Intramedullary spinal cord tumors (IMSCTs) account

Transcription

1 J Neurosurg Spine 20: , 2014 AANS, 2014 Discharge dispositions, complications, and costs of hospitalization in spinal cord tumor surgery: analysis of data from the United States Nationwide Inpatient Sample, Clinical article Mayur Sharma, M.D., M.Ch., Ashish Sonig, M.D., M.Ch., Sudheer Ambekar, M.D., M.Ch., and Anil Nanda, M.D., M.P.H. Department of Neurosurgery, Louisiana State University Health Science Center, Shreveport, Louisiana Object. The aim of this study was to analyze the incidence of adverse outcomes and inpatient mortality following resection of intramedullary spinal cord tumors by using the US Nationwide Inpatient Sample (NIS) database. The overall complication rate, length of the hospital stay, and the total cost of hospitalization were also analyzed from the database. Methods. This is a retrospective cohort study conducted using the NIS data from 2003 to Various patient-related (demographic categories, complications, comorbidities, and median household income) and hospital-related variables (number of beds, high/low case volume, rural/urban location, region, ownership, and teaching status) were analyzed from the database. The adverse discharge disposition, in-hospital mortality, and the higher cost of hospitalization were taken as the dependent variables. Results. A total of 15,545 admissions were identified from the NIS database. The mean patient age was ± years (mean ± SD), and 7938 (52%) of the patients were male. Regarding discharge disposition, 64.1% (n = 9917) of the patients were discharged to home or self-care, and the overall in-hospital mortality rate was 0.46% (n = 71). The mean total charges for hospitalization increased from $45, in 2003 to $76, in Elderly patients, female sex, black race, and lower income based on ZIP code were the independent predictors of other than routine (OTR) disposition (p < 0.001). Private insurance showed a protective effect against OTR disposition. Patients with a higher comorbidity index (OR 1.908, 95% CI ; p < 0.001) and with complications (OR 2.214, 95% CI ; p < 0.001) were more likely to have an adverse discharge disposition. Hospitals with a larger number of beds and those in the Northeast region were independent predictors of the OTR discharge disposition (p < 0.001). Admissions on weekends and nonelective admission had significant influence on the disposition (p < 0.001). Weekend and nonelective admissions were found to be independent predictors of inpatient mortality and the higher cost incurred to the hospitals (p < 0.001). High-volume and large hospitals, West region, and teaching hospitals were also the predictors of higher cost incurred to the hospitals (p < 0.001). The following variables (young patients, higher median household income, nonprivate insurance, presence of complications, and a higher comorbidity index) were significantly correlated with higher hospital charges (p < 0.001), whereas the variables young patients, nonprivate insurance, higher median household income, and higher comorbidity index independently predicted for inpatient mortality (p < 0.001). Conclusions. The independent predictors of adverse discharge disposition were as follows: elderly patients, female sex, black race, lower median household income, nonprivate insurance, higher comorbidity index, presence of complications, larger hospital size, Northeast region, and weekend and nonelective admissions. The predictors of higher cost incurred to the hospitals were as follows: young patients, higher median household income, nonprivate insurance, presence of complications, higher comorbidity index, hospitals with high volume and a large number of beds, West region, teaching hospitals, and weekend and nonelective admissions. ( Key Words spinal cord tumor surgery complications hospitalization cost Nationwide Inpatient Sample oncology Intramedullary spinal cord tumors (IMSCTs) account for 5% 10% of all spinal tumors in adults and approximately 35% in children. 12,14 The overall incidence of Abbreviations used in this paper: IMSCT = intramedullary spinal cord tumor; NIS = Nationwide Inpatient Sample; OTR = other than routine. SCTs (malignant and nonmalignant) is approximately 0.62 per 100,000 persons. 7 The majority (90%) of the IMSCTs are glial in origin, with the common histological types being ependymoma (60%) and astrocytoma (30%) This article contains some figures that are displayed in color on line but in black-and-white in the print edition. 125

2 M. Sharma et al. TABLE 1: Patient and hospital characteristics for IMSCT surgery between 2003 and 2010 in the US* Characteristic Value % of Patients age in yrs mean, median, range 44.84, 47, > sex male female race white black Hispanic Asian or Pacific Islander Native American other expected primary payer, uniform Medicare Medicaid private self-pay no charge other median HIQ for patient s ZIP code $1 38, $39,000 47, $48,000 62, $ 63, hospital size small medium large 11, hospital vol high low 10, comorbidity index low 10, high yr of treatment admission day weekday 14, weekend admission type elective 10, nonelective (continued) 126

