Resource utilization and costs for robotic-assisted and manual total knee arthroplasty – a premier healthcare database study

ABSTRACT

Introduction

The impact of robotic-assisted total knee arthroplasty (rTKA) vs. traditional, manual TKA (mTKA) on hospital costs is not well documented and is analyzed herein.

Research Design and Methods

Patients in the Premier billing Healthcare Database undergoing elective rTKA or mTKA (“index’) in the in- or outpatient setting for knee osteoarthritis between Oct 1^st, 2015, to September 30th, 2021, were identified. Variables included patient demographics and comorbidities and hospital characteristics. Matched rTKA vs. mTKA cohorts were created using direct (on provider characteristics, age, gender, race and Elixhauser index) and propensity score matching (fixation type, comorbidities). Index and 90-day inflation-adjusted costs and healthcare utilization (HCU) were analyzed for both cohorts, using generalized linear models.

Results

16,714 rTKA patients were matched to 51,199 mTKA patients. Average 90-day hospital cost reached $17,932 and were equivalent for both cohorts (rTKA vs. mTKA: $132 (95% confidence interval; -$19 to $284). There was a 2.7% (95%CI: 2.2%-3.3%) increase in home or home health discharge, and a 0.4% (95%CI: 0%-0.8%) decrease in 90-day hospital knee related re-visit in the rTKA vs. mTKA group.

Conclusions

Cost-neutrality of rTKA vs. mTKA was observed, with a potential for lowered immediate post-operative HCU in the rTKA vs. mTKA cohorts.

KEYWORDS:

• Robotic surgical procedures
• arthroplasty
• replacement
• knee
• health care costs
• health care quality
• access
• evaluation

1. Introduction

Total knee arthroplasty (TKA) is a generally successful and cost-effective procedure for the treatment of end-stage knee arthritis [1–3]. Despite overall strong outcomes, 18% of devices may fail within 25 years of implantation; this number is significant considering the increasingly young population being treated with TKA [4–6]. In addition, despite generally favorable post-operative satisfaction, for at least 80% patients, residual symptoms after TKA impact the quality of life of a large proportion of patients. As an example, 54% and 38% of patients who had TKA describe difficulty climbing stairs and getting in and out of cars, respectively [7–9].

Robotic-assisted total knee arthroplasty (rTKA) has been developed to improve patient outcomes by increasing surgical precision and reducing surgical variability compared to traditional, manual TKA (mTKA). There are currently several robotic systems available, including ROBODOC® (Curexo Technology, Fremont, CA), Mako® (Stryker Corporation, Kalamazoo, MI), Navio™ (Smith & Nephew, Cordova, TN), OMNIbotics® (Corin Group, Raynham, MA), and Rosa® (Zimmer Biomet, Warsaw, IN). Current robotic systems require the creation of a three-dimensional plan derived from a process of mapping the anatomy of the joint surfaces, with or without a preoperative computed tomography (CT) scan. In image-guided systems, the surgeon uses this virtual model to plan optimal bone resection, implant positioning, bone coverage, and limb alignment based on the patient’s unique anatomy. An intraoperative robotic device helps to execute this preoperative plan [10,11]. In imageless systems, intraoperative registration of the anatomical surfaces and kinematics after arthrotomy is used to create a 3D virtual model, develop a surgical plan, and define boundaries beyond which the bone cutting tools should not remove surface tissue [12].

Robotic TKA has been shown to deliver predictable and consistent component alignment and soft tissue balance [12]. This increased accuracy in surgical methodology may result in further improved outcomes, especially range of motion and patient satisfaction [13–16].

Most of the published medical literature evaluating rTKA suggests favorable clinical and radiographic outcomes versus mTKA [17–27]. Early studies have suggested, however, that increased operating room time, technical complexity, and costs may be limitations to rTKA adoption [28]. With limited data available on actual hospital costs, we designed our study to compare rTKA vs. mTKA in terms of operating room time, hospital length of stay (LOS), cost of the index hospitalization, and patient discharge status after TKA. Secondary objectives included all-cause and knee-related 90-day hospital-based healthcare resource utilization (HCU) and cost of hospitalizations (i.e. cost of index hospitalization + 90-day post-TKA hospital cost).

