Medicaid Payer Status Is Associated With Increased 90-Day Resource Utilization, Reoperation, and Infection Following Aseptic Revision Total Hip Arthroplasty

Abstract

Background

Prior literature has demonstrated increased resource utilization and perioperative complications in patients with a Medicaid payor status undergoing primary total hip and knee arthroplasty. This relationship has yet to be explored in patients undergoing revision total hip arthroplasty (rTHA).

Methods

The National Readmissions Database was queried from 2010 to 2015 for all patients undergoing aseptic rTHA. 90-day complication data were collected, and patients were separated into two cohorts based on insurance payor type: Medicaid and non-Medicaid. Patients were propensity score matched 2:1 on a number of comorbid and operative characteristics. The relationship between Medicaid payor status and postoperative outcomes was then assessed using binomial logistic regression analysis.

Results

3,110 Medicaid patients were identified and matched to 6,175 non-Medicaid patients. Medicaid patients had increased odds of an early prosthetic joint infection (Odds Ratio [OR] 1.29, p=0.019), superficial surgical site infection (OR: 1.48, p=0.003), and early reoperation (OR: 1.18, p=0.045). Medicaid patients also experienced higher odds of readmissions, extended length of stay, non-home discharge status, and medical complications. Finally, the Medicaid cohort had a $3,332 (95% CI: 2,412-4,253, p<0.001) increased adjusted total cost of care when compared to the non-Medicaid cohort.

Conclusion

This study identifies the Medicaid payor status as an independent risk factor for increased resource utilization, reoperation, and infection in the early postoperative period for patients undergoing rTHA. This relationship is likely due to an interplay of multiple variables, including socioeconomic status and access to care.

Level of Evidence: IV

Introduction

Total hip arthroplasty (THA) remains one of the most common and successful surgeries performed in the United States (U.S.).¹ Nearly $15 billion in annual health-care expenditures are allocated to patients undergoing primary THA, with projected growth continuing through 2030.^2-4 The expansion of primary THA has driven a concomitant increase in revision THA (rTHA) volume.^5,6 The economic impact of rTHA is substantial^5,7-10 and rTHA disproportionately consumes nearly 20% of the total Medicare expenditures dedicated to hip arthroplasty.8 With the emergence of value-based payment models aimed at curtailing these costs, preoperative identification of patients at risk of above-expected resource utilization has become increasingly important.^11-15

Following primary total joint arthroplasty, Medicaid payor status has been found to be independently associated with increased morbidity, mortality, postoperative resource utilization, and costs when compared to privately insured patients.^4,16-22 However, it is not known whether the same relationship exists in the rTHA cohort. With the growing emphasis on value, this relationship is important to define. Understanding the relationship between Medicaid payor status and resource utilization following rTHA could aid in the refinement of risk-adjustment of reimbursement models.

Risk adjustment is central to the sustainability of bundled care models to avoid unintentionally disincentivizing providers from caring for high risk patients. Clearly, this also risks exacerbating health disparities. Therefore, the purpose of this study is to compare resource utilization, complications, and readmissions between patients with and without Medicaid payor status following rTHA. We hypothesized that Medicaid payor status would be associated with increased readmission, complications, resource utilization and total cost of care when compared to matched controls.

Methods

Patient Selection

Population-level data was acquired from the Nationwide Readmission Database (NRD). The NRD is a publicly available all-payor database maintained by the Agency for Healthcare Research and Quality (AHRQ), as part of the Healthcare Cost and Utilization Project (HCUP). The database contains information on all inpatient stays occurring within a calendar year in 27 states. The database is coded such that same-state readmissions occurring within the same calendar year can be captured, even if the readmission occurred at a different hospital than the initial inpatient stay. Thus, the database allows for the capture of inpatient complications and those occurring during or causing a subsequent readmission – allowing researchers to capture major complications occurring in the 90-day postoperative period. The NRD has been used previously to study the association of Medicaid status on postoperative outcomes in a primary hip and knee arthroplasty setting.⁴

