetone, Naproxen (acid) (sodium), Rofecoxib

Anti-hypertensive medications

Amlodipine, Atenolol, Captopril, Carvedilol, Clonidine HCl, Diazoxide, Diltiazem HCl, Doxazosin mesylate, Enalapril maleate, Esmolol HCl, Felodipine, Guanfacine HCl, Hydralazine HCl, Isradipine, Labetalol HCl, Lisinopril, Losartan potassium, Metoprolol (succinate) (tartrate), Minoxidil, Nesiritide, Nicardipine HCl, Nifedipine, Nitroglycerin, Nitroprusside sodium, Papaverine HCl, Propranolol HCl, Quinapril HCl, Tolazoline HCl, Valsarta, Verapamil HCl

APPENDIX 2. CHCA-participating hospitals included in analyses

Mean annual
no. inpatients
No. index
HSP cases
Hospital name, city, and state
13,02444Children’s Hospital and Health Center, San Diego, CA
9,39825The Children’s Hospital, Denver, CO
11,84828Children’s Mercy Hospital, Kansas City, MO
9,44869Le Bonheur Children’s Medical Center, Memphis, TN
11,14545Children’s National Medical Center, Washington, DC
12,51890Children’s Hospital of Pittsburgh, Pittsburgh, PA
9,16563Children’s Memorial Hospital, Chicago, IL
8,14325Children’s Hospital Medical Center of Akron, Akron, OH
11,49274Children’s Hospital Central California, Madera, CA
6,10433Children’s Hospital of the King’s Daughters, Norfolk, VA
8,82065All Children’s Hospital, St. Petersburg, FL
8,33852Arkansas Children’s Hospital, Little Rock, AR
9,514123Children’s Hospital of Orange County, Orange, CA
6,41551Driscoll Children’s Hospital, Corpus Christi, TX
12,49153Miami Children’s Hospital, Miami, FL
14,01899Children’s Hospital, Columbus, OH
9,56969Cook Children’s Medical Center, Ft. Worth, TX
7,33951Children’s Hospital, New Orleans, LA
13,08822The Children’s Hospital of Alabama, Birmingham, AL
4,36831Children’s Healthcare Services, Omaha, NE
12,61085Children’s Hospital of Wisconsin, Milwaukee, WI
20,865133Texas Children’s Hospital, Houston, TX
15,60093Children’s Hospital Medical Center, Cincinnati, OH
10,34060St. Louis Children’s Hospital, St. Louis, MO
11,23566Children’s Hospital Los Angeles, Los Angeles, CA
20,201108Children’s Healthcare of Atlanta, Atlanta, GA
10,70341Children’s Hospital and Medical Center, Seattle, WA
19,37942The Children’s Hospital of Philadelphia, Philadelphia, PA
12,92769Children’s Hospital of Michigan, Detroit, MI
10,61025Vanderbilt Children’s Hospital, Nashville, TN
11,42730Children’s Hospital of New York-Presbyterian, New York, NY
15,36337Children’s Medical Center of Dallas, Dallas, TX
11,71935Phoenix Children’s Hospital, Phoenix, AZ
17,03424Lucile Packard Children’s Hospital at Stanford, Palo Alto, CA
9,77528Riley Hospital for Children, Indianapolis, IN


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The authors declare no conflicts of interest.


1. Ozen S, Ruperto N, Dillon M, Bagga A, Barron K, Davin JC, et al. EULAR/PReS endorsed consensus criteria* for the classification of childhood vasculitides. Ann Rheum Dis. 2006;65(7):936–41. [PMC free article] [PubMed]
2. Yang YH, Hung CF, Hsu CR, Wang LC, Chuang YH, Lin YT, et al. A nationwide survey on epidemiological characteristics of childhood Henoch-Schonlein purpura in Taiwan. Rheumatology (Oxford) 2005;44(5):618–22. [PubMed]
3. Fletcher DM. Achieving data quality. How data from a pediatric health information system earns the trust of its users. J Ahima. 2004;75(10):22–6. [PubMed]
4. Feudtner C, Levin JE, Srivastava R, Goodman DM, Slonim AD, Sharma V, et al. How well can hospital readmission be predicted in a cohort of hospitalized children? A retrospective, multicenter study. Pediatrics. 2009;123(1):286–93. [PMC free article] [PubMed]
5. Spiegelhalter DJ, Abrams KR, Myles JP. Bayesian approaches to clinical trials and health care evaluation. John Wiley & Sons; Chichester, UK: 2004.
6. Trapani S, Micheli A, Grisolia F, Resti M, Chiappini E, Falcini F, et al. Henoch Schonlein purpura in childhood: epidemiological and clinical analysis of 150 cases over a 5-year period and review of literature. Semin Arthritis Rheum. 2005;35(3):143–53. [PubMed]
7. Calvino MC, Llorca J, Garcia-Porrua C, Fernandez-Iglesias JL, Rodriguez-Ledo P, Gonzalez-Gay MA. Henoch-Schonlein purpura in children from northwestern Spain: a 20-year epidemiologic and clinical study. Medicine (Baltimore) 2001;80(5):279–90. [PubMed]
8. Dolezalova P, Telekesova P, Nemcova D, Hoza J. Incidence of vasculitis in children in the Czech Republic: 2-year prospective epidemio logy survey. J Rheumatol. 2004;31(11):2295–9. [PubMed]
9. Todd J, Bertoch D, Dolan S. Use of a large national database for comparative evaluation of the effect of a bronchiolitis/viral pneumonia clinical care guideline on patient outcome and resource utilization. Arch Pediatr Adolesc Med. 2002;156(11):1086–90. [PubMed]
10. Norton SP, Pusic MV, Taha F, Heathcote S, Carleton BC. Effect of a clinical pathway on the hospitalisation rates of children with asthma: a prospective study. Arch Dis Child. 2007;92(1):60–6. [PMC free article] [PubMed]
11. Griffin FA. Reducing surgical complications. Jt Comm J Qual Patient Saf. 2007;33(11):660–5. [PubMed]
12. Newman K, Ponsky T, Kittle K, Dyk L, Throop C, Gieseker K, et al. Appendicitis 2000: variability in practice, outcomes, and resource utilization at thirty pediatric hospitals. J Pediatr Surg. 2003;38(3):372–9. discussion -9. [PubMed]