Continued Increase in Cost of Care Despite Decrease in Stay After Posterior Spinal Fusion for Adolescent Idiopathic Scoliosis

Introduction: Previous studies have demonstrated decreased hospital length of stay (LOS) for children undergoing posterior spinal fusion (PSF) for adolescent idiopathic scoliosis (AIS). Methods: Hospitalization event data from the Kids Inpatient Database were queried for all PSF events for AIS performed in 2009, 2012, and 2016 using diagnosis and surgical codes. Data were subdivided into two groups: pre–enhanced recovery after surgery (ERAS) (2009 and 2012) and post-ERAS (2016). The primary outcome variables were LOS and total treatment charge (adjusted for 2020 inflation). Univariate and multivariate analysis were performed to identify differences in outcome variables. Results: A total of 12,010 unique hospitalization events were identified, 74% female, mean 14.3 years. There was a decrease in LOS (pre-ERAS: 5.4 ± 4.0 versus 4.3 ± 3.2 days, P < 0.0001) with an increase in adjusted total treatment charge (pre-ERAS $193,544.4 ± $108,116.1 versus $200,469.1 ± $110,112.6; P = 0.0013). Pre-ERAS, male sex, smaller hospital, and non-Medicaid insurance were predictive of longer LOS, whereas pre-ERAS, older age, non-White race, male sex, hospital outside the Northeast, and non-Medicaid insurance were predictive of higher treatment costs. Discussion: There continues to be a significant decrease in LOS for PSF hospitalization events for AIS; however, total treatment charges continue to rise. Future research should investigate potential factors influencing total treatment charges after PSF for AIS.

T he advent of enhanced recovery after surgery (ERAS) pathways has seen significant growth across the surgical specialties. These pathways have demonstrated the ability to decrease opioid medication usage and hospital length of stay (LOS). [1][2][3][4] This has been particularly true for children undergoing posterior spinal fusion (PSF) for adolescent idiopathic scoliosis (AIS) where multimodal pain management strategies have been incorporated with early mobilization and have resulted in significant decreases in opioid medication and hospital LOS without differences in postdischarge complications. [5][6][7][8][9][10][11][12][13] There has been a significant increase in publications for ERAS application in patients with AIS since the initial report in 2014. 4 The healthcare utilization and economic benefits of ERAS protocols suggest the potential for a significant societal effect [5][6][7][8]14,15 ; however, no study to date has investigated the nationwide effect of ERAS pathway utilization on the US healthcare system. This study sought to investigate the effect of ERAS pathways after PSF in AIS on LOS and hospital charges in the US healthcare system. We hypothesized that the pervasiveness of ERAS pathways would result in significant decreases in LOS and hospital charges after PSF for AIS.

Methods
This study was deemed exempt from review by the institutional review board.

Data Source
Data were reviewed from annual hospitalization events for years 2009, 2012, and 2016 from the Kids Inpatient Database (KID). KID represents the largest pediatric accessible all-payer inpatient healthcare database in the United States. KID was created and is managed by The Agency for Healthcare Research and Quality's Healthcare Cost and Utilization Project. KID sampling includes complicated and uncomplicated births, as well as other pediatric inpatient procedures from community, nonrehabilitation hospitals in the United States. The KID database contains 107 data elements, using the International Classification of Diseases, Ninth and 10th Revision, Clinical Modification (ICD-9 and ICD-10) formatted to code all the diagnoses and procedures. Databases are published on approximately 3year cycles with over three million hospital stays available for each 3-year database. KID is designed to allow accurate calculation of medical condition incidences using Healthcare Cost and Utilization Project-provided trend weights. A detailed overview of the KID design is available at https:// www.hcup-us.ahrq.gov/kidoverview.jsp.

