Center Variation in Hospital Costs for Patients Undergoing Congenital Heart Surgery
Background—Congenital heart disease consumes significant health care resources; however, there are limited data regarding factors affecting resource utilization. The purpose of this study was to evaluate variation between centers in total hospital costs for 4 congenital heart operations of varying complexity and associated factors.
Methods and Results—The Premier Database was used to evaluate total cost in children undergoing isolated atrial septal defect (ASD) repair, ventricular septal defect (VSD) repair, tetralogy of Fallot (TOF) repair, or arterial switch operation (ASO) from 2001 to 2007. Mixed models were used to evaluate the impact of center on total hospital costs adjusting for patient and center characteristics and length of stay. A total of 2124 patients were included: 719 ASD (19 centers), 792 VSD (20 centers), 420 TOF (17 centers), and 193 ASO (13 centers). Total cost increased with complexity of operation from median $12 761 (ASD repair) to $55 430 (ASO). In multivariable analysis, models that accounted for center effects versus those that did not performed significantly better for all 4 surgeries (all P≤0.01). The proportion of total cost variation explained by center was 19% (ASD repair), 11% (VSD repair), 6% (TOF repair), and 3% (ASO). Higher-volume centers had significantly lower hospital costs for ASD and VSD repair but not for TOF repair and ASO.
Conclusions—Total hospital costs varied significantly by center for all congenital heart surgeries evaluated, even after adjustment for patient and center characteristics and length of stay. Differences among centers were most prominent for lower complexity procedures.
Congenital heart disease causes an enormous health burden, affecting nearly 1% of live births and accounting for significant resource utilization, with an estimated 6 billion dollars annually in acute care costs alone.1 It has been reported that of all birth defects, congenital heart defects are associated with the highest average hospital charges, length of stay, and mortality.2 Despite this, few studies have evaluated factors affecting resource utilization in patients undergoing congenital heart surgery.3–7
Analyses to date have been limited by the use of hospital charge data (which reflect the amount the hospital billed for services rather than actual cost) and analyzing surgeries of varying levels of complexity together.3–7 This makes it difficult to discern patient and center factors associated with higher hospital costs as the driving factor in most cases is the complexity of the procedure being performed. In addition, many studies to date have focused on patients undergoing congenital heart surgery from a single center, so that there are limited data regarding differences in resource utilization between centers.5,6 If differences in resource utilization between centers do exist, this would support further evaluation of factors mediating these differences to facilitate development of initiatives aimed at cost reduction, which is of particular importance in this era of rising health care costs. Thus, the purpose of this study was to evaluate variation between centers in total hospital costs for 4 congenital heart operations of varying levels of complexity and to identify factors associated with increased hospital costs for each surgery.
The Premier Database (Premier, Inc, Charlotte, NC) is an inpatient administrative database including more than 500 US hospitals with on average more than 5 million hospital discharges per year. The database includes a wide range of hospitals with diverse characteristics in regard to geographic region, number of beds, population served, and teaching status. Both pediatric and adult data are collected. Data from both freestanding and nonfreestanding children's hospitals are included in the pediatric data. Data collected include diagnoses and procedures using International Classification of Diseases, Ninth Revision, (ICD-9) coding, demographic information, resource utilization (including total hospital costs), length of hospital stay, and in-hospital mortality. Hospital characteristics are also collected. All data received from hospitals are validated against Premier standards, and any deviations are corrected by the participating hospitals. The Premier Database has been widely used in previous health outcomes and health economics studies.8,9
Congenital heart disease treatment requires significant health care resources.
There are limited data regarding factors impacting resource utilization for patients undergoing congenital heart surgery, and whether there is significant variation in costs from center-to-center.
WHAT IS KNOWN
Total hospital costs varied significantly by center for all congenital heart surgeries evaluated, even after accounting for patient and center factors and length of stay.
Differences among centers were most prominent for lower complexity procedures.
Strategies to improve outcome and reduce cost variation may therefore be best focused on lower complexity, common operations.
