Quantifying the Opportunity Cost of Resident Involvement in Academic Orthopedic Shoulder Arthroplasty: A Matched – Pair Analysis

Abstract

Background

There is minimal work defining the economic impact of resident participation in shoulder arthroplasty. Thus, this study quantified the opportunity cost of resident participation in total shoulder arthroplasty (TSA) and hemiarthroplasty (HA) by determining differences in operative time, relative value units (RVUs)/hour, and RVUs/case.

Methods

A retrospective analysis of shoulder arthroplasty procedures were identified from the ACS-NSQIP database from 2006 to 2014 using CPT codes. Demographic, comorbidity, preoperative laboratory data and surgical procedure were used to develop matched cohorts. Mean differences in operative time, RVUs/case and RVUs/hour between attending-only (AO) cases and cases with resident involvement (RI) were examined. Cost analysis was performed to identify differences in RVUs generated per hour in dollars/case.

Results

A total of 1786 AO and 1102 RI cases were identified. With the exception of PGY-3 and PGY-4 cases, RI cases had lower mean operative times compared to AO cases. The cost of RI was highest for PGY-3 ($199.87 per case) and PGY-4 ($9 .2 9) residents with all other postgraduate years providing a cost reduction.

Discussion

Involvement of residents was associated with shorter operative times leading to a savings of $29.64 per case. Involvement of intermediate-level (PGY-3) residents were associated with increased costs that ultimately decreased as residents became more senior.

Introduction

Resident participation in surgery is a defining feature of an academic medical center. Prior studies have found that the involvement of residents in orthopedic procedures is not associated with an increase in mortality or rate of major complications.^1–5 The data is conflicting, however, when considering the impact of resident involvement on operative time. Literature examining resident involvement in total hip arthroplasty (THA), total knee arthroplasty (TKA), and studies combining multiple orthopedic procedures have found that trainee participation is associated with increased operative time.^1,5–7 In contrast, Cvetanovich and colleagues found that there was a significant decrease in operative time with resident involvement in total shoulder arthroplasty (TSA). ⁴

Recently published models predict shoulder arthroplasty volume to grow by 235% from 2 0 1 7 to 2025. ⁸ Given the increased demand and steep learning curve, it is increasingly important for residents to participate in these procedures, but the impact of that participation on operative time is unclear. ⁹ While important to further clarify the association between resident involvement in TSA and operative time, there is also the need for work evaluating the financial impact of variation in operative time due to resident participation. Extrinsic pressures, such as the shift toward value-based care and the steadily declining reimbursement rate for common orthopedic procedures necessitate a thorough understanding of a surgeon's cost structure.^10,11 One important cost to consider is opportunity cost, defined as “the potential loss or gain when one choice is made in lieu of another.” ¹² In the context of trainee participation, opportunity cost can be thought of as the amount of cases one could have completed if time had not been spent teaching. While a recent study has quantified the opportunity cost of resident participation in common sports medicine procedures, there is no work defining the opportunity cost of resident participation in shoulder arthroplasty. ¹³

Thus, the purpose of our study is to quantify the opportunity cost of resident participation in TSA and hemiarthroplasty (HA) by determining differences in operative time, relative value units (RVUs) per hour, and RVUs per case between cases with resident involvement and cases performed only by attending surgeons. Based on Beletsky et al.'s prior investigation we hypothesized that: 1) cases with resident involvement would exhibit longer operative times and decreased RVUs per hour when matched to attending-only cases and 2) operative times would increase as residents became more senior in postgraduate year (PGY), thus resulting in an inverse relationship between seniority and RVUs per hour. We suspect that increasing resident seniority will result in increased operative autonomy as reflected by prolonged operative times, thus resulting in less RVUs per hour.