2. Methods

2.1. Data source

This retrospective, cross-sectional, real-world data study identified patients undergoing TKA in the Premier Healthcare Database between October 1^st, 2015, and September 30^th, 2021. The Premier, Inc. database contains complete standardized clinical coding, including diagnosis, procedure, and hospital-prescribed medications from more than 20% of all hospital admissions throughout the United States (>1041 hospitals and hospital systems). Premier collects data from participating hospitals in its health care alliance. Although the database excludes federally funded hospitals (e.g. Veterans Affairs), the hospitals included are nationally representative based on bed size, geographic region, location (urban/rural), and teaching hospital status. The database contains a date-stamped log of all billed items by the cost-accounting department (both fixed and variable) including devices, implants, medications; laboratory, diagnostic, and therapeutic services; and primary and secondary diagnoses for each patient’s hospitalization. Identifier-linked enrollment files provide demographic and payor information. Detailed service level information for each hospital day is recorded; this includes details on medication and devices received.

2.2. Patient population

All patients ≥18 years of age with a Current Procedural Terminology (CPT) code or International Classification of Diseases, Tenth Revision (ICD-10) code indicative of elective TKA were considered for inclusion in the study. The date of the TKA was defined as the ‘index date.’ Exclusion criteria included: 1) presence on admission of one of the following diagnoses: lower leg fracture, aseptic loosening, metastatic cancer (any) or deep infection/osteomyelitis; unicompartmental or computer-assisted TKA at index; bilateral, or TKA of the contralateral knee, from 0 to 90 days post-index. Data were analyzed for the duration of the patient’s index hospitalization and 90 days post-index.

2.3. Study measures

2.3.1. Baseline demographic and clinical characteristics

Patient demographic and clinical characteristics that were evaluated included age, gender, race, marital status, payer category (Medicare, Medicaid, or commercial health insurance), hospital admission type (inpatient or outpatient service), year of TKA surgery, and patient comorbidities. The baseline comorbidity index (i.e. comorbid conditions present prior to TKA) was assessed using the Elixhauser Comorbidity Score, an aggregate measure of comorbidity created by using 31 dimensions associated with chronic disease (e.g. heart disease, cancer) and overall health conditions. Prior research has shown that increasing Elixhauser Comorbidity Scores are associated with increased healthcare utilization and greater risk of mortality [29,30]. The Functional Comorbidity Index (FCI) was also evaluated. The FCI includes 18 medical conditions and the sum of the individual condition scores yields the FCI score (range from 0 to 18) [31]. The FCI is particularly relevant for orthopedic care as it was developed as a measure of patient functioning capacity [31,32]. Additional comorbidities and health behaviors that were assessed included the presence of back or knee pain, arthritis, osteoporosis, and smoking status.

Provider variables included: 1) hospital size (defined by the number of beds); 2) teaching vs. community hospital; 3) urban vs. rural; 4) geographic location of the hospital, and 5) TKA volume. For this variable, the total count of TKA performed within each institution in the 12 months prior to a patient’s TKA was analyzed. This variable was important to evaluate whether the hospital was performing routine TKAs, or whether TKAs were rarely performed at that given site.

2.3.2. Outcomes

The procedure duration, LOS, costs for the index surgery, and hospital and discharge status were evaluated for each patient. Following discharge, occurrence and costs of all-cause and knee-related hospital visits (i.e. inpatient, or ER, or outpatient hospital visits), inpatient readmissions, and emergency room (ER) visits were analyzed for all patients. A total cost that included the index plus the 90-day post-index costs was estimated for all patients.

2.4. Statistical analyses

All study variables were analyzed descriptively. Counts and proportions (categorical variables) and means and standard deviations (continuous variables) were provided. Patients with rTKA and mTKA were matched 1:5 using a variable matching strategy. The variable matching allowed for fewer than five matches if the target number of matches was not found. This strategy has greater confounding control over the fixed-ratio matching [33,34]. Patients were matched directly on the following variables using exact matches: age categories, race, Elixhauser comorbidity score categories, year of surgery, hospital volume, hospital bed size, hospital teaching status, hospital urban or rural location, and geographic subdivision. These characteristics are important determinants of resource use [35,36]. Direct-match cohorts were further matched on propensity scores (PS) created using fixation type (cemented vs uncemented), patient demographics, health behaviors, and clinical characteristics. Greedy nearest neighbor matching technique with calipers of width equal to 0.2 of the pooled standard deviation of the logit of the PS was used. Standardized differences were evaluated to assess the balance between the groups. An absolute standardized difference <0.05 was used to indicate a negligible difference between rTKA and mTKA. HCU and costs over 90 days were compared using generalized linear models (GLM). LOS and duration of surgery were modeled using negative binomial distribution; discharge destination and health care utilization were modeled using a binomial distribution with a logit link; cost outcomes were modeled using a quasi (gamma) distribution with log link. To account for the clustering of subjects within matched sets, a robust variance estimator was used to compute 95% confidence intervals (CI). All costs were inflated to 2021 using Consumer Price Index.