3 Outcomes in spinal cord tumor surgery TABLE 1: Patient and hospital characteristics for IMSCT surgery between 2003 and 2010 in the US* (continued) Characteristic Value % of Patients discharges routine OTR median length of stay (days) Northeast 4 Midwest 4 South 5 West 5 overall 5 mean total hospital charges ($) Northeast 55, Midwest 51, South 52, West 89, overall 61, in-hospital mortality Northeast 11 of Midwest 10 of South 15 of West 35 of overall 71 of 15, overall complications neurological pulmonary cardiac renal postop bleeding/hematoma infections thromboembolism total no. of patients w/ 1 complications * Besides the inherent limitations of a retrospective study, there are some specific limitations of the NIS data set such as coding errors. Missing values in the variables were not taken into consideration for the analysis. HIQ = household income quartile. in adults, whereas astrocytomas are the most common intramedullary tumor in children. 7,12,14 With the advent of the operating microscope, microneurosurgical instruments, advanced imaging techniques, and intraoperative neurophysiological monitoring, the management of these tumors has undergone a paradigm shift from a conservative to a more aggressive approach. 1,8,12,15 Complete or near-complete excision results in long-term progressionfree survival with acceptable morbidity, especially in patients with benign or low-grade IMSCTs. 6,12 There are several hospital, locoregional, and patient factors that can affect the outcome following resection of IMSCTs. However, most of the available literature on IMSCTs is from clinical series, and the low incidence of these lesions limits the research related to the analysis of these factors. 1,4,6,8,17 To highlight this issue we analyzed the Nationwide Inpatient Sample (NIS) to study the outcomes and socioeconomic variations of IMSCT surgery. To the best of our knowledge there has been only one study in which the complications and outcomes following SCT resection were analyzed using the data compiled in the NIS database from 1993 to In that study, the adverse outcome following SCT resection was found to be associated with a higher comorbidity index and occurrence of postoperative complications. However, there is no study that has analyzed all the variables such as the hospital affiliation, complications, comorbidities, and hospital charges following resection of IMSCTs. The aim of our study was to analyze the incidence of adverse outcomes and inpatient mortality as well as the factors affecting them following resection of IMSCTs, by using the data compiled in the US NIS from 2003 to We also analyzed the impact of various patient and hospital factors on the total cost of hospitalization following this procedure. The overall complication rate and the length of the hospital stay were also analyzed using information from the database. 127

4 M. Sharma et al. TABLE 2: Patient and demographic factors affecting adverse (OTR) discharge disposition after IMSCT resection No. of Patients Routine OTR Variable Total p Value age in yrs <0.001* > sex <0.001* male female race <0.001* white black other expected primary payer, uniform <0.001* Medicare Medicaid private self-pay no charge other median HIQ for patient s ZIP code <0.001* $1 38, $39,000 47, $48,000 62, $ 63, complications <0.001* yes no ,334 comorbidity index <0.001* low ,767 high * Significant according to univariate analysis. Data Source Methods Data were obtained from the NIS of the Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality (AHRQ), which represents a 20% stratified sample of nonfederal community hospitals in the US. 9 The NIS database is the largest all-payer inpatient care database and contains data on 5 8 million hospital stays from approximately 1000 hospitals in 45 states. Defining the Study Sample The data for SCTs from 2003 to 2010 were extracted from the NIS by using the ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) diagnosis and treatment codes. All admissions with a primary diagnosis code of (malignant neoplasm of the spinal cord), (benign neoplasm of the spinal cord), and (neoplasm of uncertain behavior) were included only when they were accompanied by the primary procedure codes for spine surgery (ICD-9-CM procedure codes 03.0, 03.4, 03.09, 81.0, and ). Patients with tumors of the spinal meninges (ICD-9-CM code 225.4) and nerve sheaths (ICD-9-CM code 215) were excluded from the study. Patient and Hospital Characteristics Patient Characteristics. Patient age, race, and sex were extracted directly from the NIS. Primary payer (Medicare, Medicaid, private insurance, self-pay, no charge, or others); type of admission (emergency, urgent, elective); day of admission (weekend or weekday); and median household income for patients ZIP code (ZIP- INC_QRTL) were also included in the analysis. 128

5 Outcomes in spinal cord tumor surgery TABLE 3: Multivariate binary logistic regression analysis showing various patient and demographic factors affecting adverse (OTR) discharge disposition after IMSCT resection Variable* B Value p Value OR 95% CI age compared w/ 0- to 17-yr-old group < < > < sex (female) < race compared w/ white race black < other expected primary payer compared w/ private insurance Medicare < Medicaid < median HIQ for patient s ZIP code $38,999 compared w/ others $39,000 47, $48,000 62, $ 63, < complications (yes) < comorbidity (high) index < * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. Fig. 1. Bar graphs showing the impact of age of the patient (A); median household income quartile for the patient s ZIP code (by quartiles 1st quartile, $1 $38,999; 2nd quartile, $39,000 $47,999; 3rd quartile, $48,000 $62,999; 4th quartile, $63,000) (B); presence of complications (C); and comorbidity index (D) on adverse (OTR) discharge disposition following IMSCT surgery in the US between 2003 and