Materials and methods

Using a study design similar to Beletsky et al., ¹³ a retrospective analysis was performed using the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database from 2006 to 2014. Patients were identified using Current Procedural Terminology (CPT) codes for the common shoulder arthroplasty codes performed within academic practice settings. Cases with codes for total shoulder arthroplasty and reverse total shoulder arthroplasty (23472) and hemiarthroplasty (23470) were queried from the database. Baseline demographic variables, information on preoperative comorbidities and preoperative laboratory values were examined (Tables 1 and 2). To perform an analysis of cost, CPT codes, RVUs, operative time, presence or absence of a resident physician and PGY level were collected. Within our analysis, junior residents were defined as PGY1 and PGY2 residents while senior residents were defined as residents at the PGY3 level and higher.^13,14 All cases lacking information on attending presence or PGY-level, full demographic and laboratory data were excluded from the analysis.

Table 1.

Preoperative characteristics of unmatched and matched cohorts.

	Unmatched			Matched
	Attending (N = 1079)	Resident (N = 966)	P value	Attending (N = 796)	Resident (N = 796)	P Value
Demographic Variables
Age, yr	69.2 ± 11.0	68.4 ± 11.9	0.248	69.1 ± 10.8	70.2 ± 10.9	0.028
Weight, lbs	185.6 ± 52.7	186.3 ± 49.0	0.797	185.1 ± 52.0	186.6 ± 48.7	0.905
Height, inches	64.2 ± 15.6	64.6 ± 12.6	0.751	64.0 ± 17.0	64.3 ± 13.7	0.346
Female	625 (57.9)	569 (58.9)	0.654	457 (57.4)	467 (58.7)	0.612
Race			<0.001			0.0002
Black or African American	31 (2.9)	48 (5.0)		21 (2.6)	40 (5.0)
White	856 (79.3)	791 (81.9)		642 (80.7)	643 (80.8)
Asian	5 (0.5)	3 (0.3)		3 (0.4)	3 (0.4)
Hispanic	52 (4.8)	11 (1.1)		39 (4.9)	11 (1.4)
Other	135 (12.5)	113 (11.7)		91 (11.4)	99 (12.4)
Preoperative Comorbidities
Inpatient Stay	962 (89.2)	938 (97.1)	<0.001	763 (95.9)	768 (96.5)	0.514
ASA class			0.004			0.264
1	25 (2.3)	13 (1.3)		17 (2.1)	9 (1.1)
2	467 (43.3)	489 (50.6)		379 (47.6)	373 (46.9)
3	555 (51.4)	445 (46.1)		378 (47.5)	398 (50.0)
4	32 (3.0)	19 (2.0)		22 (2.8)	16 (2.0)
Diabetics	193 (17.9)	175 (18.1)	0.893	131 (16.5)	152 (19.1)	0.169
Smokers	104 (9.6)	107 (11.1)	0.286	71 (.9)	83 (10.4)	0.309
COPD	64 (5.9)	49 (5.1)	0.396	42 (5.3%)	42 (5.3%)	1.000
HTN	746 (69.1)	663 (68.6)	0.806	549 (69.0)	566 (71.1)	0.352
Steroid Use	44 (4.1)	58 (6.0)	0.046	35 (4.4)	24 (3.0)	0.145
Recent Weight Loss	3 (0.3)	3 (0.3)	0.892	3 (0.4)	3 (0.4)	1.000
Bleeding disorders	27 (2.5)	37 (3.8)	0 .0 8 5	21 (2.6)	11 (1.4)	0.074
Independent Functional Status	1018 (94.3)	899 (93.1)	0.232	760 (95.5)	742 (93.2)	0.518
Preoperative laboratory values
Creatinine	0.97 ± 0.5	0.96 ± 0.5	0 .5 5 0	1.0 ± 0.4	1.0 ± 0.6	0.167
WBC, 1000 cells/mm3	7.3 ± 2.3	7.1 ± 2.2	0.006	7.02 ± 2.3	7.2 ± 2.2	0.671
Hematocrit, %	39.9 ± 4.8	39.7 ± 4.5	0.609	39.89 ± 4.7	40.0 ± 4.3	0.706
Platelet, 1000 cells/mm3	248.3 ± 73.3	240.9 ± 72.7	0.029	244.8 ± 69.3	245.9 ± 71.4	0.895

Data presented as mean ± standard deviation or count (percentage).

ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; WBC, white blood cell; HTN, hypertension.

*Statistically significance at α = 0.05.

Table 2.

Mean operative time stratified by current procedural terminology code.

		Mean Operative Time, min
CPT code	Procedure	Difference	Attending	Resident	P value
23472	Total Shoulder Arthroplasty	+ 8.9	125.6	116.7	<0.0001
23470	Hemiarthroplasty	−18.4	106.1	124.5	0.0018

*Statistically significance at α = 0.05.

Statistical and cost analyses

Two equally sized cohorts were created with a matching algorithm using the nearest-neighbor method. Cohorts were created based on complete sets of available demographic, preoperative comorbidity data, CPT codes and laboratory values. ¹⁵ The initial cohort contained cases in which only attendings were involved and coded in the NSQIP databases as “attending in OR” or “attending alone” cases. The matching cohort was composed of cases in which residents were present and coded as “attending and resident in OR” cases. ¹⁶ Bivariate analyses were performed using Wilcoxon Rank Sum test for non-parametric continuous variables, Student t tests for parametric continuous variables and χ2 testing to examine differences in baseline categorical demographic, comorbidity and laboratory variables prior to and after matching. Operative time and RVU-per-hour values were stratified by CPT code to quantify relative effect sizes and statistically significant relationships in operative time and RVU-per-hour between attending and resident cohorts. Statistical significance was defined as p < 0 .0 5 . A post-hoc analysis was performed to identify the source of difference between significant categorical demographic variables. A Bonferroni correction was applied for any analysis with repeated measures.

Resident participation was stratified by PGY-level in order to examine if significant differences existed in RVU-per-hour by each level of training. The same comparison was also made between junior and senior residents. Differences in continuous variables were evaluated using Kruskal-Wallis or Welch 2-sided t-tests where appropriate. Post hoc comparisons were made using the Dwass, Steel, Critchlow-Fligner Method. The primary outcome, cost of resident involvement, was determined by comparing the differences in operative time between attending-only and resident-involved cases using previously described methods. (Figure 1)^13,17 The generic 2020 Medicare Conversion Factor ($36.09/RVU) was used to translate the lost RVUs per hour into cost data correlating with the cost of resident education. A geographic pricing cost index of 1.0, the national average, was used across cost calculations in the study. All statistical analyses were performed using SAS University Edition (Version 15.1, Cary, NC).

Figure 1.

Formulas used to estimate cost of resident involvement (RVU, relative value unit).

Results

A total of 1786 attending only and 1102 resident cases were identified prior to matching. Significant differences were found between postoperative inpatient admissions, race, steroid use, white blood cell (WBC) count, platelets and American Society of Anesthesiologist (ASA) classification (p < 0.001). Following matching, significant differences remained between race (p = 0.0002), and age (69.1 ± 10.8 vs 70.2 ± 10.9, p = 0.028) in attending only and resident cases (n = 831 for each cohort). Post-hoc analyses revealed significant difference between populations of Hispanic patients in attending-only and resident cases (p < 0 .0 0 5) (Table 1)

Mean operative times were significantly longer in attending-only total shoulder arthroplasties compared to those with resident involvement (1 2 5 .6 min vs 1 1 6 .7 min, p < 0 .0 0 1). Conversely, operative times for hemiarthroplasties were longer when residents were involved compared to attending-only settings (1 2 4 .5 min vs 1 0 6 .1 min, p = 0 .0 0 1 8). (Table 2) When stratifying operative time by PGY-level, interns had the shortest operative time compared to all other PGY years while residents at the PGY-3 level had the longest mean operative time (p < 0 .0 0 0 1). Cases with PGY-3 residents had statistically significant longer operative times when compared with interns (p < 0 .0 1), PGY-2 (p < 0 .0 0 1), PGY-4 (p < 0 .0 1), PGY-5 (p < 0 .0 0 1) and PGY-6 (p < 0 .0 0 1) residents. Full data on mean operative times, RVUs per hour and cost as a function of PGY level, can be found in Table 3.