2.4.1. Sensitivity analysis

Sensitivity analysis was carried out to obtain a stricter match between the rTKA and mTKA. In sensitivity analyses, all comorbidities that showed a greater than a 0.2%-point absolute difference between the matched rTKA and mTKA cohorts were identified. These comorbidities were included for direct-matching, in addition to previously mentioned variables that were matched directly. PS-matching was then carried out to match on the remaining variables.

3. Results

A total of 36,197 patients from the Premier Healthcare Database received rTKA and 1,009,835 patients received mTKA between 1 October 2015, to 30 September 2021. Figure 1 presents the attrition flow for rTKA and mTKA using the inclusion and exclusion criteria defined above. Before matching, 25,206 rTKA and 688,315 mTKA patients were identified. After DM-PSM, 16,714 patients with rTKA and 51,199 patients with mTKA were included in the analyses.

Figure 1. Attrition flow diagram of patients included in the analysis.

3.1. Patient baseline demographic and clinical characteristics

Baseline demographic and clinical characteristics for patients with rTKA and mTKA before and after matching are presented in Table 1. Before DM-PSM, a greater proportion of patients with rTKA were married, commercially insured, had their TKA procedure performed in 2019–2021 usually in inpatient setting, compared to mTKA.

Table 1. Baseline characteristics of patients undergoing rTKA and mTKA surgery before and after DM-PSM.

	Before DM-PSM			After DM-PSM
Patient Characteristics	rTKA	mTKA	Standardized Differences	rTKA	mTKA	Standardized Differences
N	25,206	688,315		16,714	51,199
Age in years, N(%)
18–54	10.9	9.7	0.072	9.1	9.1	0
55–64	30.3	28.2		30.3	30.3
65–74	38.7	40.2		40.1	40.1
75 plus	20.1	21.9		20.5	20.5
Gender, %
Males	40.1	38.4	0.035	39.4	40.2	0.016
Race, %
White	83.5	83.1	0.016	87.9	87.9	0
Black	8.3	8.7		6.4	6.4
Asian	1.4	1.3		0.8	0.8
Other	6.9	6.9		4.9	4.9
Marital Status, %
Married	61.9	59.6	0.047	64.7	60.8	0.083
Single	30.9	32.9		29.3	32.2
Other or Unknown	7.2	7.5		6.0	7.0
Payor category, %
Medicare	56.0	60.8	0.112	57.1	58.7	0.056
Commercial	34.8	30.6		35.1	32.8
Medicaid	3.7	4.2		2.7	3.3
Other	5.4	4.4		5.1	5.2
Patient Type, %
Inpatient	87.4	82.1	0.147	86.4	87.1	0.021
Year of TKA surgery, %
2015	1.5	5.6	0.755	1.9	1.9	0
2016	5.1	20.9		6.6	6.6
2017	9.5	20.6		11.8	11.8
2018	19.4	18.0		20.8	20.8
2019	29.3	17.6		29.1	29.1
2020	24.3	12.4		22.8	22.8
2021	10.8	4.9		7.1	7.1
Additional Comorbidities, %
Arthritis	99.5	97.6	0.163	99.4	99.0	0.050
Smoking	31.4	31.4	0.001	31.1	31.7	0.013
Knee pain	10.0	7.2	0.099	8.3	7.8	0.018
Osteoporosis	4.3	4.1	0.009	4.6	4.1	0.024
Back pain	3.4	3.4	0.003	3.3	3.4	0.003
Functional Comorbidity Index, %
0–2	32.2	31.9	0.005	32.6	32.1	0.013
4-Mar	44.5	44.7		45.0	45.6
5 +	23.3	23.3		22.4	22.2
Elixhauser Comorbidity Index, %
0	13.8	13.1	0.044	12.9	12.9	0
2-Jan	53.0	51.8		55.2	55.2
4-Mar	27.1	28.1		26.8	26.8
5 +	6.2	7.0		5.0	5.0

AIDS, acquired immunodeficiency syndrome; DM-PSM, direct-matching and propensity score-matching; Elix, Elixhauser; FCI, functional comorbidity index; HIV, human immunodeficiency virus; mTKA, manual total knee arthroplasty; rTKA, robotic total knee arthroplasty.

