Neurosurg Focus 37 (5):E5, 2014 ©AANS, 2014
Cerebrospinal fluid shunt placement in the pediatric population: a model of hospitalization cost Sandi K. Lam, M.D., M.B.A.,1,2 Visish M. Srinivasan, M.D.,1,2 Thomas G. Luerssen, M.D.,1,2 and I-Wen Pan, Ph.D.1,2 Division of Pediatric Neurosurgery, Texas Children’s Hospital; and 2Department of Neurosurgery, Baylor College of Medicine, Houston, Texas
1
Object. There have been no large-scale analyses on cost drivers in CSF shunt surgery for the treatment of pediatric hydrocephalus. The objective of this study was to develop a cost model for hospitalization costs in pediatric CSF shunt surgery and to examine risk factors for increased costs. Methods. Data were extracted from the Kids’ Inpatient Database (KID) of the Healthcare Cost and Utilization Project. Children with initial CSF shunt placement in the 2009 KID were examined. Patient charge was converted to cost using a cost-to-charge ratio. The factors associated with costs of CSF shunt hospitalizations were examined, including patient demographics, hospital characteristics, and clinical data. The natural log transformation of cost per inpatient day (CoPID) was analyzed. Three multivariate linear regression models were used to characterize the cost. Variance inflation factor was used to identify multicollinearity for each model. Results. A total of 2519 patients met the inclusion criteria and were included in study. Average cost and length of stay (LOS) for initial shunt placement were $49,317 ± $74,483 (US) and 18.2 ± 28.5 days, respectively. Cost per inpatient day was $4249 ± $2837 (median $3397, range $80–$22,263). The average number of registered nurse (RN) full-time equivalents (FTEs) per 1000 adjusted inpatient days was 5.8 (range 1.6–10.8). The final model had the highest adjusted coefficient of determination (R2 = 0.32) and was determined to be the best among 3 models. The final model showed that child age, hydrocephalus etiology, weekend admission, number of chronic diseases, hospital type, number of RN FTEs per 1000 adjusted inpatient days, number of procedures, race, insurance type, income level, and hospital regions were associated with CoPID. Conclusions. A patient’s socioeconomic status, such as race, income level, and insurance, in addition to hospitalrelated factors such as number of hospital RN FTEs, hospital type, and US region, could affect the costs of initial CSF shunt placement, in addition to clinical factors such as hydrocephalus origin and LOS. To create a cost model of initial CSF shunt placement in the pediatric population, consideration of such nonclinical factors may be warranted. (http://thejns.org/doi/abs/10.3171/2014.8.FOCUS14454)
Key Words • pediatric • ventriculoperitoneal shunt • cost analysis • cerebrospinal fluid shunt • hydrocephalus
P
lacement of ventriculoperitoneal shunts is one of the most common procedures in pediatric neurosurgery and is a significant health care burden. Shunt failure and shunt infections are common reasons for readmissions and revision surgeries.5 Total hospital charges for treatment of pediatric hydrocephalus in the US are estimated to be between $1.4 and $2.0 billion each year.35 Shunt admissions account for 0.6% of all pediatric hospital admissions, but consist of 3.1% of all pediatric hospital charges,35 suggesting that children with hydrocephalus use a disproportionate number of hospital days and resources in the US.
Abbreviations used in this paper: CoPID = cost per inpatient day; FTE = full-time equivalent; ICD = International Classification of Diseases; IVH = intraventricular hemorrhage; KID = Kids’ Inpatient Database; LOS = length of stay; RN = registered nurse; VIF = variance inflation factor.
Neurosurg Focus / Volume 37 / November 2014
Studies of costs in adult and pediatric neurosurgery also show that CSF shunt-related problems are among the most common causes for hospital readmissions within 30 days.30 In pediatric CSF shunt surgery, 5,6 reoperations and readmissions have come under scrutiny.40 Several patientor hospital-level factors have been reported to be associated with inpatient costs, such as length of stay (LOS), age, sex, comorbidities, case mix, and use of facilities and services.4,17,26 Hospital LOS is often highly correlated with inpatient costs for patients treated using inpatient surgical procedures.4,26 There is increasing pressure from public and private payers on medical service providers for cost reduction. Payment strategies are shifting from traditional fee-forservice to bundled schemes. The bundled payment of an episode begins with hospital admission to the end of inpatient stay; all expenses are included, such as facility, professional (i.e., services provided by health profes1
S. K. Lam et al. sionals), and prescription drug expenses. This idea has gained traction with Medicare and private plans and has been applied to adult inpatient services for years, as in the Centers for Medicare and Medicaid Services diagnosisrelated groups and Medicare Severity diagnosis-related groups.3,7,39 Risk-adjusted models and hierarchical coexisting conditions models have been adapted by the Centers for Medicare and Medicaid Services for adjusting payment of Medicare services.27 However, there is no existing risk adjustment for pediatric payment models. Furthermore, studies have shown that existing hierarchical coexisting conditions models may underpredict the actual costs associated with treating adult patients with multiple comorbid conditions.24 Previous reports show mean charges for initial CSF shunt placement increased from $28,370 to $40,260 in 2003 dollars between 1997 and 2003, with a stable mean LOS (6.8 days).35 From the payer and societal perspectives, it is important to investigate factors associated with the inpatient costs for CSF shunt placement to characterize the economic burden and target areas for containment and improvement. There are currently no large-scale analyses on cost drivers in shunt surgery for hydrocephalus treatment. We aim to use the national Kids’ Inpatient Database (KID) to develop a cost model of pediatric CSF shunt surgery hospitalization costs and thus examine risk factors for increased costs. In doing so, this study considers the complexity of the factors associated with inpatient costs, which may show some of the risk in bundling payment models.
Methods Data Source
Data were extracted from the KID, one of a family of administrative databases developed by the Healthcare Cost and Utilization Project sponsored by the US Agency for Healthcare Research and Quality. Patient discharge data includes patient demographics, admission type and source, diagnostic and procedural codes from the International Classification of Diseases (ICD) Ninth Revision, LOS, disposition, and payer data. The number of participating states was 44 in 2009, with data from 4100 hospitals. The KID contains information from 2–3 million pediatric discharges, weighted to represent 6.5–7.5 million national discharges. The basic unit of analysis is a patient discharge, rather than an individual patient.
Study Sample
Children (0–20 years old) with initial CSF shunt placement in the 2009 KID were selected as follows: hydrocephalus-related admission with any procedure code for intracranial ventricular shunt placement (02.31–02.35), excluding those with any procedure code for replacement of ventricular shunt (02.42, 54.95) or removal of ventricular shunt (02.43), and any diagnosis code for shunt malfunction (996.2) or CSF shunt infection (996.63).34,35 Patient records with missing insurance status, age, sex, or cost-to-charge ratio at their treatment hospital were also excluded.
2
Cost Estimation
The KID provides two types of cost-to-charge ratios. Patient charges were converted to costs by either hospitalspecific cost-to-charge ratio (all-payer inpatient cost-tocharge ratio) if available, or the group-weighted average cost-to-charge ratio in the small minority who were missing all-payer inpatient cost-to-charge ratio values.12 To remove the effect of LOS on total costs, we used cost per inpatient day (CoPID) instead of total inpatient cost to investigate factors associated with costs in the study. In addition, we used the natural logarithmic (log) transformation of CoPID instead of the original form. The original CoPID was skewed to the right and not normally distributed (Fig. 1 upper) and thus not appropriate for a regression model. Therefore, we used the natural logarithmic transformation of CoPID, which fits the normal distribution almost perfectly (Fig. 1 lower) and fulfills the assumptions of the regression model. Use of log transformation of such data are reported to produce more precise and unbiased estimation.19 Cost was characterized by nonnegative measurements of the outcomes and a positively skewed empirical distribution. While diseases with specific ictus (such as stroke, acute cardiac events, and ruptured aneurysms) may incur costs at the beginning of hospitalization and necessitate examining costs on certain days, the pediatric hydrocephalus population may incur costs over a different time frame. For instance, CSF shunt placement may occur after a long neonatal intensive care unit stay, after a period of observation after myelomeningocele closure, or soon after presentation with an obstructive tumor. This variability is an important distinction to many conditions studied in adult populations, and supports the use of CoPID rather than costs on specific hospital days.
Patient Demographics and Clinical Characteristics
The following patient demographics associated with CoPID were examined: age (0–1 year and > 1 year), sex (male/female), insurance status (private, Medicaid, and other), race (White, Black, Hispanic, Asian and others, and unspecified), and median household income in quartiles for a patient’s zip code of residence (zip-code income level). Patient were assigned into 1 of 9 hydrocephalus etiologies based on ICD-9 Clinical Modification diagnosis codes as follows, based on previously published methodologies:34,35 myelomeningocele (653.7, 655.0, 741.xx), intraventricular hemorrhage (IVH; 772.1x), congenital hydrocephalus (742.3), meningitis (320–322, 326), CNS tumor (191.xx- 194.xx, 198.3, 198.4, 225.0–225.2, 225.8– 225.9, 227.4, 237.0–237.1, 237.5–237.7, 239.6–239.7), trauma (767.4, 851.xx-854.xx, 995.55), communicating hydrocephalus (331.3), or obstructive hydrocephalus (331.4). The number of chronic conditions, weekend or weekday admission, and mortality were also variables included in the analysis. Hospital Characteristics
Hospital characteristics used in the study included region, number of registered nurse (RN) full-time equivalents (FTEs) per 1000 adjusted inpatient days (RN FTEs), Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization
Fig. 1. Distributions for original CoPID (upper), which is not normally distributed, and the natural log transformation of CoPID (Ln_CoPID; lower), which fits a normal distribution.
and hospital type. Hospital type was categorized into government owned, nonprofit non-children’s hospitals, nonprofit children’s hospitals, nonprofit children’s units in a hospital, investor-owned private hospitals, and unspecified. Teaching status, hospital size, and hospital volume were not included in the analysis due to high collinearity with either hospital type or RN FTEs. Three-Model Comparison and Factor Selection
Several state Medicaid programs including Texas,
Neurosurg Focus / Volume 37 / November 2014
Arizona, and Florida currently reimburse or plan to reimburse using a prospective payment and pricing system based on all patient-refined diagnosis-related groups.2,13,37 All patient-refined diagnosis-related groups are an expansion of the diagnosis-related groups to include non-Medicare populations such as pediatric patients with adjusted severity of illness.3,28 The data elements used by the all patient-refined diagnosis-related groups include principal diagnosis, secondary diagnosis, procedures, age, sex, and discharge disposition. The initial model (Model I) was 3
S. K. Lam et al. based on this concept with consideration of geographical variation (http://www.dartmouthatlas.org) and hospital types.21 To build upon Model I, hospital operational factors such as RN FTEs were added into Model II. Nurse-to-patients ratios10 are reported to increase costs but to improve quality of care. Finally, nonclinical socioeconomic covariates, such as race, insurance status, and income level, were taken into consideration for investigating their impacts on costs in Model 3. The study sample includes all elective (69%) and nonelective (31%) cases. Statistical Analysis
Simple and multivariate linear regression models were used to build the cost model. Covariates with a p value > 0.2 in simple linear regression were excluded from the multivariate regression model. We compared 3 different models to understand the effect of additional covariates on the cost model; from there, we selected the final model. First, we modeled cost with age, sex, weekend admission, hydrocephalus etiology, number of chronic conditions, number of surgical procedures, hospital type, and region. For the next model, the number of RN FTEs per 1000 adjusted inpatient days was added into the list of variables. Finally, for the third model, we added race, insurance status, and zip-code income level in addition to the variables used in the first 2 models. Variance inflation factor (VIF) was used to identify multicollinearity for the multivariate regression model. No VIF greater than 10 was accepted. All statistical analyses were performed using SAS (version 9.3, SAS Institute, Inc.) and Stata (version 13, StataCorp). Test results with a p value < 0.05 were considered statistically significant.
