Factors Associated With Increased Resource Utilization for Congenital Heart Disease
http://www.100md.com
《小儿科》
the Department of Cardiology, Children’s Hospital, Boston, Massachusetts
ABSTRACT
Objective. To identify patient, institutional, and regional factors that are associated with high resource utilization for congenital heart surgery.
Methods. We used hospital discharge data from the Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID) year 2000 (data from 27 states). Patients who had congenital heart surgery and were younger than 18 years were identified using International Classification of Diseases, Ninth Revision, Clinical Modification codes. High resource utilization admissions were defined as those in the highest decile for total hospital charges. Univariate and multivariate analyses with and without deaths were used to determine demographic and hospital predictors for cases of high resource use. Case-mix severity was approximated using Risk Adjustment for Congenital Heart Surgery risk groups. Regional and state differences were also examined.
Results. Among 10569 cases of congenital heart surgery identified, median total hospital charges were $53828. Statewide differences in the number of high resource use admissions were present; California, Colorado, Florida, Hawaii, Pennsylvania, and Texas were more likely to have high resource use cases, and Maine and South Carolina were less likely. Subsequent analyses were performed adjusting for baseline state effects. Multivariate analyses using generalized estimating equations models revealed Risk Adjustment for Congenital Heart Surgery risk category (odds ratio [OR]: 1.66–14.1), age (OR: 3.81), prematurity (OR: 4.85), the presence of other major noncardiac structural anomalies (OR: 2.53), Medicaid insurance (OR: 1.48), and admission during a weekend (OR: 1.62) to be independent predictors of a higher odds of high cost cases. Although some institutional differences were noted in univariate analyses, gender, race, bed size, teaching and children’s hospital status, hospital ownership, and hospital volume of cardiac cases were not independently associated with greater odds of high resource utilization.
Conclusions. States varied in the frequency of high resource utilization for congenital heart surgery. Patients who had greater disease complexity, younger age, prematurity, other anomalies, and Medicaid and were admitted during a weekend were more likely to result in high resource utilization. Institutions of various types did not differ in high cost admissions, regardless of children’s hospital or teaching status.
Key Words: congenital heart disease Kids’ Inpatient Database resource utilization hospital charges risk adjustment
Abbreviations: HCUP, Healthcare Cost and Utilization Project KID, Kids’ Inpatient Database SID, State Inpatient Database ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification RACHS-1, Risk Adjustment for Congenital Heart Surgery ROC, receiver operating characteristic OR, odds ratio CI, confidence interval
Resource utilization for children with complex disease is largely undescribed or quantified yet is important to every stakeholder involved in the delivery of health care. Although only 1% of children have a complex disease, these children consume a disproportionately large share of available resources. Children who are born with congenital heart disease are a population associated with numerous hospitalizations, services of a multidisciplinary team of specialists, and use of both advanced and innovative technology.
A description of current hospital resource utilization for patients with congenital heart disease and a better understanding of determinants of increased resource utilization during surgical admission would assist stakeholders who are engaged in current pediatric health care delivery in shaping future health care reform. The purpose of this study was to examine current hospital resource use from a national perspective and to identify independent predictors of increased resource use for patients with congenital heart disease during surgical admissions.
METHODS
Data Source
Data were obtained from the Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID) 2000. The KID consists of a stratified random sample of 2516833 discharges from 2784 institutions in 27 states (Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Iowa, Kansas, Kentucky, Maine, Maryland, Massachusetts, Missouri, North Carolina, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Texas, Utah, Virginia, Washington, Wisconsin, and West Virginia). The KID 2000 used the American Hospital Association’s definition of hospitals to identify all nonfederal, short-term, general, and other specialty hospitals. Pediatric hospitals, academic medical centers, and specialty hospitals were included. The database does not include all admissions from participating institutions but instead includes a 10% sample of uncomplicated in-hospital births from these institutions and an 80% sample of other pediatric discharges (age <21 years). For obtaining information that is nationally representative, the sample is weighted to represent the population of pediatric discharges from all community, nonrehabilitation hospitals in the United States that were open for any part of calendar year 2000. To protect confidentiality, the KID does not contain specific patient or hospital identifiers.
Case Selection
Cases <18 years of age were identified as all hospital discharges with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes indicating surgical repair of a congenital heart defect.1 To focus on structural repairs only, patients with codes for cardiac transplantation were eliminated, as were premature infants and newborns who were 30 days of age and underwent ligation of patent ductus arteriosus only.
Definition of Resource Use
Total hospital charges accrued during hospitalization, available for most patients in the KID 2000, were used as a surrogate for resource utilization. The distribution of total hospital charges for congenital heart surgery cases was examined. An arbitrary cut point was made at the 90th percentile; cases >90th percentile were labeled as especially high resource utilization.
Statistical Methods
Variables that were examined as potential predictors of high resource use included demographic information (age, gender, race, insurance status, and day of admission), clinical indicators (prematurity and the presence of other, noncardiac structural anomalies), and hospital characteristics (hospital bed size, location, teaching status, children’s hospital status, and volume of cardiac cases performed by institution). Geographic variables including region (Northeast, South, Midwest, and West) and state (27 states) were also examined.