After DM-PSM, all standardized mean differences were lower than 0.05, suggesting appropriate matching based on all included variables. The mean age was ~67 years old, ~60% of the patients were ≥65 years old, ~60% of the patients were female, the majority of patients were White (~88%), ~65% of patients were married, and ~31% of patients were smokers. Most patients had Medicare (~57%) or private commercial health insurance (~35%) and were admitted as hospital inpatients (~87%) for their TKA. A greater proportion of TKAs was performed in 2019 (~29%) or 2020 (~23%) vs. prior years. Commonly observed comorbidities among patients included uncomplicated hypertension (~59%), obesity (~29%), hypothyroidism (~17%), chronic pulmonary disease (~15%), uncomplicated diabetes (~14%), and depression (~14%) (Supplementary Table 1).

3.2. Provider baseline characteristics

Baseline provider characteristics are presented in Table 2. Before DM-PSM, a greater proportion of hospitals performing rTKA vs. mTKA had a pre-index TKA volume of <274, a size ≥ 100 beds, were conducted in the New England or West South-Central geographic regions.

Table 2. Provider characteristics of patients undergoing rTKA and mTKA surgery before and after DM-PSM.

	Before DM-PSM			After DM-PSM
Provider Characteristics	rTKA	mTKA	Standardized Differences	rTKA	mTKA	Standardized Differences
N	25,206	688,315		16,714	51,199
Provider volume annual, %
1–273	44.1	35.1	0.192	41.4	41.4	0
274–479	25.4	27.4		24.5	24.5
480–906	27.4	33.2		31.7	31.7
907 plus	3.1	4.3		2.4	2.4
Urban or Rural, %
Urban	87.5	87.6	0.005	90.0	90.0	0
Teaching, %
Yes	42.2	42.1	0.003	43.6	43.6	0
Beds group, %
000–099	5.2	9.7	0.177	4.9	4.9	0
100–199	19.8	18.5		19.1	19.1
200–299	19.5	19.2		18.2	18.2
300–399	19.4	17.1		21.7	21.7
400–499	11.5	11.8		10.0	10.0
500+	24.5	23.7		26.3	26.3
US Geographic Subdivision, %
East North Central	14.9	17.9	0.503	16.8	16.8	0
East South Central	9.5	9.2		8.4	8.4
Middle Atlantic	15.8	12.1		16.1	16.1
Mountain	6.1	3.9		5.5	5.5
New England	9.0	4.2		6.9	6.9
Pacific	7.0	9.7		7.6	7.6
South Atlantic	18.6	26.1		21.1	21.1
West North Central	1.7	8.3		1.1	1.1
West South Central	17.3	8.6		16.6	16.6
US Census Region, %			0.281			0
Midwest	16.7	26.2		17.9	17.9
Northeast	24.8	16.3		22.9	22.9
South	45.4	43.9		46.1	46.1
West	13.1	13.6		13.0	13.0
Fixation Method, %
Uncemented	24.8	4.1	0.616	12.1	11.8	0.008

DM-PSM, direct-matching and propensity score-matching; mTKA, manual total knee arthroplasty; rTKA, robotic total knee arthroplasty.

After DM-PSM, provider characteristics were similar for rTKA and mTKA and were as follows: most surgeries took place in an urban setting (~90%), ~44% were performed at teaching hospitals, ~ 26% in hospitals with 500 beds or more. Approximately 46% of all surgeries were conducted in the South census region and ~23% in the Northeast census region. The fixation method was also balanced with uncemented proportion of 12.1% for rTKA vs 11.8% for mTKA cohort (SMD = 0.008).

3.3. Resource use and costs

Hospital costs for the index- and post-index care were not different between patients treated with robotic vs. manual TKA (Table 3). Total costs over 90 days averaged approximately $17.9 K; excluding costs unrelated to knee care, costs averaged approximately 17.3 K. Compared to manual surgery, the duration of robotic surgery was longer by 11 minutes (95%CI: 9–13), resulting in increased operating room costs (an increase of $513 (95%CI: $466 to $559). However, LOS was shorter for patients with rTKA (0.2 days (0.17 to 0.22 days – 95%CI)) resulting in lower room and board costs for robotic vs. manual TKA (difference of $168 (95% CI: $129 to $206)).

