Surgeon, Not Institution, Case Volume Is Associated with Limb Outcomes After Lower Extremity Bypass for Critical Limb Ischemia in the Vascular Quality Initiative

Abstract

Objectives

Studies from large administrative databases have demonstrated associations between institutional case volume and outcomes after lower extremity bypass (LEB). We hypothesized that increased institutional and surgeon volume would be associated with improved outcomes after LEB. Using a national, prospectively collected clinical database, the objective of this study was to determine the effects of both surgeon and institutional volume on outcomes after LEB.

Methods

The Vascular Quality Initiative (VQI) was queried to identify all LEB for critical limb ischemia or claudication between 2004–2014. Average annual case volume was calculated by dividing an institution or surgeon’s total LEB volume by the number of years they reported to VQI. Institutional and surgeon volumes were analyzed as continuous variables to determine the impact of volume on major adverse cardiac events (MACE), major adverse limb events (MALE), graft patency, and amputation free survival. Hierarchical regression models were used with cases clustered by surgeon and center. Time-dependent outcomes were evaluated with multivariable shared frailty Cox proportional hazards models.

Results

From 2004–2014, 14,678 LEB operations were performed at 114 institutions by 587 surgeons. Average annual institutional volume ranged from 1.0–137.5 LEBs per year, with a median of 26.9 [IQR 14–45.3]. Average annual surgeon volume ranged from 1–52 LEBs per year with a median of 5.7 [IQR 2.5–9.3]. Institutional LEB volume was not associated with major adverse cardiac or limb events, nor with loss of patency. However, average annual surgeon volume was independently associated with reduced MALE and improved primary patency. Institutional and surgeon volume did not predict MACE.

Conclusions

In contradistinction to previous studies, there was no relationship in this study between institutional LEB volume and outcomes after LEB. However, greater average annual surgeon volume was associated with improved primary patency and decreased risk of major adverse limb events. Open LEB remains a safe and effective procedure for limb salvage. Limb-related outcomes in CLI and claudication will be optimized if surgeons maintain adequate volume of LEB.

Introduction

Procedural volume has been positively associated with patient outcomes across a wide variety of procedures and institutional settings, including vascular surgery.^1–8 This association is studied most often related to postoperative mortality, but it has been described for more disease-specific outcomes as well, including stroke following carotid endarterectomy⁹ and amputation following lower extremity arterial bypass.^10,11 Traditionally institutional volume has been the primary independent variable but several studies examining the role of surgeon volume have been conducted as well.^12–15

The reasons for the volume-outcome association are thought to be several, and include both institutional and provider factors. It has been shown that much of the difference at the institutional level is in the capacity to rescue patients after a complication rather than in lower overall complication rates.^16,17 More technically demanding operations seem to show a stronger volume-outcome relationship than those less technically challenging.^1,17 Lower extremity bypass actually describes a collection of several different procedures of varying complexity and technical difficulty, and this heterogeneity has been problematic for studies examining the volume-outcome relationship in LEB.^10,18

Notably, the majority of studies demonstrating these relationships have largely been conducted using administrative or billing databases. This was necessary given the need to measure a large number of patients across multiple institutions, and the lack of adequate breadth in most clinical databases. However, the national Vascular Quality Initiative is especially well-suited to address these questions as it captures de-identified clinical data, including short and long-term follow-up, at hundreds of centers across the country, and both centers and surgeons are tracked using anonymized identifiers.¹⁹ The aim of this study was to investigate the impact of institutional and surgeon volume on outcomes after lower extremity bypass for critical limb ischemia (CLI) and claudication using a prospective multi-institutional clinical database with long-term follow-up on a large sample of patients.

Methods

Sample

The Society of Vascular Surgeons’ Vascular Quality Initiative (VQI) is a collection of over 10 individual procedural registries that prospectively collect patient and clinical data and capture both short and long-term outcomes from over 220 institutions across the country. De-identified data from the national VQI open LEB registry were provided by the VQI, and included operations performed between 2004 through 2014. Exclusion criteria included patients under age 18 and those undergoing LEB for acute limb ischemia. Institutional Review Board approval and informed consent was waived given the de-identified nature of the data.