Case-mix severity was approximated using the Risk Adjustment for Congenital Heart Surgery (RACHS-1) risk categories.2,3 The RACHS-1 method was originally developed to adjust for differences in case mix when comparing in-hospital mortality among groups of patients. For applying this method, cases are assigned to 1 of the 6 predefined risk categories on the presence or absence of specific diagnosis and procedure ICD-9 codes, in which risk category 1 has the lowest risk for death and risk category 6 has the highest (Appendix 1). Cases with combinations of cardiac surgical procedures (eg, coarctation of the aorta and ventricular septal defect closure) are placed in the risk category corresponding to the single highest risk procedure.
Demographic, clinical, hospital, and geographic factors associated with high resource utilization were identified in univariate analyses using generalized estimating equations models to adjust for intra-institutional correlation among cases. In initial analyses, significant baseline differences in the proportion of cases with high resource use were identified for 8 states. Additional models were constructed to include indicator variables representing these states. The area under the receiver operating characteristic (ROC) curve was calculated for each model to quantify how well each factor was able to discriminate between cases that were and were not high resource users. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for each factor. After baseline state effects were adjusted for, the single factor that provided the most additional predictive information was retained in the model; the remaining factors then were reevaluated. This process was repeated until none of the remaining factors contributed significantly to the prediction of high resource use. Analyses were repeated excluding in-hospital deaths. All data were analyzed using SAS 8.2 (SAS Institute, Cary, NC) statistical software.
RESULTS
Summary of Cases of Congenital Heart Surgery
Of 2 516 833 pediatric discharges, 12717 cases with codes indicating congenital heart surgery were identified. Of these, 11381 cases met inclusion criteria (ie, age 18 years, cardiac transplants, and premature infants and newborns who were 30 days of age and had patent ductus arteriosus ligation as their only cardiac procedure were eliminated); 10602 (from 198 institutions) had total hospital charges reported. The number of cases by state ranged from 7 to 2394 (Fig 1). The median total hospital charge for this group was $51125, with a minimum of $195 and a maximum of $1000000. The 90th percentile cutpoint defining the especially high resource users was $192272 (Fig 2). Total hospital charges varied among the 27 states; California ($74294) had the highest median total hospital charges, and Maryland had the lowest ($19722; Fig 1). Descriptive information and differences in regional, demographic, clinical, and institutional variables associated with a higher proportion of cases with high resource use are shown in Table 1.
Multivariate Predictors of High Resource Use
Effects of State
Using New York State as a baseline, California, Colorado, Florida, Hawaii, Pennsylvania, and Texas had a significantly higher and Maine and South Carolina a significantly lower likelihood of cases with especially high resource use. Indicator variables representing each of these 8 states versus all others combined were included in additional models (Table 2).
Patient Demographic and Clinical Characteristics
Adjusting for baseline state differences in multivariate models, cases that were younger than 1 year had a higher odds of being in the high resource group as patients who were reported as premature or having other, noncardiac structural anomalies. Cases that had higher RACHS-1 risk categories or were unable to be categorized were more likely to use increased resources. Cases that were admitted during a weekend were more likely to be in the high resource group as were cases that were reported to have Medicaid or "other" insurance (Table 2).
After state and these additional patient characteristics were adjusted for, race, surgery at a children’s hospital, hospital cardiac volume, and geographic location were not associated with higher resource use. Findings were similar when in-hospital deaths were excluded.
Model Performance
The final model achieved an area under the ROC curve of 0.837 in discriminating cases of high versus non–high resource use.
DISCUSSION
The results of our population-based analysis of children who underwent surgical repair of congenital heart disease using the KID 2000 data set highlights significant geographic variation in resource utilization. Our results also reveal especially high resource utilization to be found within a subset of children who underwent congenital heart repair rather than the population as a whole.
Past research in this field has been limited to the investigation of the direct costs of a single complex cardiac defect and its surgical management, practice pattern, and cost variations among 8 centers that care for children and adults with congenital heart disease and cost of intensive care and post–intensive care stay for critically ill children.4–7 In addition, 1 previous study described a single institutional experience of cost of surgical care and predictors for higher cost in congenital heart repair.8 The researchers reported age of child; complexity of defect; presence of other, noncardiac anomalies or syndromes; and length of stay to be associated with increased financial risk.
Actual cost measures are very difficult to determine and were not available in our data source. Because our goal was to investigate especially high resource utilization, total hospital charges were used as a surrogate. The assumption that cases with especially high total hospital charges also had especially high resource utilization and therefore an overall higher cost seems reasonable, because most health care resources result in a higher charge to payers. Use of this reasonable surrogate allowed us to examine resource use for a rare, complex, pediatric condition from a national perspective. However, because charging structures are complex and do not accurately reflect costs, we chose to structure analyses to identify cases of especially high resource use only, ie, those in the upper 10th percentile for charges; we believed that attempting to identify predictors of actual charges used as a continuous variable was beyond the validity of our surrogate. It is interesting that our "cutpoint" demarcating high resource use was at $192272, suggesting that we have identified cases that are especially expensive. In fact, our high resource group accounted for >40% of the total charges accrued for all of these admissions.
Identification and profiling of increased resource utilization during surgical admission for children with congenital heart disease adds to the knowledge of a population that is associated with repeated hospitalizations, use of advanced technology, and use of innovative medical therapies. Our findings suggest that there is a subset of children who are especially high users of hospital resources and that state influences are also present.