Results
A total of 2519 patients were included in this study. More than half of the patients (61.1%) were under 1 year of age (Table 1). Payers included Medicaid (50.1%) and private insurers (43.4%). Only 1.4% of patients died during the CSF shunt hospital admission. The most common hydrocephalus etiologies were obstructive hydrocephalus (33.1%), congenital hydrocephalus (20.0%), myelomeningocele (14.9%), tumor (8.7%), and IVH (6.8%). The most common hospital type was children’s unit in a nonprofit hospital (33.6%), followed by nonprofit children’s hospital (26.3%) and government-owned hospital (12.4%). Average cost and LOS for initial shunt placement were $49,317 ± $74,483 and 18.2 ± 28.5 days, respectively. Average CoPID was $4249 ± $2837 (median $3397, range $80–$22,263). Average number of RN FTEs per 1000 adjusted inpatient days was 5.8 (range 1.6–10.8 RN FTEs; Table 2). By hydrocephalus etiology, IVH and patients with meningitis had the longest average LOSs (76 ± 44 days and 45 ± 30 days, respectively) and total costs were $176,073 ± $130,844 and $106,159 ± $78,458, respectively, although their CoPIDs were the lowest and the second lowest ($2438 ± $1224, and $2453 ± $1084, respectively; Tables 2 and 3). Conversely, communicating hydrocepha4
TABLE 1: Patient- and hospital-level characteristics Variable patient characteristics age (yrs) 0 1–20 sex female male race White Black Hispanic Asian & others unspecified insurance type private Medicaid others income level (%) 0–25 25–50 50–75 75–100 unspecified clinical characteristics admission at weekend yes no death during hospital admission yes no hydrocephalus etiology communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others hospital characteristics hospital type government nonprofit, non-children’s nonprofit, children’s nonprofit, children’s unit in a hospital private, investor-owned unspecified
No. of Patients
%
1538 981
61.1 38.9
1151 1368
45.7 54.3
1129 294 457 211 428
44.8 11.7 18.1 8.4 17.0
1094 1261 164
43.4 50.1 6.5
771 657 563 481 47
30.6 26.1 22.4 19.1 1.9
317 2202
12.6 87.4
36 2483
1.4 98.6
137 833 503 220 171 376 40 75 164
5.4 33.1 20.0 8.7 6.8 14.9 1.6 3.0 6.5
312 278 663 847 57 362
12.4 11.0 26.3 33.6 2.3 14.4 (continued)
Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization TABLE 1: Patient- and hospital-level characteristics (continued) Variable
No. of Patients
%
hospital characteristics (continued) hospital region Northeast Midwest South West
296 648 1011 564
11.8 25.7 40.1 22.4
lus had the lowest average LOS (8 ± 14 days) and lowest total cost ($25,145 ± $36,498), but the highest CoPID ($5222 ± $3167). Patients under 1 year of age had higher mean total costs and LOSs than patients over 1 year of age (Table 3). Male, black, and Medicaid patients had mean higher total costs than female, white, and private insurance patients. Mean total costs and LOS were greater for patients admitted on the weekend ($70,984 and 27 days, respectively) than those admitted on a weekday ($46,308 and 17 days, respectively). Patients with eventual inpatient hospital death incurred more total costs and higher LOSs ($145,441 ± $150,128, and 41 ± 40 days, respectively). Patients treated in nonprofit children’s hospitals had higher total costs ($60,315 ± $92,637) than other types of hospitals. Also, patients who were treated in the West region had the highest total costs ($55,692 ± $75,957), with West and Midwest regions having shorter LOS than the South and Northeast regions. Except for those patients who died in the hospital (p = 0.54) during CSF shunt admission, all potential covariates with a p value < 0.2 were included in multivariate models. To discriminate the marginal effect of the covariates of
interest, such as number of RN FTEs per 1000 inpatient days, race, insurance, and zip-code income level, we compared the results of 3 models to observe the variation of coefficients and the change of the adjusted coefficient of determination (R2, Table 4). We found the parsimonious model (Model I) did not best represent the variation of CoPID (adjusted R2 = 0.29) compared with Model II (adjusted R2 = 0.30) and the final model (adjusted R2 = 0.32). Model I showed that nonprofit children’s hospitals have higher CoPID than government-owned hospitals. However, after adjusting for the number of RN FTEs per 1000 adjusted inpatient days, the differences of CoPID between government-owned and nonprofit children’s hospitals became nonsignificant. The final model had the highest adjusted R2 value (0.32). Children 1 year and older and those who were treated at a hospital with more RN FTEs had higher CoPIDs than children under 1 year and treated at hospitals with lower RN FTEs. Also, children who were admitted to the hospital on weekend days; had a hydrocephalus etiology of congenital hydrocephalus, IVH, myelomeningocele, meningitis, or trauma; had a higher number of chronic diseases; underwent a greater number of procedures; and were treated at nonprofit nonchildren’s hospitals or at investor-owned hospitals had significantly lower CoPIDs than the reference groups (Table 4). Also, Black children had significantly lower CoPID than White children. Patients with Medicaid had significantly higher CoPIDs than those with private insurance. Lastly, patients who lived in the highest income level neighborhood by zip codes had higher CoPIDs than patients from the lowest income level neighborhoods. All 3 multivariate models waived the risk of collinearity (VIF > 10). Average VIF in each model was smaller than 3 (the final model was less than 2), and the highest VIF was less than 5.
TABLE 2: Hospital cost and other continuous covariates Variables
Mean
SD
Median
Min
Max
LOS (days) total cost ($) CoPID ($) ln(CoPID)* no. of chronic conditions no. of RN FTEs per 1000 adjusted inpatient days no. of surgical procedures CoPID by hydrocephalus etiology ($) communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others
18.2 49,317 4249 8 3.1 5.8 4 5222 4939 4119 4461 2438 3346 2453 3682 4710
28.5 74,483 2837 0.6 2.1 1.8 4 3167 3023 2790 2727 1224 2179 1084 3004 2951
8.1 20,462 3397 8 3 5.5 2 4458 4068 3286 3674 2253 2811 2266 2823 3768
1 626 80 4 0 1.6 1 1347 545 676 1132 80 202 816 761 535
263 774,316 22,263 10 15 10.8 30 17,787 19,431 22,263 19,445 10,530 16,313 5794 20,463 17,079
* ln(CoPID) = natural log transformation of CoPID.