For the purposes of this study, the NRD was queried from 2010-2015 for patients undergoing rTHA using International Classification of Disease Ninth Revision-Clinical Modification (ICD-9) procedural codes 00.70 (revision of hip replacement, both acetabular and femoral components), 00.71 (revision of hip replacement, acetabular component), 00.72 (revision of hip replacement, femoral component), 00.73 (revision of hip replacement, acetabular liner and/or femoral head only), and 81.53 (revision of hip replacement, not otherwise specified). As the database does not allow for patient follow-up between calendar years, quarter 4 of each year was excluded as 90-day follow-up for these patients could not be ensured. Additionally, the study period was stopped in the 3rd quarter of 2015 to prevent cross over with ICD-10 coding which could compromise cohort homogeneity. Furthermore, we excluded those undergoing revision for a prosthetic joint infection (ICD-9 code 996.66) as these patients are known to have different risk profiles and postoperative course then those undergoing revision for aseptic indications. Furthermore, those under 18 years old and those with missing baseline information or payor status (described below) were also excluded.

The primary payor for the index operation was identified. The national readmission database categorizes the payor as either Medicare, Medicaid, private insurance, self-pay, no charge, or other (includes Worker’s Compensation, the Civilian Health and Medical Program of the Uniformed Services, and the Civilian Health and Medical Program of the Department of Veterans Affairs). Two cohorts of patients were then created, those with a Medicaid payor and those with other payor types (those with any of the other categories above, per precedence).⁴

Baseline demographic and operative characteristics were also identified. This included age, sex, surgery type (both components, acetabulum only, femur only, or isolated modular component exchange), zip-code income quartile (i.e. local prosperity as a composite estimate of socioeconomic status), hospital funding (private or public), and patient smoking status. Additionally, comorbid information was collected using the Elixhauser comorbidity index as done in a previous study.²³ The number of comorbidities were tallied and quantified as the total number of comorbidities present. These were grouped as follows: 0, 1, 2, 3+. Each comorbidity was weighted equally.

Outcomes of Interest

Postoperative complications were identified using ICD-9 diagnosis and procedure codes, as well as variables unique to the NRD (which included discharge status, length of stay (LOS), and hospital charges). Medical complications were included if they occurred during the initial inpatient stay or if they prompted readmission. Readmission for prosthetic joint infection or repeat revision surgery were counted separately. Extended LOS was defined as those patients staying in the hospital for greater than 4 days. Hospital charges were converted to cost with the use of cost-to-charge ratios provided by the NRD. Additionally, cost was adjusted for inflation to 2015 dollars using the US Consumer Price Index.²⁴

Statistical Analysis

This investigation utilized chi-square or independent sample T-tests to perform univariate analysis. Our raw data demonstrated inherent differences in baseline variables between Medicaid and non-Medicaid patients (Table 1). To adequately control for these baseline characteristics, 2-to-1 propensity score matching was subsequently utilized to match two non-Medicaid controls to one Medicaid patient.

Table 1.

Medicaid Payor Status in Revision THA by Patient Demogrpahics and Comorbidities, Unmatched and Matched Cohorts

	Unmtached				Matched
Characteristic	Non-Medicaid	Medicaid	P-Value	Stand. Diff.	Non-Medicaid	Medicaid	P-Value	Stand. Diff.
Total	71,783 (95.85)	3,111 (4.15)	<0.001	1.316	6,175 (66.51)	3,110 (33.49)		0.005
Age, yrs (mean ± SD)	67.81 ± 12.68	52.27 ± 10.83			52.22 ± 11.17	52.28 ± 10.82	0.812
Sex			0.093	0.031				0.014
Male	29,692 (41.36)	1,334 (42.88)			2,691 (43.58)	1,334 (42.89)	0.530
Female	42,091 (58.64)	1,777 (57.12)			3,484 (56.42)	1,776 (57.11)
Elixhauser			0.130	0.028				0.025
0	10,446 (14.55)	484 (15.56)		0.006	904 (14.64)	483 (15.53)	0.717	0.003
1	10,446 (14.55)	734 (23.59)		0.037	1,465 (23.72)	734 (23.60)		0.003
2	10,446 (14.55)	692 (22.28)		0.016	1,383 (22.40)	693 (22.28)		0.013
3+	10,446 (14.55)	1,200 (38.57)			2,423 (39.24)	1,200 (38.59)
Component Replaced			<0.001	0.028				0.040
Both Components	36,983 (51.62)	1,774 (57.02)		0.006	3,641 (58.96)	1,773 (57.01)	0.141	0.003
Acetabulum	11,335 (15.79)	468 (15.04)		0.051	923 (14.95)	468 (15.05)		0.011
Femoral	12,532 (17.46)	484 (15.56)		0.083	937 (15.17)	484 (15.56)		0.046
Liner or Head-Ball Exchange	10,933 (15.23)	385 (12.38)			674 (10.91)	385 (12.38)
Private Hospital			<0.001	0.202				0.052
No	7,716 (10.75)	553 (17.78)			975 (15.79)	552 (17.75)	0.016
Yes	64,067 (89.25)	2,558 (82.22)			5,200 (84.21)	2,558 (82.25)
Income Quartile of Patient Zipcode			<0.001	0.395				0.010
1	14,576 (20.31)	1,179 (37.90)		0.051	2,308 (37.38)	1,178 (37.88)	0.761	0.003
2	17,149 (23.89)	812 (26.10)		0.124	1,620 (26.23)	812 (26.11)		0.010
3	19,277 (26.85)	670 (21.54)		0.357	1,305 (21.13)	670 (21.54)		0.022
4	20,781 (28.95)	450 (14.46)			942 (15.26)	450 (14.47)
Smoking Status			<0.001	0.582				0.049
No	65,100 (90.69)	2,118 (68.08)			4,345 (70.36)	2,118 (68.10)	0.025
Yes	6,683 (9.31)	993 (31.92)			1,830 (29.64)	992 (31.90)