Eight of the 27 states were noted to be significantly predictive of resource use. Cases from California, Colorado, Florida, Hawaii, Pennsylvania, and Texas had higher odds of being high resource utilization cases, and cases from South Carolina and Maine had significantly lower odds of being high resource utilization cases. Increased or decreased cost of living in certain states, such as Maine or California, may also affect cost of health care delivery, leading to geographic variation in total hospital charges. However, it is not certain at this time why geographic differences of high magnitude exist; this will be an area of our future research.
Our results suggest RACHS-1 risk category and younger age to be highly predictive of increased resource utilization. We also found clinical descriptors such as prematurity and presence of other, noncardiac structural anomalies to be highly predictive of resource utilization. Although there is a paucity of information with which to compare our findings, work by Ungelieder et al8 focusing on identification of patients associated with high-cost surgical admissions at a single institution produced similar findings. Our findings would further support their conclusions that there are clinical predictors of financial risk that may facilitate implementation of risk adjustments for payers and for strategic resource allocation within institutions.
Although the timing of hospital admission was noted to be a significant independent predictor of high resource utilization, we do not have a conclusive explanation for this. However, patients who were admitted on a weekend may possibly be cases that were not diagnosed prenatally, or required emergent surgical intervention, or were stabilized at time of admission and waited for surgery to be performed during a weekday.
Patients who were reported to have Medicaid or other insurance rather than private insurance were found to be independently predictive of high resource utilization. However, this variable added the least to the discrimination between our resource groups in the final model. The structure of the KID did not specify timing of recorded status, ie, admission versus discharge; therefore, it cannot be concluded whether these cases had Medicaid as a primary insurer on the basis of socioeconomic status or became Medicaid cases as a result of accruing significant hospital charges.
Demone et al9 previously reported Medicaid as an independent predictor of pediatric cardiac surgical outcomes. Their study examining the effect of insurance type on mortality for congenital heart surgery revealed that not only do Medicaid patients have a higher risk for death as compared with commercial or managed care patients, but also differences were present within institutions identified as low, average, and high mortality, suggesting that the adverse effect of Medicaid may be attributable to differential referral among patients who are treated at similar institutions. These findings when combined with our findings in resource utilization suggest that differences in care may exist and should be an area of continued study.
Although some institutional differences were noted in univariate analyses, bed size, teaching and children’s hospital status, and hospital volume of cardiac cases were not independently associated with greater numbers of cases of high resource use. Our findings are not consistent with what has been previously reported on pediatric conditions in teaching facilities10 and may be attributable to the limited number of children’s hospital and nonteaching facilities that perform congenital heart repair available in the data set. As is the current trend, most pediatric cardiac surgical programs exist in teaching facilities with a children’s unit, and it would be difficult to find an adequate number of nonteaching facilities for examination.
Limitations
Although the KID 2000 is considered to provide a representative sample of pediatric hospitalizations, discharge information originates from only 27 states. The KID data set did not contain unique patient identifiers or record linkage numbers, thereby making it impossible to identify discharges as individual patients. Missing data and coding errors are universal limitations in using large administrative data sets. For example, the data element "race" was missing for 19% of our overall sample. We therefore could not conclude with certainty that race was not a factor in high resource utilization. At the onset of our study, a limitation that was foreseen was the lack of detailed clinical information, which is another limitation of all studies that use administrative databases in outcomes research.11 In this population, birth weight and organ function have been cited as predictors of initial clinical outcomes as might also reflect resource utilization.12
Although our developed model of factors associated with increased resource utilization achieved a final area under the ROC curve of 0.837, it has not been validated in a second data set. Despite these limitations, our analyses included a large number of cases of complex congenital heart surgery from all regions of the country and yielded a number of important conclusions.
CONCLUSION
States varied in the frequency of cases of especially high resource use for congenital heart surgery, with some having a higher- and some having a lower-than-expected number. Patients who had greater disease complexity, younger age, prematurity, other anomalies, and Medicaid and were admitted during a weekend were more likely to result in especially high resource use. Institutions of various types did not differ in frequency of especially high resource use, regardless of children’s hospital or teaching status. Although demographic characteristics and disease complexity are fixed and offer no easy solutions for health care reform, this examination has identified a high magnitude of statewide variation and may reflect opportunities for more efficient care to be instituted.