6 M. Sharma et al. TABLE 4: Patient and demographic factors affecting inpatient mortality after IMSCT resection No. Who Died During Hospitalization No Yes Variable Total p Value age 0.001* > sex male female race white black other expected primary payer, uniform <0.001* Medicare Medicaid private self-pay no charge other median HIQ for patient s ZIP code <0.001* $1 38, $39,000 47, $48,000 62, $ 63, complications 0.001* yes no 12, ,339 comorbidity index <0.001* low 10, ,777 high * Significant according to univariate analysis. The following variables were recoded into 5 different categories to further aid in our analysis. 1) Age of the patient was categorized into 0 17, 18 44, 45 64, and > 64 years. 2) Race was recoded into white, black, and other for the analysis. 3) Comorbidities were classified using the Charlson Comorbidity Index and dichotomized as low (0 or 1) versus high ( 2) for analysis. 5 4) Complications were extracted from the NIS based on the ICD-9-CM diagnosis codes: a) neurological ( , ); b) pulmonary ( , 997.3); c) cardiac (997.1, 410); d) renal and urinary (584, 997.5); e) postoperative bleeding or hematoma (998.1, ); f) infections (998.0, 998.3, 998.5, , ); and g) deep vein thrombosis and pulmonary embolism (415, 415.1, , , , 451, , 451.8, 451.9, 453, , 453.8, 453.9). Patients were then dichotomized into groups with and without complications. 5) Type of admission was recoded as elective or nonelective for the analysis. Missing values in the variables were not taken into consideration for the analysis. Hospital Characteristics. Region of the hospital (Northeast, Midwest, South, or West); hospital location (rural or urban); teaching status; hospital size by number of beds (small, medium, or large); and hospital control (government, private nonprofit, private, and investor owned) these variables were recoded into government and private for the analysis. The annual number of cases at each hospital (hospital volume) was calculated using the hospital identification number (HOSPID). Percentiles were calculated and hospitals in which the number of cases was greater than the 90th percentile were categorized into high-volume centers. Such high-volume centers were recognized for each year and merged together for the final analysis. This variable was coded as high- and low-volume centers. 130

7 Outcomes in spinal cord tumor surgery TABLE 5: Multivariate binary logistic regression analysis showing various patient and demographic factors affecting inpatient mortality after IMSCT resection Variable* B Value p Value OR 95% CI age compared w/ 0- to 17-yr-old group > sex (female) race compared w/ white race black other expected primary payer compared w/ private insurance Medicare Medicaid < median HIQ for patient s ZIP code $38, compared w/ others $39,000 47, $48,000 62, $ 63, complications (yes) comorbidity (high) index < * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. Fig. 2. Graphs showing that the in-hospital mortality increased from 0.1% in 2003 to 0.9% in 2010, with 0% in 2005 (A). Bar graphs showing the impact of hospital region (B), age of patient in years (C), and type of insurance (D) on in-hospital mortality. 131

