Drivers of Cost in Primary Single-Level Lumbar Fusion Surgery

Abstract

Study Design:

Retrospective cohort.

Objectives:

Allocating cost is challenging with traditional hospital accounting. Time-driven activity-based costing (TDABC) is an efficient method to accurately assign cost. We sought to characterize the variation in direct total hospital cost (THC) among both lumbar fusion approaches and surgeons.

Methods:

Patients were treated with single-level anterior interbody (ALIF), lateral interbody (LLIF), transforaminal interbody (TLIF), instrumented posterolateral (PLF) or in-situ fusion (ISF) for degenerative disease. Process maps were developed for preoperative, intraoperative and postoperative care. THC was composed of implant, medication, other supply, and personnel costs. Linear regression and descriptive statistics were used to analyze THC variation.

Results:

A total of 696 patients underwent surgery by 8 surgeons. Approximately 50% of THC variation was associated with procedure choice while patient characteristics explained 10%. Implants (including biologics) accounted for 45% of cost. With reference to PLF, THC ranged from 0.6x (ISF) to 1.7x (LLIF). Implant cost ranged from 2.5x reference (LLIF) to 0.1x (ISF). There was a 1.7x difference between the highest THC surgeon and the lowest. The fusion type with the highest THC variation was TLIF. The surgeon with the highest TLIF THC was 1.5x more expensive than the surgeon with the lowest.

Conclusions:

Surgeon-based choices have the greatest effect on THC variation and represent the largest opportunities for cost savings. Primary single-level lumbar fusion THC is driven primarily by fusion type. Implants, including biologics, account for nearly half this cost. Future work should incorporate outcomes data to characterize the differential value conferred by higher THC fusions.

Introduction

The cost of health care in the United States leads developed nations by a wide margin. ¹ Despite being more expensive, the outcomes purchased with these expenditures lag behind lower cost nations. ² As such, there is increasing attention paid toward measuring and understanding value in healthcare. In an effort to control expenditures, providers are increasingly placed at risk for the cost and outcomes associated with the care they deliver. This is reflected in the various bundled payment models now common in, for example, total joint replacement and cardiac care. These value-based reimbursement programs are more challenging to implement in spine surgery due to a variety of factors including poor specificity in spine diagnosis coding, difficulty defining homogeneous patient populations and case complexity.^3,4 On the other hand, spine surgery is a natural focal point in this trend toward delivering higher value. The number of elective lumbar fusions performed increased 276% between 2002 and 2014 with costs increasing accordingly. ⁵ Healthcare spending for low back and neck pain rank third-highest, behind diabetes and ischemic heart disease. ⁶ Developing a deep understanding of the cost of spine surgery will be critical in navigating these emerging value-based reimbursement models.

Historically, understanding the cost of healthcare has been hindered by complexity, inaccurate allocation of indirect costs and flawed costing systems. ⁷ This leads to difficulty capturing the true cost of a patient’s episode of care. Time-driven activity-based costing (TDABC) is an innovative method to accurately assign cost that has been well described in both the business and healthcare literature.^8-10 Costs of otherwise difficult to quantify resources, such as operating room nursing care, can be allocated based on the cost per unit time and quantity of time consumed. Using TDABC, the cost structure of spine surgery can be more accurately captured, facilitating more informed decision making and value analysis.

In this study we sought to characterize the direct cost of inpatient care associated with single-level primary lumbar fusion using TDABC. Our primary goal was to identify patient demographic and surgical case characteristics associated with higher cost. Our secondary goal was to identify areas of variation between surgeons and fusion techniques. By defining this variation, we hope to identify opportunities for improvement.