APPENDIX 1: INDIVIDUAL PROCEDURES BY RISK CATEGORY
Risk Category 1
Atrial septal defect surgery (including atrial septal defect secundum, sinus venosus atrial septal defect, and patent foramen ovale closure)
Aortopexy
Patent ductus arteriosus surgery at age >30 days
Coarctation repair at age >30 days
Partially anomalous pulmonary venous connection surgery
Risk Category 2
Aortic valvotomy or valvuloplasty at age >30 days
Subaortic stenosis resection
Pulmonary valvotomy or valvuloplasty
Pulmonary valve replacement
Right ventricular infundibulectomy
Pulmonary outflow tract augmentation
Repair of coronary artery fistula
Atrial septal defect and ventricular septal defect repair
Atrial septal defect primum repair
Ventricular septal defect repair
Ventricular septal defect closure and pulmonary valvotomy or infundibular resection
Ventricular septal defect closure and pulmonary artery band removal
Repair of unspecified septal defect
Total repair of tetralogy of Fallot
Repair of total anomalous pulmonary veins at age >30 days
Glenn shunt
Vascular ring surgery
Repair of aortopulmonary window
Coarctation repair at age 30 days
Repair of pulmonary artery stenosis
Transection of pulmonary artery
Common atrium closure
Left ventricular to right atrial shunt repair
Risk Category 3
Aortic valve replacement
Ross procedure
Left ventricular outflow tract patch
Ventriculomyotomy
Aortoplasty
Mitral valvotomy or valvuloplasty
Mitral valve replacement
Valvectomy of tricuspid valve
Tricuspid valvotomy or valvuloplasty
Tricuspid valve replacement
Tricuspid valve repositioning for Ebstein anomaly at age >30 days
Repair of anomalous coronary artery without intrapulmonary tunnel
Repair of anomalous coronary artery with intrapulmonary tunnel (Takeuchi)
Closure of semilunar valve, aortic or pulmonary
Right ventricular to pulmonary artery conduit
Left ventricular to pulmonary artery conduit
Repair of double outlet right ventricle with or without repair of right ventricular obstruction
Fontan procedure
Repair of transitional or complete atrioventricular canal with or without valve replacement
Pulmonary artery band
Repair of tetralogy of Fallot with pulmonary atresia
Repair of cor triatriatum
Systemic to pulmonary artery shunt
Atrial switch operation
Arterial switch operation
Reimplantation of anomalous pulmonary artery
Annuloplasty
Repair of coarctation and ventricular septal defect closure
Excision of intracardiac tumor
Risk Category 4
Aortic valvotomy or valvuloplasty at age 30 days
Konno procedure
Repair of complex anomaly (single ventricle) by ventricular septal defect enlargement
Repair of total anomalous pulmonary veins at age 30 days
Atrial septectomy
Repair of transposition, ventricular septal defect, and subpulmonary stenosis (Rastelli)
Atrial switch operation with ventricular septal defect closure
Atrial switch operation with repair of subpulmonary stenosis
Arterial switch operation with pulmonary artery band removal
Arterial switch operation with ventricular septal defect closure
Arterial switch operation with repair of subpulmonary stenosis
Repair of truncus arteriosus
Repair of hypoplastic or interrupted arch without ventricular septal defect closure
Repair of hypoplastic or interrupted aortic arch with ventricular septal defect closure
Transverse arch graft
Unifocalization for tetralogy of Fallot and pulmonary atresia
Double switch
Risk Category 5
Tricuspid valve repositioning for neonatal Ebstein anomaly at age 30 days
Repair of truncus arteriosus and interrupted arch
Risk Category 6
Stage 1 repair of hypoplastic left heart syndrome (Norwood operation)
Stage 1 repair of nonhypoplastic left heart syndrome conditions
Damus-Kaye-Stansel procedure
FOOTNOTES
Accepted Dec 29, 2004.
No conflict of interest declared.
REFERENCES
Jenkins KJ, Gauvreau K. Center-specific differences in mortality: preliminary analyses using the Risk Adjustment in Congenital Heart Surgery (RACHS-1) method. J Thorac Cardiovasc Surg. 2002;124 :97 –104
Jenkins KJ, Gauvreau K, Newburger JW, Kyn LB, Iezzoni LI, Mayer JE. Validation of relative value scale for congenital heart operations. Ann Thorac Surg. 1998;66 :860 –869
Jenkins KJ, Gauvreau K, Newburger JW, Spray TL, Moller JH, Iezzoni LI. Consensus-based method for risk adjustment for surgery for congenital heart disease. J Thorac Cardiovasc Surg. 2002;123 :110 –118
Williams DL, Gelijns AC, Moskowitz AJ, et al. Hypoplastic left heart syndrome: valuing the survival. J Thorac Cardiovasc Surg. 2000;119 :720 –731
Kanter RK. Post-intensive care unit pediatric hospital stay and estimated costs. Crit Care Med. 2000;28 :220 –223
Chalom R, Raphaely RC, Costarino AT Jr. Hospital costs of pediatric intensive care. Crit Care Med. 1999;27 :2079 –2085
Garson A Jr, Allen HD, Gersony WM, et al. The cost of congenital heart disease in children and adults. A model for multicenter assessment of price and practice variation. Arch Pediatr Adolesc Med. 1994;148 :1039 –1045
Ungerleider RM, Bengur AR, Kessenich AL, et al. Risk factors for higher cost in congenital heart operations. Ann Thorac Surg. 1997;64 :44 –48; discussion 49
DeMone JA, Gonzalez PC, Gauvreau K, Piercey GE, Jenkins KJ. Risk of death for Medicaid recipients undergoing congenital heart surgery. Pediatr Cardiol. 2003;24 :97 –102
Srivastava R, Homer CJ. Length of stay for common pediatric conditions: teaching versus nonteaching hospitals. Pediatrics. 2003;112 :278 –281
Iezzoni LI. Using administrative data to study persons with disabilities. Milbank Q. 2002;80 :347 –379
Iannettoni MD, Bove EL, Mosca RS, et al. Improving results with first-stage palliation for hypoplastic left heart syndrome. J Thorac Cardiovasc Surg. 1994;107 :934 –940(Jean Anne Connor, DNSc, R)
ABSTRACT
Objective. To identify patient, institutional, and regional factors that are associated with high resource utilization for congenital heart surgery.