Patient Selection
Hospitalization event data from the KID database for surgical intervention for AIS were identified by ICD codes. Databases are published on approximately 3year cycles, with the 2016 database representing the most current available data. Data were selected to review for year groups 2009, 2012, and 2016. For data from 2009 and 2012 data sets, ICD-9 code 737.30 was used, and ICD-10 code M41.12X was used for 2016 data. ICD procedural codes were used to include all hospitalizations corresponding to a PSF procedure (8104, 8105, 8106, 8107, 8108, 8451, 8452, and 8459). Hospitalization events were excluded from inclusion if they were ,10 years of age at the time or surgery or .18 years. All hospitalization events corresponding to an anterior spinal fusion were also excluded or demonstrated a nonidiopathic scoliosis diagnosis.
The ERAS pathway was first introduced to the pediatric spinal deformity community in 2014 and consists of preferential admission to the hospital floor with early Foley and drain discontinuation, mobilization the day after surgery, combined with early institution of oral feeding, and transition to oral pain medications. 4 As such, patient data were subdivided into two cohorts: (1) 2009 and 2012 and (2) 2016. Data including national rates, hospital costs, hospital charges, and LOS were investigated. Costs reflect the expenses incurred in the production of hospital services, whereas charges represent hospital billing. Potential confounding variables for hospital charges and LOS were collected, including age at the time of surgery, ethnicity, obesity diagnosis, and sex. Additional variables unique to patient socioeconomic status, region of treatment, hospital setting and size, and patient insurance type were also collected.
Statistical Analysis SAS 9.4 was used for all statistical analyses. The significance level was set at 0.05. Descriptive statistics were calculated for all qualitative and quantitative variables. The outcomes of interest were LOS and total treatment charge. Total treatment charge was adjusted for inflation Dr. Shaw or an immediate family member serves as a board member, owner, officer, or committee member of the NASS and AAOS. Dr. Ange or an immediate family member serves as a paid consultant to DePuy and OrthoPediatrics; has received research or institutional support from OrthoPediatrics; and serves as board member for the Journal of Pediatric Orthopaedics, POSNA, Spine Journal, Journal of Spine Deformity, and the Scoliosis Research Society. Dr. Fletcher or an immediate family member is a member of a speakers' bureau or has made paid presentations on behalf of NuVasive, Zimmer, Medtronic, and OrthoPediatrics; serves as a paid consultant to Medtronic; serves as a board member, owner, officer, or committee member of the SRS and POSNA; and serves on the editorial board for the JPOSNA. Neither of the following authors nor any immediate family member has received anything of value from or has stock or stock options held in a commercial company or institution related directly or indirectly to the subject of this article: Dr. Ange and Dr. George. rates and normalized to 2020 rates. Independent variables of interest were age, race, sex, obese (comorbidity or no comorbidity), geographic region of hospital, hospital bedsize (small, medium, and large [the definition of each bedsize category varies based on the geographic region of the hospital, in location, and teaching status]), hospital location and teaching institution (rural, urban nonteaching, or urban teaching), insurance type (Medicare, Medicaid, private insurance, self-pay, no charge, or other), median household income by zip code (0 to 25th percentile, 26th to 50th percentile, 51st to 75th percentile, or 76th to 100th percentile), and volume (number of discharges in the sample for the stratum). Individuals from 2009 and 2012 were combined into one cohort (pre-ERAS), whereas the 2016 was defined as the post-ERAS cohort. Because of the non-normal nature of the data, Mann-Whitney U tests were calculated to examine differences in LOS and total treatment charge across the two cohorts. To examine the differences in the two cohorts on the outcomes while controlling for the independent variables of interest, individual multiple linear regression models were calculated. Each independent variable was first examined in a bivariate model on each of the outcomes. Independent variables that were significant at the 0.1 alpha level in a simple linear model were used to build a multivariable linear regression model. Akaike Information Criteria (AIC) were examined for each possible model. Smaller AIC values indicate a better fit, so the model with the lowest AIC was determined to be the best model for each outcome variable.

Results
A total of 12,010 hospitalization events for PSF were identified, corresponding to a diagnosis of AIS over the 3-   year groups. Table 1  A summary of univariate comparisons (Mann-Whitney U tests) to identify differences in hospital LOS and total adjusted treatment costs according to the pre-ERAS or post-ERAS year groups is presented in Table 2. Hospital LOS was significantly lower in the post-ERAS cohort (4.3 The results of the simple linear regression results (bivariate models) on total treatment charges, adjusted for 2020 inflation rates, are summarized in Table 4. Of the tested variables, only hospital size and presence of comorbid obesity were not statistical predictive of changes in adjusted treatment charges. The top five predictive variables for greater adjusted treatment charges include surgery in an urban teaching hospital (slope = 223688, P , 0.0001), surgery performed outside the Northeast United States (slope = 15602,