Methods

Cases

A retrospective review of all patients undergoing single-level lumbar fusion surgery at a musculoskeletal specialty hospital from January 1, 2015 to December 31, 2017 was performed. Patients undergoing single-level, primary lumbar fusions with or without instrumentation, with or without interbody (regardless of approach), with or without concomitant decompression for degenerative disease were included. Patients with non-degenerative pathology (i.e. tumor or trauma), multilevel procedures, prior lumbar fusion and revision fusions were excluded. All subjects were categorized by their fusion surgery technique: Lateral lumbar interbody (LLIF), anterior lumbar interbody (ALIF), transforaminal lumbar interbody (TLIF), instrumented posterolateral lumbar (PLF) and non-instrumented or in situ (ISF). All fusion techniques utilized instrumentation with the exception of ISF. To reduce the impact of learning curve effects, we excluded surgeons who had been in practice for less than 4 years and/or performed less than 40 cases of any fusion technique that represented 20% or more of their volume.^11-14

Cost

The cost of care at our institution is calculated with TDABC using the Avant-garde Health (Boston, MA) platform. TDABC was initially conceived to address challenges associated with traditional costing methodologies used broadly in business. ¹⁵ More recently, TDABC has been applied to healthcare as an alternative to legacy systems that use ratio of costs to charges (RCC) and allocations based on Relative Value Units (RVU). Current systems used to calculate health care costs are deeply flawed. Both RCC and RVU costing make faulty assumptions that sacrifice accuracy for the sake of simplifying implementation. For example, RCC allocates costs in proportion to charges for individual services. This incorrectly assumes the cost of care is proportional to charges. It is well-known that hospital charges for individual items and services are generally unrelated to their true costs. The RVU method involves dividing total spending by total RVUs generated. This method ignores the cost of non-reimbursed services and assumes every RVU consumes the same set of resources.

Rather than a top down approach, TDABC begins by identifying all steps involved in a particular care pathway. For surgical treatment, this creates a process map for preoperative, intraoperative and postoperative care. Within each step, this map identifies the supply resources utilized and amount of time that personnel are involved with patient care. The latter is extracted from case-level electronic medical record data to calculate on a patient-by-patient basis the amount of time different personnel spent providing care. Personnel cost is determined by using the cost per minute of each personnel type (the capacity cost rate) and the number of minutes each personnel type is involved in care. The capacity cost rate is calculated by dividing personnel expenses by personnel available capacity. Personnel expenses include salary and fringe benefits. Available capacity is the total number of minutes that person is available to work per year excluding non-productive time (e.g. breaks, continuing education). Actual purchase prices were used for supply costs. Our institution utilizes reference pricing (i.e. a single price that applies to all vendors) for common, more commoditized implants such as pedicle screws. Niche products with unique features are priced separately.

The current study considered direct personnel and supply costs. Total direct hospital cost included 7 categories: implant (including biologics), medication, non-implant non-medication (e.g. drapes), surgery personnel (including preoperative), post-anesthesia care unit personnel, inpatient personnel and other. Costs were converted from dollar amounts to standardized mean scores for analysis and presentation.

Statistical Analysis

Chi-square tests of independence and Fisher’s exact test were used to compare patient demographic and surgery case characteristics across fusion techniques. Multivariable linear regression was used to evaluate the relationship between patient demographic and surgery case characteristics with cost variability across subjects. Potential predictor variables included sex, age, body mass index (BMI), American Society for Anesthesiologists (ASA) score, smoking history, insurance type, fusion technique, fusion level, and intraoperative variables such as durotomy, revision decompression and cyst presence. Model fit was assessed by observing the change in explained variance (R ² ) as individual covariates were added to an intercept-only model.

Post-hoc descriptive analyses were conducted to examine areas of cost variance between both different fusion techniques and surgeons. Average costs by cost category were presented as a proportion of overall costs by fusion technique. To describe the relative difference in magnitude between fusion techniques, mean cost is presented for each procedure as a percentage of mean PLF costs as a reference. Standardized costs are presented for each surgeon, and relative category costs are compared between surgeons with the highest and lowest average TLIF costs. All statistical analyses were performed in SAS v9.4 (SAS Institute, Cary, NC). This study received approval from the New England Baptist Hospital Institutional Review Board.