WHAT THE STUDY ADDS
Patients 0 to 18 years of age undergoing congenital heart surgery from 2001 to 2007 at a Premier center were eligible for inclusion. As previously described, Risk Adjustment in Congenital Heart Surgery, version 1 (RACHS-1) methodology was used to identify patients undergoing 4 surgeries of varying levels of complexity: secundum atrial septal defect repair, ventricular septal defect repair, tetralogy of Fallot repair (excludes tetralogy of Fallot with pulmonary atresia), and the arterial switch operation for transposition of the great arteries.3,10 Only patients undergoing these procedures in isolation were included, and those undergoing other concurrent cardiac surgical procedures were excluded. The complexity of the 4 included procedures as assessed by the Aristotle Basic Complexity Score ranged from 3 for atrial septal defect repair, 6 for ventricular septal defect repair, 8 for tetralogy of Fallot repair, to 10 for the arterial switch operation.11 Centers with fewer than 5 cases during the study period for a specific procedure were excluded from analysis of that procedure. Only survivors to hospital discharge were included in the analysis. This study was approved by the Duke Institutional Review Board with a waiver of informed consent.
Data collection included patient age (captured in years), sex, race/ethnicity, prematurity/low birth weight (neonates only), presence of genetic syndrome or major noncardiac abnormality, and payor (private, government, other) as previously described.10,12 Center characteristics included center region, location (urban versus rural), teaching hospital status, and children's hospital status (freestanding versus nonfreestanding). Center average annual volume over the study period for the procedure of interest was also calculated.
The primary outcome was total hospital cost. The Premier Database collects data directly from the cost-accounting systems at participating hospitals. In the Premier Database, total hospital costs include direct or variable costs (including direct costs for in-hospital services such as procedures, room and board, professional fees for services performed by hospital staff, and pharmacy costs) and overhead or fixed costs for in-hospital services during the stay. Premier also maps individual costs to a standardized list of cost elements (eg, room and board, pharmacy, laboratory, etc) for each participating center. Physician professional fees are not captured by the Premier Database. Given that the study period included several years of data (2001 to 2007), all cost figures were discounted to 2001 dollars.
Patient and center characteristics and total hospital costs for each of the 4 surgeries were summarized as median and interquartile range for continuous variables and frequencies and percentages for categorical variables. Total hospital costs for each center included in the analysis were also graphed by box-and-whisker plots for each of the 4 procedures.
To evaluate the impact of individual center on total hospital costs and the contribution of various patient and center factors, mixed models were used. Total cost was log-transformed for analysis because of the skewed distribution of cost data. Six random-effect models were developed for each operation and compared to assess model performance with the inclusion of various patient and center factors. Model 1 included center as the only random effect and was not adjusted for other factors. Models 2 to 5 all included center as a random effect and adjusted for various other factors as fixed effects. Model 2 included patient factors as fixed effects (age, sex, race, genetic syndrome, major noncardiac abnormality, low birth weight, and insurance status); model 3 included patient factors (as noted above) and length of stay as fixed effects; model 4 included center factors only (children's hospital status, urban versus rural location, teaching status, region of country, and center volume for procedure of interest) as fixed effects; and model 5 included all factors as fixed effects (patient factors, center factors, and length of stay). The sixth and final model included all patient factors, center factors, and length of stay as fixed effects but removed the random effect of center. This was done to evaluate the impact of center, accounting for all other patient and center factors. Models were compared using likelihood ratio testing and Akaike information criterion (AIC), with smaller AIC representing a better model.13 The proportion of total cost variation explained by the random center effect was also calculated for each model.
Parameter estimates and standard errors are reported for center factors from model 5 (which adjusts for patient factors and length of stay) to assess the independent association of various center factors with total costs. Estimates from this model of log cost were transformed back to actual cost, using the smearing estimate.
Finally, the cost savings per patient if all centers were to perform as well as the lowest cost quartile were estimated from the mixed models (model 5) for each surgery. The estimated cost for each patient was calculated by transforming log cost to cost using the smearing estimate. Then, while the fixed effects were kept constant, the hospital effect was replaced with the average hospital effect of the lowest cost quartile and estimated cost per patient recalculated. All analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, NC). A probability value <0.05 was considered statistically significant.