SD: Standard Deviation, Stand. Diff.: Standardized Difference, DM: Diabetes Mellitus; *Lower standardized difference suggests a smaller difference between cohorts, value <0.10 indicates a negligable difference

In order to calculate propensity scores, binary logistic regression was utilized, and the payor status (Medicaid) served as the outcome variable. Our study identified several confounding variables (Table 1), which were included in the propensity score model. Non-Medicaid patients were then matched 2-to-1 to Medicaid patients using a greedy matching algorithm based on propensity score. This study utilized caliper matching, in which non-Medicaid patients within a certain caliper width of the propensity score of a Medicaid patient had potential to be matched. Our caliper width was set at 0.20 standard deviations of the logit of the propensity score. Patients were matched at random when multiple control subjects had propensity scores that were equally close to an exposed subject.

The propensity score distribution between matched (Figure 1B) and unmatched (Figure 1A) data sets were compared to assess covariate balance. The propensity score distributions of Medicaid patients and non-Medicaid patients in the matched data set most closely resembled one another. Standardized differences between covariates before and after matching were then analyzed, with a standardized difference of <0.10 suggesting negligible difference between control and exposure groups.

Figure 1.

Propensity-score Distribution in Unmatched and Matched Revision THA Data Sets: Non-Medicaid and Medicaid Patients. (A) Propensity Score Distribution (Density) of Unmatched Datase. (B) Propensity Score Distribution (Density) of Matched Dataset.

Finally, to isolate the effect of Medicaid payor status within the parameters of this study, binomial logistic regression was then utilized to control for any remaining cohort differences. Total healthcare costs between Medicaid and non-Medicaid patients were compared using a generalized linear model with gamma distribution and a logarithmic link function. Table 1 demonstrates the covariates in these models. This study utilized SAS (version 9.4, Cary, NC) for all statistical analysis, and generated Figure 1 with R (version 4.0.2).

Results

Baseline Demographics and Matching

A total of 74,894 patients in the NRD were identified as undergoing aseptic rTHA between 2010-2015 (Table 1). This included 71,783 (95.9%) non-Medicaid patients and 3,111 (4.2%) Medicaid patients. At baseline, there were multiple differences between payor cohorts in the unmatched dataset. Non-Medicaid patients were on average 15 years older than the Medicaid patients (67.81 vs. 52.27, p<0.001). Likewise, Medicaid patients were more likely to be smokers (31.9% vs. 9.31%, p<0.001), be of the lowest zip-code income quartile (37.90% vs. 20.31%), undergo both component revision (57.02% vs. 51.62%, p<0.001), and undergo revision at a publicly funded hospital (17.78% vs. 10.75%, p<0.001). Many of these variables had large standardized differences, indicating cohort heterogeneity.

2-to-1 propensity score matching produced cohorts of 6,175 (66.5%) non-Medicaid patients and 3,110 (33.5%) Medicaid patients. All standardized differences in the matched dataset met a criterion of <0.10 (all standardized differences <0.052, Table 1) indicating a successful match.