Methods. We used hospital discharge data from the Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID) year 2000 (data from 27 states). Patients who had congenital heart surgery and were younger than 18 years were identified using International Classification of Diseases, Ninth Revision, Clinical Modification codes. High resource utilization admissions were defined as those in the highest decile for total hospital charges. Univariate and multivariate analyses with and without deaths were used to determine demographic and hospital predictors for cases of high resource use. Case-mix severity was approximated using Risk Adjustment for Congenital Heart Surgery risk groups. Regional and state differences were also examined.
Results. Among 10569 cases of congenital heart surgery identified, median total hospital charges were $53828. Statewide differences in the number of high resource use admissions were present; California, Colorado, Florida, Hawaii, Pennsylvania, and Texas were more likely to have high resource use cases, and Maine and South Carolina were less likely. Subsequent analyses were performed adjusting for baseline state effects. Multivariate analyses using generalized estimating equations models revealed Risk Adjustment for Congenital Heart Surgery risk category (odds ratio [OR]: 1.66–14.1), age (OR: 3.81), prematurity (OR: 4.85), the presence of other major noncardiac structural anomalies (OR: 2.53), Medicaid insurance (OR: 1.48), and admission during a weekend (OR: 1.62) to be independent predictors of a higher odds of high cost cases. Although some institutional differences were noted in univariate analyses, gender, race, bed size, teaching and children’s hospital status, hospital ownership, and hospital volume of cardiac cases were not independently associated with greater odds of high resource utilization.
Conclusions. States varied in the frequency of high resource utilization for congenital heart surgery. Patients who had greater disease complexity, younger age, prematurity, other anomalies, and Medicaid and were admitted during a weekend were more likely to result in high resource utilization. Institutions of various types did not differ in high cost admissions, regardless of children’s hospital or teaching status.
Key Words: congenital heart disease Kids’ Inpatient Database resource utilization hospital charges risk adjustment
Abbreviations: HCUP, Healthcare Cost and Utilization Project KID, Kids’ Inpatient Database SID, State Inpatient Database ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification RACHS-1, Risk Adjustment for Congenital Heart Surgery ROC, receiver operating characteristic OR, odds ratio CI, confidence interval
Resource utilization for children with complex disease is largely undescribed or quantified yet is important to every stakeholder involved in the delivery of health care. Although only 1% of children have a complex disease, these children consume a disproportionately large share of available resources. Children who are born with congenital heart disease are a population associated with numerous hospitalizations, services of a multidisciplinary team of specialists, and use of both advanced and innovative technology.
A description of current hospital resource utilization for patients with congenital heart disease and a better understanding of determinants of increased resource utilization during surgical admission would assist stakeholders who are engaged in current pediatric health care delivery in shaping future health care reform. The purpose of this study was to examine current hospital resource use from a national perspective and to identify independent predictors of increased resource use for patients with congenital heart disease during surgical admissions.
METHODS
Data Source
Data were obtained from the Healthcare Cost and Utilization Project (HCUP) Kids’ Inpatient Database (KID) 2000. The KID consists of a stratified random sample of 2516833 discharges from 2784 institutions in 27 states (Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Iowa, Kansas, Kentucky, Maine, Maryland, Massachusetts, Missouri, North Carolina, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Texas, Utah, Virginia, Washington, Wisconsin, and West Virginia). The KID 2000 used the American Hospital Association’s definition of hospitals to identify all nonfederal, short-term, general, and other specialty hospitals. Pediatric hospitals, academic medical centers, and specialty hospitals were included. The database does not include all admissions from participating institutions but instead includes a 10% sample of uncomplicated in-hospital births from these institutions and an 80% sample of other pediatric discharges (age <21 years). For obtaining information that is nationally representative, the sample is weighted to represent the population of pediatric discharges from all community, nonrehabilitation hospitals in the United States that were open for any part of calendar year 2000. To protect confidentiality, the KID does not contain specific patient or hospital identifiers.
Case Selection
Cases <18 years of age were identified as all hospital discharges with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes indicating surgical repair of a congenital heart defect.1 To focus on structural repairs only, patients with codes for cardiac transplantation were eliminated, as were premature infants and newborns who were 30 days of age and underwent ligation of patent ductus arteriosus only.
Definition of Resource Use
Total hospital charges accrued during hospitalization, available for most patients in the KID 2000, were used as a surrogate for resource utilization. The distribution of total hospital charges for congenital heart surgery cases was examined. An arbitrary cut point was made at the 90th percentile; cases >90th percentile were labeled as especially high resource utilization.
Statistical Methods
Variables that were examined as potential predictors of high resource use included demographic information (age, gender, race, insurance status, and day of admission), clinical indicators (prematurity and the presence of other, noncardiac structural anomalies), and hospital characteristics (hospital bed size, location, teaching status, children’s hospital status, and volume of cardiac cases performed by institution). Geographic variables including region (Northeast, South, Midwest, and West) and state (27 states) were also examined.