Neurosurg Focus / Volume 37 / November 2014
5
S. K. Lam et al. TABLE 3: Total cost and LOS among different groups Total Cost ($) Variable patient characteristics age (yrs) 0 1–20 sex female male race White Black Hispanic Asian and others unspecified insurance type private Medicaid others income level (%) 0–25 25–50 50–75 75–100 unspecified clinical characteristics admission at weekend yes no died yes no hydrocephalus etiology communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others hospital characteristics hospital type government owned nonprofit, non-children’s nonprofit, children’s nonprofit, children’s units in hospital private, investor-owned unspecified
LOS (days)
Mean ± SD
Median (range)
Mean ± SD
Median (range)
54,675 ± 81,103 41,165 ± 63,124
24,702 (626–774,316) 16,342 (1089–663,477)
22 ± 33 12 ± 19
10 (1–263) 4 (1–196)
45,448 ± 68,864 52,749 ± 79,488
19,403 (1003–774,316) 21,884 (626–699,750)
16 ± 27 20 ± 30
6 (1–263) 7 (1–200)
42,169 ± 65,012 69,243 ± 107,267 54,228 ± 70,651 55,853 ± 81,069 46,585 ± 70,647
17,135 (1003–572,395) 26,194 (626–774,316) 26,837 (3314–531,071) 25,785 (2984–465,403) 20,225 (4599–663,477)
16 ± 26 29 ± 39 19 ± 26 21 ± 32 16 ± 25
5 (1–200) 11 (1–263) 9 (1–168) 8 (1–173) 6 (1–140)
46,002 ± 73,666 52,450 ± 76,594 48,824 ± 69,013
17,800 (1003–663,477) 23,236 (626–774,316) 22,491 (4646–413,141)
15 ± 26 21 ± 31 17 ± 26
5 (1–200) 9 (1–263) 7 (1–118)
56,493 ± 91,045 45,702 ± 62,481 45,017 ± 69,173 48,133 ± 69,042 50,913 ± 55,905
22,233 (1003–774,316) 19,187 (626–494,339) 18,301 (3878–515,200) 20,645 (1089–475,324) 27,004 (3562–282,038)
22 ± 34 18 ± 28 16 ± 25 15 ± 24 19 ± 21
9 (1–263) 7 (1–178) 6 (1–173) 5 (1–186) 9 (1–77)
70,984 ± 93,960 46,308 ± 71,225
34,950 (4114–774,316) 18,855 (626–699,750)
27 ± 35 17 ± 27
13 (1–263) 6 (1–196)
145,441 ± 150,128 48,021 ± 72,368
95,608 (9156–699,750) 20,093 (626–774,316)
41 ± 40 18 ± 28
30 (1–148) 7 (1–263)
25,145 ± 36,498 35,938 ± 67,202 34,798 ± 52,458 68,618 ± 67,435 176,073 ± 130,844 39,851 ± 34,830 106,159 ± 78,458 72,723 ± 86,756 22,552 ± 31,290
13,495 (3878–220,734) 13,675 (1089–699,750) 16,686 (3314–494,339) 48,366 (2984–465,403) 148,486 (5263–774,316) 31,798 (1003–268,169) 107,098 (7495–405,547) 35,439 (3484–515,200) 12,554 (626–257,138)
8 ± 14 12 ± 24 14 ± 23 20 ± 20 76 ± 44 15 ± 14 45 ± 30 25 ± 27 8 ± 13
3 (1–97) 3 (1–200) 5 (1–186) 13 (1–142) 79 (1–263) 13 (1–99) 39 (5–126) 16 (1–116) 3 (1–85)
57,221 ± 88,195 47,620 ± 64,230 60,315 ± 92,637 39,928 ± 54,694 31,319 ± 41,571 49,138 ± 75,542
20,688 (3561–606,378) 22,434 (3442–423,516) 24,698 (4949–774,316) 18,361 (1445–414,924) 17,937 (626–184,413) 19,241 (2984–494,339)
21 ± 32 22 ± 30 18 ± 30 17 ± 26 16 ± 24 17 ± 28
8 (1–196) 10 (1–168) 6 (1–263) 7 (1–186) 8 (1–131) 5 (1–171) (continued)
6
Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization TABLE 3: Total cost and LOS among different groups (continued) Total Cost ($)
LOS (days)
Variable
Mean ± SD
Median (range)
Mean ± SD
Median (range)
hospital characteristics (continued) hospital region Northeast Midwest South West
50,914 ± 68,859 49,856 ± 77,638 45,187 ± 74,036 55,692 ± 75,957
20,931 (1445–423,516) 20,711 (4027–663,477) 18,621 (626–774,316) 23,790 (3802–475,324)
19 ± 30 17 ± 27 20 ± 31 17 ± 25
8 (1–196) 6 (1–200) 7 (1–263) 7 (1–168)
Discussion
Using KID, the largest all-payer national database for pediatric discharge data, we identified factors associated with costs in initial CSF shunt placement. We conducted a comparison of 3 different models to determine the best model of CSF shunt hospitalization cost. To our knowledge, this is the first such cost analysis study. The contributions of each factor associated with the hospitalization episodes’ costs were quantified in the model. We highlight the methodology used for this cost analysis. The study considers the complexity of the factors associated with inpatient costs, which may show some of the risk for bundling payment models. Cost Measurement
The association of LOS with total inpatient costs is well established. However, other factors contributing to inpatient costs remain unclear. As expected, our study found LOS was highly correlated with inpatient cost of shunt placement (correlation coefficient = 87.6%). If we only included LOS in the total cost model, the adjusted R2 could be as high as 82.16. Meanwhile, many potential determinants could impact both inpatient costs and LOS, such as hydrocephalus etiology, number of chronic conditions, and age. The estimation of conventional multivariate-regression modeled cost with LOS and all other possible covariates will be biased due to collinearity between LOS and other covariates. Alternatively, a simultaneous equations model system for both cost and LOS is an option. However, we could not find a strong instrumental variable that correlated with LOS and did not correlate with total costs. Therefore, we used CoPID to identify the other determinants of cost except for LOS. The results should be interpreted in the context of LOS. Socioeconomic factors were significant contributors in the final model in predicting cost of CSF shunt hospitalizations. Racial and socioeconomic disparity has been noted in surgical care and outcomes.6,36 Black children and Hispanic children had higher total cost of CSF shunt hospitalization, longer LOS, and lower CoPID when compared with White children, controlling for other factors such as hospital type, insurance, income, number of chronic conditions, and number of surgical procedures. The etiology of hydrocephalus was a significant predictor of cost associated with pediatric CSF shunt placement for total hospital charges and CoPID. Previous studNeurosurg Focus / Volume 37 / November 2014
ies have mentioned this relationship but have not attempted to study costs.20,25,38 Etiologies with long LOSs such as IVH and meningitis had high total costs but relatively low CoPID. This result suggests a longer length of lower cost care leading up to or associated with shunt placement, which describes the pattern of a perinatal admission with IVH of prematurity, or intravenous antibiotics with meningitis. Hydrocephalus etiology is a logical predictor of survival,16 which was found to be associated with CoPID and total cost. Total costs and CoPID varied with the type of hospital. The KID reports data on 4 types of hospitals: children’s nonprofit hospital, children’s unit in nonprofit (primarily adult) hospital, investor-owned hospital, and governmentowned hospital. Studies in other medical21 and surgical disciplines have used the KID to examine charges by hospital type, such as for pyloromyotomy,29 where Raval et al. found that children’s hospitals had lower complication rates and shorter LOSs, but higher charges than general hospitals (although lower charges than children’s units in general hospitals). We found 34% of CSF shunt patients at nonprofit children’s hospitals were transfers, compared with 20% at other hospitals. This may suggest patients treated in nonprofit children’s hospitals had higher clinical complexity. The National Association of Children’s Hospitals and Related Institutions recommends that “all children need children’s hospitals.”23 Because of the unique mission of children’s hospitals, higher costs have been reported. Freestanding children’s hospitals comprise approximately 1% of US hospitals, but account for 39% of pediatric admissions, 49% of pediatric inpatient days, and 59% of costs.21 Subspecialty care, laboratories, radiology, and ancillary services designed for children are housed in freestanding children’s hospitals, where they are recognized to be higher cost centers.21 Hospital location was also found to be a significant predictor of cost, as with other reported studies in the pediatric and adult neurosurgical literature.32 In keeping with our findings, other studies have shown that centers in the West and Midwest have significantly higher costs when compared with the Northeast.29,32 The effect of regional differences on spending is widely reported across specialties (http://www.dartmouthatlas.org).4 This variation may identify opportunities for minimizing the regional disparities and aiming utilization to be at the level of the lower cost regions. It would be important to assess whether this goal can be safely achieved. This study cannot assess this question, and further research is warranted. 7
S. K. Lam et al. TABLE 4: Comparison of cost models* Variable
Model I
Model II
Final Model
age (1–20 yrs) female admission on weekend day hydrocephalus etiology (ref: communicating hydrocephalus) obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others no. of chronic conditions no. of procedures hospital type (ref: government owned) nonprofit, non-children’s nonprofit, children’s nonprofit, children’s unit in a hospital investor-owned, private unspecified region (ref.: Northeast) Midwest South West no. of RN FTEs per 1000 adjusted inpatient days race (ref: White) Black Hispanic Asian and others unspecified insurance type (ref: private) Medicaid other income level (ref: 0%–25%) 25–50% 50–75% 75–100% unspecified constant (in the multivariate linear regression model) no. of cases R2
0.21 (0.00) 0.03 (0.19) −0.16 (0.00)
0.22 (0.00) 0.03 (0.21) −0.17 (0.00)
0.20 (0.00) 0.03 (0.18) −0.17 (0.00)
−0.04 (0.42) −0.17 (0.00) −0.07 (0.32) −0.35 (0.00) −0.28 (0.00) −0.43 (0.00) −0.23 (0.00) −0.12 (0.08) −0.02 (0.01) −0.03 (0.00)
−0.06 (0.28) −0.18 (0.00) −0.08 (0.21) −0.35 (0.00) −0.29 (0.00) −0.44 (0.00) −0.23 (0.00) −0.13 (0.05) −0.02 (0.00) −0.03 (0.00)
−0.05 (0.30) −0.18 (0.00) −0.09 (0.15) −0.34 (0.00) −0.29 (0.00) −0.44 (0.00) −0.21 (0.01) −0.13 (0.05) −0.02 (0.01) −0.03 (0.00)
−0.21 (0.00) 0.20 (0.00) −0.08 (0.12) −0.41 (0.00) 0.15 (0.15)
−0.17 (0.00) 0.12 (0.09) −0.08 (0.13) −0.36 (0.00) 0.14 (0.11)
−0.19 (0.00) 0.10 (0.12) −0.09 (0.06) −0.38 (0.00) 0.13 (0.13)
0.10 (0.18) −0.15 (0.03) 0.13 (0.08)
0.05 (0.52) −0.18 (0.01) 0.09 (0.21) 0.05 (0.01)
0.06 (0.44) −0.15 (0.03) 0.10 (0.18) 0.05 (0.01) −0.07 (0.04) 0.00 (0.90) −0.03 (0.47) 0.02 (0.71) 0.10 (0.00) 0.06 (0.26)
8.18 (0.00) 2519 0.29
7.98 (0.00) 2519 0.30
0.01 (0.84) 0.03 (0.42) 0.09 (0.00) −0.08 (0.37) 7.91 (0.00) 2519 0.32
* All data given as coefficient of the covariates in the regression model, or β (p value). Significant p values indicated in boldface. ref = reference group.
Other hospital operational factors also came into consideration. Higher RN FTEs contribute to CoPID in our cost model; this may reflect acuity of care and staffing requirements. The optimal nurse staffing ratio for providing quality care has been debated.8,9,18 Studies from the 8
US and Europe demonstrate that nurse staffing correlates with adult inpatient mortality outcomes,1,15,31 and higher ratios may also increase costs.9,14,33 Weekend admissions had higher CoPID than weekday admissions. Both delay in care and delay in discharges have been reported in Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization relation with weekend admissions across disciplines.11,22 These ideas may motivate a closer look in staffing, operations, and efficiency. Study Limitations
Limitations of this study include the inherent nature of KID data. As with any database, coding inaccuracies may exist, and we are unable to independently verify diagnoses or procedures by chart review. There are ongoing validation studies in pediatric neurosurgery. Details about clinical indications, causality, and disease severity are not known. Specific complications and outcomes are not included in this initial study, although further work is much needed to examine quality, clinical outcomes, return to the emergency room, readmission, and reoperation. While bed assignment is important to consider in costs, due to limitations in KID data, we are unable to identify or separate the total LOS in each location (intensive care unit vs ward). Cost-to-charge ratios help in converting charge to cost. This is a surrogate for cost, but not reflective of actual cost data. The idea of the cost model for pediatric CSF shunting is in its nascent stages. The KID database does not have the level of detail in variables required to build a predictive plug-and-play model and equation. Further work with more databases is needed. In the future, predictive costing models will likely be possible and may become the norm. Covering risk in such costing models may depend on volume and efficiency based on iteration of quality. The advantage may be with higher volume centers. There is a potential application for more delineation in bundle pricing as we further understand the differences in hydrocephalus subgroups. As such, there may be even greater potential with the advent of ICD-10 and its new level of detail. Although our models cannot account for all facets of variations in CSF shunt-related costs due to limitations in the type and granularity of variables in KID, this study shows a direction that is needed in neurosurgery subspecialty care: the ability to engage in health care economics conversations. This type of model can be a starting tool for examining areas for improvement, minimizing preventable variance in costs, creating data-driven efficiency optimization, and recognizing populations at higher risk for resource-intensive hospitalizations. We need to develop further understanding into the complexity of factors associated with costs and the implications of bundling payment models.