Definitions

Low, medium, and high surgeon volume was based on the lowest 25^th percentile, 25^th to 75^th percentile, and highest 75^th percentile, rounded to the nearest round number of cases. Low, medium, and high center volume was similarly determined. Complex operations were defined as those that included any tibial or more distal target, those with the inflow vessel below the SFA, and a femoral to below-knee popliteal bypass with anything other than prosthetic or single-segment saphenous vein. Major adverse cardiac events (MACE) included death, myocardial infarction (MI), and stroke; MI and stroke were only captured out to 30 days. Major adverse limb events (MALE) included major amputation or graft revision, either open or endovascular, of the bypass. Primary patency is patency of the original bypass without any intervention required to maintain patency. Primary assisted patency includes primary patency as well as a bypass that was revised due to stenosis, but was not completely occluded. Secondary patency includes bypasses in the aforementioned categories as well as those bypasses that are re-canalized and patent after occlusion.

Study Design

This study design accounts for the clustered, non-independent nature of outcomes by using hierarchical multi-level mixed-effects regression models. In the case of time-to-event analyses, the standard errors are adjusted using a shared frailty methodology that also accounts for clustering.

Statistical Analysis

Preoperative demographic variables and comorbidities, as well as outcomes including mortality, amputation, patency, MACE and MALE were analyzed. Limb outcomes (MALE and amputation-free survival) were evaluated separately in patients with CLI and those with claudication. Differences in categorical variables were tested using Pearson’s chi-squared test, and in continuous variables using the Wilcoxon rank-sum test. Measures of central tendency are presented as medians with interquartile range. Statistical significance for all tests was two-tailed and set at α=0.05. For the purposes of modeling, volume was treated as a continuous function and tested using both restricted cubic spline and linear functions, and then compared. These functions were not significantly different, so volume was modeled as a linear function. Multivariable models included all patient and operative characteristics that were significant at the p<=0.2 level on univariate analysis for the outcome of interest. Analysis was performed using Stata v14.0 (Stata Corp, College Station, TX).

Results

Sample

From 2004–2014, the VQI contains 20,672 LEB operations. After excluding operations for acute limb ischemia and asymptomatic patients, as well as excluding centers with less than 50% long-term follow-up as defined by VQI, the sample for this study was 14,678 bypasses performed at 114 institutions by 587 surgeons. Overall, 67% of the patients were male with a median age of 67 [IQR 59–75] (Table I). Hypertension and smoking were the most common comorbidities at 88% and 84% of the study population, respectively. Almost 70% of bypass patients had critical limb ischemia defined as either rest pain or tissue loss (Rutherford 4–6), with the remainder being claudicants (30%). The median preoperative ABI was 0.48 [IQR 0.33–0.64]. Approximately 69% of patients had at least 9 months of follow-up recorded in VQI, with an overall median follow-up time of 314 days [IQR 28–393].

Table 1.

Demographics of study population. IQR - interquartile range, BMI – Body Mass Index, Hgb – hemoglobin

Factor	Value
N	14,678
Age, median (IQR)	67 (59, 75)
Male	9,825 (67.0%)
Caucasian	11,883 (81.0%)
Black	1835 (12.5%)
Hispanic Ethnicity	660 (4.5%)
BMI, median (IQR)	26.9 (23.4, 31.0)
Any Smoking	12,331 (84.1%)
Hypertension	12,913 (88.0%)
Any Diabetes	7,377 (50.3%)
Hgb A1C, median (IQR)	7.2 (6.3, 8.6)
Coronary Disease	4,650 (31.7%)

Volumes

Average annual institutional volume ranged from 1.0–137.5 LEBs per year, with a median of 26.9 [IQR 14–45.3]. Average annual surgeon volume ranged from 1–52 LEBs per year with a median of 5.7 [IQR 2.5–9.3]. Center volume cut-offs for the low, medium, and high volume groups were approximately <14, 14–45, and >450 cases per year. For surgeons, the volume cut-offs were <2, 2–10, and >10 bypass operations per year. Overall, 695 operations were performed at low-volume centers, 5,298 at moderate volume centers, and 8,149 at centers in the highest volume quartile. By surgeon volume, 355 operations were performed by surgeons in the lowest volume quartile, 5,167 in the moderate volume group, and 8,620 in the high-volume group. All strata of surgeon volume were represented in each category of institutional volume. Overall, 6,339 operations (43%) were classified as complex (Table II).

Table 2.

Distribution of simple vs complex operations by center and surgeon volume.