Case-mix severity was approximated using the Risk Adjustment for Congenital Heart Surgery (RACHS-1) risk categories.2,3 The RACHS-1 method was originally developed to adjust for differences in case mix when comparing in-hospital mortality among groups of patients. For applying this method, cases are assigned to 1 of the 6 predefined risk categories on the presence or absence of specific diagnosis and procedure ICD-9 codes, in which risk category 1 has the lowest risk for death and risk category 6 has the highest (Appendix 1). Cases with combinations of cardiac surgical procedures (eg, coarctation of the aorta and ventricular septal defect closure) are placed in the risk category corresponding to the single highest risk procedure.
Demographic, clinical, hospital, and geographic factors associated with high resource utilization were identified in univariate analyses using generalized estimating equations models to adjust for intra-institutional correlation among cases. In initial analyses, significant baseline differences in the proportion of cases with high resource use were identified for 8 states. Additional models were constructed to include indicator variables representing these states. The area under the receiver operating characteristic (ROC) curve was calculated for each model to quantify how well each factor was able to discriminate between cases that were and were not high resource users. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for each factor. After baseline state effects were adjusted for, the single factor that provided the most additional predictive information was retained in the model; the remaining factors then were reevaluated. This process was repeated until none of the remaining factors contributed significantly to the prediction of high resource use. Analyses were repeated excluding in-hospital deaths. All data were analyzed using SAS 8.2 (SAS Institute, Cary, NC) statistical software.
RESULTS
Summary of Cases of Congenital Heart Surgery
Of 2 516 833 pediatric discharges, 12717 cases with codes indicating congenital heart surgery were identified. Of these, 11381 cases met inclusion criteria (ie, age 18 years, cardiac transplants, and premature infants and newborns who were 30 days of age and had patent ductus arteriosus ligation as their only cardiac procedure were eliminated); 10602 (from 198 institutions) had total hospital charges reported. The number of cases by state ranged from 7 to 2394 (Fig 1). The median total hospital charge for this group was $51125, with a minimum of $195 and a maximum of $1000000. The 90th percentile cutpoint defining the especially high resource users was $192272 (Fig 2). Total hospital charges varied among the 27 states; California ($74294) had the highest median total hospital charges, and Maryland had the lowest ($19722; Fig 1). Descriptive information and differences in regional, demographic, clinical, and institutional variables associated with a higher proportion of cases with high resource use are shown in Table 1.
Multivariate Predictors of High Resource Use
Effects of State
Using New York State as a baseline, California, Colorado, Florida, Hawaii, Pennsylvania, and Texas had a significantly higher and Maine and South Carolina a significantly lower likelihood of cases with especially high resource use. Indicator variables representing each of these 8 states versus all others combined were included in additional models (Table 2).
Patient Demographic and Clinical Characteristics
Adjusting for baseline state differences in multivariate models, cases that were younger than 1 year had a higher odds of being in the high resource group as patients who were reported as premature or having other, noncardiac structural anomalies. Cases that had higher RACHS-1 risk categories or were unable to be categorized were more likely to use increased resources. Cases that were admitted during a weekend were more likely to be in the high resource group as were cases that were reported to have Medicaid or "other" insurance (Table 2).
After state and these additional patient characteristics were adjusted for, race, surgery at a children’s hospital, hospital cardiac volume, and geographic location were not associated with higher resource use. Findings were similar when in-hospital deaths were excluded.
Model Performance
The final model achieved an area under the ROC curve of 0.837 in discriminating cases of high versus non–high resource use.
DISCUSSION
The results of our population-based analysis of children who underwent surgical repair of congenital heart disease using the KID 2000 data set highlights significant geographic variation in resource utilization. Our results also reveal especially high resource utilization to be found within a subset of children who underwent congenital heart repair rather than the population as a whole.
Past research in this field has been limited to the investigation of the direct costs of a single complex cardiac defect and its surgical management, practice pattern, and cost variations among 8 centers that care for children and adults with congenital heart disease and cost of intensive care and post–intensive care stay for critically ill children.4–7 In addition, 1 previous study described a single institutional experience of cost of surgical care and predictors for higher cost in congenital heart repair.8 The researchers reported age of child; complexity of defect; presence of other, noncardiac anomalies or syndromes; and length of stay to be associated with increased financial risk.
Actual cost measures are very difficult to determine and were not available in our data source. Because our goal was to investigate especially high resource utilization, total hospital charges were used as a surrogate. The assumption that cases with especially high total hospital charges also had especially high resource utilization and therefore an overall higher cost seems reasonable, because most health care resources result in a higher charge to payers. Use of this reasonable surrogate allowed us to examine resource use for a rare, complex, pediatric condition from a national perspective. However, because charging structures are complex and do not accurately reflect costs, we chose to structure analyses to identify cases of especially high resource use only, ie, those in the upper 10th percentile for charges; we believed that attempting to identify predictors of actual charges used as a continuous variable was beyond the validity of our surrogate. It is interesting that our "cutpoint" demarcating high resource use was at $192272, suggesting that we have identified cases that are especially expensive. In fact, our high resource group accounted for >40% of the total charges accrued for all of these admissions.
Identification and profiling of increased resource utilization during surgical admission for children with congenital heart disease adds to the knowledge of a population that is associated with repeated hospitalizations, use of advanced technology, and use of innovative medical therapies. Our findings suggest that there is a subset of children who are especially high users of hospital resources and that state influences are also present.