Conclusions
A patient’s socioeconomic status, such as race, income level, and insurance, in addition to hospital-related factors such as number of hospital RN FTEs, hospital type, and US region, could affect the costs of initial CSF shunt placement in addition to clinical factors such as hydrocephalus etiology and LOS. To create a cost model of initial CSF shunt placement in the pediatrics population, consideration of such nonclinical factors is warranted. Neurosurg Focus / Volume 37 / November 2014
Disclosure The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Author contributions to the study and manuscript preparation include the following. Conception and design: Pan, Lam. Acquisition of data: Pan. Analysis and interpretation of data: Pan. Drafting the article: Pan, Lam, Srinivasan. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Pan. Statistical analysis: Pan. Administrative/technical/ material support: Pan, Lam, Luerssen. Study supervision: Pan, Lam. References 1. Aiken LH, Sloane DM, Bruyneel L, Van den Heede K, Griffiths P, Busse R, et al: Nurse staffing and education and hospital mortality in nine European countries: a retrospective observational study. Lancet 383:1824–1830, 2014 2. Arizona Health Care Cost Containment System: AHCCCS Transition to DRG-Based Payment. (http://www.azahcccs. gov/commercial/ProviderBilling/DRGBasedPayments.aspx) [Accessed September 3, 2014] 3. Averill RF, McCullough EC, Goldfield N, Hughes JS, Bonazelli J, Bentley L, et al: 3MTM APR DRG Classification System, Version 31.0 (effective 10/01/2013). Healthcare Cost and Utilization Project. (http://www.hcup-us.ahrq.gov/ db/nation/nis/grp031_aprdrg_meth_ovrview.pdf) [Accessed September 1, 2014] 4. Bekelis K, Missios S, Labropoulos N: Cerebral aneurysm coiling: a predictive model of hospitalization cost. J Neurointerv Surg [epub ahead of print], 2014 5. Buchanan CC, Hernandez EA, Anderson JM, Dye JA, Leung M, Buxey F, et al: Analysis of 30-day readmissions among neurosurgical patients: surgical complication avoidance as key to quality improvement. Clinical article. J Neurosurg 121: 170–175, 2014 6. Chern JJ, Bookland M, Tejedor-Sojo J, Riley J, Shoja MM, Tubbs RS, et al: Return to system within 30 days of discharge following pediatric shunt surgery. Clinical article. J Neurosurg Pediatr 13:525–531, 2014 7. de Brantes F, Rosenthal MB, Painter M: Building a bridge from fragmentation to accountability—the Prometheus Payment model. N Engl J Med 361:1033–1036, 2009 8. Duffin C: Nurse-to-patient ratios must increase to improve safety. Nurs Older People 24:6–7, 2012 9. Garretson S: Nurse to patient ratios in American health care. Nurs Stand 19:33–37, 2004 10. Gever M: Improving the quality of care: the continuing debate over nurse-patient ratios. National Conference of State Legislatures. (http://www.ncsl.org/print/health/shn/shn535.pdf) [Accessed September 3, 2014] 11. Goldstein SD, Papandria DJ, Aboagye J, Salazar JH, Van Arendonk K, Al-Omar K, et al: The “weekend effect” in pediatric surgery—increased mortality for children undergoing urgent surgery during the weekend. J Pediatr Surg 49:1087– 1091, 2014 12. Healthcare Cost Utilization Project: Cost-to-Charge Ratio Files: 2009 Kids’ Inpatient Database (KID) User Guide. (http://www.hcup-us.ahrq.gov/db/state/CCR2009KIDUserGuide.pdf) [Accessed September 3, 2014] 13. Healthcare Financial Management Association: Medicaid DRGs. Get ready, ‘cuz here they come! HFMA Florida Chapter. (http://www.floridahfma.org/presentations.lib/items/medi caid-drgs/MEDICAID%20DRGS.pdf) [Accessed September 1, 2014] 14. Howell AM, Panesar SS, Burns EM, Donaldson LJ, Darzi A: Reducing the burden of surgical harm: a systematic review
9
S. K. Lam et al. of the interventions used to reduce adverse events in surgery. Ann Surg 259:630–641, 2014 15. Kane RL, Shamliyan TA, Mueller C, Duval S, Wilt TJ: The association of registered nurse staffing levels and patient outcomes: systematic review and meta-analysis. Med Care 45: 1195–1204, 2007 16. Kulkarni AV, Riva-Cambrin J, Butler J, Browd SR, Drake JM, Holubkov R, et al: Outcomes of CSF shunting in children: comparison of Hydrocephalus Clinical Research Network cohort with historical controls. Clinical article. J Neurosurg Pediatr 12:334–338, 2013 17. Lave JR, Lave LB: Hospital cost functions. Annu Rev Public Health 5:193–213, 1984 18. Lewis KK, New England Public Policy Center: Nurse-to-patient ratios: research and reality. Issue Brief (Mass Health Policy Forum) (25):1–19, 2005 19. Manning WG, Mullahy J: Estimating log models: to transform or not to transform? J Health Econ 20:461–494, 2001 20. McGirt MJ, Leveque JC, Wellons JC III, Villavicencio AT, Hopkins JS, Fuchs HE, et al: Cerebrospinal fluid shunt survival and etiology of failures: a seven-year institutional experience. Pediatr Neurosurg 36:248–255, 2002 21. Merenstein D, Egleston B, Diener-West M: Lengths of stay and costs associated with children’s hospitals. Pediatrics 115: 839–844, 2005 22. Nandyala SV, Marquez-Lara A, Fineberg SJ, Schmitt DR, Singh K: Comparison of perioperative outcomes and cost of spinal fusion for cervical trauma: weekday versus weekend admissions. Spine (Phila Pa 1976) 38:2178–2183, 2013 23. National Association of Children’s Hospitals and Related Institutions: All Children Need Children’s Hospitals. (http:// www.upstate.edu/gch/education/allchildren.pdf) [Accessed September 3, 2014] 24. Noyes K, Liu H, Temkin-Greener H: Medicare capitation model, functional status, and multiple comorbidities: model accuracy. Am J Manag Care 14:679–690, 2008 25. Patwardhan RV, Nanda A: Implanted ventricular shunts in the United States: the billion-dollar-a-year cost of hydrocephalus treatment. Neurosurgery 56:139–145, 2005 26. Polverejan E, Gardiner JC, Bradley CJ, Holmes-Rovner M, Rovner D: Estimating mean hospital cost as a function of length of stay and patient characteristics. Health Econ 12: 935–947, 2003 27. Pope GC, Kautter J, Ellis RP, Ash AS, Ayanian JZ, Lezzoni LI, et al: Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev 25:119– 141, 2004 28. Quinn K: New directions in Medicaid payment for hospital care. Health Aff (Millwood) 27:269–280, 2008 29. Raval MV, Cohen ME, Barsness KA, Bentrem DJ, Phillips JD, Reynolds M: Does hospital type affect pyloromyotomy outcomes? Analysis of the Kids’ Inpatient Database. Surgery 148:411–419, 2010
10
30. Rolston JD, Han SJ, Lau CY, Berger MS, Parsa AT: Frequency and predictors of complications in neurological surgery: national trends from 2006 to 2011. Clinical article. J Neurosurg 120:736–745, 2014 31. Sasichay-Akkadechanunt T, Scalzi CC, Jawad AF: The relationship between nurse staffing and patient outcomes. J Nurs Adm 33:478–485, 2003 32. Sharma M, Sonig A, Ambekar S, Nanda A: Discharge dispositions, complications, and costs of hospitalization in spinal cord tumor surgery: analysis of data from the United States Nationwide Inpatient Sample, 2003–2010. Clinical article. J Neurosurg Spine 20:125–141, 2014 33. Sherenian M, Profit J, Schmidt B, Suh S, Xiao R, Zupancic JA, et al: Nurse-to-patient ratios and neonatal outcomes: a brief systematic review. Neonatology 104:179–183, 2013 34. Simon TD, Hall M, Riva-Cambrin J, Albert JE, Jeffries HE, Lafleur B, et al: Infection rates following initial cerebrospinal fluid shunt placement across pediatric hospitals in the United States. Clinical article. J Neurosurg Pediatr 4:156–165, 2009 35. Simon TD, Riva-Cambrin J, Srivastava R, Bratton SL, Dean JM, Kestle JR: Hospital care for children with hydrocephalus in the United States: utilization, charges, comorbidities, and deaths. J Neurosurg Pediatr 1:131–137, 2008 36. Stone ML, LaPar DJ, Mulloy DP, Rasmussen SK, Kane BJ, McGahren ED, et al: Primary payer status is significantly associated with postoperative mortality, morbidity, and hospital resource utilization in pediatric surgical patients within the United States. J Pediatr Surg 48:81–87, 2013 37. Texas Medicaid & Healthcare Partnership: All Patient Refined – Diagnosis Related Groups (APR-DRG). (http://www. tmhp.com/Pages/Medicaid/Hospital_APR-DRG.aspx) [Accessed September 3, 2014] 38. Tuli S, Drake J, Lawless J, Wigg M, Lamberti-Pasculli M: Risk factors for repeated cerebrospinal shunt failures in pediatric patients with hydrocephalus. J Neurosurg 92:31–38, 2000 39. White C, Reschovsky JD, Bond AM: Inpatient hospital prices drive spending variation for episodes of care for privately insured patients. NIHCR Research Brief No. 14. National Institute for Health Care Reform. (http://www.nihcr.org/ Episode-Spending-Variation) [Accessed September 1, 2014] 40. Wrubel DM, Riemenschneider KJ, Braender C, Miller BA, Hirsh DA, Reisner A, et al: Return to system within 30 days of pediatric neurosurgery. Clinical article. J Neurosurg Pediatr 13:216–221, 2014 Manuscript submitted July 16, 2014. Accepted August 19, 2014. Please include this information when citing this paper: DOI: 10.3171/2014.8.FOCUS14454. Address correspondence to: I-Wen Pan, Ph.D., Department of Neurosurgery, Texas Children’s Hospital, 6701 Fannin St., CCC Ste. 1230-01, Houston, TX 77030. email: [email protected].