Table 3.

Mean Operative Time, Relative Value Units per Hour, and Cost Across PGYs.

	Operative Time, min	RVUs per Hour	Cost, $
Attending Only	122.1 ± 46.6	12.6 ± 6.7
PGY-1 (n = 1 6)	101.9 ± 31.5	13.2 ± 4.7	-$152.74
PGY-2 (n = 6 5)	115.9 ± 53.3	13.5 ± 6.1	-$47.26
PGY-3 (n = 1 0 3)	148.5 ± 64.1	9.7 ± 3.6	+ $199.87
PGY-4 (n = 1 2 3)	123.3 ± 46.3	12.2 ± 4.7	+ $9.29
PGY-5 (n = 2 0 3)	103.9 ± 40.4	15.3 ± 15.7	-$137.70
PGY-6 (n = 2 8 6)	116.6 ± 51.9	13.2 ± 5.2	-$41.46

Data presented as mean ± standard deviation or mean.

PGY, postgraduate year; RVU, relative value unit.

*Statistically significance at α = 0.05.

Analysis of RVUs per hour, found that PGY-5 resident involvement was associated with greater values compared to than attending-only cases (p < 0.0001), PGY-3 (p < 0.0001), and PGY4 (p = 0.0013) cases with PGY-6 level resident involvement also being associated with greater values compared to PGY-3 (p < 0.0001) and attending cases (p = 0.031). Residents at the PGY-3 level had significantly (p < 0.01) lower RVU per hour values compared to all levels of resident involvement with the exception of interns. Across individual PGY levels, PGY-5 residents had the lowest associated cost estimate per case (-$137.70), while PGY-3 resident involvement was found have the highest opportunity cost estimate ( + $199.87). Overall, when comparing differences in RVUs per hour, cases with resident involvement were estimated to result in a savings of $29.64 per case. Ultimately, no significant differences were found in operative time or RVUs per hour between junior and senior residents (Table 4).

Table 4.

Comparison of Junior and Senior Residents.

	Junior Residents (PGY1-2, n = 1 8 4)	Senior Residents (PGY 3 + , n = 7 1 5)	P value
Operative time, min	113.1 ± 49.9	118.8 ± 51.8	0.342
RVUs per hour	13.4 ± 5.8	13.1 ± 9.4	0.572

Data presented as mean ± standard deviation or mean.

RVU, relative value unit.

*Statistically significance at α = 0.05.

Discussion

This study sought to quantify the cost of resident involvement in shoulder arthroplasty procedures performed at academic orthopedic centers. The primary finding of the study was that resident involvement resulted in decreased operative times for all levels except residents at the PGY-3 and PGY-4 training level. Significant increases in mean operative were seen with resident involvement in hemiarthroplasty procedures, while significantly decreased operative times were seen with total shoulder arthroplasty. In terms of RVU-per-hour, PGY-3 resident involvement resulting in lower RVU-per hour values while the greatest values were seen in cases with PGY-5 involvement. Overall, resident involvement resulted in average savings of $26.67 per case; however, PGY-3 level resident involvement was associated with the largest opportunity cost of $199.87 per case.