8 M. Sharma et al. TABLE 6: Hospital factors affecting adverse (OTR) discharge disposition after IMSCT resection Variable Outcome Factors No. of Patients Routine OTR p Value hospital size <0.001* small medium large hospital vol high low hospital control government private teaching status of hospital nonteaching teaching location of hospital rural urban hospital region <0.001* Northeast Midwest South West admission day <0.001* weekday weekend admission status <0.001* elective nonelective * Significant according to univariate analysis. The primary adverse end points were either inhospital death (DIED) or discharge to facilities other than home/self-care (DISPUniform). As described in our study, discharge disposition to home was considered to be routine disposition as mentioned in the NIS, and discharges to all other places (Transfer to Short-Term Hospital, Skilled Nursing Facility, Intermediate Care Facility, Another Type of Facility, Home Health Care, Against Medical Advice, and Died) were categorized as other than routine (OTR) disposition. 18 Cost of the hospitalization was calculated using the variable Total charges (TOTCHG) in the NIS database. Percentiles of the cost were calculated and recoded as high cost ( 75th percentile, $74,284) and low cost (< 75th percentile) for the analysis. Statistical Analysis The statistical analysis was accomplished using SPSS v21 (IBM, Inc.) and Microsoft Excel. Chi-square and Pearson correlation tests were used to look for the association between the categorical variables. Comparisons were considered significant only if the p value was < Univariate analysis was used to identify the covariates that might affect the outcome. Multivariate logistic regression analysis was used for the multivariate analysis to generate the probability value, odds ratio, and confidence interval. Variables with p < 0.1 in the univariate analysis were considered for the multivariate analysis. Discharge disposition, hospital charges, complications, and mortality were the dependent variables used in the analysis. Extrapolation from the sampled data set to the nationwide population was performed using a weighting variable provided to the data set in SPSS. Results A total of 15,545 admissions of patients with SCTs that had undergone surgical management were identified from the NIS database between 2003 and The patient characteristics are given in Table 1. The mean patient age was ± years (mean ± SD; range 1 90 years), and 73.6% (n = 11,362) of the patients were between 18 and 64 years of age. Fifty-two percent (n = 7938) of the patients were males, and white was the dominant race (72.9%, n = 8637). Private insurance was the primary payer in 64.2% (n = 9973), Medicare in 18.3% (n = 2843), and Medicaid in 10.1% (n = 1562) of the patients. For the factor median household income for the patient s ZIP code, 20.5% (n = 3189) of the patients were in the 1st quartile and 28.2% (n = 4380) were in the 4th quartile. The majority (70.6%, n = 10,818) of the patients had a low comorbidity index, and 4.2% (n = 547) of the patients experienced one or more complications following the procedure, with the most common being neurological deficits (1.1%, n = 148). The annual case volume increased from 2033 in 2003 to 2108 in The majority of the admissions were on weekdays (95.7%, n = 14,871) and the type was elective (77.1%, n = 10,192). The median length of stay was higher in the South and West regions as compared with the Northeast and Midwest regions (5 days vs 4 days), with the overall median stay being 5 days. Overall 64.1% (n = 9917) of the patients were discharged to home or self-care. The inpatient hospital mortality rate increased from 0.1% in 2003 to 0.9% in 2010, with none in Patient Characteristics and Adverse (OTR) Discharge Disposition Overall 35.9% (n = 5563) of the patients had an adverse discharge disposition. In univariate analysis, advanced age, female sex, and black race showed a significant correlation with the adverse discharge disposition (p < 0.05). Similarly, private insurance, higher median household income for the patient s ZIP code, lower comorbidity index, and absence of complications showed a negative correlation with OTR discharges (p < 0.05). In multivariate regression analysis, adverse discharge disposition was 2 times more likely in patients with a higher comorbidity index than in those with a lower index (OR 1.908, 95% CI ; p < 0.001). Similarly, 132

9 Outcomes in spinal cord tumor surgery TABLE 7: Multivariate binary logistic regression analysis showing various hospital factors affecting adverse (OTR) discharge disposition after IMSCT resection Variable* B Value p Value OR 95% CI hospital size compared w/ small medium < large < high hospital vol (yes) government hospital control teaching status of hospital (yes) urban location of hospital (yes) hospital region compared w/ West Northeast Midwest < South < weekend admission day (yes) elective admission status (yes) < * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. Fig. 3. Bar graphs showing the impact of hospital factors: hospital region (A), number of beds (B), day of admission (C), and type of admission (D) on adverse discharge disposition following IMSCT surgery in the US between 2003 and

10 M. Sharma et al. TABLE 8: Hospital factors affecting in-hospital patient mortality after IMSCT resection No. Who Died During Hospitalization Variable No Yes Total p Value hospital size small medium large 11, ,693 hospital vol 0.016* low 10, ,928 high hospital control government 13, ,465 private teaching status of hospital nonteaching teaching 12, ,552 location of hospital <0.001* rural urban 15, ,105 hospital region <0.001* Northeast Midwest South West admission day 0.004* weekday 14, ,822 weekend admission status 0.008* elective 10, ,169 nonelective * Significant according to univariate analysis. TABLE 9: Multivariate binary logistic regression analysis showing various hospital factors affecting in-patient mortality after IMSCT resection Variable* B Value p Value OR 95% CI hospital size compared w/ small medium large high hospital vol (yes) government hospital control teaching status of hospital (yes) urban location of hospital (yes) < hospital region compared w/ West Northeast Midwest South weekend admission day (yes) elective admission status (yes) * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. 134