Neurosurg Focus / Volume 37 / November 2014
Cerebrospinal fluid shunt placement in the pediatric population: a model of hospitalization cost Sandi K. Lam, M.D., M.B.A.,1,2 Visish M. Srinivasan, M.D.,1,2 Thomas G. Luerssen, M.D.,1,2 and I-Wen Pan, Ph.D.1,2 Division of Pediatric Neurosurgery, Texas Children’s Hospital; and 2Department of Neurosurgery, Baylor College of Medicine, Houston, Texas
1
Object. There have been no large-scale analyses on cost drivers in CSF shunt surgery for the treatment of pediatric hydrocephalus. The objective of this study was to develop a cost model for hospitalization costs in pediatric CSF shunt surgery and to examine risk factors for increased costs. Methods. Data were extracted from the Kids’ Inpatient Database (KID) of the Healthcare Cost and Utilization Project. Children with initial CSF shunt placement in the 2009 KID were examined. Patient charge was converted to cost using a cost-to-charge ratio. The factors associated with costs of CSF shunt hospitalizations were examined, including patient demographics, hospital characteristics, and clinical data. The natural log transformation of cost per inpatient day (CoPID) was analyzed. Three multivariate linear regression models were used to characterize the cost. Variance inflation factor was used to identify multicollinearity for each model. Results. A total of 2519 patients met the inclusion criteria and were included in study. Average cost and length of stay (LOS) for initial shunt placement were $49,317 ± $74,483 (US) and 18.2 ± 28.5 days, respectively. Cost per inpatient day was $4249 ± $2837 (median $3397, range $80–$22,263). The average number of registered nurse (RN) full-time equivalents (FTEs) per 1000 adjusted inpatient days was 5.8 (range 1.6–10.8). The final model had the highest adjusted coefficient of determination (R2 = 0.32) and was determined to be the best among 3 models. The final model showed that child age, hydrocephalus etiology, weekend admission, number of chronic diseases, hospital type, number of RN FTEs per 1000 adjusted inpatient days, number of procedures, race, insurance type, income level, and hospital regions were associated with CoPID. Conclusions. A patient’s socioeconomic status, such as race, income level, and insurance, in addition to hospitalrelated factors such as number of hospital RN FTEs, hospital type, and US region, could affect the costs of initial CSF shunt placement, in addition to clinical factors such as hydrocephalus origin and LOS. To create a cost model of initial CSF shunt placement in the pediatric population, consideration of such nonclinical factors may be warranted. (http://thejns.org/doi/abs/10.3171/2014.8.FOCUS14454)
Key Words • pediatric • ventriculoperitoneal shunt • cost analysis • cerebrospinal fluid shunt • hydrocephalus
P
lacement of ventriculoperitoneal shunts is one of the most common procedures in pediatric neurosurgery and is a significant health care burden. Shunt failure and shunt infections are common reasons for readmissions and revision surgeries.5 Total hospital charges for treatment of pediatric hydrocephalus in the US are estimated to be between $1.4 and $2.0 billion each year.35 Shunt admissions account for 0.6% of all pediatric hospital admissions, but consist of 3.1% of all pediatric hospital charges,35 suggesting that children with hydrocephalus use a disproportionate number of hospital days and resources in the US.
Abbreviations used in this paper: CoPID = cost per inpatient day; FTE = full-time equivalent; ICD = International Classification of Diseases; IVH = intraventricular hemorrhage; KID = Kids’ Inpatient Database; LOS = length of stay; RN = registered nurse; VIF = variance inflation factor.
Neurosurg Focus / Volume 37 / November 2014
Studies of costs in adult and pediatric neurosurgery also show that CSF shunt-related problems are among the most common causes for hospital readmissions within 30 days.30 In pediatric CSF shunt surgery, 5,6 reoperations and readmissions have come under scrutiny.40 Several patientor hospital-level factors have been reported to be associated with inpatient costs, such as length of stay (LOS), age, sex, comorbidities, case mix, and use of facilities and services.4,17,26 Hospital LOS is often highly correlated with inpatient costs for patients treated using inpatient surgical procedures.4,26 There is increasing pressure from public and private payers on medical service providers for cost reduction. Payment strategies are shifting from traditional fee-forservice to bundled schemes. The bundled payment of an episode begins with hospital admission to the end of inpatient stay; all expenses are included, such as facility, professional (i.e., services provided by health profes1
S. K. Lam et al. sionals), and prescription drug expenses. This idea has gained traction with Medicare and private plans and has been applied to adult inpatient services for years, as in the Centers for Medicare and Medicaid Services diagnosisrelated groups and Medicare Severity diagnosis-related groups.3,7,39 Risk-adjusted models and hierarchical coexisting conditions models have been adapted by the Centers for Medicare and Medicaid Services for adjusting payment of Medicare services.27 However, there is no existing risk adjustment for pediatric payment models. Furthermore, studies have shown that existing hierarchical coexisting conditions models may underpredict the actual costs associated with treating adult patients with multiple comorbid conditions.24 Previous reports show mean charges for initial CSF shunt placement increased from $28,370 to $40,260 in 2003 dollars between 1997 and 2003, with a stable mean LOS (6.8 days).35 From the payer and societal perspectives, it is important to investigate factors associated with the inpatient costs for CSF shunt placement to characterize the economic burden and target areas for containment and improvement. There are currently no large-scale analyses on cost drivers in shunt surgery for hydrocephalus treatment. We aim to use the national Kids’ Inpatient Database (KID) to develop a cost model of pediatric CSF shunt surgery hospitalization costs and thus examine risk factors for increased costs. In doing so, this study considers the complexity of the factors associated with inpatient costs, which may show some of the risk in bundling payment models.
Methods Data Source
Data were extracted from the KID, one of a family of administrative databases developed by the Healthcare Cost and Utilization Project sponsored by the US Agency for Healthcare Research and Quality. Patient discharge data includes patient demographics, admission type and source, diagnostic and procedural codes from the International Classification of Diseases (ICD) Ninth Revision, LOS, disposition, and payer data. The number of participating states was 44 in 2009, with data from 4100 hospitals. The KID contains information from 2–3 million pediatric discharges, weighted to represent 6.5–7.5 million national discharges. The basic unit of analysis is a patient discharge, rather than an individual patient.
Study Sample
Children (0–20 years old) with initial CSF shunt placement in the 2009 KID were selected as follows: hydrocephalus-related admission with any procedure code for intracranial ventricular shunt placement (02.31–02.35), excluding those with any procedure code for replacement of ventricular shunt (02.42, 54.95) or removal of ventricular shunt (02.43), and any diagnosis code for shunt malfunction (996.2) or CSF shunt infection (996.63).34,35 Patient records with missing insurance status, age, sex, or cost-to-charge ratio at their treatment hospital were also excluded.
2
Cost Estimation
The KID provides two types of cost-to-charge ratios. Patient charges were converted to costs by either hospitalspecific cost-to-charge ratio (all-payer inpatient cost-tocharge ratio) if available, or the group-weighted average cost-to-charge ratio in the small minority who were missing all-payer inpatient cost-to-charge ratio values.12 To remove the effect of LOS on total costs, we used cost per inpatient day (CoPID) instead of total inpatient cost to investigate factors associated with costs in the study. In addition, we used the natural logarithmic (log) transformation of CoPID instead of the original form. The original CoPID was skewed to the right and not normally distributed (Fig. 1 upper) and thus not appropriate for a regression model. Therefore, we used the natural logarithmic transformation of CoPID, which fits the normal distribution almost perfectly (Fig. 1 lower) and fulfills the assumptions of the regression model. Use of log transformation of such data are reported to produce more precise and unbiased estimation.19 Cost was characterized by nonnegative measurements of the outcomes and a positively skewed empirical distribution. While diseases with specific ictus (such as stroke, acute cardiac events, and ruptured aneurysms) may incur costs at the beginning of hospitalization and necessitate examining costs on certain days, the pediatric hydrocephalus population may incur costs over a different time frame. For instance, CSF shunt placement may occur after a long neonatal intensive care unit stay, after a period of observation after myelomeningocele closure, or soon after presentation with an obstructive tumor. This variability is an important distinction to many conditions studied in adult populations, and supports the use of CoPID rather than costs on specific hospital days.
Patient Demographics and Clinical Characteristics
The following patient demographics associated with CoPID were examined: age (0–1 year and > 1 year), sex (male/female), insurance status (private, Medicaid, and other), race (White, Black, Hispanic, Asian and others, and unspecified), and median household income in quartiles for a patient’s zip code of residence (zip-code income level). Patient were assigned into 1 of 9 hydrocephalus etiologies based on ICD-9 Clinical Modification diagnosis codes as follows, based on previously published methodologies:34,35 myelomeningocele (653.7, 655.0, 741.xx), intraventricular hemorrhage (IVH; 772.1x), congenital hydrocephalus (742.3), meningitis (320–322, 326), CNS tumor (191.xx- 194.xx, 198.3, 198.4, 225.0–225.2, 225.8– 225.9, 227.4, 237.0–237.1, 237.5–237.7, 239.6–239.7), trauma (767.4, 851.xx-854.xx, 995.55), communicating hydrocephalus (331.3), or obstructive hydrocephalus (331.4). The number of chronic conditions, weekend or weekday admission, and mortality were also variables included in the analysis. Hospital Characteristics
Hospital characteristics used in the study included region, number of registered nurse (RN) full-time equivalents (FTEs) per 1000 adjusted inpatient days (RN FTEs), Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization
Fig. 1. Distributions for original CoPID (upper), which is not normally distributed, and the natural log transformation of CoPID (Ln_CoPID; lower), which fits a normal distribution.
and hospital type. Hospital type was categorized into government owned, nonprofit non-children’s hospitals, nonprofit children’s hospitals, nonprofit children’s units in a hospital, investor-owned private hospitals, and unspecified. Teaching status, hospital size, and hospital volume were not included in the analysis due to high collinearity with either hospital type or RN FTEs. Three-Model Comparison and Factor Selection
Several state Medicaid programs including Texas,
Neurosurg Focus / Volume 37 / November 2014
Arizona, and Florida currently reimburse or plan to reimburse using a prospective payment and pricing system based on all patient-refined diagnosis-related groups.2,13,37 All patient-refined diagnosis-related groups are an expansion of the diagnosis-related groups to include non-Medicare populations such as pediatric patients with adjusted severity of illness.3,28 The data elements used by the all patient-refined diagnosis-related groups include principal diagnosis, secondary diagnosis, procedures, age, sex, and discharge disposition. The initial model (Model I) was 3
S. K. Lam et al. based on this concept with consideration of geographical variation (http://www.dartmouthatlas.org) and hospital types.21 To build upon Model I, hospital operational factors such as RN FTEs were added into Model II. Nurse-to-patients ratios10 are reported to increase costs but to improve quality of care. Finally, nonclinical socioeconomic covariates, such as race, insurance status, and income level, were taken into consideration for investigating their impacts on costs in Model 3. The study sample includes all elective (69%) and nonelective (31%) cases. Statistical Analysis
Simple and multivariate linear regression models were used to build the cost model. Covariates with a p value > 0.2 in simple linear regression were excluded from the multivariate regression model. We compared 3 different models to understand the effect of additional covariates on the cost model; from there, we selected the final model. First, we modeled cost with age, sex, weekend admission, hydrocephalus etiology, number of chronic conditions, number of surgical procedures, hospital type, and region. For the next model, the number of RN FTEs per 1000 adjusted inpatient days was added into the list of variables. Finally, for the third model, we added race, insurance status, and zip-code income level in addition to the variables used in the first 2 models. Variance inflation factor (VIF) was used to identify multicollinearity for the multivariate regression model. No VIF greater than 10 was accepted. All statistical analyses were performed using SAS (version 9.3, SAS Institute, Inc.) and Stata (version 13, StataCorp). Test results with a p value < 0.05 were considered statistically significant.