Payor Status and Resource Utilization

Resource utilization metrics were compared between Medicaid and non-Medicaid patients, including readmission, discharge status, extended LOS, and hospital costs during index admission (Table 2). 30-day and 90-day readmission rates were 9.9% and 17.5% for non-Medicaid patients and 12.5% and 21.4% for Medicaid patients, respectively (p-value <0.001 and < 0.001). Likewise, rates of an extended length of stay (LOS) for non-Medicaid and Medicaid patients were 39.8% and 52.8%, respectively. Similarly, rates of non-home discharge were 21.2% and 26.5%, respectively (p-value <0.01). Cost of the initial inpatient stay was $18,506 for non-Medicaid patients and $20,674 for Medicaid patients (p-value < 0.001). Using multivariate analysis that controlled for demographic and comorbid data, Medicaid patients demonstrated a 1.30 times increased odds of 30-day readmission (95% confidence interval [95% CI]: 1.14-1.49, p-value <0.001) when compared to non-Medicaid patients. Similarly, Medicaid patients had a 1.29 times increased odds of 90-day readmission (95% CI 1.14-1.44, p-value <0.001), a 1.78 times increased odds of an extended LOS (95% CI 1.62-1.95, p-value < 0.001), and a 1.42 times increased odds of a non-home discharge destination (95% CI 1.271.58, p-value < 0.001) when compared to non-Medicaid patients (Table 3). Additionally, there was a $3,332 (CI:2,412-4,253; p-value <0.01) adjusted cost increase when comparing Medicaid patients to non-Medicaid patients.

Table 2.

Univariate Analysis of 90-Day Complications in Matched Cohorts, rTHA

Characteristic	Non-Medicaid	Medicaid	P-Value
30-Day Readmission	614 (9.94)	390 (12.54)	<0.001
90-Day Readmission	1,080 (17.49)	665 (21.38)	<0.001
Extended LOS (>4 Days)	2,455 (39.76)	1,642 (52.80)	<0.001
Non-Home Discharge	1,306 (21.16)	825 (26.54)	<0.001
Any Wound Infection	319 (5.17)	211 (6.78)	0.002
Early Prosthetic Joint Infection	224 (3.63)	145 (4.66)	0.016
Superficial SSI	137 (2.22)	100 (3.22)	0.004
Wound Dehisence	149 (2.41)	83 (2.67)	0.456
Early Reoperation	417 (6.75)	248 (7.97)	0.031
Myocardial Infarction	20 (0.32)	11 (0.35)	0.814
Pneumonia	242 (3.92)	157 (5.05)	0.013
Deep Vein Thrombosis	93 (1.51)	72 (2.32)	0.005
Acute Kidney Injury	248 (4.02)	164 (5.27)	0.006
Urinary Tract Infection	312 (5.05)	218 (7.01)	<0.001
C. difficile Infection	32 (0.52)	27 (0.87)	0.045
Cost [USD, IQR]	18,506 [12,927]	20,674 [15,389]	<0.001

LOS: Length of Stay, SSI; Surgical Site Infection

Table 3.

Multivariate Analysis of 90-Day Complications in Matched Cohorts, rTHA

Characteristic	Odds Ratio (95% CI)	P-Value
30-Day Readmission	1.30 (1.14-1.49)	<0.001
90-Day Readmission	1.29 (1.15-1.44)	<0.001
Extended LOS (>4 Days)	1.78 (1.62-1.95)	<0.001
Non-Home Discharge	1.42 (1.27-1.58)	<0.001
Any Wound Infection	1.34 (1.11-1.60)	0.002
Early Prosthetic Joint Infection	1.29 (1.04-1.60)	0.019
Superficial SSI	1.48 (1.14-1.92)	0.003
Wound Dehisence	1.10 (0.84-1.45)	0.474
Early Reoperation	1.18 (1.01-1.39)	0.045
Myocardial Infarction	1.10 (0.52-2.32)	0.789
Pneumonia	1.32 (1.07-1.63)	0.008
Deep Vein Thrombosis	1.577 (1.15-2.15)	0.004
Acute Kidney Injury	1.37 (1.18-1.69)	0.003
Urinary Tract Infection	1.45 (1.21-1.75)	<0.001
C. difficile Infection	1.74 (1.04-2.91)	0.036
Adjusted Cost Difference [USD]	+$3,332 (2,412-4,253)	<0.001