Results

A total of 696 cases performed by 8 surgerons were identified for review. Among fusion techniques, subjects differed in terms of age (P < 0.001), ASA class (P = .007), insurance (P < 0.001), incidental durotomy (P = 0.022), revision decompression (P = 0.047) and fusion level (P < 0.001) and blood transfusion (P = 0.034) but were largely similar in terms of gender, BMI, ethnicity, smoking status, performance of a cyst excision and future lumbar surgery (Table 1).

Table 1.

Study Population and Surgery Characteristics.

	ALIF		ISF		LLIF		PLF		TLIF
	n = 73		n = 29		n = 60		n = 220		n = 314		P-value
Female (n, %)	34	47%	19	66%	30	50%	111	50%	163	52%	0.533
Age (mean, sd)	51.2	12.2	75.8	11.6	61.7	11.0	67.0	10.0	60.5	11.0	<.0001
Age (n, %)
≤45	25	34%	1	3%	4	7%	6	3%	32	10%
46-55	19	26%	0	0%	19	32%	24	11%	66	21%
56-65	19	26%	2	7%	11	18%	49	22%	101	32%
>65	10	14%	26	90%	26	43%	141	64%	115	37%	<.0001
BMI (n, %)
<25	16	22%	7	24%	19	32%	50	23%	64	20%
25 - 29.9	29	40%	9	31%	22	37%	74	34%	135	43%
30-34.9	21	29%	11	38%	13	22%	59	27%	70	22%
≥35	7	10%	2	7%	6	10%	37	17%	45	14%	0.284
Ethnicity (n, %)
White	66	90%	27	93%	57	95%	199	90%	290	92%
Other/unknown	7	10%	2	7%	3	5%	21	10%	24	8%	0.785
Smoking status (n, %)
Never smoker	36	49%	12	41%	31	52%	102	46%	152	48%
Former smoker	30	41%	16	55%	26	43%	108	49%	140	45%
Current smoker	7	10%	1	3%	3	5%	10	5%	22	7%	0.720
ASA class (n, %)
1	5	7%	0	0%	4	7%	11	5%	16	5%
2	64	88%	17	59%	48	80%	163	74%	237	75%
3	4	5%	12	41%	8	13%	46	21%	61	19%	0.007
Insurance (n, %)
Medicare	10	14%	24	83%	23	38%	121	55%	102	32%
Commercial	24	33%	3	10%	26	43%	52	24%	111	35%
Other	39	53%	2	7%	11	18%	47	21%	101	32%	<.0001
Cyst (n, %)	0	0%	0	0%	0	0%	8	4%	6	2%	0.311
Durotomy (n, %)	1	1%	4	14%	0	0%	12	5%	10	3%	0.022
Revision decompression (n, %)	2	3%	2	7%	0	0%	20	9%	20	6%	0.047
Blood transfusion (n, %)	0	0%	3	10%	3	5%	20	9%	25	8%	0.034
Future surgery (n, %)	6	8%	2	7%	4	7%	21	10%	15	5%	0.308
Fusion level (n, %)
L1-L2	0	0%	0	0%	0	0%	3	1%	1	0%
L2-L3	0	0%	0	0%	8	13%	13	6%	11	4%
L3-L4	0	0%	4	14%	14	23%	38	17%	27	9%
L4-L5	7	10%	23	79%	37	62%	142	65%	193	61%
L5-S1	66	90%	2	7%	1	2%	24	11%	82	26%	<.0001

Multivariable linear regression modeling demonstrated that approximately 50% of the variation in cost could be explained by procedure choice alone (Table 2). Including patient demographic and surgery case characteristics as covariates provided minimal additional explanatory value (ΔR ² = 0.02). A model excluding fusion technique revealed that patient characteristics (sex, age, BMI, ASA score, smoking history, and insurance type) accounted for only 10% of the overall cost variance.

Table 2.

Results of 3 Regression Models Relating Patient and Surgery Characteristics to Total Direct Costs.