	Low Vol Centers			Med Vol Centers			High Vol Centers
Surg Vol	Low	Mod	High	Low	Mod	High	Low	Mod	High
Simple	60 (60.0%)	285 (62.9%)	115 (70.6%)	92 (59.7%)	1546 (53.8%)	1463 (59.7%)	82 (67.8%)	1115 (55.1%)	3581 (56.4%)
Complex	40 (40.0%)	168 (37.1%)	48 (29.4%)	62 (40.3%)	1325 (46.2%)	986 (40.3%)	39 (32.2%)	907 (44.9%)	2764 (43.6%)
Total	100	453	163	154	2871	2449	121	2022	6345

Outcomes

The 30-day mortality for all patients was 2.2%, with an increase to 19.% at one year of those patients with confirmed one-year follow-up (. Within the first year after bypass, 21.9% of patients underwent any amputation; this increases to 23.2% in patients with CLI. Primary patency at one year was 75%, and secondary patency was 84.4% at one year. Overall, over a third of patients experience a postoperative complication (Table III). In unadjusted analyses, there are no clear trends that apply to all strata of surgeon and center volumes. By strata, MACE appears to vary as a function of surgeon volume in medium- and high-volume centers (Table IV). There was also a significant unadjusted relationship between MACE and surgeon volume (OR 0.94 per 5 case increase, 95% CI 0.89–0.99) but this association was not substantiated in the mixed-effects multivariable model (Table V) with volume as a continuous variable (OR 1.00 per 5 case increase, 95% CI 0.94–1.06), and there was no effect of center volume (OR 0.93 per 20 case increase, 95% CI 0.86–1.02). Primary, but not secondary, patency was also improved with increasing surgeon volume without any effect by center volume in both univariate and multivariable analyses (Table V). In patients with CLI, there was an increased hazard of amputation or death as a function of increasing surgeon volume in both univariate and multivariable models (Adjusted OR, AOR 1.03 per 5 case increase, 95% CI 1.00–1.06, p=0.04). However, there was no change in amputation-free survival in claudicants as a function of either surgeon or center volume. Freedom from MALE in patients with CLI was improved as a function of surgeon (AOR 0.95 per 5 case increase, 95% CI 0.92–0.99) but not center (AOR 1.01 per 20 case increase, 95% CI 0.96–1.07). Patients with claudication also saw improvement in freedom from MALE with increasing surgeon (AOR 0.93, 95% CI 0.88–0.98) but not center (AOR 0.99, 95% CI 0.92–1.06) case volume.

Table 3.

Indications for and complications of lower extremity bypass, entire study population. LOS – length of stay, IQR – Interquartile range, ABI – Ankle-brachial index, CHF – congestive heart failure, AKI – Acute kidney injury, MI – myocardial infarction.

Factor	Value
N	14,678
Indication
Claudication	4,454 (30.3%)
Rest Pain	3,879 (26.4%)
Tissue Loss	6,345 (43.2%)
Post-op LOS, median (IQR)	4 (3, 7)
Pre-op ABI, median (IQR)	.48 (.33,.64)
Complex	6,339 (43.2%)
30 day Mortality	265 (2.2%)
1 year Mortality	1,440 (19.2%)
Any amputation at 1 year	1,285 (21.9%)
Post-op MI	455 (3.1%)
Wound Infection	530 (3.6%)
Graft Infection	49 (0.3%)
Transfusion units, median (IQR)	0 (0, 1)
Dysrhythmia	625 (4.3%)
CHF Exacerbation	359 (2.5%)
Respiratory (Pneumonia or ventilator requirement)	327 (2.3%)
AKI	675 (4.7%)
Stroke	78 (0.7%)
Patency at Discharge
Primary	13,763 (94.5%)
Primary Assisted	14,148 (96.4%)
Secondary	14,429 (98.3%)
Occluded	182 (1.3%)
Return to Operating Room	1,574 (10.8%)
Any Postop Complication	4,537 (35.8%)

Table 4.

Major outcomes by surgeon and center volume strata. MACE – Major adverse cardiac events, MALE – Major adverse limb events.