To date, most investigations on orthopedic resident involvement have focused on the risk of subsequent complications and operative time^2,4,7,18,19 while fewer studies have looked at the potential economic impact of resident participation.^13,20 Within academic medicine, a balance is weighed between educating trainees, while also successfully and efficiently completing a surgical procedure without negatively impacting surgical outcomes. As a result, attending surgeons in academic centers are presented with the challenge of how to best incorporate and educate residents in the operating room. Despite these challenges, resident involvement could portend several benefits from an efficiency and financial standpoint, in addition to the positive impacts on orthopedic care in the perioperative setting. ²¹

Our study showed procedures involving interns, or first-year surgical residents, resulted in the lowest opportunity cost (-$152.74), while PGY3 residents were associated with the largest opportunity cost ( + $1 9 9 .8 7) compared to all other years of training. Ultimately, having first year residents was found to be most beneficial from an operative efficiency standpoint. At this early stage, interns are traditionally afforded minimal operative autonomy, however, are invested learners and knowledgeable assistants likely contributing to much shorter operative times when compared to attending-only cases with varying levels of assistance. ²² At this stage, operative education is focused on understanding the anatomy and steps of the procedure rather than more complex technical aspects. As surgical training progresses, operative autonomy typically increases and, with it, greater expectations in terms of understanding the procedure at hand. More autonomy is typically given to senior levels residents as they are expected to showcase greater levels of competence and efficiency. The resultant decrease in operative time could be interpreted as a proxy for an increase in surgical skill. This explanation is also supported by prior studies on surgical training reporting a decrease in operative times as PGY level increases, thus suggesting increasing efficiency with further development of both surgical skills and confidence within the operating room.^23,24 Additionally, Su et al.'s investigation on the educational experience of treating supracondylar humerus fractures supports this claim that the time dedicated to teaching trainees decreased with increased exposure to the procedure at hand. ²⁵ Although trainees at the PGY-6 level were found to have longer average operative times compared to PGY-5 residents, this may be a factor of case complexity as revision procedures with aberrant anatomy or pathologies with significant deformity could be deemed as more difficult and thus require increased operative times. Similar reasoning could be used to explain the prolonged operative times identified between hemiarthroplasties and shoulder arthroplasties when comparing residents to attending-only cases. Hemiarthroplasties are often performed in the setting of trauma for proximal humerus fractures, potential nerve injury precluding arthroplasty or, historically, rotator cuff arthropathy. Additionally, particular attention needs to be paid to proper reduction of the tuberosities which is critical to the success of the procedure. ²⁶ In such cases, surgeon experience may play a role in the efficient completion of the surgery.

Cvetanovich et al. previously reported a significant decrease in operative time with shoulder arthroplasty procedures involving residents. ⁴ Our findings largely agree with these prior findings with the exception of PGY-3 and PGY-4 level resident involvement representing an inflection point in both average operative time and opportunity cost with shoulder arthroplasty procedures. Third-year residents, although often viewed as intermediate or senior residents by the ACGME, represent a transition point in surgical training where residents incur a steep learning curve as they oftentimes adjust from primarily consult management and direct patient care tasks, to an increased operative experience with greater expectations surgically and progressive autonomy with routine procedures. This shift to a new role could subsequently be a driver for an increase in operative time as increased instruction may be required.

To date, a multitude of studies have reinforced the safety of resident involvement.^2,4,7,18,19 Our investigation provides additional support for the inclusion of residents based on the potential financial implications but also provides an opportunity to asses and overcome potential pitfalls in residency education that might decrease efficiency in the operating room. As PGY-3 level residents incurred the greatest opportunity cost, at the transition from PGY-2 to PGY-3, there may be a need to implement additional interventions to ensure trainees are confident, comfortable and adept at shoulder arthroplasty procedures. Currently, several efforts have been proposed, and integrated, into orthopedic residencies to improve surgical education and clinical competency. Given previously cited concerns about work hour restrictions, cost-related implications and the growing complexity of minimally invasive procedures, discussion of additional learning modalities such as surgical video reviews, ²⁷ virtual simulators,^28,29 and cadaveric dissections ³⁰ have been proposed as adjuncts to surgical education. Further, the concept of mental rehearsal has also been advocated as a helpful tool to accelerate learning, compartmentalize learning and permit recall of surgical steps in the absence of the risk for consequences within the operating room. ³¹ Although currently validated for arthroscopic procedures by Kovacevic et al. such interventions have yet to be evaluated fully for shoulder arthroplasty but could serve as beneficial adjuncts. ³²