	Model 1		Model 2		Model 3
	Std-β	P	Std-β	P	Std-β	P
Intercept	0.00	<.0001	-	<.0001	0.00	<.0001
Sex, female	−0.07	0.072	-	-	−0.02	0.373
Age (per 5 years)	−0.30	<.0001	-	-	−0.07	0.059
ASA score
ASA 1	−0.04	0.282	-	-	−0.03	0.240
ASA 2	REF	REF	-	-	REF	REF
ASA 3	0.05	0.190	-	-	0.10	0.001
Smoking history
Never	REF	REF	-	-	REF	REF
Former	0.02	0.644	-	-	0.01	0.605
Current	0.03	0.416	-	-	0.04	0.146
Insurance
Commercial	REF	REF	-	-	REF	REF
Medicare	0.03	0.570	-	-	0.05	0.255
Other	0.10	0.018	-	-	0.11	0.001
Procedure
PLF	-	-	0.00	<.0001	REF	REF
ALIF	-	-	0.47	<.0001	0.44	<.0001
TLIF	-	-	0.27	<.0001	0.25	<.0001
LLIF	-	-	0.54	<.0001	0.55	<.0001
ISF	-	-	−0.26	<.0001	−0.26	<.0001
R 2	0.10		0.50		0.52

Across the entire cohort, implants (including biologics) accounted for 47.9% of cost, ranging from 37.9% among PLFs to 60.9% among ALIFs (Table 3). Surgery personnel were the next largest expense (25.8%) followed by inpatient personnel (15.0%) and medication and other non-implant supply (8.0%).

Table 3.

Component Costs as a Proportion of Total Direct costs, by Fusion Type.

Component cost	All Cases (n = 696)	Procedure
		LLIF (n = 60)	ALIF (n = 73)	TLIF (n = 314)	PLF (n = 220)	ISF (n = 29)
Surgery personnel	25.8%	20.2%	20.1%	25.1%	31.5%	39.5%
PACU personnel	1.4%	1.0%	1.1%	1.4%	1.7%	3.2%
Inpatient personnel	15.0%	9.2%	10.8%	15.0%	18.5%	33.5%
Implant	47.9%	55.2%	60.9%	49.6%	37.9%	6.6%
Medication	0.7%	0.4%	0.5%	0.8%	1.0%	1.4%
Non-med, non-implant supply	7.3%	12.8%	5.3%	6.3%	7.3%	11.9%
Other day-of costs	1.8%	1.3%	1.4%	1.8%	2.1%	3.9%

When comparing overall cost among fusion techniques, LLIF was highest in mean cost, followed by ALIF, TLIF, PLF and ISF (Figure 1). Medication, non-implant non-medication supply and other supplies were combined into “Medication and other supplies” because medication and other supply costs were comparatively small. Relative to average PLF cost, the average cost for specific fusion techniques ranged from 54% (ISF) to 168% (LLIF). This represents a 310% difference between the highest and lowest cost fusion techniques. The cost category with the largest variation among fusion techniques was implant (including biologics), where LLIF was most expensive (245% of PLF) and ISF was least (9% of PLF).

Figure 1.

Total direct cost, by fusion type, broken down by component.

Fusion technique utilization varied widely from surgeon to surgeon (Table 4). For example, 49% of single-level lumbar fusions performed by Surgeon E were LLIF while Surgeon A performed 49% TLIFs. The standardized average cost across surgeons was significantly different (P < 0.0001, Figure 2). At the extremes, the average cost of single-level lumbar fusions performed by Surgeon E was 129% the overall average while that of Surgeon A was 77%. This represents a 168% difference between the highest and lowest cost surgeons. For contrast, the variation for total hip replacement between 12 surgeons at our institution during the same time period was lower: the highest cost surgeon was 111% the overall average compared to 91% for the lowest.

Table 4.

Fusion Type Utilization by Surgeon.