Table 3. Healthcare resource utilization and costs for rTKA and mTKA from index surgery through 90 days post-index.

	rTKA	mTKA	Difference:
			rTKA – mTKA
N	16,714	51,199
Index Hospitalization
Length of stay in days, mean (95%CI)	1.598 (1.578 to 1.618)	1.799 (1.787 to 1.812)	−0.2 (−0.22 to −0.17)
Surgical time in mins, mean (95%CI)	153 (152 to 155)	143 (142 to 143)	11 (9 to 13)
Cost of index surgery ($), mean (95%CI)	17,005 (16,906 to 17,104)	16,740 (16,683 to 16,796)	266 (152 to 379)
Operating room cost ($), mean (95%CI)	5,506 (5,465 to 5,547)	4,994 (4,971 to 5,016)	513 (466 to 559)
Room and board cost ($), mean (95%CI)	2,055 (2,022 to 2,088)	2,223 (2,203 to 2,242)	−168 (−206 to −129)
Discharged to home with or without home health agency, %	91.0 (90.5 to 91.5)	88.2 (88 to 88.5)	2.8 (2.2 to 3.3)
Discharged home without home health, %	43.0 (42.2 to 43.7)	38.0 (37.6 to 38.4)	5.0 (4.1 to 5.8)
Discharged to skilled nursing facility or other facility, %	9.0 (8.6 to 9.4)	11.8 (11.5 to 12.1)	−2.8 (−3.3 to −2.3)
Follow-up Period (from Discharge to 90 days post-discharge)
Any hospital revisit*, %	28.7 (28 to 29.3)	29.2 (28.8 to 29.6)	−0.6 (−1.4 to 0.2)
Any inpatient readmission, %	3.4 (3.2 to 3.7)	3.9 (3.7 to 4)	−0.4 (−0.8 to −0.1)
Any ER revisit, %	5.4 (5.1 to 5.8)	6.3 (6.1 to 6.5)	−0.9 (−1.3 to −0.5)
Knee-related hospital revisit*, %	7.0 (6.6 to 7.4)	7.4 (7.2 to 7.6)	−0.4 (−0.8 to 0)
Knee-related inpatient readmission, %	1.8 (1.6 to 2)	2.1 (2 to 2.2)	−0.3 (−0.5 to −0.1)
Costs for any hospital revisit ($), mean (95%CI)	994 (920 to 1,069)	1,127 (1,082 to 1,173)	−133 (−220 to −46)
Cost for any inpatient readmission ($), mean (95%CI)	501 (437 to 565)	590 (550 to 630)	−89 (−165 to −14)
Cost for any ER revisit ($), mean (95%CI)*	44 (40 to 48)	47 (45 to 49)	−3 (−8 to 1)
Costs for knee-related hospital revisit ($), mean (95%CI)	460 (404 to 515)	538 (503 to 572)	−78 (−143 to −13)
Costs for knee-related inpatient readmission ($), mean (95%CI)*	286 (238 to 335)	342 (311 to 372)	−55 (−113 to 2)
Total Inpatient Costs: Including Index and Up to 90 days post-discharge – excluding SNF and home health costs
All care costs ($)*	17,999 (17,868 to 18,131)	17,867 (17,792 to 17,942)	132 (−19 to 284)
Knee-related care costs ($)	17,465 (17,346 to 17,584)	17,277 (17,209 to 17,345)	188 (51 to 325)

Note: Differences were significant (p < 0.05) for all variables except for the ones marked with *

Differences in outcomes were identified by discharge disposition: a larger proportion of robotics patients were discharged to home (91.0% home discharge for rTKA, an increase of 2.8% (95%CI: 2.2% – 3.3%) from mTKA), and 2.8% fewer patients were discharged to skilled nursing homes. The 90-day post-index period showed consistently and fewer readmissions in terms of percentage of patients with revisits and the cost of revisits for patients treated with rTKA vs. mTKA. The percentage of patients requiring any inpatient readmission decreased by 0.4% (0.1–0.8%) in the rTKA vs. mTKA group, and specifically, knee-related readmissions decreased by 0.3% (0.1–0.5%). Emergency room visits also decreased significantly by 0.9% (0.5%-1.3%) in the rTKA vs. mTKA cohorts. The hospital inpatient costs for post-index care were, therefore, lower in the rTKA vs. mTKA group.

In sensitivity analyses, the rTKA and mTKA cohorts were directly matched on additional variables including congestive heart failure, peripheral vascular disorders, hypertension, uncomplicated diabetes, renal failure, fluid and electrolyte disorders, deficiency anemia, depression, arthritis, osteoporosis and marital status. After DM-PSM, 10,622 patients with rTKA and 31,327 patients with mTKA were included in the sensitivity analyses. The results in the sensitivity analyses were similar to the main analyses (Table 4). Patients with rTKA at index had significantly lower LOS, higher proportion discharged home and lower proportion discharged to a skilled nursing facility or other inpatient facility. Resource use and costs during the 90-day period after the TKA were lower for patients who received rTKA vs. patients who received mTKA. However, overall knee-related care cost was not statistically significantly different between rTKA and mTKA.