Limitations

There are several limitations to note from our study. The data utilized in this investigation was gathered from the NSQIP database which compiles data from multiple academic hospitals contributing to a national registry of surgical procedures. Further, our study can only be generalizable to institutions with similar practice models and may not apply to private practice or smaller community settings. With regards to academic practice models, it is likely that attending-only cases did include additional assistants in the operating room; however, due to limitations in coding, we are unable to examine more granular variables such as the availability and competency of qualified assistants that can further influence operative times. Further, we are unable to comment on the effect of overlapping surgery on operative time. In the setting of multiple operative rooms, an unequal distribution of supervision and assistance might have been divided between rooms with junior residents and senior residents which could have affected overall operative time. Also, attending comfort with working with trainees is variable and due to the limitations of the database used, we are unable to make comparisons between operative times and RVU-based data for the same surgeons in the presence or absence of resident assistance. As this investigation reviewed cases between 2006 and 2 0 1 4, when the NSQIP-ACS database recorded resident involvement, we are unable to comment on if the observed differences in operative time have stayed consistent with the advent of modern advancement s in arthroplasty implants and techniques. A firm understanding of anatomy and techniques to manage remnant rotator cuff and capsular tissue to facilitate exposure remains paramount to performing total shoulder arthroplasty, yet one could argue that with the introduction of implants that tout improved ease of use and speed compared to earlier generations of implants, operative times would be shorter across both resident and attending-only cases. During the time in which resident involvement data was recorded by the NSQIP database, both reverse and anatomic total shoulder arthroplasties were included under the CPT code 2 3 4 7 2 . As such, the ability to differentiate operative times and RVU related values between shoulder arthroplasty procedures was limited. Coding limitations prevented any delineation of revision cases, specific indications for arthroplasty or complexity of cases which could influence surgical times and subsequently, opportunity cost calculations. In terms of cost, RVUs were selected as a proxy for estimating costs since the NSQIP database lacks hospital payment data and discreet operative costs. As a result, we are unable to make claims or comparisons between cost estimates and reimbursement rates. In order to perform an investigation such as this, a large nationwide database was utilized with the hopes of obtaining sufficient sample sizes to detect differences in operative time, RVUs/hour and RVUs. Despite this, post hoc analyses revealed our study was relatively underpowered to detect significance amongst the small, observed differences. However, detection of such differences would be relatively unfeasible as sample sizes > 2 0000 surgical cases would be needed. If these sample sizes are not currently contained in a large, nationwide database, it would suggest that significantly more data would need to be collected in the future would be needed to fully assess these differences. Despite this limitation, the current findings offer merit in terms of the potential cost savings and fiduciary implications of resident involvement within orthopedic surgical training programs. Lastly, PGY-6 residents were viewed as senior residents in this investigation even though in some practice settings, they may be considered as attending-level providers. It would also be worth noting that there was a skew of the cases by resident level with more senior PGY levels examined compared to younger PGY levels within the database. Despite the aforementioned limitations, this study is strengthened by the use of a matched- pair design to limit potential confounders to cost estimates between cases with and without resident involvement. Additionally, it provides an argument for the inclusion of residents in shoulder arthroplasty procedures as our findings show that their involvement can result in increased surgical efficiency, in addition to the well-known benefits of resident involvement perioperatively, and subsequently lower opportunity costs. ²¹

Conclusions

Involvement of residents was associated with shorter operative times leading to a savings of $26.67 per case. Involvement of intermediate-level (PGY-3) residents were associated with increased costs that ultimately decreased as residents became more senior.