	Surgeon
Procedure	A	B	C	D	E	F	G	H	Total N
ALIF	7%	24%	15%	8%	10%	1%	7%	0%	73
ISF	8%	7%	2%	0%	5%	1%	2%	7%	29
LLIF	1%	3%	9%	2%	49%	0%	9%	0%	60
PLF	36%	19%	53%	39%	9%	25%	44%	25%	220
TLIF	49%	47%	21%	51%	28%	73%	38%	68%	314
Total N	132	119	111	88	80	77	45	44	696

Figure 2.

Standardized total direct cost, by surgeon (all fusion techniques).

In terms of fusion technique, TLIFs demonstrated the highest cost variation. The surgeon with the highest average TLIF cost was 126% the average TLIF cost while the lowest was 83% (Figure 3). This represents a 151% difference between the highest and lowest cost TLIF surgeon. Implant and non-medication non-implant supply costs were more expensive in the highest cost surgeon but inpatient personnel cost was lower. The lowest cost TLIF surgeon performed TLIFs at nearly the same cost as that of the average cost for PLF across all surgeons (101% average PLF cost).

Figure 3.

Total direct cost, overall and for the highest and lowest cost TLIF surgeons, broken down by component.

Discussion

This study was motivated by the belief that understanding institution level variation in cost and utilization will be critical to success in value-based reimbursement programs in spine surgery. By taking advantage of TDABC to more accurately allocate costs, variation in fusion type utilization was identified as the primary source of potentially modifiable cost variation and cost excess. Significant variation in cost existed across surgeons and across fusion techniques. Regression analysis revealed that half of this variation was due to the choice of fusion technique alone.

While previous studies have evaluated hospital and surgeon level effects on variation in payments, variation in cost has not been as well described.^16,17 Twitchell et al retrospectively performed a cost analysis of 276 open and minimally invasive (MIS) 1- and 2- level lumbar fusions. ¹⁸ Length of stay, MIS and number of operated levels were found to predict cost in a multivariate analysis. Supplies and implants (55%) and facility (36%) categories were found to account for most of the direct cost. This is consistent with the findings of the current study where implants accounted for 45% and personnel 42%. Sivaganesan et al identified various patient- and case-related factors associated with variation in 90-day direct cost of care for elective lumbar laminectomy-fusion surgery. ¹⁹ They identified a high degree of variation in cost. Regression tree analysis identified 19 variables that accounted for roughly 65% of this variation. An analysis of variation by fusion technique or surgeon however was not reported.

High practice variability in the treatment of lumbar degenerative disease across the country has been previously reported.^20,21 But stakeholders (including policy makers, payers, healthcare facilities and surgeons) hoping to successfully implement value-based payment models cannot make informed decisions without more institution- and surgeon-level cost insight, particularly in the setting of such variability. Chotai et al studied the 90-day cost of performing lumbar laminectomy-fusion surgery at a single institution. ²² Significant differences in cost between 7 surgeons were identified. The authors concluded that targeting modifiable factors tied to how surgeons practice may increase the value of lumbar laminectomy-fusion. Multilevel surgeries were included in this study but an analysis of variation by fusion technique was not reported.

Not surprisingly, in the current study, we identified a number of significant differences in the patient populations across fusion techniques. This is almost certainly a reflection of surgeon choices based on patient characteristics. For example, disease at L5-S1 is more amenable to ALIF than at L2-3 but could also be treated with other fusion techniques. In situ fusions are commonly reserved for older patients with more co-morbidities. Durotomies are more likely in approaches involving direct decompression and older patients. Linear regression modeling, however, revealed that fusion technique dominated patient characteristics in explaining cost. But among patient characteristics, age had the most significant impact on cost. This appears to be primarily a result of age-related procedure selection bias (i.e. younger patients tended to undergo more expensive procedures such as ALIF and LLIF while older patients tended to undergo the less expensive ISF). Implant cost, which is closely linked to the type of fusion selected, was the single largest cost category for all instrumented fusion techniques.

