The cost of medical care in general, and spinal surgery specifically, has dramatically increased over the last decade, with rising costs associated with cervical, lumbar, and deformity surgery.1–3 Recently, Cutler and Ghosh4 showed that spinal conditions ranked number 10 on the list of the 17 most expensive conditions among Medicare patients 65 years of age and older.
With the rising cost of healthcare, new methods of delivering and paying for care are being developed to help create a fiscally sustainable system in the United States. The Bundled Payments for Care Improvement (BPCI) Initiative was created by the Center for Medicare and Medicaid Services (CMS) in order to increase value by both improving quality and decreasing the cost of care.5 Under this new system, CMS would provide a single payment to cover the entire episode of care, including all of the associated fees related to that episode. The transfer of risk away from payers and onto hospitals and providers would potentially drive down costs to create a more financially sustainable model. To what extent this would succeed in driving down the cost of care while maintaining or improving quality (and thereby increasing value), however, remains unproven.
Our larger, urban, academic institution entered into the BPCI initiative for diagnosis related groups (DRG) 459 and 460 (spinal fusion except cervical with and without major complication or comorbidity) from 2013 to 2014. Within the four different payment models, we entered into Model 2. Under this system, the episode of care includes all payments made from 72 hours prior to surgery extending up to 90 days after discharge. In this paper, we present a retrospective analysis of the financial, patient, and procedural data from our 2-year experience in which we failed to result in a net savings over the episode.
MATERIALS AND METHODS
Between January 2013 and December 2014, all Medicare patients at our institution undergoing non-cervical posterior spinal fusion between one and eight levels were enrolled into the BPCI initiative. Under the BPCI initiative, all patients classified under DRG 459 and 460 were enrolled. DRG 459 includes all patients undergoing spinal fusion (except cervical), between one and eight levels, performed through a posterior approach with one or more major complication or comorbidity (MCC). DRG 460 includes a similar cohort of patients, but without MCC. DRG 459 and 460 specifically apply to surgeries performed through a posterior approach only, which includes transforaminal lumbar interbody fusion (TLIF) and posterior lumbar interbody fusion (PLIF). Anterior, lateral, or combined approach surgeries are excluded from these DRGs and are given a different designation. Procedures performed for infection, tumor, deformity, or including more than eight levels are also classified within a separate DRG.
Based upon a 2009 to 2012 cohort of patients from DRG 459 and 460, CMS set a target price for each DRG based on the previous average Medicare payments. Under the BPCI program, CMS continues to pay for all expenses on a fee-for-service basis. However, all Medicare payments for the BPCI initiative are reconciled quarterly against this target price. Based on performance compared with the target price, the hospital either pays the difference or receives a payment from CMS of the difference. CMS tracked all patients enrolled into the program, including all readmissions, within 90 days at both our institution and all outside hospitals.
CMS provided the financial data for the cohort including all costs incurred during the 90 days bundle. This data set was used to calculate payments per DRG, length of stay (LOS), readmission rates, discharge disposition, and the costs associated with discharge disposition. We compared the BPCI cohort with a baseline cohort of patients who underwent non-cervical spinal fusions and were assigned to DRGs 459 or 460 from 2009 to 2012 prior to the initiation of the BPCI initiative upon which our target price was based.
Through electronic hospital records, the baseline retrospective cohort and the BPCI cohort were analyzed to compare differences in surgical practices. Number of levels fused, primary versus revision surgery, and use of interbody implants were compared between the two groups.
Statistical analysis was performed with the statistical software SPSS v23.0 (IBM Corp, Armonk, NY). Mann–Whitney U tests were used to compare mean cost and LOS between groups. Chi-square tests were used to evaluate differences in readmission rates, discharge disposition, and case complexity. Statistical significance was set at P < 0.05 for all comparisons.
Over the 2-year study period, 350 patients were enrolled into the BPCI initiative at our institution (21 patients for DRG 459 and 329 patients for DRG 460). The baseline group contained 518 patients (35 patients for DRG 459 and 483 patients for DRG 460).
Average Medicare Payments
Mean 90 day Medicare payment for all patients in the BPCI group was not significantly different than the total baseline group ($52,655 ± 27,028 vs. $48,913 ± 24,764, P = 0.360). Likewise, for DRG 459 with MCC, mean total payment did not differ between BPCI and baseline cohorts ($76,866 ± 26,579 vs. $90,032 ± 44,626, P = 0.226). For DRG 460 without MCC, the mean total payment per episode in the BPCI group was greater than the mean total payment in the baseline group ($51,105 ± 26,347 vs. $45,934 ± 19,638, P = 0.001). The differences in payments between groups are summarized in Table 1.