11 Outcomes in spinal cord tumor surgery TABLE 10: Difference between the mean hospital charges among various groups (Kruskal-Wallis test) in patients who underwent IMSCT resection Variable Mean Total Charges ($) p Value yr of treatment <0.001* , , , , , , , , complications <0.001* yes 80, no 63, comorbidity index <0.001* low 54, high 77, hospital size <0.001* small 58, medium 54, large 62, hospital control <0.001* government 61, private 58, teaching status of hospital <0.001* nonteaching 53, teaching 62, location of hospital <0.001* rural 37, urban 61, hospital region <0.001* Northeast 55, Midwest 51, South 52, West 89, admission day <0.001* weekday 59, weekend 90, admission status <0.001* elective 49, nonelective 71, * Significant. OTR discharges were twice as likely in patients with complications (OR 2.214, 95% CI ; p < 0.001). Patients in the middle-aged and elderly groups were more likely to have adverse outcomes when compared with those < 17 years of age (OR > 1, p < 0.001). Female sex and black race independently predicted for OTR disposition (OR and 1.956, respectively; p < in both instances). Private insurance showed a protective effect against OTR disposition as compared with Medicare and Medicaid (p < 0.001). Patients with income of $38,999 based on their ZIP code had a significantly higher OTR discharge disposition compared with those in the income categories $48,000 (OR 0.8, p < [univarate analysis]) (Tables 2 and 3, Fig. 1). Patient Characteristics and In-Hospital Mortality Rate The overall in-hospital mortality rate was 0.46% (n = 71). Of the various factors examined, the presence of complications (OR , 95% CI , p < 0.001) and higher comorbidity index (OR 9.68, p < 0.001) were significantly associated with in-hospital mortality on univariate analysis. Similarly, private insurance had shown a protective effect on in-hospital mortality as compared with Medicare (OR 3.085, p < 0.001) and Medicaid (OR 6.745, p < 0.001). In-hospital mortality was significantly higher in the young patients (0 17 years) when compared with the patients in the other age groups (18 44 and years OR 0.441, p = and OR 0.250, p < 0.001, respectively); however, this difference has not reached statistical significance when compared with patients in the elderly age group (> 64 years). Moreover, inpatient mortality was significantly higher in young patients (0 17 years) compared with the rest of the population (0.95% vs 0.39%, p = 0.005). There was no difference in the mortality rate between patients with median household income $38,999 and those with median income $63,000. Male sex and white race showed trends toward higher mortality rates, but the difference did not reach statistical significance. There was no in-hospital mortality encountered in black patients and in those with a median income between $48,000 and $62,999. Of the various factors examined, young age of the patient, insurance (Medicaid), and higher comorbidity index were the independent predictors of in-hospital mortality (p < 0.05) (Tables 4 and 5, Fig. 2). Hospital Characteristics and Adverse Discharge Disposition We found that 75.4% (n = 11,728) of the patients had undergone surgery in large hospitals, and that 70.7% (n = 10,982) of the patients were treated in hospitals with low case volumes. Large hospitals showed a significant adverse discharge disposition when compared with small and medium-sized hospitals (p < 0.05), and this variable is also shown to be an independent predictor of OTR discharge disposition (OR 1.778, p < 0.001). Although hospitals with higher case volumes had a lower number of patients with OTR discharge disposition, this difference did not reach statistical significance (p = 0.44). Similarly, ownership, teaching status, and location of the hospital had no influence on the adverse discharge disposition (p > 0.05). Admissions on weekends and nonelective admission were independent predictors of OTR discharge disposition (OR 1.308, 95% CI , p < 0.05; OR 1.955, 95% CI , p < 0.001, respectively). Hospital regions showed a significant correlation with the adverse discharge disposition, with the highest (42.19%) be- 135

12 M. Sharma et al. Fig. 4. A: Graph showing that the mean hospital charges increased by > $30,000 from 2003 to Bar graphs showing the impact of hospital region (B), presence of complications (C), and comorbidity index (D) on the total cost of hospitalization. ing in the Northeast region and the lowest in the Midwest region (32.65%), and the difference was found to be statistically significant (OR , 95% CI , p < 0.001) (Tables 6 and 7, Fig. 3). Hospital Characteristics and In-Hospital Mortality Rates Of the various factors examined, hospitals with higher case volumes (OR 2.821, p = 0.006), hospitals in rural locations (OR 0.065, p < 0.001), and nonelective admission status (OR 2.469, p = 0.007) were independent predictors of in-hospital mortality. Although mortality was higher in larger and private hospitals when compared with the smaller and government-owned hospitals, this difference did not reach statistical significance (p = and p = 0.386, respectively). In-hospital mortality was thrice as likely in patients with weekend than in those with weekday admissions (OR 3.241, 95% CI , p = 0.004). There was no significant difference between the teaching and nonteaching hospitals in terms of mortality (p = 0.759). In-hospital mortality was thrice as likely in hospitals in the West region when compared with hospitals in the South region (OR 3.438, 95% CI , p < 0.001) (Tables 8 and 9, Fig. 2). Costs of Hospitalization The mean total charges of hospitalization increased from $45, in 2003 to $76, in The overall mean expenditure was $61, ± $61, Table 10 depicts the difference of means in the expenditure in different groups and this was shown to be significant (p < 0.001) by using nonparametric tests. The difference in mean hospitalization charges between large versus small hospitals ($62, vs $58,152.64), teaching versus nonteaching hospitals ($62, vs $53,205.99), urban versus rural ($61, vs $37,667.01), government versus private (61, vs 58,218.51), and West versus South region ($89, vs $52,846.35) was significant (p < 0.001) (Table 10, Fig. 4). As mentioned in the Methods, this variable was dichotomized into high ( 75th percentile) and low cost. On univariate and multivariate regression analyses the following factors were found to be independent predictors of higher hospital cost: larger hospital size (OR 1.223, p = 0.035), hospitals with higher case volumes (OR 1.387, p < 0.001), teaching hospitals (OR 2.582, p < 0.001), urban location (OR 4.978, p < 0.001), and hospitals in the West region compared with those in the South region (OR 0.526, p < 0.001). Surprisingly, there was no significant difference in the hospital costs between government-owned and privately owned hospitals (p = 0.717). Admission on weekends and nonelective admissions showed a significant correlation with higher cost on both univariate and multivariate analysis, 136