A total of 2124 survivors to hospital discharge were included: 719 patients undergoing atrial septal defect repair (n=19 centers), 792 patients undergoing ventricular septal defect repair (n=20 centers), 420 patients undergoing tetralogy of Fallot repair (n=17 centers), and 193 patients undergoing the arterial switch operation (n=13 centers). Patient and center characteristics for each of the 4 surgical groups are displayed in Table 1. Of note, there was a wider range of center annual volume for the lower complexity surgeries (atrial septal defect repair, and ventricular septal defect repair) compared with the higher complexity surgeries (tetralogy of Fallot repair, and arterial switch operation).
Unadjusted total hospital costs for each of the 4 surgeries are displayed in Table 2. As expected, total hospital costs increased with increasing complexity of operation from a median of $12 761 for atrial septal defect repair to a median of $55 430 for the arterial switch operation. For each of the 4 surgeries, room and board accounted for the largest proportion of total costs (Table 2), and the upper cost quartile had longer median length of stay compared with the lower cost quartile (atrial septal defect repair 4 days versus 3 days, ventricular septal defect repair 6 days versus 4 days, tetralogy of Fallot repair 12 days versus 7 days, and arterial switch operation 22 days versus 15 days). Box-and-whisker plots displaying unadjusted total hospital costs by center for each of the 4 surgeries are displayed in the Figure(A through D). Of note, although only survivors to hospital discharge were included in the analysis, we also examined the mortality rate (before excluding those who died) by center cost quartiles for each operation. Mortality rates tended to be higher in the upper versus lower cost quartile for the higher complexity operations: tetralogy of Fallot repair (5% versus 2%) and arterial switch operation (15% versus 1%) and similar in the upper versus lower cost quartile for the lower complexity operations: atrial septal defect repair (1% versus 1%) and ventricular septal defect repair (0.5% versus 1%).
Results from mixed models are displayed in Table 3. For all 4 surgeries, the proportion of total cost variation explained by center decreased as patient and center factors and length of stay were added into the models (comparing model 1 through model 5). Adjusting for patient and center factors and length of stay, models that accounted for center effects (model 5) versus those that did not (model 6) performed significantly better for all 4 surgeries, indicating that center is a significant factor affecting total hospital costs (P-value from likelihood ratio testing comparing model 5 versus model 6: atrial septal defect repair, P<0.0001; ventricular septal defect repair, P<0.0001; tetralogy of Fallot repair, P<0.0001; and arterial switch operation, P=0.01). For all surgeries, AIC was also smaller for model 5 (which included center effects) versus model 6, indicating better model performance. Accounting for all patient and center factors and length of stay (model 5), the proportion of total cost variation explained by center decreased as the complexity of surgery increased: 19% for atrial septal defect repair, 11% for ventricular septal defect repair, 6% for tetralogy of Fallot repair, and 3% for arterial switch operation.
The association of various center factors with adjusted total hospital costs for each of the 4 surgeries from model 5 (which accounts for patient characteristics, length of stay, and center effects) is displayed in Table 4. Higher center volume was associated with significantly lower hospital costs. This effect was only present for lower complexity surgeries (atrial septal defect and ventricular septal defect repair). The estimated average difference in adjusted total hospital costs between the upper and lower center volume quartiles for atrial septal defect repair was $1782 and for ventricular septal defect repair, $1196.
Finally, the average adjusted cost savings per patient if all centers were to perform as well as the lowest cost quartile were estimated from the models for each surgery: atrial septal defect repair, $3741; ventricular septal defect repair, $6323; tetralogy of Fallot repair, $5789; and arterial switch operation, $12 846.
In this multicenter analysis, we found that total hospital costs varied significantly by center for all 4 congenital heart operations evaluated. These differences persisted even after adjustment for patient characteristics and length of stay and were most prominent for lower complexity operations.