Table 4. Sensitivity analysis for healthcare resource utilization and costs for rTKA and mTKA from index surgery through 90 days post-index*.

	rTKA	mTKA	Difference:
			rTKA – mTKA
N	10,622	31,327
Index Hospitalization
Length of stay in days, mean (95%CI)	1.623 (1.603 to 1.643)	1.825 (1.813 to 1.838)	−0.2 (−0.225 to −0.179)
Surgical time in mins, mean (95%CI)	153 (152 to 155)	144 (143 to 144)	10 (8 to 11)
Cost of index surgery ($), mean (95%CI)*	17,060 (16,958 to 17,162)	17,071 (17,013 to 17,130)	−12 (−129 to 106)
Operating room cost ($), mean (95%CI)	5,502 (5462 to 5542)	5,005 (4983 to 5027)	497 (451 to 543)
Room and board cost ($), mean (95%CI)	2,089 (2056 to 2121)	2,254 (2235 to 2273)	−165 (−203 to −128)
Discharged to home with or without home health agency, %	91 (90.5 to 91.4)	88.2 (87.9 to 88.4)	2.8 (2.3 to 3.3)
Discharged home without home health, %	42.6 (41.8 to 43.3)	37.9 (37.5 to 38.3)	4.6 (3.8 to 5.5)
Discharged to skilled nursing facility or other facility, %	9 (8.6 to 9.5)	11.8 (11.6 to 12.1)	−2.8 (−3.3 to −2.3)
Follow-up Period (from Discharge to 90 days post-discharge)
Any hospital revisit*, %	28.6 (28 to 29.3)	29.2 (28.8 to 29.6)	−0.5 (−1.3 to 0.2)
Any inpatient readmission, %	3.4 (3.2 to 3.7)	4 (3.8 to 4.2)	−0.6 (−0.9 to −0.3)
Any ER revisit, %	5.4 (5.1 to 5.8)	6.3 (6.1 to 6.5)	−0.9 (−1.3 to −0.5)
Knee-related hospital revisit, %	7 (6.6 to 7.3)	7.6 (7.4 to 7.8)	−0.6 (−1.1 to −0.2)
Knee-related inpatient readmission, %	1.8 (1.6 to 2)	2.2 (2.1 to 2.4)	−0.4 (−0.7 to −0.2)
Costs for any hospital revisit ($), mean (95%CI)	983 (910 to 1056)	1,172 (1126 to 1217)	−189 (−275 to −103)
Cost for any inpatient readmission ($), mean (95%CI)	496 (434 to 559)	622 (582 to 662)	−125 (−199 to −52)
Cost for any ER revisit ($), mean (95%CI)*	44 (40 to 48)	47 (45 to 49)	−3 (−7 to 2)
Costs for knee-related hospital revisit ($), mean (95%CI)	453 (401 to 505)	561 (528 to 594)	−108 (−170 to −47)
Costs for knee-related inpatient readmission ($), mean (95%CI)	280 (235 to 325)	360 (331 to 390)	−80 (−134 to −26)
Total Inpatient Costs: Including Index and Up to 90 days post-discharge – excluding SNF and home health costs
All care costs ($)	18,043 (17,909 to 18,176)	18,243 (18,166 to 18,320)	−200 (−354 to −46)
Knee-related care costs ($)*	17,512 (17,392 to 17,633)	17,632 (17,563 to 17,701)	−120 (−259 to 19)

Note: Differences were significant (p < 0.05) for all variables except for the ones marked with *

mTKA, manual total knee arthroplasty; rTKA, robotic total knee arthroplasty.

* Sensitivity analyses involved directly matched on additional variables including congestive heart failure, peripheral vascular disorders, hypertension, uncomplicated diabetes, renal failure, fluid and electrolyte disorders, deficiency anemia, depression, arthritis, osteoporosis and marital status in addition to age categories, Elixhauser comorbidity score, year of surgery, hospital volume, hospital bed size, hospital teaching status, hospital urban or rural location, and geographic subdivision. Patients were matched on other demographic and clinical characteristics based on propensity scores.