13 Outcomes in spinal cord tumor surgery TABLE 11: Patient and demographic factors affecting the total hospital charges after IMSCT resection Variable No. of Patients Low Cost (<75th percentile) High Cost ( 75th percentile) p Value age <0.001* > sex male female race <0.001* white black other expected primary payer, uniform <0.001* Medicare Medicaid private self-pay no charge other median HIQ for patient s ZIP code <0.001* $1 38, $39,000 47, $48,000 62, $ 63, complications <0.001* yes no comorbidity index <0.001* low high * Significant according to univariate analysis. and were found to be independent predictors of higher cost incurred to the hospitals (OR 2.539, p < 0.001; OR 2.614, p < 0.001) (Tables 11 14). Among patient factors, younger age group, black race, nonprivate insurance, and patients with a median household income in the 4th quartile were strong independent predictors of increased hospital charges (p < 0.001). Similarly, the presence of any complication was also found to be an independent predictor of higher hospital charges (OR 1.761, p < 0.001), and patients with a higher comorbidity index were twice as likely to incur higher hospital charges than were those with a lower comorbidity index (OR 1.949, p < 0.001) (Fig. 4). Discussion We analyzed a total of 15,545 admissions for surgically treated SCT between 2003 and In a previous study, Patil et al. 16 reported the incidence of complications and outcomes following SCT surgery between 1993 and 2002; however, there is no study focusing on the demographic and hospital-related factors that can affect the discharge disposition and cost of hospitalization. The incidence of adverse discharge disposition (OTR discharge) increased from 20.6% in to 35.9% in ; this is in contrast to the in-hospital mortality rates, which decreased from 0.55% to 0.46% during the same treatment period. 16 Similarly, compared with the results of an earlier study of surgically treated SCTs in the US between 1993 and 2002, the total complication rate and mean length of stay decreased from 17.5% and 7.2 days, respectively, in to 4.2% and 6.39 days in The number of patients has remained the same during the study period despite an increase in the number of reporting hospitals, which is in accordance with the previous study

14 M. Sharma et al. TABLE 12: Multivariate binary logistic regression analysis showing various patient and demographic factors affecting the total cost of hospitalization after IMSCT resection Variable* B Value p Value OR 95% CI age compared w/ 0- to 17-yr-old group < < > < sex (female) race compared w/ white race black other < expected primary payer compared w/ private insurance Medicare Medicaid < median HIQ for patient s ZIP code $38, compared w/ others $39,000 47, $48,000 62, $ 63, < complications (yes) < comorbidity (high) index < * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. Patient Factors and Outcome In our study, favorable discharge disposition was seen in young, Caucasian, male patients with private insurance, median household income in the 4th quartile, low comorbidity index, and without postoperative complications. Similarly, Patil et al. 16 reported fewer adverse discharge dispositions in young patients with a low comorbidity index and without complications; however, there was no difference in the outcome between the sexes. Our analysis showed that the following variables young patients, those with nonprivate insurance, median household income in the 4th quartile, and high comorbidity index were the independent predictors of in-hospital mortality (p < 0.001). In contrast to our findings, age and comorbidities were not correlated with the mortality rate in another study. 16 The increased mortality rate in young patients can be attributed to more aggressive resection because of technical advances and thereby an increase in the rate of postoperative complications as well as mortality. However, due to the limitations of the NIS database it is not possible to evaluate the specific factors leading to such a finding. Patil et al. have shown that patients with a single complication were 4 times more likely to die and 2 times more likely to have an adverse discharge disposition compared with those without complications. In our study, inhospital mortality was seen in 1.86% of the patients with complications and in 0.48% of the patients without complications, which was statistically significant (p < 0.05). Therefore, avoidance or prompt diagnosis and management of these complications cannot be overemphasized. Apart from these demographic factors, other described factors that can affect the outcome, such as preoperative neurological status, tumor grade, histological type, and extent of resection cannot be evaluated due to the limitations inherent in the NIS database. 1,6,10,12,16,17 Hospital Factors and Outcome Among the various hospital factors, large size, hospitals in the Northeast region, weekend admission, and nonelective admission were the independent predictors of adverse discharge disposition. Hospitals with higher case volumes have significantly lower rates of mortality and adverse discharge dispositions; this concept has been explained in earlier studies. 2,3,11 However, in our study, hospitals with a larger number of beds and higher case volumes were 1.7 and 2.8 times more likely to have adverse discharge disposition and in-hospital mortality compared with the small hospitals and those with low case volumes, respectively. This may be due to the fact that the complicated and difficult surgeries are more likely to be performed in larger hospitals and those with higher case volumes and consequently worse outcomes. The Northeast had worse discharge outcomes compared with other regions, whereas the West had the highest in-hospital mortality when compared with other regions. Consistent with our previous study, nonelective admissions were 2.4 times more likely to result in inhospital mortality. 18 On analysis we found that 81.33% of the patients with weekday admissions and 92.1% of those with elective admissions underwent surgery on the day of admission, compared with 17.52% of those with weekend admissions and 34.3% of those with nonelective admissions (p < 0.05). Interestingly, 0.9% of the pa- 138