Eight of the 27 states were noted to be significantly predictive of resource use. Cases from California, Colorado, Florida, Hawaii, Pennsylvania, and Texas had higher odds of being high resource utilization cases, and cases from South Carolina and Maine had significantly lower odds of being high resource utilization cases. Increased or decreased cost of living in certain states, such as Maine or California, may also affect cost of health care delivery, leading to geographic variation in total hospital charges. However, it is not certain at this time why geographic differences of high magnitude exist; this will be an area of our future research.
Our results suggest RACHS-1 risk category and younger age to be highly predictive of increased resource utilization. We also found clinical descriptors such as prematurity and presence of other, noncardiac structural anomalies to be highly predictive of resource utilization. Although there is a paucity of information with which to compare our findings, work by Ungelieder et al8 focusing on identification of patients associated with high-cost surgical admissions at a single institution produced similar findings. Our findings would further support their conclusions that there are clinical predictors of financial risk that may facilitate implementation of risk adjustments for payers and for strategic resource allocation within institutions.
Although the timing of hospital admission was noted to be a significant independent predictor of high resource utilization, we do not have a conclusive explanation for this. However, patients who were admitted on a weekend may possibly be cases that were not diagnosed prenatally, or required emergent surgical intervention, or were stabilized at time of admission and waited for surgery to be performed during a weekday.
Patients who were reported to have Medicaid or other insurance rather than private insurance were found to be independently predictive of high resource utilization. However, this variable added the least to the discrimination between our resource groups in the final model. The structure of the KID did not specify timing of recorded status, ie, admission versus discharge; therefore, it cannot be concluded whether these cases had Medicaid as a primary insurer on the basis of socioeconomic status or became Medicaid cases as a result of accruing significant hospital charges.
Demone et al9 previously reported Medicaid as an independent predictor of pediatric cardiac surgical outcomes. Their study examining the effect of insurance type on mortality for congenital heart surgery revealed that not only do Medicaid patients have a higher risk for death as compared with commercial or managed care patients, but also differences were present within institutions identified as low, average, and high mortality, suggesting that the adverse effect of Medicaid may be attributable to differential referral among patients who are treated at similar institutions. These findings when combined with our findings in resource utilization suggest that differences in care may exist and should be an area of continued study.
Although some institutional differences were noted in univariate analyses, bed size, teaching and children’s hospital status, and hospital volume of cardiac cases were not independently associated with greater numbers of cases of high resource use. Our findings are not consistent with what has been previously reported on pediatric conditions in teaching facilities10 and may be attributable to the limited number of children’s hospital and nonteaching facilities that perform congenital heart repair available in the data set. As is the current trend, most pediatric cardiac surgical programs exist in teaching facilities with a children’s unit, and it would be difficult to find an adequate number of nonteaching facilities for examination.
Limitations
Although the KID 2000 is considered to provide a representative sample of pediatric hospitalizations, discharge information originates from only 27 states. The KID data set did not contain unique patient identifiers or record linkage numbers, thereby making it impossible to identify discharges as individual patients. Missing data and coding errors are universal limitations in using large administrative data sets. For example, the data element "race" was missing for 19% of our overall sample. We therefore could not conclude with certainty that race was not a factor in high resource utilization. At the onset of our study, a limitation that was foreseen was the lack of detailed clinical information, which is another limitation of all studies that use administrative databases in outcomes research.11 In this population, birth weight and organ function have been cited as predictors of initial clinical outcomes as might also reflect resource utilization.12
Although our developed model of factors associated with increased resource utilization achieved a final area under the ROC curve of 0.837, it has not been validated in a second data set. Despite these limitations, our analyses included a large number of cases of complex congenital heart surgery from all regions of the country and yielded a number of important conclusions.
CONCLUSION
States varied in the frequency of cases of especially high resource use for congenital heart surgery, with some having a higher- and some having a lower-than-expected number. Patients who had greater disease complexity, younger age, prematurity, other anomalies, and Medicaid and were admitted during a weekend were more likely to result in especially high resource use. Institutions of various types did not differ in frequency of especially high resource use, regardless of children’s hospital or teaching status. Although demographic characteristics and disease complexity are fixed and offer no easy solutions for health care reform, this examination has identified a high magnitude of statewide variation and may reflect opportunities for more efficient care to be instituted.