Results
A total of 2519 patients were included in this study. More than half of the patients (61.1%) were under 1 year of age (Table 1). Payers included Medicaid (50.1%) and private insurers (43.4%). Only 1.4% of patients died during the CSF shunt hospital admission. The most common hydrocephalus etiologies were obstructive hydrocephalus (33.1%), congenital hydrocephalus (20.0%), myelomeningocele (14.9%), tumor (8.7%), and IVH (6.8%). The most common hospital type was children’s unit in a nonprofit hospital (33.6%), followed by nonprofit children’s hospital (26.3%) and government-owned hospital (12.4%). Average cost and LOS for initial shunt placement were $49,317 ± $74,483 and 18.2 ± 28.5 days, respectively. Average CoPID was $4249 ± $2837 (median $3397, range $80–$22,263). Average number of RN FTEs per 1000 adjusted inpatient days was 5.8 (range 1.6–10.8 RN FTEs; Table 2). By hydrocephalus etiology, IVH and patients with meningitis had the longest average LOSs (76 ± 44 days and 45 ± 30 days, respectively) and total costs were $176,073 ± $130,844 and $106,159 ± $78,458, respectively, although their CoPIDs were the lowest and the second lowest ($2438 ± $1224, and $2453 ± $1084, respectively; Tables 2 and 3). Conversely, communicating hydrocepha4
TABLE 1: Patient- and hospital-level characteristics Variable patient characteristics age (yrs) 0 1–20 sex female male race White Black Hispanic Asian & others unspecified insurance type private Medicaid others income level (%) 0–25 25–50 50–75 75–100 unspecified clinical characteristics admission at weekend yes no death during hospital admission yes no hydrocephalus etiology communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others hospital characteristics hospital type government nonprofit, non-children’s nonprofit, children’s nonprofit, children’s unit in a hospital private, investor-owned unspecified
No. of Patients
%
1538 981
61.1 38.9
1151 1368
45.7 54.3
1129 294 457 211 428
44.8 11.7 18.1 8.4 17.0
1094 1261 164
43.4 50.1 6.5
771 657 563 481 47
30.6 26.1 22.4 19.1 1.9
317 2202
12.6 87.4
36 2483
1.4 98.6
137 833 503 220 171 376 40 75 164
5.4 33.1 20.0 8.7 6.8 14.9 1.6 3.0 6.5
312 278 663 847 57 362
12.4 11.0 26.3 33.6 2.3 14.4 (continued)
Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization TABLE 1: Patient- and hospital-level characteristics (continued) Variable
No. of Patients
%
hospital characteristics (continued) hospital region Northeast Midwest South West
296 648 1011 564
11.8 25.7 40.1 22.4
lus had the lowest average LOS (8 ± 14 days) and lowest total cost ($25,145 ± $36,498), but the highest CoPID ($5222 ± $3167). Patients under 1 year of age had higher mean total costs and LOSs than patients over 1 year of age (Table 3). Male, black, and Medicaid patients had mean higher total costs than female, white, and private insurance patients. Mean total costs and LOS were greater for patients admitted on the weekend ($70,984 and 27 days, respectively) than those admitted on a weekday ($46,308 and 17 days, respectively). Patients with eventual inpatient hospital death incurred more total costs and higher LOSs ($145,441 ± $150,128, and 41 ± 40 days, respectively). Patients treated in nonprofit children’s hospitals had higher total costs ($60,315 ± $92,637) than other types of hospitals. Also, patients who were treated in the West region had the highest total costs ($55,692 ± $75,957), with West and Midwest regions having shorter LOS than the South and Northeast regions. Except for those patients who died in the hospital (p = 0.54) during CSF shunt admission, all potential covariates with a p value < 0.2 were included in multivariate models. To discriminate the marginal effect of the covariates of
interest, such as number of RN FTEs per 1000 inpatient days, race, insurance, and zip-code income level, we compared the results of 3 models to observe the variation of coefficients and the change of the adjusted coefficient of determination (R2, Table 4). We found the parsimonious model (Model I) did not best represent the variation of CoPID (adjusted R2 = 0.29) compared with Model II (adjusted R2 = 0.30) and the final model (adjusted R2 = 0.32). Model I showed that nonprofit children’s hospitals have higher CoPID than government-owned hospitals. However, after adjusting for the number of RN FTEs per 1000 adjusted inpatient days, the differences of CoPID between government-owned and nonprofit children’s hospitals became nonsignificant. The final model had the highest adjusted R2 value (0.32). Children 1 year and older and those who were treated at a hospital with more RN FTEs had higher CoPIDs than children under 1 year and treated at hospitals with lower RN FTEs. Also, children who were admitted to the hospital on weekend days; had a hydrocephalus etiology of congenital hydrocephalus, IVH, myelomeningocele, meningitis, or trauma; had a higher number of chronic diseases; underwent a greater number of procedures; and were treated at nonprofit nonchildren’s hospitals or at investor-owned hospitals had significantly lower CoPIDs than the reference groups (Table 4). Also, Black children had significantly lower CoPID than White children. Patients with Medicaid had significantly higher CoPIDs than those with private insurance. Lastly, patients who lived in the highest income level neighborhood by zip codes had higher CoPIDs than patients from the lowest income level neighborhoods. All 3 multivariate models waived the risk of collinearity (VIF > 10). Average VIF in each model was smaller than 3 (the final model was less than 2), and the highest VIF was less than 5.
TABLE 2: Hospital cost and other continuous covariates Variables
Mean
SD
Median
Min
Max
LOS (days) total cost ($) CoPID ($) ln(CoPID)* no. of chronic conditions no. of RN FTEs per 1000 adjusted inpatient days no. of surgical procedures CoPID by hydrocephalus etiology ($) communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others
18.2 49,317 4249 8 3.1 5.8 4 5222 4939 4119 4461 2438 3346 2453 3682 4710
28.5 74,483 2837 0.6 2.1 1.8 4 3167 3023 2790 2727 1224 2179 1084 3004 2951
8.1 20,462 3397 8 3 5.5 2 4458 4068 3286 3674 2253 2811 2266 2823 3768
1 626 80 4 0 1.6 1 1347 545 676 1132 80 202 816 761 535
263 774,316 22,263 10 15 10.8 30 17,787 19,431 22,263 19,445 10,530 16,313 5794 20,463 17,079
* ln(CoPID) = natural log transformation of CoPID.