15 Outcomes in spinal cord tumor surgery TABLE 13: Hospital factors affecting the total hospital cost in patients who underwent IMSCT resection Variable Low Cost (<75th percentile) No. of Patients High Cost ( 75th percentile) p Value hospital size <0.001* small medium large hospital control government private hospital vol <0.001* high low teaching status of hospital <0.001* nonteaching teaching location of hospital <0.001* rural urban 11, hospital region <0.001* Northeast Midwest South West admission day <0.001* weekday 11, weekend admission status <0.001* elective nonelective * Significant according to univariate analysis. TABLE 14: Multivariate binary logistic regression analysis showing various hospital factors affecting the total cost of hospitalization after IMSCT resection Variable* B Value p Value OR 95% CI hospital size compared w/ small medium large high hospital vol (yes) < government hospital control < teaching status of hospital (yes) < urban location of hospital (yes) < hospital region compared w/ West Northeast Midwest < South < weekend admission day (yes) < elective admission status (yes) < * Dependent variable was discharge disposition (outcome) and the various factors mentioned were covariates. Significant. 139

16 M. Sharma et al. tients with nonelective admissions underwent surgery at another center 1 day prior to the admission at the index center, compared with 0.2% of the patients with elective admissions (p < 0.05), whereas none of those with weekend admissions underwent surgery prior to the admission at the index center. Therefore, a higher adverse outcome in patients with nonelective admissions can be attributed to the fact that a higher proportion of such patients were surgically treated at other centers and then transferred to the index center for further management. However, due to the limitations of the NIS database it was not possible to stratify the admissions according to the preoperative neurological status and to evaluate the factors leading to a delay in the surgical management of nonelective and weekend admissions. Because the majority of patients with weekend admissions underwent surgery between Days 1 and 3 of admission, this factor can be attributed as one of those responsible for the unfavorable outcome. Rural location was also an independent predictor of mortality (OR < 1, p < 0.001). Hospital Factors and Costs Various patient and hospital factors that can increase the total cost of hospitalization were analyzed. The cost of IMSCT surgery has increased from $27,223 in to $61, in , a 2.5-fold increase. 16 The West region had the highest and the Midwest had the lowest cost of hospitalization (p < 0.001). This is in concordance with the previous study in which it was found that the West region had the highest hospital cost. 18 Teaching hospitals were 2.5 times (p < 0.001) more likely to have a higher cost of hospitalization than were nonteaching hospitals, which is in accordance with other studies. 13,19 Similarly, hospitals with a large number of beds, urban location, and weekend and nonelective admissions were the independent predictors of higher hospital charges. Among patient factors, the highest odds ratio for higher hospital charges was seen in patients with a higher comorbidity index (OR 2.117, p < 0.001), followed by nonprivate insurance (OR 1.824, p < 0.001) and presence of complications (OR 1.694, p < 0.001). In their study Patil et al. showed that the presence of a single complication increased the mean total cost by $10,000, and with 3 complications this cost increased by 3-fold. In our study the presence of a single complication increased the total charges by > $20,000. Limitations of the Study Apart from the inherent limitations of a retrospective study, there are some specific limitations of the NIS data set such as coding errors and underreporting of events because of the inability to include outpatients. The NIS database may also underestimate mortality and overestimate morbidity rates because the outcome was measured at the time of discharge. Some patients may need to undergo repeat resections for recurrent IMSCTs; therefore, it is possible that such patients were counted twice in this study, underestimating the cost and overestimating the total number of discharges. Limitations specific to this study include the preoperative neurological status as well as the severity of postoperative complications, which cannot be coded from the data set. Similarly, the extent of tumor resection cannot be coded separately from the data set, and therefore routine and OTR discharge dispositions cannot be equated to the level of resections. Moreover, the outcome measures such as the long-term outcome, recurrence-free survival, quality of life, and functional status cannot be evaluated from the NIS database, thereby necessitating further prospective studies. Nonetheless, it helps to understand the regional trends, short-term outcomes, and practice patterns in the US. Conclusions This is a nationwide study of patients who were surgically treated for IMSCTs within the US between 2003 and There are several patient and hospital factors that can affect the discharge disposition, in-hospital mortality, and total cost of hospitalization. Our study provides an overview of the outcome, the total cost of hospitalization, and the regional variations after surgery for IMSCTs. Elderly age group, lower median household income, nonprivate insurance, a higher comorbidity index, presence of complications, larger hospital size, Northeast region, and weekend and nonelective admissions were the independent predictors of adverse discharge disposition. Young age group, higher median household income, nonprivate insurance, presence of complications, a higher comorbidity index, high-volume and larger-size hospitals, West region, teaching hospitals, and weekend and nonelective admissions were the predictors of higher cost incurred at the hospitals. 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: Nanda, Sharma. Acquisition of data: Sharma. Analysis and interpretation of data: Sharma, Sonig. Drafting the article: Sharma. 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: Nanda. Statistical analysis: Sharma, Sonig, Ambekar. Administrative/technical/material support: Nanda. Study supervision: Nanda. References 1. Bansal S, Suri A, Borkar SA, Kale SS, Singh M, Mahapatra AK: Management of intramedullary tumors in children: analysis of 82 operated cases. Childs Nerv Syst 28: , Barker FG II: Craniotomy for the resection of metastatic brain tumors in the U.S., : decreasing mortality and the effect of provider caseload. Cancer 100: , Barker FG II, Carter BS, Ojemann RG, Jyung RW, Poe DS, McKenna MJ: Surgical excision of acoustic neuroma: patient outcome and provider caseload. Laryngoscope 113: , Chandy MJ, Babu S: Management of intramedullary spinal cord tumours: review of 68 patients. Neurol India 47: , Charlson ME, Pompei P, Ales KL, MacKenzie CR: A new 140