During the BPCI period, average payment per episode for DRG 459 ($76,866) was less than the target price established by CMS ($82,919). In this group we were able to effect an average savings of $6053 per patient. Averaged over the 21 patients in this group, the sum of these savings was $127,113. For DRG 460, the average payment per episode during the BPCI period ($51,105) was greater than the established target price ($46,803). For the patients in DRG 460 there was a per-patient average loss of $4302, and a total loss of $1.42 million. There was an overall net loss of $1.29 million for the 350 patients enrolled in the BPCI initiative.
Length of Stay
Mean LOS for all patients in the BPCI group was less than mean LOS for the baseline group. (4.58 ± 2.51 days vs. 5.13 ± 3.75 days, P = 0.009) (Figure 1). For DRG 459, there was no difference in LOS between groups (7.67 ± 4.73 vs. 10.83 ± 7.70, P = 0.960). For DRG 460, the decrease in mean LOS for the BPCI cohort compared with the baseline cohort did not reach statistical significance (4.38 ± 2.16 vs. 4.72 ± 2.89, P = 0.055). The differences in LOS between groups are summarized in Table 1.
Forty-five of the 350 total patients in the BPCI group were readmitted during the 90 days episode of care, for an overall readmission rate of 12.9% during the BPCI period. This was not a statistically significant decrease in readmissions compared with the baseline rate of 14.5% (P = 0.497). Of the 21 DRG 459 patients during the BPCI period, five (23.8%) were readmitted. This decrease from the baseline DRG 459 readmission rate of 31.4% did not reach statistical significance (P = 0.054). For DRG 460, there was no difference in readmission rates between groups (12.2% vs. 13.3%, P = 0.647).
There was a significant difference in discharge disposition between the BPCI group and the baseline group (P < 0.001). In the BPCI cohort, a greater percentage of patients were discharged home with home health aide (53.4% vs. 34.0%) while a lower percentage of patients were discharged to inpatient rehabilitation (14.0% vs. 32.0%) compared with the baseline cohort. However, more patients in the BPCI cohort were discharged to skilled nursing facility (SNF) (16.9% vs. 10.0%), and fewer patients were discharged home with self-care only (15.4% vs. 24.0%; Figure 2).
There were large cost variations between different discharge dispositions. One patient was discharged to a long-term care facility with a Medicare payment of $27,024. Average cost of inpatient rehabilitation was $23,194 and discharges to SNF averaged $17,726. Discharge home with home health aide was the least expensive disposition with an average cost of $3789 (Figure 3).
There were important differences between the baseline cohort and the BPCI cohort with respect to the types of cases performed. In the baseline cohort, 11% of patients underwent a revision/refusion procedure, while 14% of patients in the BPCI cohort were revision/refusion cases (P = 0.19). 11% of patients underwent four to eight levels fusions in the baseline cohort, the remainder undergoing one to three levels fusions. The number of longer fusions increased during the 2 years under the BPCI Initiative, with 15% of patients undergoing four to eight levels fusions, although this did not reach statistical significance (P = 0.08). The number of revision four to eight levels fusions increased from 2% during the baseline period to 4% in the BPCI period, although this did not reach significance (P = 0.08).
The largest change noted was due to a trend towards increasing number of interbody procedures utilizing TLIF. From 2009 to 2012, only 2% of patients had anterior interbody support used in the fusion. During the BPCI period, this increased to 16% (P < 0.001).
We considered the revision, four to eight levels fusions, or TLIF cases to be considered “complex” cases, whereas the shorter primary fusions without anterior interbody devices placed were not. If analyzed in this way, 23% of the baseline cohort cases would be considered complex cases, whereas 45% of the cases in the BPCI cohort were complex (P < 0.001). See Figure 4.