Neurosurg Focus / Volume 37 / November 2014
5
S. K. Lam et al. TABLE 3: Total cost and LOS among different groups Total Cost ($) Variable patient characteristics age (yrs) 0 1–20 sex female male race White Black Hispanic Asian and others unspecified insurance type private Medicaid others income level (%) 0–25 25–50 50–75 75–100 unspecified clinical characteristics admission at weekend yes no died yes no hydrocephalus etiology communicating hydrocephalus obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others hospital characteristics hospital type government owned nonprofit, non-children’s nonprofit, children’s nonprofit, children’s units in hospital private, investor-owned unspecified
LOS (days)
Mean ± SD
Median (range)
Mean ± SD
Median (range)
54,675 ± 81,103 41,165 ± 63,124
24,702 (626–774,316) 16,342 (1089–663,477)
22 ± 33 12 ± 19
10 (1–263) 4 (1–196)
45,448 ± 68,864 52,749 ± 79,488
19,403 (1003–774,316) 21,884 (626–699,750)
16 ± 27 20 ± 30
6 (1–263) 7 (1–200)
42,169 ± 65,012 69,243 ± 107,267 54,228 ± 70,651 55,853 ± 81,069 46,585 ± 70,647
17,135 (1003–572,395) 26,194 (626–774,316) 26,837 (3314–531,071) 25,785 (2984–465,403) 20,225 (4599–663,477)
16 ± 26 29 ± 39 19 ± 26 21 ± 32 16 ± 25
5 (1–200) 11 (1–263) 9 (1–168) 8 (1–173) 6 (1–140)
46,002 ± 73,666 52,450 ± 76,594 48,824 ± 69,013
17,800 (1003–663,477) 23,236 (626–774,316) 22,491 (4646–413,141)
15 ± 26 21 ± 31 17 ± 26
5 (1–200) 9 (1–263) 7 (1–118)
56,493 ± 91,045 45,702 ± 62,481 45,017 ± 69,173 48,133 ± 69,042 50,913 ± 55,905
22,233 (1003–774,316) 19,187 (626–494,339) 18,301 (3878–515,200) 20,645 (1089–475,324) 27,004 (3562–282,038)
22 ± 34 18 ± 28 16 ± 25 15 ± 24 19 ± 21
9 (1–263) 7 (1–178) 6 (1–173) 5 (1–186) 9 (1–77)
70,984 ± 93,960 46,308 ± 71,225
34,950 (4114–774,316) 18,855 (626–699,750)
27 ± 35 17 ± 27
13 (1–263) 6 (1–196)
145,441 ± 150,128 48,021 ± 72,368
95,608 (9156–699,750) 20,093 (626–774,316)
41 ± 40 18 ± 28
30 (1–148) 7 (1–263)
25,145 ± 36,498 35,938 ± 67,202 34,798 ± 52,458 68,618 ± 67,435 176,073 ± 130,844 39,851 ± 34,830 106,159 ± 78,458 72,723 ± 86,756 22,552 ± 31,290
13,495 (3878–220,734) 13,675 (1089–699,750) 16,686 (3314–494,339) 48,366 (2984–465,403) 148,486 (5263–774,316) 31,798 (1003–268,169) 107,098 (7495–405,547) 35,439 (3484–515,200) 12,554 (626–257,138)
8 ± 14 12 ± 24 14 ± 23 20 ± 20 76 ± 44 15 ± 14 45 ± 30 25 ± 27 8 ± 13
3 (1–97) 3 (1–200) 5 (1–186) 13 (1–142) 79 (1–263) 13 (1–99) 39 (5–126) 16 (1–116) 3 (1–85)
57,221 ± 88,195 47,620 ± 64,230 60,315 ± 92,637 39,928 ± 54,694 31,319 ± 41,571 49,138 ± 75,542
20,688 (3561–606,378) 22,434 (3442–423,516) 24,698 (4949–774,316) 18,361 (1445–414,924) 17,937 (626–184,413) 19,241 (2984–494,339)
21 ± 32 22 ± 30 18 ± 30 17 ± 26 16 ± 24 17 ± 28
8 (1–196) 10 (1–168) 6 (1–263) 7 (1–186) 8 (1–131) 5 (1–171) (continued)
6
Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization TABLE 3: Total cost and LOS among different groups (continued) Total Cost ($)
LOS (days)
Variable
Mean ± SD
Median (range)
Mean ± SD
Median (range)
hospital characteristics (continued) hospital region Northeast Midwest South West
50,914 ± 68,859 49,856 ± 77,638 45,187 ± 74,036 55,692 ± 75,957
20,931 (1445–423,516) 20,711 (4027–663,477) 18,621 (626–774,316) 23,790 (3802–475,324)
19 ± 30 17 ± 27 20 ± 31 17 ± 25
8 (1–196) 6 (1–200) 7 (1–263) 7 (1–168)
Discussion
Using KID, the largest all-payer national database for pediatric discharge data, we identified factors associated with costs in initial CSF shunt placement. We conducted a comparison of 3 different models to determine the best model of CSF shunt hospitalization cost. To our knowledge, this is the first such cost analysis study. The contributions of each factor associated with the hospitalization episodes’ costs were quantified in the model. We highlight the methodology used for this cost analysis. The study considers the complexity of the factors associated with inpatient costs, which may show some of the risk for bundling payment models. Cost Measurement
The association of LOS with total inpatient costs is well established. However, other factors contributing to inpatient costs remain unclear. As expected, our study found LOS was highly correlated with inpatient cost of shunt placement (correlation coefficient = 87.6%). If we only included LOS in the total cost model, the adjusted R2 could be as high as 82.16. Meanwhile, many potential determinants could impact both inpatient costs and LOS, such as hydrocephalus etiology, number of chronic conditions, and age. The estimation of conventional multivariate-regression modeled cost with LOS and all other possible covariates will be biased due to collinearity between LOS and other covariates. Alternatively, a simultaneous equations model system for both cost and LOS is an option. However, we could not find a strong instrumental variable that correlated with LOS and did not correlate with total costs. Therefore, we used CoPID to identify the other determinants of cost except for LOS. The results should be interpreted in the context of LOS. Socioeconomic factors were significant contributors in the final model in predicting cost of CSF shunt hospitalizations. Racial and socioeconomic disparity has been noted in surgical care and outcomes.6,36 Black children and Hispanic children had higher total cost of CSF shunt hospitalization, longer LOS, and lower CoPID when compared with White children, controlling for other factors such as hospital type, insurance, income, number of chronic conditions, and number of surgical procedures. The etiology of hydrocephalus was a significant predictor of cost associated with pediatric CSF shunt placement for total hospital charges and CoPID. Previous studNeurosurg Focus / Volume 37 / November 2014
ies have mentioned this relationship but have not attempted to study costs.20,25,38 Etiologies with long LOSs such as IVH and meningitis had high total costs but relatively low CoPID. This result suggests a longer length of lower cost care leading up to or associated with shunt placement, which describes the pattern of a perinatal admission with IVH of prematurity, or intravenous antibiotics with meningitis. Hydrocephalus etiology is a logical predictor of survival,16 which was found to be associated with CoPID and total cost. Total costs and CoPID varied with the type of hospital. The KID reports data on 4 types of hospitals: children’s nonprofit hospital, children’s unit in nonprofit (primarily adult) hospital, investor-owned hospital, and governmentowned hospital. Studies in other medical21 and surgical disciplines have used the KID to examine charges by hospital type, such as for pyloromyotomy,29 where Raval et al. found that children’s hospitals had lower complication rates and shorter LOSs, but higher charges than general hospitals (although lower charges than children’s units in general hospitals). We found 34% of CSF shunt patients at nonprofit children’s hospitals were transfers, compared with 20% at other hospitals. This may suggest patients treated in nonprofit children’s hospitals had higher clinical complexity. The National Association of Children’s Hospitals and Related Institutions recommends that “all children need children’s hospitals.”23 Because of the unique mission of children’s hospitals, higher costs have been reported. Freestanding children’s hospitals comprise approximately 1% of US hospitals, but account for 39% of pediatric admissions, 49% of pediatric inpatient days, and 59% of costs.21 Subspecialty care, laboratories, radiology, and ancillary services designed for children are housed in freestanding children’s hospitals, where they are recognized to be higher cost centers.21 Hospital location was also found to be a significant predictor of cost, as with other reported studies in the pediatric and adult neurosurgical literature.32 In keeping with our findings, other studies have shown that centers in the West and Midwest have significantly higher costs when compared with the Northeast.29,32 The effect of regional differences on spending is widely reported across specialties (http://www.dartmouthatlas.org).4 This variation may identify opportunities for minimizing the regional disparities and aiming utilization to be at the level of the lower cost regions. It would be important to assess whether this goal can be safely achieved. This study cannot assess this question, and further research is warranted. 7
S. K. Lam et al. TABLE 4: Comparison of cost models* Variable
Model I
Model II
Final Model
age (1–20 yrs) female admission on weekend day hydrocephalus etiology (ref: communicating hydrocephalus) obstructive hydrocephalus congenital hydrocephalus tumor IVH myelomeningocele meningitis trauma others no. of chronic conditions no. of procedures hospital type (ref: government owned) nonprofit, non-children’s nonprofit, children’s nonprofit, children’s unit in a hospital investor-owned, private unspecified region (ref.: Northeast) Midwest South West no. of RN FTEs per 1000 adjusted inpatient days race (ref: White) Black Hispanic Asian and others unspecified insurance type (ref: private) Medicaid other income level (ref: 0%–25%) 25–50% 50–75% 75–100% unspecified constant (in the multivariate linear regression model) no. of cases R2
0.21 (0.00) 0.03 (0.19) −0.16 (0.00)
0.22 (0.00) 0.03 (0.21) −0.17 (0.00)
0.20 (0.00) 0.03 (0.18) −0.17 (0.00)
−0.04 (0.42) −0.17 (0.00) −0.07 (0.32) −0.35 (0.00) −0.28 (0.00) −0.43 (0.00) −0.23 (0.00) −0.12 (0.08) −0.02 (0.01) −0.03 (0.00)
−0.06 (0.28) −0.18 (0.00) −0.08 (0.21) −0.35 (0.00) −0.29 (0.00) −0.44 (0.00) −0.23 (0.00) −0.13 (0.05) −0.02 (0.00) −0.03 (0.00)
−0.05 (0.30) −0.18 (0.00) −0.09 (0.15) −0.34 (0.00) −0.29 (0.00) −0.44 (0.00) −0.21 (0.01) −0.13 (0.05) −0.02 (0.01) −0.03 (0.00)
−0.21 (0.00) 0.20 (0.00) −0.08 (0.12) −0.41 (0.00) 0.15 (0.15)
−0.17 (0.00) 0.12 (0.09) −0.08 (0.13) −0.36 (0.00) 0.14 (0.11)
−0.19 (0.00) 0.10 (0.12) −0.09 (0.06) −0.38 (0.00) 0.13 (0.13)
0.10 (0.18) −0.15 (0.03) 0.13 (0.08)
0.05 (0.52) −0.18 (0.01) 0.09 (0.21) 0.05 (0.01)
0.06 (0.44) −0.15 (0.03) 0.10 (0.18) 0.05 (0.01) −0.07 (0.04) 0.00 (0.90) −0.03 (0.47) 0.02 (0.71) 0.10 (0.00) 0.06 (0.26)
8.18 (0.00) 2519 0.29
7.98 (0.00) 2519 0.30
0.01 (0.84) 0.03 (0.42) 0.09 (0.00) −0.08 (0.37) 7.91 (0.00) 2519 0.32
* All data given as coefficient of the covariates in the regression model, or β (p value). Significant p values indicated in boldface. ref = reference group.
Other hospital operational factors also came into consideration. Higher RN FTEs contribute to CoPID in our cost model; this may reflect acuity of care and staffing requirements. The optimal nurse staffing ratio for providing quality care has been debated.8,9,18 Studies from the 8
US and Europe demonstrate that nurse staffing correlates with adult inpatient mortality outcomes,1,15,31 and higher ratios may also increase costs.9,14,33 Weekend admissions had higher CoPID than weekday admissions. Both delay in care and delay in discharges have been reported in Neurosurg Focus / Volume 37 / November 2014
Cost model of CSF shunt hospitalization relation with weekend admissions across disciplines.11,22 These ideas may motivate a closer look in staffing, operations, and efficiency. Study Limitations
Limitations of this study include the inherent nature of KID data. As with any database, coding inaccuracies may exist, and we are unable to independently verify diagnoses or procedures by chart review. There are ongoing validation studies in pediatric neurosurgery. Details about clinical indications, causality, and disease severity are not known. Specific complications and outcomes are not included in this initial study, although further work is much needed to examine quality, clinical outcomes, return to the emergency room, readmission, and reoperation. While bed assignment is important to consider in costs, due to limitations in KID data, we are unable to identify or separate the total LOS in each location (intensive care unit vs ward). Cost-to-charge ratios help in converting charge to cost. This is a surrogate for cost, but not reflective of actual cost data. The idea of the cost model for pediatric CSF shunting is in its nascent stages. The KID database does not have the level of detail in variables required to build a predictive plug-and-play model and equation. Further work with more databases is needed. In the future, predictive costing models will likely be possible and may become the norm. Covering risk in such costing models may depend on volume and efficiency based on iteration of quality. The advantage may be with higher volume centers. There is a potential application for more delineation in bundle pricing as we further understand the differences in hydrocephalus subgroups. As such, there may be even greater potential with the advent of ICD-10 and its new level of detail. Although our models cannot account for all facets of variations in CSF shunt-related costs due to limitations in the type and granularity of variables in KID, this study shows a direction that is needed in neurosurgery subspecialty care: the ability to engage in health care economics conversations. This type of model can be a starting tool for examining areas for improvement, minimizing preventable variance in costs, creating data-driven efficiency optimization, and recognizing populations at higher risk for resource-intensive hospitalizations. We need to develop further understanding into the complexity of factors associated with costs and the implications of bundling payment models.