Figure 2
Graph showing the summary of total hospital charge, adjusted for inflation to the 2020 year rates, after posterior spinal fusion for adolescent idiopathic scoliosis according to year group.
P , 0.0001), male sex (slope = 13516, P , 0.0001), non-White race (slope = 8406, P , 0.0001), and post-ERAS (slope = 6294, P = 0.0008). The final model for LOS included ERAS year groups, race, hospital bedsize, insurance type, and sex. The final model for adjusted total treatment charge included race, sex, hospital region, hospital bedsize, hospital location and teaching, and median household income by zip code. These variables were entered into a multivariable linear regression for hospital LOS and total adjusted treatment charges (Table 5). Independent factors predictive of shorter hospital LOS included post-ERAS (P , 0.0001), larger hospital bedsize (P = 0.0028), and female sex (P , 0.0001). Independent factors predictive of lower total adjusted treat charge included female sex (P , 0.0001), hospital located in the Northeast (P , 0.0002), smaller hospital size (P = 0.0002), nonurban teaching hospital (P , 0.001), and Caucasian race (P , 0.001).

Discussion
The implementation of ERAS pathways after PSF for AIS has been shown to decrease hospital LOS and result in lower treatment charges. In this study using a nationwide database, surgeries performed before the advent of ERAS pathways demonstrated longer hospital LOS but were also associated with lower adjusted total treatment charges. Independent risk factors for shorter hospital LOS included surgery during the post-ERAS cohort, female sex, and larger hospital bedsize. Surgery performed in the Northeastern United States, nonurban teaching hospital setting, female sex, and Caucasian patient were predictive of lower adjusted treatment charge. The length of hospital stay after PSF has significantly changed since the procedure was initially introduced for the treatment of scoliosis. No longer is LOS measured in weeks 16 but is now measured in days. [5][6][7][8]11,14 The introduction of ERAS protocols for AIS surgery, including multimodal pain management strategies, has been shown to produce substantial reductions in LOS but also improved pain control, faster postsurgical recovery, and lower opioid medication requirement. 4,5,17 As the adoption of ERAS protocols has shown growing implementation since their introduction, 17 we hypothesized that hospital LOS would significantly decrease between year cohorts and was supported with a mean 1.1 day decrease in hospital LOS in the post-ERAS cohort (4.3 6 3.2 days versus 5.4 6 4.0 days, P , 0.0001).
Additional factors were found to influence the duration of hospital LOS, including male sex, non-White race, non-Medicaid insurance, and smaller hospital size, all being associated with longer LOS. The role of sex on hospital LOS after surgery has been previously reported following various orthopaedic surgeries. 18 Specific to PSF for AIS, few previous studies have focused on sex differences in LOS. Elsamadicy et al 19 reported that female sex was an independent predictor of prolonged hospital LOS using the KID database. In contrast to these findings and using a substantially larger population, we found that male sex was not only predictive of longer hospital LOS but was also predictive of greater total adjusted treatment charges.
Treatment costs after PSF for AIS have been of growing concern given continued and persistent increases in mean hospital charges over time. 20,21 Vigneswaran et al 20  treatment charges in the current study. In fact, the adjusted treatment charges significantly increased from a mean of $167,846.2 in the pre-ERAS cohort to $185,105.4, following the rising trends identified in previous studies. 20,21 One possible explanation for this continued rise may be found in the cost and utilization of spinal implants. Kamerlink et al 22 assessed the hospital cost and charge data of 125 consecutive spinal fusion cases, including anterior and posterior procedures, to treat AIS. The mean charge varied based on the Lenke curve classification but was reported at a mean $73,843 per case. The largest contributor to treatment cost was spinal implants, accounting for 29% of the overall cost, followed by intensive care unit/patient room costs (22%), operating room use (9.9%), and bone graft (6%).
Over time, there has been a trend toward the use of a greater number of spinal implants, especially pedicle screws, during PSF for AIS. 4,21,23 The Minimize Implants Maximize Outcomes Study Group has investigated the influence of implant density on curve correction and clinical outcomes for children with AIS in the United States treated at 12 pediatric institutions. Higher implant densities (.1.8 screws per vertebral level fused) have been shown to produce slight but significant increases in curve corrections (69% versus 66%) for main thoracic curve patterns with similar but slightly improved clinical outcomes on patient-reported outcome measures in comparison to low implant density constructs (,1.4 screws per vertebral level fused). 24 Given these slight improvements in deformity correction and outcomes; however, the question becomes whether these improvements are cost-effective for high-density implant constructs. High-density implant constructs for PSF result in significantly higher costs, 25 and surgical implants have demonstrated continued increase in cost over time. Martin et al 21 reviewed drivers of treatment charges related to PSF for AIS over a 4-year period, finding an annual increase in implant costs of 27.6% each year. The percentage of the total treatment charge dictated by implant cost increased from 28% to 53% over that 4year period. Larson et al 26 sought to investigate the potential mean reduction in treatment cost that could be achieved with the use of a lower-density implant construct in terms of cost and complications associated with malpositioned screws. The lower-density construct was predicted to result in a mean of 3.2 fewer screws per patient with a potential annual cost saving of between $11 and 20 million dollars, representing between 4% and 7% reduction in treatment costs.