LOS: Length of Stay, SSI; Surgical Site Infection; CI: confidence interval

Payor Status and Surgical Complications

Surgical complications, including wound infection, prosthetic joint infection, and early reoperation, were also higher in Medicaid patients compared to non-Medicaid patients (Table 2). For instance, Medicaid patients had higher incidence of PJI (3.6% v.4.67%, p=0.016; non-Medicaid v. Medicaid), superficial surgical site infection (2.2% v. 3.2%, p=0.004), and early reoperation (6.75% v. 7.97%, p=0.031). This relationship was conserved on multivariate analysis, where Medicaid patients had increased odds of an early prosthetic joint infection (Odds Ratio [OR] 1.29, CI:1.04-1.60, p=0.019), superficial SSI (OR: 1.48, CI:1.14-1.92, p=0.003), and early reoperation (OR: 1.18, CI:1.01-1.39, p=0.045) (Table 3). However, Medicaid payor status did not impact the odds of developing wound dehiscence (OR: 1.10, CI:0.84-1.45, p=0.474).

Payor Status and Medical Complications

On univariate analysis, Medicaid payor status was also associated with most medical complications queried, including pneumonia, deep vein thrombosis (DVT), or acute kidney injury (Table 2). On multivariate analysis this equated to a 1.32 (CI:1.07-1.63,p=0.008) times increased odds of pneumonia, a 1.57 (CI:1.15-2.15, p=0.004) times increased odds of deep vein thrombosis, a 1.37 (CI:1.18-1.69, p=0.003) times increased odds of acute kidney injury, and a 1.45 (CI:1.21-1.75, p<0.001) times odds of developing a urinary tract infection (Table 3). No difference in rates of myocardial infarction were seen (OR:1.10, CI:0.52-2.32, p=0.789).

Discussion

As a result of the Affordable Care Act (ACA) Medicaid has undergone a substantial expansion. Additionally, the incidence of rTHA continues to rise.^2,3 The relationship between Medicaid payor status and outcomes following rTHA was previously unknown. However, the results of this investigation indicated that Medicaid payor status is associated with increased odds of PJI, revision surgery, medical complications, and DVT after rTHA. To our knowledge, this is the first study that performs a population-level analysis to examine the association of payor status with postoperative complications and resource utilization metrics following rTHA. Prior literature investigating similar parameters has been limited by less representative populations, combining hip and knee arthroplasty data, and/or does not focus on revision arthroplasty despite its unique risk and cost profile from primary reconstruction.^{4,15,18,19,25,26}

In our study, both univariate and multivariate analysis demonstrated significant differences in 90-day outcomes between matched Medicaid and non-Medicaid cohorts. Prior studies have demonstrated that while Medicaid patients are younger in age, they often have higher prevalence of medical comorbidities.^18,21 Our study design addressed these known differences (i.e., smoking history and medical comorbidities) through propensity score matching. However, given the granularity of our population-level data, we are unable to assess the disease severity of these matched factors. While this represents an obvious limitation in our work, it is also highlighting a relationship which may be driving a component of the risk observed in patients with Medicaid payor status. Transcending the boundaries of specialty and disease metrics, Medicaid patients often have more severe disease and more disease sequelae across many phases of medical care.^27-36 This has routinely been attributed to a lack of access to quality care, a disparity that is well documented for patients on need-based federally-funded health insurance.^37,38 This represents a complex issue, as it is clear that payor status is not a direct harbinger of an unsatisfactory outcome, but rather represents a meaningful composite marker of a constellation of risk factors that sum to drive our findings. However, when combined with the advanced cost associated with the perioperative care of rTHA patients, this threatens to disrupt the goals of systematic changes.^7-10