The high cost variation at our institution was largely driven by fusion technique. There was a 310% difference between the average cost of the most expensive fusion technique (LLIF) compared to the least expensive (ISF). This is directionally intuitive given the higher cost of many biologics and implants associated with less invasive approaches such as LLIF and MIS TLIF as well as the absence of implants (and in general, biologics) with ISF. The high relative cost of implants and biologics also tends to offset the near-term cost benefits of less invasive surgery, such as reduced length of stay and reduced post-operative pain. Widening the scope of the analysis to include post-acute care may alter the patterns observed but in doing so would also alter the perspective of the analysis away from the hospital, for which this study was intended.

The significant cost variation exhibited by surgeons in this study was considerably more variable than arthroplasty surgeons performing total hip replacement during the same time period at our institution. The average cost of arthroplasty surgeons ranged from 91% to 111% of the overall average while spine surgeons ranged from 77% to 129%. This level of variation exists despite surgeons treating single-level degenerative lumbar conditions within the same patient population at a single facility. To some extent the variation in cost in spine surgery may be expected based on the higher degree of heterogeneity inherent in spine care. Our institution’s reference pricing policy, with exceptions for niche products, results in a wide range of available implants and biologics, which at least in part facilitates variation in cost based on surgeon preference. These effects are reflected in the secondary analysis of TLIF, the most common fusion technique in our study with the highest variation in cost. Within TLIF, individual surgeon average cost ranged from 83% to 126% of the TLIF average. The highest cost TLIF surgeon, however, performs primarily MIS TLIFs. His TLIF cost pattern is consistent with many of the touted tradeoffs associated with MIS approaches, i.e. higher cost implants, shorter length of stay (as reflected by his lower inpatient personnel cost).

These findings suggest that surgeon-dependent factors, specifically the choice of fusion technique and in turn the implants and biologics used, drive highly variable single-level lumbar fusion cost. Whether this is based on a surgeon’s treatment philosophy, skill set or other factors, this is an important insight for institutions interested in understanding their cost profile. Efforts to reduce cost should therefore include approaches geared toward reducing this variability. In a value-based reimbursement environment, higher cost interventions must be associated with incremental outcome benefit to justify their utilization. It is certainly conceivable that a more expensive fusion technique or a higher cost surgeon may yield clinically superior outcomes in a set of as yet undefined circumstances. Ultimately, clinical outcomes must be combined with accurate cost data to determine the relative value of these interventions. Analyses of this nature will likely need to occur on the institution level as the skill sets and preferences of staff surgeons will vary from facility to facility.

Limitations of this study include a degree of heterogeneity of lumbar disease within the study population, although this effect is likely mitigated by the primarily elective adult degenerative practices of the surgeons studied. Furthermore, this study is retrospective and nonrandomized; surgeon treatment selection bias is inherent to this study. This bias is likely a major source of the inter-surgeon variability observed. This study was performed from the perspective of the hospital. Broader analyses of cost, including post-acute care, rehabilitation and readmission costs would be more appropriate for alternate perspectives, such as payer or policy maker. In addition, studying a single institution obviously limits the generalizability of these findings. However, the purpose of this study was to demonstrate the value of TDABC analysis in identifying true and accurate cost patterns within any given institution. Disparate costing methodologies and variation in supply/implant pricing as well as labor costs among institutions are some of the most challenging obstacles to accurate multicenter cost analyses. Differences in implant cost across institutions and among vendors are significant, particularly at higher volumes. ²³ Focusing on a single institution allows for uniformity of unit costs in terms of labor, implants, and other supplies. We anticipate that individual institutions will need to conduct facility-specific analyses to account for their unique characteristics including implant pricing, staff surgeon behavior and patient population. We hope that this study may provide both motivation to do so and an example of the variation and opportunities that can be expected.

Conclusion

Single-level primary lumbar spine fusion cost is driven primarily by surgeon-based choices (e.g. fusion technique, implants). Implants, including biologics, account for nearly 50% of cost. There is high variation in both the techniques of fusions employed by surgeons as well as average cost by surgeon. These elements of variation highlight the importance of understanding institution-specific cost structures to identify the primary opportunities for lowering cost. Future work will incorporate outcomes data to characterize the differential value conferred by fusion techniques associated with higher cost.