Previous studies have identified variations in LOS and hospital charges based on geographic, hospital, and patient demographic variables after treatment of AIS. [27][28][29] Menger et al 29 reviewed the NIS database for children with AIS undergoing surgery, indicating that lower volume centers were associated with higher treatment charges but with shorter LOS after surgery. The current data found that higher volume centers demonstrated no difference in hospital LOS but were associated with higher adjusted treatment charges after PSF. However, larger hospital bedsize was predictive of both shorter hospital LOS despite being also predictive of greater total adjusted treatment costs.
In addition, the geographic location of surgery also influenced on total adjusted treatment charges with surgeries performed in the Northeastern United States having lower adjusted treatment charges than the remainder of the United States, despite no difference in hospital LOS. Daffner et al 27  Children undergoing surgery in the Western United States were identified as having the highest hospital charges, with the lowest charges seen in the Southern United States. In addition, hospital LOS was highest in the Midwest at a mean 6.5 days and lowest in the South at 5.2 mean days. The current data suggest that across the United States, geographic differences in LOS have abated but geographic treatment charge differences do remain prevalent. This could be in part explained by the type of hospital performing the surgeries by location as teaching hospitals were identified as having higher treatment charges, which has been supported in the literature. 28 The current study has several limitations that require recognition. As a longitudinal analysis of a blinded administrative inpatient database, causation of the identified trends is unable to be investigated or determined. There are numerous additional, inherent limitations with the use of a large database such as KIDS. There is a lack of clinical detail regarding the hospitalization events, ranging from curve magnitude, classification, treatment approach, and instrumentation density, all of which can have significant implications on both treatment cost and hospital LOS. 22,24 In addition, the data set identifies surgical treatment using ICD procedural codes rather than the more commonly reported Current Procedural Terminology (CPT) codes. The current analysis is reliant on the accuracy of coding for diagnosis and treatment. Given the presence of coding inaccuracies as well as the potential for data transfer errors or exclusions in large database studies, this potential source of error cannot be overlooked. 30 Given the utilization of a less familiar coding system for procedure codes, this may have an additional effect on coding accuracy.
Given the restraints of the current data set, we were unable to isolate and identify individual hospitals to identify them by the presence or absence of postoperative ERAS protocols. The surgeries performed included a nationwide collection of surgeons, introducing variation in treatment and surgical technique and the potential for indication bias. In addition, this investigation has been performed using data generated within 2 years of the first publication of ERAS pathways for AIS. To date, the 2016 data set is the most recent package available for analysis, which allows for only 1 year of potential post-ERAS data for analysis. Additional, more recent data would further assist the current analysis regarding the potential effect of ERAS protocols. Given the ongoing trend in decreased hospital LOS identified in previous studies, 20 this difference in LOS may be a continued reflection of this trend rather than a true depiction of the nationwide effect of ERAS protocols.
In conclusion, using a large nationwide database of pediatric hospitalization events occurring in the United States, there continues to be a significant reduction in hospital LOS after PSF hospitalization events for AIS which persists since the introduction of ERAS protocols; however, the effect of ERAS protocols was not able to be isolated by the current data set. Interestingly, over the assessed time periods, the adjusted total treatment charges were found to increase rather than decrease as has been shown in previous hospital-based studies. Geographic and hospital setting variables do have a significant influence on both hospital LOS and total adjusted treatment charge after surgery. Future research should further investigate factors influencing this rise in adjusted total treatment charges to optimize the costeffectiveness of PSF.