	Low Vol Centers			Med Vol Centers			High Vol Centers
	Low Vol Surgeons	Mod Vol Surgeons	High Vol Surgeons	Low Vol Surgeons	Mod Vol Surgeons	High Vol Surgeons	Low Vol Surgeons	Mod Vol Surgeons	High Vol Surgeons
N	55	440	74	100	3484	2865	68	2244	7576
MACE	3.6%	3.4%	5.4%	8.0%	4.7%	3.4%	7.4%	4.7%	3.7%
Primary Patency (1 yr)	71.0%	65.9%	74.2%	81.5%	70.9%	77.5%	84.7%	74.1%	76.4%
Secondary Patency (1 yr)	87.5%	81.9%	82.4%	88.1%	84.6%	85.5%	90.2%	84.0%	84.0%
Limb outcomes in critical limb ischemia
Freedom from MALE	61.8%	63.4%	63.3%	71.6%	63.0%	65.3%	66.9%	65.1%	67.4%
Amputation-Free Survival	73.4%	75.6%	85.3%	76.5%	77.1%	77.5%	75.4%	78.2%	76.4%
Limb outcomes in claudication
Freedom from MALE	77.3%	76.9%	84.2%	71.6%	73.5%	80.6%	91.4%	77.2%	81.1%
Amputation-Free Survival	97.2%	97.3%	94.3%	92.9%	93.7%	94.5%	94.6%	94.6%	94.0%

Table 5.

Results of univariate and multivariable models using volume as a linear function.

Outcome	Events at 1 Year†	Univariate Odds & Hazard Ratios (95% Confidence Intervals)		Multivariable Odds & Hazard Ratios (95% Confidence Intervals)
		Surgeon volume (per 5 cases)	Center volume (per 20 cases)	Surgeon volume (per 5 cases)	Center volume (per 20 cases)
MACE	4.17%	0.94 (0.89–0.99)* p=0.03	0.95 (0.86–1.05)	1.00 (0.94–1.06)	0.93 (0.86–1.02)
Primary patency	75.0%	0.96 (0.94–0.98)* p=0.001	0.96 (0.90–1.03)	0.96 (0.93–0.99) *p=0.008	0.97 (0.90–1.04)
Secondary patency	84.40%	1.00 (0.97–1.03)	1.01 (0.93–1.11)	1.00 (0.96–1.04)	1.00 (0.91–1.10)
Limb outcomes in critical limb ischemia
Amputation-free survival	77.0%	1.02 (1.00–1.04)* p=0.043	1.04 (0.98–1.09)	1.03 (1.00–1.06)* p=0.04	0.97 (0.92–1.03)
Freedom from MALE	65.8%	0.96 (0.94–0.99)* p=0.003	0.98 (0.93–1.03)	0.95 (0.92–0.99) *p=0.007	1.01 (0.96–1.07)
Limb outcomes in claudication
Amputation-free survival	94.2%	1.03 (0.98–1.07)	1.02 (0.93–1.13)	1.06 (0.99–1.12)	0.98 (0.87–1.10)
Freedom from MALE	78.8%	0.93 (0.89–0.97)* p=0.001	0.94 (0.88–1.01)	0.93 (0.88–0.98)* p=0.01	0.99 (0.92–1.06)

^†Events for MACE are at 30 days or inhospital. MACE – Major adverse cardiac events, MALE – Major adverse limb events. Variables included in the multivariable model were sex, ethnicity, body mass index, age, transfer status, smoking history, hypertension, diabetes mellitus, congestive heart failure, chronic obstructive pulmonary disease, end-stage renal disease, creatinine, coronary artery disease, ASA class, hemoglobin, history of major amputation, preoperative statin, preoperative beta-blocker, indication for bypass, urgency, estimated blood loss, and complex vs non-complex case.

Discussion

This study demonstrates that surgeon, but not institutional, volume is inversely related to adverse limb events after lower extremity bypass. Moreover, we find no effect of volume on major adverse cardiac events within 30 days of surgery in adjusted analyses. There is a small but significant reduction in amputation-free survival with increasing surgeon volume in patients with CLI. Compared to the PREVENT III trial data, VQI demonstrates a lower rate of MACE (8.0 vs 4.0%), but this may be for several reasons, including the fact that MI and stroke outcomes are limited to in-hospital events in VQI and do not capture out to 30 days.^20,21 Additionally, the PREVENT III population was restricted to patients with CLI, who may represent a higher-risk cohort than the claudicants.