Congenital heart defects, while relatively rare, consume a large share of health care resources.1,2 Children with congenital heart disease often require numerous hospitalizations, which may include advanced surgical or interventional therapies, care by a multidisciplinary team of specialists, frequent imaging and other diagnostic testing, drug and device therapy, and lifelong outpatient follow-up. Over the past several decades, survival after congenital heart surgery has improved significantly, such that evaluation of other outcomes including resource utilization, quality of life, complications, and neurodevelopmental function has now received greater attention.14 Previous studies have suggested that inpatient costs account for the majority of resource utilization in terms of direct medical costs for patients undergoing congenital heart surgery, as opposed to outpatient services.15 Similar to others, we also found that the vast majority of inpatient costs were related to room and board charges and that higher costs were associated with longer length of stay.7 Differences in length of stay likely are related to both patient preoperative, operative, and postoperative factors as well as center-related factors.
We are not aware of previous studies evaluating variation in hospital costs between centers caring for patients undergoing congenital heart surgery. In a large study evaluating in-hospital charges in more than 10 000 patients undergoing congenital heart surgery in 2000, Connor et al3 demonstrated significant variation in hospital charges between the 27 different US states included in the analysis. Variation between centers was not evaluated. Investigators also found that, as expected, higher complexity cases were associated with increased hospital charges.3
In the present study, we chose to perform separate analyses of 4 common surgeries to evaluate variation between institutions and important factors affecting resource utilization, across varying levels of surgical complexity. We also evaluated hospital costs rather than charges to assess the actual cost of the services. Interestingly, we found that although total hospital costs varied significantly by center for all operations evaluated, these differences were most prominent for the lower complexity procedures (atrial septal defect and ventricular septal defect repair). The reasons for this are unclear; however, it may be possible that for the higher complexity cases, the complexity of the procedure or patient-specific factors outweigh any center-related factors affecting cost. In addition, we found that for the lower complexity surgeries, higher center surgical volume was associated with lower total hospital costs, whereas this association was not seen for the higher complexity surgeries. In evaluating other center-level factors affecting hospital costs, we did not find that teaching or children's hospitals were associated with reduced hospital costs, similar to other studies.3
There may be several reasons for the association of center volume with lower hospital costs. It may be that patients at higher-volume centers have fewer postoperative complications. Although were not able to evaluate postoperative complications specific to congenital heart surgery in this study utilizing administrative data, it has previously been shown that complications after congenital heart surgery are associated with increased resource utilization.4 In addition, it may also be possible that larger, more experienced centers have greater resources and standardized protocols of care to facilitate postoperative recovery. The implementation of multidisciplinary standardized care protocols has been associated with decreased length of stay and reduced hospital charges for patients undergoing congenital heart surgery.16–18 Davis et al16 showed that implementation of a standardized care plan for 65 consecutive children undergoing atrial septal defect repair at their institution was associated with significant reductions in intensive care unit and total hospital length of stay, as well as a reduction in total hospital charges. Fernandes et al17 evaluated a more diverse population of 175 patients undergoing congenital heart surgery and also found that patients treated according to a standardized recovery protocol versus conventional care had reduced length of stay and reduced hospital costs. Standardized management has also been shown to improve outcome in high risk cohorts as well, such as single-ventricle patients undergoing the Norwood operation.19 Interestingly, we found that the association of center volume with cost was only significant for the lower complexity surgeries. This may be due to a wider range of center surgical volume for the lower complexity surgeries, compared with the higher complexity surgeries.
The results of this study have several implications in this era of rising health care costs. Our data, along with the previous Connor et al3 analysis, suggest significant variation in resource utilization among different US states and different centers caring for patients undergoing congenital heart surgery. Further analysis of factors associated with outcome and cost in this population may elucidate best practices that can both improve outcome and reduce cost across different centers and regions. Our data suggest that although higher complexity operations account for greater relative cost, there is greater variation in cost from center to center for lower complexity operations. In addition, although our data indicate that the estimated cost savings per patient if all centers performed as well as the lower cost quartile is greater for the higher complexity procedures, the total cost savings across the board probably would be greater for the lower complexity surgeries such as atrial septal defect and ventricular septal defect repair, which are >2 times more common than tetralogy of Fallot repair and the arterial switch operation.20 Thus, initial efforts to reduce costs may be best focused on these lower complexity, but common, congenital heart operations.