The BPCI initiative for lower extremity total joint arthroplasty (TJA) has achieved significant successes, with institutions demonstrating decreases in overall episode costs.6,7 An analysis of 184 hospitals participating in the BPCI program demonstrated that the bundled payment system led to decreased Medicare spending for TJA without decreasing quality of outcomes.8 Additionally, implementation of BPCI Model 2 for TJA was successful at our own institution during the same time period as our engagement in the spine BPCI program. We saved an average of approximately $7000 per episode for the lower extremity joint bundles.9 Programs designed to increase efficiency and improve patient safety under a bundled payment system have led to decreased hospital length of stay, increased percentage of patients discharged to home, and ultimately overall reduced episode cost.10–13
Despite our success with the TJA bundles, we were unable to achieve the cost savings required to be successful in the BPCI program for non-cervical spinal fusion. When compared with the baseline cohort, the patients enrolled in the BPCI cohort achieved a shorter length of stay, fewer discharges to expensive inpatient rehabilitation facilities, and increased discharges with home health aid. Nonetheless, average cost per episode increased by an average of $3742 and the larger DRG 460 group without MCC was found to have a significantly greater episode cost during the BPCI period.
We believe the difficulty in achieving cost savings was due in large part to an increasing case complexity in the BPCI cohort when compared with the baseline cohort. Due in part to changes in referral patterns at our institution during the study period, there was an increased number of fusion levels, increased number of revision cases, and increased utilization of interbody fusion (mostly TLIF) in the BPCI cohort. The data obtained from CMS do not include details of the cost components comprising the overall total episode cost. We were therefore unable to analyze the specific variables driving the increased cost of the BPCI group. However, interbody fusions and greater number of operated levels have both been associated with higher costs.14,15 Given the greater percentage of complex cases in the BPCI group, these variables may explain the lack of cost savings compared with the baseline pre-BPCI group.
The changing practice patterns at our institution are ubiquitous across the Medicare population, with complex spinal arthrodesis and the associated costs drastically increasing. An analysis of the National Inpatient Sample from 1998 to 2008 demonstrated that the annual number of spinal fusions increased by 2.4-fold (137%), while mean hospital charges for spinal fusion increased 3.3-fold.16 Deyo et al 17 looked at Medicare claims from 2002 to 2007. They found that during those years, the rates of laminectomy and simple fusions decreased while complex fusion surgery increased 15-fold. Additionally, hospital charges increased 40%. Bae looked at the National Inpatient Sample to analyze trends in lumbar spinal stenosis from 2004 to 2009.18 They found that simple fusions increased from 21.5% to 31.2%, the use of bone morphogenic protein (BMP) doubled from 14.5% to 33.0%, and the use of interbody devices increased from 28.5% to 45.1%.
Based on our experience in the BPCI initiative, we believe the current DRG system appears not well-suited to managing the increasing case complexity and heterogeneity present in modern spinal surgery. A simulation of episode of care costs based on DRG demonstrated wide variability between patients within a single DRG.19 DRGs 459 and 460 are particularly predisposed to wide cost variation as they do not differentiate between a single level or eight-level fusion.
Wright et al 20 demonstrated that cost variation among spinal procedures was significantly higher than among TJA DRGs. In the case of spine procedures, the DRGs contain an overly varied case load to adequately control for costs, partially due to failure to account for surgical approach. While the differences between primary and revision TJA is easy to recognize and separate into different DRGs, the distinction between primary and revision spinal surgery is more complicated. Whether the focus should be on primary versus revision or on the specific surgical technique or invasiveness employed, the status quo of the current DRG system has significant consequences for spine surgery.
While the use of the more complex surgical techniques adopted during the BPCI period are thought to yield better clinical outcomes with greater cost-effectiveness ratios, the total expenditures for the episode are also increased.14 Our results demonstrate that changes in clinical practice can lead to episode costs that exceed target prices set based on historical data. This highlights the financial risk imposed by a bundled payment system for hospitals that undergo a change in practice or adopt innovate but costly treatment options. Additional flexibility in the target price based on case complexity and advancements of surgical technique would likely prevent this unintended consequence of the bundled payment system.
There are several limitations to our study. First, we were unable to analyze specific cost components, as discussed above. Future studies should include a more detailed cost analysis to determine the variables responsible for cost disparities. This should include an evaluation of any coexisting diseases that may influence the overall cost of a patient's hospitalization. Second, as we present data from only a single institution, it is possible that our experience does not necessarily reflect the case load and experience at other institutions. Third, the study was performed retrospectively. Although much of the data is provided by CMS, the operative data are subject to selection bias, and it is possible that there are other considerations that may account for increased costs that may not have been considered.