According to a systematic review by Awopetu et al., nine studies have examined the effect of hospital volume on mortality after LEB. They report that five of those studies demonstrated a positive volume-outcome relationship, whereas the other four did not find any evidence of a volume effect on postoperative mortality yielding a meta-analysis that supports the effect of increased volume leading to decreased mortality, but this finding is also associated with significant heterogeneity and thus must be interpreted with caution.²² In the same review, seven studies examined the rates of amputation at low vs. high volume hospitals, three of which were suitable for inclusion in the meta-analysis. The pooled effect estimate was an odds ratio of 0.88 with low heterogeneity favoring high volume hospitals. However, of the seven total studies only three showed a volume-outcome relationship.²² Our study does not demonstrate any effect of either institutional or surgeon volume on amputation-free survival, but does show an effect of surgeon volume on major adverse limb events, which includes major amputation.

The median annual surgeon volume for LEB in the VQI is surprisingly low at less than one bypass every two months, which could be explained several ways. First, it is possible that the VQI is capturing a number of general and/or cardiac surgeons who rarely perform LEB as part of their daily practice. Second, there may be clerical issues with the tracking of individual surgeons in the VQI resulting in erroneous additional records. Third, the overall rate of open operations has decreased markedly over time as utilization of endovascular therapies has risen. Finally, the VQI is designed to capture 100% of eligible cases at an institution, and billing records are audited on occasion to ensure compliance; nevertheless, it may be that not all LEBs are not being captured, leading to falsely reduced volumes. In a study that examined hospital discharge records in Florida in the early 1990s, the median surgeon volume was considerably higher at 21 LEBs annually, but this challenging to interpret given the relative paucity of endovascular intervention in that era.²³ Kantonen and colleagues used a threshold of 10 LEBs per year to distinguish between high and low volume surgeons based on a national study conducted in the United Kingdom, also in the early 1990s, that demonstrated higher primary amputation rates in surgeons who performed fewer than 10 LEBs annually, though interestingly the rates of limb salvage, secondary amputation, length of stay, and mortality were not different between strata of surgical volumes in that study.^24,25 Benchmarking our surgeon volume data against more current databases is an important next step in terms of validating our findings. It would also potentially highlight an opportunity to improve data fidelity in VQI if need be, though as a self-selecting sample it may be difficult to directly compare VQI with HCUP/NIS or CMS data.

The decrease in amputation-free survival with increasing surgeon volume is, at face value, counter-intuitive. However, based on the MACE data, this finding is driven by amputations rather than overall survival, and VQI does not distinguish planned versus unplanned amputations. In other words, if revascularization is the precursor operation to permit healing of an amputation, there is no mechanism in VQI to record this as a planned procedure. This results in a database that may unfairly penalize surgeons for amputations that do not in fact represent a treatment failure. Additionally based on the above AFS data, it becomes clear that the reduction in MALE is driven predominately by primary patency and the reduced need for interventions to achieve secondary patency. This reduction likely results in cost savings to both patients and the healthcare system. It also is consistent with the hypothesis that the technical skill of the surgeon, which is associated with operative volume in other studies, may have a significant impact on limb outcomes after bypass. This is likely not reflected in amputation rates due to the equivalency of secondary patency across surgeon and center volumes.

The implications of these results, if validated, are broad. Due to the linear nature of the relationship we describe, it is difficult to meet the demand for specific threshold numbers: how many is enough? This applies to training and required case volume, to the ongoing national debate regarding centralization of specialized surgical care, and to individual surgeons who are now being asked about case volumes to determine hospital privileges. Moreover, while the rates of secondary patency are equivalent across center and surgeon volumes, the rates of primary patency are not, and this is associated with real costs both to the healthcare system and to patients in terms of additional procedures required to maintain graft patency over time.

This study is limited in several ways. First, it suffers the same risk of bias as all retrospective non-randomized observational studies. The risk of bias is compounded by the selection bias inherent in the VQI: centers voluntarily participate in the VQI generally, and in individual surgical modules specifically, and the data are self-reported by surgeons and centers, thus introducing another element of potential bias. It is certainly possible that these decisions may be influenced by the cost of participation as well as an institution’s perception of its performance. This selection bias also makes VQI unsuitable for following trends in providers or institutions over time, as the early adopters were higher volume tertiary centers but the diversity of participating institutions has broadened considerably over the last decade.

Conclusion

In conclusion, this study shows that increasing surgeon volume is independently associated in a linear fashion with lower rates of adverse limb events and higher rates of primary patency. Our analysis does not substantiate previous studies that find a relationship between institutional volume and outcomes; center volume is not related to short-term adverse cardiac events, limb outcomes, or overall survival in the VQI.