APPENDIX 1: INDIVIDUAL PROCEDURES BY RISK CATEGORY
Risk Category 1
Atrial septal defect surgery (including atrial septal defect secundum, sinus venosus atrial septal defect, and patent foramen ovale closure)
Aortopexy
Patent ductus arteriosus surgery at age >30 days
Coarctation repair at age >30 days
Partially anomalous pulmonary venous connection surgery
Risk Category 2
Aortic valvotomy or valvuloplasty at age >30 days
Subaortic stenosis resection
Pulmonary valvotomy or valvuloplasty
Pulmonary valve replacement
Right ventricular infundibulectomy
Pulmonary outflow tract augmentation
Repair of coronary artery fistula
Atrial septal defect and ventricular septal defect repair
Atrial septal defect primum repair
Ventricular septal defect repair
Ventricular septal defect closure and pulmonary valvotomy or infundibular resection
Ventricular septal defect closure and pulmonary artery band removal
Repair of unspecified septal defect
Total repair of tetralogy of Fallot
Repair of total anomalous pulmonary veins at age >30 days
Glenn shunt
Vascular ring surgery
Repair of aortopulmonary window
Coarctation repair at age 30 days
Repair of pulmonary artery stenosis
Transection of pulmonary artery
Common atrium closure
Left ventricular to right atrial shunt repair
Risk Category 3
Aortic valve replacement
Ross procedure
Left ventricular outflow tract patch
Ventriculomyotomy
Aortoplasty
Mitral valvotomy or valvuloplasty
Mitral valve replacement
Valvectomy of tricuspid valve
Tricuspid valvotomy or valvuloplasty
Tricuspid valve replacement
Tricuspid valve repositioning for Ebstein anomaly at age >30 days
Repair of anomalous coronary artery without intrapulmonary tunnel
Repair of anomalous coronary artery with intrapulmonary tunnel (Takeuchi)
Closure of semilunar valve, aortic or pulmonary
Right ventricular to pulmonary artery conduit
Left ventricular to pulmonary artery conduit
Repair of double outlet right ventricle with or without repair of right ventricular obstruction
Fontan procedure
Repair of transitional or complete atrioventricular canal with or without valve replacement
Pulmonary artery band
Repair of tetralogy of Fallot with pulmonary atresia
Repair of cor triatriatum
Systemic to pulmonary artery shunt
Atrial switch operation
Arterial switch operation
Reimplantation of anomalous pulmonary artery
Annuloplasty
Repair of coarctation and ventricular septal defect closure
Excision of intracardiac tumor
Risk Category 4
Aortic valvotomy or valvuloplasty at age 30 days
Konno procedure
Repair of complex anomaly (single ventricle) by ventricular septal defect enlargement
Repair of total anomalous pulmonary veins at age 30 days
Atrial septectomy
Repair of transposition, ventricular septal defect, and subpulmonary stenosis (Rastelli)
Atrial switch operation with ventricular septal defect closure
Atrial switch operation with repair of subpulmonary stenosis
Arterial switch operation with pulmonary artery band removal
Arterial switch operation with ventricular septal defect closure
Arterial switch operation with repair of subpulmonary stenosis
Repair of truncus arteriosus
Repair of hypoplastic or interrupted arch without ventricular septal defect closure
Repair of hypoplastic or interrupted aortic arch with ventricular septal defect closure
Transverse arch graft
Unifocalization for tetralogy of Fallot and pulmonary atresia
Double switch
Risk Category 5
Tricuspid valve repositioning for neonatal Ebstein anomaly at age 30 days
Repair of truncus arteriosus and interrupted arch
Risk Category 6
Stage 1 repair of hypoplastic left heart syndrome (Norwood operation)
Stage 1 repair of nonhypoplastic left heart syndrome conditions
Damus-Kaye-Stansel procedure
FOOTNOTES
Accepted Dec 29, 2004.
No conflict of interest declared.
REFERENCES
Jenkins KJ, Gauvreau K. Center-specific differences in mortality: preliminary analyses using the Risk Adjustment in Congenital Heart Surgery (RACHS-1) method. J Thorac Cardiovasc Surg. 2002;124 :97 –104
Jenkins KJ, Gauvreau K, Newburger JW, Kyn LB, Iezzoni LI, Mayer JE. Validation of relative value scale for congenital heart operations. Ann Thorac Surg. 1998;66 :860 –869
Jenkins KJ, Gauvreau K, Newburger JW, Spray TL, Moller JH, Iezzoni LI. Consensus-based method for risk adjustment for surgery for congenital heart disease. J Thorac Cardiovasc Surg. 2002;123 :110 –118
Williams DL, Gelijns AC, Moskowitz AJ, et al. Hypoplastic left heart syndrome: valuing the survival. J Thorac Cardiovasc Surg. 2000;119 :720 –731
Kanter RK. Post-intensive care unit pediatric hospital stay and estimated costs. Crit Care Med. 2000;28 :220 –223
Chalom R, Raphaely RC, Costarino AT Jr. Hospital costs of pediatric intensive care. Crit Care Med. 1999;27 :2079 –2085
Garson A Jr, Allen HD, Gersony WM, et al. The cost of congenital heart disease in children and adults. A model for multicenter assessment of price and practice variation. Arch Pediatr Adolesc Med. 1994;148 :1039 –1045
Ungerleider RM, Bengur AR, Kessenich AL, et al. Risk factors for higher cost in congenital heart operations. Ann Thorac Surg. 1997;64 :44 –48; discussion 49
DeMone JA, Gonzalez PC, Gauvreau K, Piercey GE, Jenkins KJ. Risk of death for Medicaid recipients undergoing congenital heart surgery. Pediatr Cardiol. 2003;24 :97 –102
Srivastava R, Homer CJ. Length of stay for common pediatric conditions: teaching versus nonteaching hospitals. Pediatrics. 2003;112 :278 –281
Iezzoni LI. Using administrative data to study persons with disabilities. Milbank Q. 2002;80 :347 –379
Iannettoni MD, Bove EL, Mosca RS, et al. Improving results with first-stage palliation for hypoplastic left heart syndrome. J Thorac Cardiovasc Surg. 1994;107 :934 –940(Jean Anne Connor, DNSc, R)