While our institution was successful in achieving significant cost savings for the lower extremity arthroplasty bundles, we were unable to achieve a similar cost savings for spine fusions despite employing the same tools. The heterogeneity of patients and procedures encompassed within DRGs 459 and 460 combined with increasing surgical complexity likely prevented our institution from achieving the target price for an episode of care.
As providers take on more of the risk and financial burdens of patient care, physicians, and hospitals may find that providing care becomes financially untenable. The answer to this problem is therefore for providers (physicians and hospitals) and payers (CMS) to work to create appropriate DRGs that can better anticipate the cost of each episode in order to create a financially sustainable model. This work has already begun for the lower extremity arthroplasty DRGs, where primary and revision arthroplasty are now encompassed within unique DRGs. But without similar adjustments in the spine fusion DRGs with the goal of achieving more consistent reimbursements, the current model will remain financially unsustainable.
At the same time, providers must continue to perform high quality research on new expensive equipment and techniques to justify their use in a new era of value-based care. The variability in how spine fusions are performed—BMP versus no BMP, TLIF versus no TLIF, MIS versus open—makes the spine fusion bundle model particularly challenging. A movement to standardize practice in an attempt to improve outcomes as well as control costs will help limit this cost variability.
1. Marquez-Lara A, Nandyala SV, Fineberg SJ, et al. Current trends in demographics, practice, and in-hospital outcomes in cervical spine surgery: a national database analysis between 2002 and 2011. Spine (Phila Pa 1976)
2. Zygourakis CC, Liu CY, Keefe M, et al. Analysis of national rates, cost
, and sources of cost
variation in adult spinal deformity. Neurosurgery
3. Zygourakis CC, Liu CY, Wakam G, et al. Geographic and hospital variation in cost
of lumbar laminectomy and lumbar fusion for degenerative conditions. Neurosurgery
4. Cutler DM, Ghosh K. The potential for cost
savings through bundled episode payments. N Engl J Med
6. Althausen PL, Mead L. Bundled payments for care improvement: lessons learned in the first year. J Orthop Trauma
2016; 30 (suppl):S50–S53.
7. Froemke CC, Wang L, DeHart ML, et al. Standardizing care and improving quality under a bundled payment
initiative for total joint arthroplasty. J Arthroplasty
8. Dummit LA, Kahvecioglu D, Marrufo G, et al. Association between hospital participation in a medicare bundled payment
initiative and payments and quality outcomes for lower extremity joint replacement episodes. JAMA
9. Iorio R, Bosco J, Slover J, et al. Single institution early experience with the bundled payments for care improvement initiative. J Bone Joint Surg Am
10. Bolz NJ, Iorio R. Bundled payments: our experience at an Academic Medical Center. J Arthroplasty
11. Doran JP, Zabinski SJ. Bundled payment
initiatives for Medicare and non-Medicare total joint arthroplasty patients at a community hospital: bundles in the real world. J Arthroplasty
12. Dundon JM, Bosco J, Slover J, et al. Improvement in total joint replacement quality metrics: year one versus year three of the bundled payments for care improvement initiative. J Bone Joint Surg Am
13. Iorio R, Clair AJ, Inneh IA, et al. Early results of medicare's bundled payment
initiative for a 90-day total joint arthroplasty episode of care. J Arthroplasty
14. Bydon M, Macki M, Abt NB, et al. The cost
-effectiveness of interbody fusions versus posterolateral fusions in 137 patients with lumbar spondylolisthesis. Spine J
15. Twitchell S, Karsy M, Reese J, et al. Assessment of cost
drivers and cost
variation for lumbar interbody fusion procedures using the Value
Driven Outcomes database. Neurosurg Focus
16. Rajaee SS, Bae HW, Kanim LE, et al. Spinal fusion in the United States: analysis of trends from 1998 to 2008. Spine (Phila Pa 1976)
17. Deyo RA, Mirza SK, Martin BI, et al. Trends, major medical complications, and charges associated with surgery for lumbar spinal stenosis in older adults. JAMA
18. Bae HW, Rajaee SS, Kanim LE. Nationwide trends in the surgical management of lumbar spinal stenosis. Spine (Phila Pa 1976)
19. Ugiliweneza B, Kong M, Nosova K, et al. Spinal surgery: variations in health care costs and implications for episode-based bundled payments. Spine (Phila Pa 1976)
20. Wright DJ, Mukamel DB, Greenfield S, et al. Cost
variation within spinal fusion payment groups. Spine (Phila Pa 1976)