Conclusions
A patient’s socioeconomic status, such as race, income level, and insurance, in addition to hospital-related factors such as number of hospital RN FTEs, hospital type, and US region, could affect the costs of initial CSF shunt placement in addition to clinical factors such as hydrocephalus etiology and LOS. To create a cost model of initial CSF shunt placement in the pediatrics population, consideration of such nonclinical factors is warranted. Neurosurg Focus / Volume 37 / November 2014
Disclosure The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Author contributions to the study and manuscript preparation include the following. Conception and design: Pan, Lam. Acquisition of data: Pan. Analysis and interpretation of data: Pan. Drafting the article: Pan, Lam, Srinivasan. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Pan. Statistical analysis: Pan. Administrative/technical/ material support: Pan, Lam, Luerssen. Study supervision: Pan, Lam. References 1. Aiken LH, Sloane DM, Bruyneel L, Van den Heede K, Griffiths P, Busse R, et al: Nurse staffing and education and hospital mortality in nine European countries: a retrospective observational study. Lancet 383:1824–1830, 2014 2. Arizona Health Care Cost Containment System: AHCCCS Transition to DRG-Based Payment. (http://www.azahcccs. gov/commercial/ProviderBilling/DRGBasedPayments.aspx) [Accessed September 3, 2014] 3. Averill RF, McCullough EC, Goldfield N, Hughes JS, Bonazelli J, Bentley L, et al: 3MTM APR DRG Classification System, Version 31.0 (effective 10/01/2013). Healthcare Cost and Utilization Project. (http://www.hcup-us.ahrq.gov/ db/nation/nis/grp031_aprdrg_meth_ovrview.pdf) [Accessed September 1, 2014] 4. Bekelis K, Missios S, Labropoulos N: Cerebral aneurysm coiling: a predictive model of hospitalization cost. J Neurointerv Surg [epub ahead of print], 2014 5. Buchanan CC, Hernandez EA, Anderson JM, Dye JA, Leung M, Buxey F, et al: Analysis of 30-day readmissions among neurosurgical patients: surgical complication avoidance as key to quality improvement. Clinical article. J Neurosurg 121: 170–175, 2014 6. Chern JJ, Bookland M, Tejedor-Sojo J, Riley J, Shoja MM, Tubbs RS, et al: Return to system within 30 days of discharge following pediatric shunt surgery. Clinical article. J Neurosurg Pediatr 13:525–531, 2014 7. de Brantes F, Rosenthal MB, Painter M: Building a bridge from fragmentation to accountability—the Prometheus Payment model. N Engl J Med 361:1033–1036, 2009 8. Duffin C: Nurse-to-patient ratios must increase to improve safety. Nurs Older People 24:6–7, 2012 9. Garretson S: Nurse to patient ratios in American health care. Nurs Stand 19:33–37, 2004 10. Gever M: Improving the quality of care: the continuing debate over nurse-patient ratios. National Conference of State Legislatures. (http://www.ncsl.org/print/health/shn/shn535.pdf) [Accessed September 3, 2014] 11. Goldstein SD, Papandria DJ, Aboagye J, Salazar JH, Van Arendonk K, Al-Omar K, et al: The “weekend effect” in pediatric surgery—increased mortality for children undergoing urgent surgery during the weekend. J Pediatr Surg 49:1087– 1091, 2014 12. Healthcare Cost Utilization Project: Cost-to-Charge Ratio Files: 2009 Kids’ Inpatient Database (KID) User Guide. (http://www.hcup-us.ahrq.gov/db/state/CCR2009KIDUserGuide.pdf) [Accessed September 3, 2014] 13. Healthcare Financial Management Association: Medicaid DRGs. Get ready, ‘cuz here they come! HFMA Florida Chapter. (http://www.floridahfma.org/presentations.lib/items/medi caid-drgs/MEDICAID%20DRGS.pdf) [Accessed September 1, 2014] 14. Howell AM, Panesar SS, Burns EM, Donaldson LJ, Darzi A: Reducing the burden of surgical harm: a systematic review
9
S. K. Lam et al. of the interventions used to reduce adverse events in surgery. Ann Surg 259:630–641, 2014 15. Kane RL, Shamliyan TA, Mueller C, Duval S, Wilt TJ: The association of registered nurse staffing levels and patient outcomes: systematic review and meta-analysis. Med Care 45: 1195–1204, 2007 16. Kulkarni AV, Riva-Cambrin J, Butler J, Browd SR, Drake JM, Holubkov R, et al: Outcomes of CSF shunting in children: comparison of Hydrocephalus Clinical Research Network cohort with historical controls. Clinical article. J Neurosurg Pediatr 12:334–338, 2013 17. Lave JR, Lave LB: Hospital cost functions. Annu Rev Public Health 5:193–213, 1984 18. Lewis KK, New England Public Policy Center: Nurse-to-patient ratios: research and reality. Issue Brief (Mass Health Policy Forum) (25):1–19, 2005 19. Manning WG, Mullahy J: Estimating log models: to transform or not to transform? J Health Econ 20:461–494, 2001 20. McGirt MJ, Leveque JC, Wellons JC III, Villavicencio AT, Hopkins JS, Fuchs HE, et al: Cerebrospinal fluid shunt survival and etiology of failures: a seven-year institutional experience. Pediatr Neurosurg 36:248–255, 2002 21. Merenstein D, Egleston B, Diener-West M: Lengths of stay and costs associated with children’s hospitals. Pediatrics 115: 839–844, 2005 22. Nandyala SV, Marquez-Lara A, Fineberg SJ, Schmitt DR, Singh K: Comparison of perioperative outcomes and cost of spinal fusion for cervical trauma: weekday versus weekend admissions. Spine (Phila Pa 1976) 38:2178–2183, 2013 23. National Association of Children’s Hospitals and Related Institutions: All Children Need Children’s Hospitals. (http:// www.upstate.edu/gch/education/allchildren.pdf) [Accessed September 3, 2014] 24. Noyes K, Liu H, Temkin-Greener H: Medicare capitation model, functional status, and multiple comorbidities: model accuracy. Am J Manag Care 14:679–690, 2008 25. Patwardhan RV, Nanda A: Implanted ventricular shunts in the United States: the billion-dollar-a-year cost of hydrocephalus treatment. Neurosurgery 56:139–145, 2005 26. Polverejan E, Gardiner JC, Bradley CJ, Holmes-Rovner M, Rovner D: Estimating mean hospital cost as a function of length of stay and patient characteristics. Health Econ 12: 935–947, 2003 27. Pope GC, Kautter J, Ellis RP, Ash AS, Ayanian JZ, Lezzoni LI, et al: Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev 25:119– 141, 2004 28. Quinn K: New directions in Medicaid payment for hospital care. Health Aff (Millwood) 27:269–280, 2008 29. Raval MV, Cohen ME, Barsness KA, Bentrem DJ, Phillips JD, Reynolds M: Does hospital type affect pyloromyotomy outcomes? Analysis of the Kids’ Inpatient Database. Surgery 148:411–419, 2010
10
30. Rolston JD, Han SJ, Lau CY, Berger MS, Parsa AT: Frequency and predictors of complications in neurological surgery: national trends from 2006 to 2011. Clinical article. J Neurosurg 120:736–745, 2014 31. Sasichay-Akkadechanunt T, Scalzi CC, Jawad AF: The relationship between nurse staffing and patient outcomes. J Nurs Adm 33:478–485, 2003 32. Sharma M, Sonig A, Ambekar S, Nanda A: Discharge dispositions, complications, and costs of hospitalization in spinal cord tumor surgery: analysis of data from the United States Nationwide Inpatient Sample, 2003–2010. Clinical article. J Neurosurg Spine 20:125–141, 2014 33. Sherenian M, Profit J, Schmidt B, Suh S, Xiao R, Zupancic JA, et al: Nurse-to-patient ratios and neonatal outcomes: a brief systematic review. Neonatology 104:179–183, 2013 34. Simon TD, Hall M, Riva-Cambrin J, Albert JE, Jeffries HE, Lafleur B, et al: Infection rates following initial cerebrospinal fluid shunt placement across pediatric hospitals in the United States. Clinical article. J Neurosurg Pediatr 4:156–165, 2009 35. Simon TD, Riva-Cambrin J, Srivastava R, Bratton SL, Dean JM, Kestle JR: Hospital care for children with hydrocephalus in the United States: utilization, charges, comorbidities, and deaths. J Neurosurg Pediatr 1:131–137, 2008 36. Stone ML, LaPar DJ, Mulloy DP, Rasmussen SK, Kane BJ, McGahren ED, et al: Primary payer status is significantly associated with postoperative mortality, morbidity, and hospital resource utilization in pediatric surgical patients within the United States. J Pediatr Surg 48:81–87, 2013 37. Texas Medicaid & Healthcare Partnership: All Patient Refined – Diagnosis Related Groups (APR-DRG). (http://www. tmhp.com/Pages/Medicaid/Hospital_APR-DRG.aspx) [Accessed September 3, 2014] 38. Tuli S, Drake J, Lawless J, Wigg M, Lamberti-Pasculli M: Risk factors for repeated cerebrospinal shunt failures in pediatric patients with hydrocephalus. J Neurosurg 92:31–38, 2000 39. White C, Reschovsky JD, Bond AM: Inpatient hospital prices drive spending variation for episodes of care for privately insured patients. NIHCR Research Brief No. 14. National Institute for Health Care Reform. (http://www.nihcr.org/ Episode-Spending-Variation) [Accessed September 1, 2014] 40. Wrubel DM, Riemenschneider KJ, Braender C, Miller BA, Hirsh DA, Reisner A, et al: Return to system within 30 days of pediatric neurosurgery. Clinical article. J Neurosurg Pediatr 13:216–221, 2014 Manuscript submitted July 16, 2014. Accepted August 19, 2014. Please include this information when citing this paper: DOI: 10.3171/2014.8.FOCUS14454. Address correspondence to: I-Wen Pan, Ph.D., Department of Neurosurgery, Texas Children’s Hospital, 6701 Fannin St., CCC Ste. 1230-01, Houston, TX 77030. email: [email protected].
Neurosurg Focus / Volume 37 / November 2014