17 Outcomes in spinal cord tumor surgery method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40: , Constantini S, Miller DC, Allen JC, Rorke LB, Freed D, Epstein FJ: Radical excision of intramedullary spinal cord tumors: surgical morbidity and long-term follow-up evaluation in 164 children and young adults. J Neurosurg 93 (2 Suppl): , Duong LM, McCarthy BJ, McLendon RE, Dolecek TA, Kruchko C, Douglas LL, et al: Descriptive epidemiology of malignant and nonmalignant primary spinal cord, spinal meninges, and cauda equina tumors, United States, Cancer 118: , Guidetti B, Mercuri S, Vagnozzi R: Long-term results of the surgical treatment of 129 intramedullary spinal gliomas. J Neurosurg 54: , Healthcare Cost and Utilization Project: Overview of the Nationwide Inpatient Sample (NIS). ( ahrq.gov/nisoverview.jsp) [Accessed September 26, 2013] 10. Jallo GI, Freed D, Epstein F: Intramedullary spinal cord tumors in children. Childs Nerv Syst 19: , Kalkanis SN, Eskandar EN, Carter BS, Barker FG II: Microvascular decompression surgery in the United States, 1996 to 2000: mortality rates, morbidity rates, and the effects of hospital and surgeon volumes. Neurosurgery 52: , Kothbauer KF: Neurosurgical management of intramedullary spinal cord tumors in children. Pediatr Neurosurg 43: , McGuire KJ, Chacko AT, Bernstein J: Cost-effectiveness of teaching hospitals for the operative management of hip fractures. Orthopedics 34:e598 e601, Mechtler LL, Nandigam K: Spinal cord tumors: new views and future directions. Neurol Clin 31: , Morota N, Deletis V, Constantini S, Kofler M, Cohen H, Epstein FJ: The role of motor evoked potentials during surgery for intramedullary spinal cord tumors. Neurosurgery 41: , Patil CG, Patil TS, Lad SP, Boakye M: Complications and outcomes after spinal cord tumor resection in the United States from 1993 to Spinal Cord 46: , Samii M, Klekamp J: Surgical results of 100 intramedullary tumors in relation to accompanying syringomyelia. Neurosurgery 35: , Sonig A, Khan IS, Wadhwa R, Thakur JD, Nanda A: The impact of comorbidities, regional trends, and hospital factors on discharge dispositions and hospital costs after acoustic neuroma microsurgery: a United States nationwide inpatient data sample study ( ). Neurosurg Focus 33(3):E3, Taylor DH Jr, Whellan DJ, Sloan FA: Effects of admission to a teaching hospital on the cost and quality of care for Medicare beneficiaries. N Engl J Med 340: , 1999 Manuscript submitted March 19, Accepted September 24, Please include this information when citing this paper: published online November 29, 2013; DOI: / SPINE Address correspondence to: Anil Nanda, M.D., M.P.H., Department of Neurosurgery, Louisiana State University Health Sciences, 1501 Kings Highway, Shreveport, LA ananda@ lsuhsc.edu. 141