4. Discussion

Our study evaluated the costs of rTKA vs. mTKA in a large database of > 1000 US hospitals, including more than 67,000 matched patients. Existing published retrospective database analyses evaluating rTKA and mTKA have used Medicare fee-for-service data, which limits the analysis to patients over the age of 65 who are publicly insured [37,38]. In contrast, our study is payer-agnostic, including all patients over the age of 18 and patients with both public insurance (Medicare ~57% and Medicaid ~3%) and private insurance (private commercial ~35%). Hospitals also bear the added costs of performing robotic TKA (vs manual). Findings from this study, therefore, constitute benchmarks for hospital-based resource use and costs associated with rTKA compared to mTKA from a nationally representative sample of patients across diverse hospitals in the United States.

In our matched analyses, the overall hospital costs of care from index to 90 days post-index were not different between rTKA vs. mTKA. Knee-related costs were higher for rTKA vs mTKA in our primary analysis, but not in our sensitivity analysis, where matching included a larger number of confounders. For all other outcomes, findings were consistent between our primary and sensitivity analysis. Length of stay was shorter, and more patients were discharged home following rTKA vs. mTKA. This finding suggests that rTKA patients had uncomplicated care and fast recovery. The duration of surgery, however, was 7.7% (11 minutes) longer with rTKA, which is aligned with the additional steps required for robotic use. From a financial standpoint, costs for room and board during the index hospitalization were lower with rTKA whereas operating room costs were higher with rTKA, vs. mTKA. These differences however balanced each other out.

Discharge disposition was also different between rTKA and mTKA: approximately 3% fewer patients were discharged to a skilled nursing facility (SNF), and about 5% more patients were discharged to home without home health in the rTKA vs. mTKA arm. These significant differences were not captured as cost savings in our analysis because our database only includes hospital costs, and costs of SNF/non-hospital admissions were not accessible for this research.

Over the 90 days post-index, rates of revisits and readmissions, both all-cause and knee-related, were also significantly lower for inpatient and ER setting in the rTKA vs. mTKA cohorts. These differences suggest potential early clinical advantages in patients treated with rTKA vs. mTKA. Interestingly, the overall costs were not different between rTKA and mTKA, possibly due to the fact that inpatient admissions were still relatively infrequent (4.0% or less), and small changes, once averaged across the entire population, did not reach significance.

The differences in post-operative healthcare revisits and readmissions may be explained by the following: Compared with manual surgery, rTKA provides improved precision of soft-tissue dissection and fewer soft-tissue releases, better visualization of the surgical field, and overall greater surgical accuracy [10]. Other studies have also demonstrated that rTKA systems are associated with reduced postoperative pain and have improved functional outcomes compared to mTKA across the patient care pathway [15,16,39–41]. These attributes may explain why, in our study, fewer patients return to the hospital for follow-up visits and readmissions. Other studies have shown that rTKA may lead to more reproducible bone preparation, component positioning, lower leg alignment, and soft tissue balancing [26,39–42]. These findings may contribute to less complicated recovery and increased gait quality, thus greater functional outcomes [27,38,43]. Our findings suggest that, at cost neutrality between mTKA and rTKA, rTKA may result in reduced post-operative care requirements.

The observed variability in healthcare resource use and cost among patients receiving rTKA and mTKA suggests that there are still opportunities for improving care and reducing costs of patients who need TKA. Robotic solutions have become increasingly popular for TKA; however, some robotic systems are associated with limitations including the complexity of their use and training requirements [41], additional operating room time [41,44], limitations of cut-block dependent or burr-based robotic systems [45], large footprints which can be cumbersome in the operating room, need for additional personnel in the operating room, and additional processes required during set-up and during procedures such as CT upload, registration, and milling [41]. The inefficiencies of some robotic systems may result in the full economic benefits of robotics not being fully realized.

This study represents the largest economic analysis of robotic versus manual TKA. The healthcare resource utilization and costs for rTKA and mTKA observed in this study are in alignment with previously published studies of smaller cohorts. A retrospective database evaluation by Cool et al. (2019) [38] using Medicare data between 1 January 2016, and 31 March 2017, evaluated healthcare resource utilization and costs of rTKA (n = 519) vs. mTKA (n = 2,595) up to 90 days post-surgery. Study findings showed that rTKA was associated with lower 30-, 60-, and 90-day postoperative costs and health care utilization, including ER services, readmissions, and home health visits. The Cool et al. (2019) [38] study results did find more pronounced differences for some resource use and cost study endpoints, which may be due to their focus on the payer vs. the hospital perspective.