In this analysis, we used a large administrative database to evaluate variation in hospital costs. Although administrative data sources contain valuable resource utilization information that makes these types of analyses possible, they may lack detailed clinical data in some cases. It is also possible that certain unmeasured confounders may be present; however, we were able to adjust for important patient and center factors demonstrated to affect resource utilization in previous analyses using both administrative and clinical data. Although the Premier database includes a diverse variety of hospitals in terms of region of the country, teaching status, and center volume, not all US children's hospitals are represented, and thus our results may not be generalizable across all centers. A strength of the present analysis was the use of cost data rather than charges to better reflect the cost of the services being provided. However, these data probably are an underestimate of true total costs as they do not account for physician professional fees, and other costs relevant to the family and society such as transportation to the hospital, lodging, child care for other siblings, and loss of income and productivity if it is necessary to take time off from the workplace. We were also unable to account for the cost of previous or subsequent surgical procedures the patient may have had, such as tetralogy of Fallot patients, who under went initial palliation with a Blalock-Taussig shunt before complete repair. In addition, the relatively small number of patients and procedures in certain groups, such as those undergoing the arterial switch operation may have limited our power to detect differences in this group. Finally, although we used widely recognized methodology to identify operations of interest in the database, it is possible that coding errors may exist in administrative data sources.10 The similarity of our cost data with that from other studies using clinical data sources, and expected increase in total cost with increased complexity of operations in our analysis, support the validity of these data.5–7
In this multicenter evaluation of hospital costs in patients undergoing congenital heart surgery, we found significant variation from center to center even after accounting for important patient factors and length of stay. These differences were most pronounced for lower complexity surgeries, and hospital surgical volume appeared to play a role. Further evaluation of patient outcomes in conjunction with cost and associated factors may elucidate strategies to both improve outcome and reduce cost variation across centers. Our data would suggest that these efforts may be best focused first on lower complexity, common, operations, where cost variation across centers is most pronounced and the overall expected cost reduction would be greatest given the higher frequency of these types of surgeries. In addition, the recent linking of clinical registry data with administrative data in the congenital heart surgery population will enable analyses such as those evaluating costs (captured in the administrative database) associated with specific postoperative complications (captured in the registry) and thus the cost-savings that may be expected from quality improvement initiatives to reduce postoperative morbidities and complications.21
Sources of Funding
Dr Pasquali received grant support from the National Heart, Lung, and Blood Institute (1K08HL103631-01) and from the American Heart Association Mid-Atlantic Affiliate Clinical Research Program. This study was supported in part by cooperative agreement HS016957 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality.
- Received August 20, 2010.
- Accepted January 26, 2011.
- © 2011 American Heart Association, Inc.
- Gray DT,
- Pourmoghadam K,
- Weis J,
- Hsu A,
- Moore J,
- Jacobs J,
- Steiner C,
- Jacobs M
- Connor JA,
- Gauvreau K,
- Jenkins KJ
- Kim MH,
- Klingman D,
- Lin J,
- Pathak P,
- Battleman D
- Al-Radi OO,
- Harrell FE,
- Caldarone CA,
- McCrindle BW,
- Jacobs JP,
- Williams MG,
- VanArsdell GS,
- Williams WG
- Pasquali SK,
- Hall M,
- Slonim AD,
- Jenkins KJ,
- Marino BS,
- Cohen MS,
- Shah SS
- Fernandes AM,
- Mansur AJ,
- Canêo LF,
- Lourenço DD,
- Piccioni MA,
- Franchi SM,
- Afiune CM,
- Gadioli JW,
- Oliveira Sde A,
- Ramires JA
- Jacobs JP,
- Jacobs ML,
- Mavroudis C,
- Lacour-Gayet FG,
- Tchervenkov CI
- Pasquali SK,
- Jacobs JP,
- Shook GJ,
- O'Brien SM,
- Hall M,
- Jacobs ML,
- Welke KF,
- Gaynor JW,
- Peterson ED,
- Shah SS,
- Li JS