There are many socioeconomic factors which may further elucidate the discrepant outcomes observed between the Medicaid and privately insured patient. The Medicaid population is known to more frequently have limited access to care, lack of home support for perioperative rehabilitation, lower income status and lower levels of health literacy.^37,38 While we have identified Medicaid payor status as a risk factor for increased 90-day complications, resource utilization and readmission rates, this association needs to be further explored. The Medicaid population often has inferior social and fiscal determinants of health, which may increase their predisposition for postoperative complications and readmissions. For instance, this population has previously been shown to have a limited ability to obtain appropriate transportation to their clinical visits.^37,38 Therefore, patients undergoing rTHA may be physically unable to attend their postoperative visits and physical therapy appointments due to transportation constraints. Furthermore, the vast majority of surgical candidates in an orthopaedic preoperative clinic consists of referrals from a primary care provider. Due to the limited number of primary care providers, long wait times for appointments and reduced willingness of physicians to care for the Medicaid patient population,³⁷ Medicaid patients may often present with advanced or neglected hip pathologies. Further, unlike its Medicare counterpart, the Medicaid program currently offers minimal funding for inpatient rehabilitation programs which may lower the recovery ceiling in these patients.

While the ACA enables previously uninsured patients to gain insured access to elective and semi-elective adult reconstructive procedures, the increased resource utilization and 90-day complications associated with Medicaid payor status demonstrated in this study have significant implications on access to care. As alternate payment models (APMs), such as bundled payment models, become more prevalent, hospitals and providers inherently benefit by reducing costs without compromising quality of care.^11,12 These findings are concerning as current payment models have failed to adjust reimbursements despite previous evidence demonstrating increased rates of inpatient mortality as well as inferior peri- and post-operative outcomes associated with Medicaid payor status.^16-20

In this study, Medicaid patients demonstrated significantly higher 30- and 90-day readmission rates, extended LOS, non-home discharge and as a result, increased adjusted cost differences ($+3,332) when compared to non-Medicaid patients undergoing rTHA. In bundled payment models, readmission and discharge to a rehabilitation facility consume a substantial portion of the allocated episode of care resources.^39,40 Furthermore, our study identified higher rates of wound infection, superficial SSI, early reoperation rates and medical complications among Medicaid patients. These complications, in combination or alone, dramatically increase both the surgical morbidity and costs per episode of care, for patients with Medicaid payor status. As has been demonstrated in the primary arthroplasty setting,⁴ this has the undesired potential to disincentivize surgeons and institutions to assume care for a costlier cohort with inferior outcomes. Unfortunately, this is not theoretical, as it is known that patients with commercial insurance have a higher likelihood of having their insurance accepted by orthopaedic surgeons than those with Medicaid.⁴¹ This is compounded by the fact that Medicaid reimburses less than private insures.⁴² Given this, it is our hope hope that this data is used to drive policy that will appropriately risk adjust reimbursement models. Ultimately, this approach will have an intended long-term goal of diminishing the clear distinction between the Medicaid and non-Medicaid patient that currently exists and allow for more equitable access to care.

This study has several strengths and limitations. A key strength of this study lies in the large, nationally representative sample derived from a nationwide database. Furthermore, the statistical methodology used in this study is robust, controlling for cohort heterogeneity to the greatest extent possible. Still, there are limitations which must be addressed. First, as with any analysis of a large database, we are reliant on complete and accurate coding of procedures and complications. Second, while our statistical methodology is robust and controls for cohort heterogeneity to the greatest extent possible, the potential for unmeasured confounding persists. For instance, as mentioned previously, we are unable to control for some components of socioeconomic status as well as comorbid disease severity. These include factors such as education level, employment status, personal income level, among others. Other factors such as surgical technique, surgeon experience and case complexity of a heterogeneous procedure were not incorporated in our analysis. Still, we used zip code income quartile as a surrogate for socioeconomic status and controlled for the presence of many comorbidities, therefore isolating payor status to the greatest extent possible.

Conclusion

In this study, Medicaid patients undergoing rTHA demonstrated increased 90-day readmission rates, all-cause 90-day morbidity, hospital length of stay, resource utilization and total cost of care when compared to matched non-Medicaid patients. As such, Medicaid likely represents a reliable index of these patients’ constellations of social determinants of health. As APMs, such as bundled payment models, continue to increase in prevalence, concerns over disparate access to care for this population may be perpetuated and exacerbated. In order to provide equitable care for this vulnerable patient population and avoid financial penalties for surgeons and hospitals, risk adjustment models should account for Medicaid payor status. Unlike primary THA, much of rTHA is typically, at best, semi-elective, and may be obligate to restoring a functional hip. While it is clear the Medicaid population needs high quality, comprehensive care, this must not come at a penalty to surgeons and hospitals.