Additional strengths of this study include the consideration of both all-cause and knee-related resource utilization and costs in rTKA and mTKA, Furthermore, the current study adjusted for an array of Elixhauser comorbidities and the Functional comorbidity index and was able to adjust for hierarchical confounding factors such as hospital volume, hospital bed size, hospital teaching status, hospital urban or rural location, and geographic subdivision. This study also carried out sensitivity analyses to confirm the results with stricter control of comorbidities to mitigate any residual confounding, demonstrating the robustness of the findings.

A limitation of the current study is that it only captures outcomes and resource utilization in the Premier Healthcare Database for the index hospitalization and for up to 90 days post-TKA. The resource use and costs beyond 90 days post-TKA were not evaluated, which could give important information for outcomes and resource utilization over a longer period of time. In addition, other important costs such as costs for SNF, home health, and other services (such as any preoperative imaging), were not available for analysis. Our study also reflects only robots that were used from 10/2015 to 09/2021 in the United States and may be specific to those technologies.

Another limitation of this study is that it uses billing data, which are not collected specifically for research purposes. Under-reported or missing diagnoses for variables such as smoking are based on patients’ choice not to seek care or access challenges. Billing data may also have clerical inaccuracies, recording bias secondary to financial incentives, and temporal changes in billing codes [46,47]. Costs from the Premier Healthcare Database do not include capital expenditure costs or operating expenses such as maintenance. However, robots are placed on varying contract terms, including no upfront fee for volume-based contracts. In this scenario, there would be no upfront cost to the hospital. The study also included both fixed and variable costs in cost calculation and amortization of capital expenditures would be part of variable costs thereby minimizing the impact of these limitations.

This study is also limited in that the findings from the Premier Healthcare Database may not be generalizable to all hospitalized patients with mTKA and rTKA, particularly those in other countries. Also, Patients treated in non-Premier Inc. hospitals for post-op care will be lost to follow-up in this study. The study has tried to capture as many patient, provider, surgeon and procedure characteristics as available; however, there might be other unmeasured variables (e.g. surgical technique, administrative policy, implants used) that may lead to residual confounding. The study has limitations associated with observational study design and as such the study showcases association in outcomes instead of causal relationship. Despite these limitations, this study provides an informative perspective of the hospital-based experience of care that patients had with rTKA and mTKA up to 90 days post-surgery.

5. Conclusions

These hospital data for more than 67,000 patients, under the limitation of the study (including unavailability of the purchase cost of robot), showed that rTKA was cost-neutral with mTKA over 90 days and may result in a greater discharge to home with fewer post-operative readmissions and revisits, compared to mTKA. These data demonstrate that the adoption of rTKA may benefit patients without adding costs to hospital systems. The economic benefit of rTKA will continue to increase with ongoing improvements in procedure efficiency and standardization.

Declaration of interest

LG, ASC, KM, AG, CEH and DH are employees of Johnson & Johnson (JnJ) and own stock in the company and participated in the design of the study and in the collection, analysis, and interpretation of the data and in the writing the manuscript. SK was a contractor with JnJ at the time of the study. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewers Disclosure

Peer reviewers on this manuscript have no relevant financial relationships or otherwise to disclose.

IRB Approval

The use of Premier Healthcare Database was reviewed by the New England Institutional Review Board (IRB) and was determined to be exempt from broad IRB approval, as this research project did not involve human subject research and used data from an anonymous, de-identified, administrative claims database compliant with the Health Insurance Portability and Accountability Act of 1996.

Availability of Data and Materials

The data for these analyses were made available to the authors by third-party licenses from PREMIER (https://products.premierinc.com/applied-sciences/solutions/applied-research-and-analytics), a data provider in the US. Under the licensing agreement, the authors cannot provide raw data themselves. Other researchers could access the data by purchase through PREMIER, and the inclusion criteria specified in the Methods section would allow them to identify the same cohort of patients we used for these analyses.

Author’s Contributions

TBA, ASC, CEH, AG, KM, DPH and LG participated in the design of the study and in the collection, analysis, and interpretation of the data and in the writing the manuscript. KS, AG and CEH completed all programming for the research. TBA and DPH co-led the study design, reviewed all the data analyses and clinical interpretation, and led the final manuscript preparation.

Acknowledgments

The authors thank Natalie Edwards of Health Services Consulting Corporation, Boxborough, MA, USA for editorial assistance with the manuscript and Anshu Gupta of Johnson and Johnson, NJ, USA for re-doing the analysis to address reviewers’ comments.

Supplementary Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17434440.2023.2185135

Additional information

Funding

This study was funded by Johnson & Johnson (JnJ).