Variation in the Use of Lower Extremity Vascular Procedures for Critical Limb Ischemia
Background—Many believe that variation in vascular practice may affect limb salvage rates in patients with severe peripheral arterial disease. However, the extent of variation in procedural vascular care obtained by patients with critical limb ischemia (CLI) remains unknown.
Methods and Results—By using Medicare 2003 to 2006 data, we identified all patients with CLI who underwent major lower extremity amputation in the 306 hospital referral regions described in the Dartmouth Atlas of Healthcare. For each patient, we studied the use of lower extremity vascular procedures (open surgery or endovascular intervention) in the year before amputation. Our main outcome measure was the intensity ofvascular care, defined as the proportion of patients in the hospital referral region undergoing a vascular procedure in the year before amputation. Overall, 20 464 patients with CLI underwent major lower extremity amputations during the study period, and collectively underwent 25 800 vascular procedures in the year before undergoing amputation. However, these procedures were not distributed evenly: 54% of patients had no vascular procedures performed in the year before amputation, 14% underwent 1 vascular procedure, and 32% underwent >1 vascular procedure. In the regions in the lowest quintile of vascular intensity, vascular procedures were performed in 32% of patients. Conversely, in the regions in the highest quintile of vascular intensity, revascularization was performed in 58% of patients in the year before amputation (P<0.0001). In analyses accounting for differences in age, sex, race, and comorbidities, patients in high-intensity regions were 2.4 times as likely to undergo revascularization in the year before amputation than patients in low-intensity regions (adjusted odds ratio, 2.4; 95% CI, 2.1–2.6; P<0.001).
Conclusions—Significant variation exists in the intensity of vascular care provided to patients in the year before major amputation. In some regions, patients receive intensive care, whereas in other regions, far less vascular care is provided. Future work is needed to determine the association between intensity of vascular care and limb salvage.
Lower extremity peripheral arterial disease (PAD) manifests in its most severe form, with limb-threatening rest pain and ulceration, as “critical limb ischemia” (CLI) in nearly 1 million patients.1–7 Estimates of the economic burden of CLI patients alone exceed $3.1 billion annually in Medicare, because of the high incidence of limb loss and the need for major amputation.8 These costs are driven, at least in part, by the increase in the use of catheter-based endovascular interventions as a major component of vascular care in recent years.9–11
Despite the increase in health care resources dedicated to PAD, major lower extremity amputation remains common, and the incidence of major amputation varies according to several factors, such as patient age, race, and socioeconomic status, and local health care environment.12–16 For example, the major amputation rate in Corpus Christi, TX (4.4 amputations per 1000), is 10 times higher than the rate of amputation in Grand Junction, CO (0.4 amputations per 1000). Research in the Dartmouth Atlas of Health Care has attributed some of these regional differences in amputation to patient-level factors, such as 5-fold higher amputation rates among blacks and those of low socioeconomic status.12 However, because patient-level differences do not fully explain the variation in amputation rates across regions, we hypothesized that the intensity of vascular care may vary across regions as well.
For this reason, we sought to characterize the variation in invasive treatments provided to patients with CLI across the United States. Because the extent of PAD affects the threshold for use of invasive vascular procedures, we chose to specifically focus on patients with PAD who had undergone major limb amputation. We chose this group because, by definition, they had PAD, which can definitively be classified as CLI in the year before their amputation. Therefore, in this group of patients with a similar extent of PAD, we sought to examine the presence and extent of variation in procedural vascular care.
Major lower extremity amputation occurs commonly, especially among black patients, those of low socioeconomic status, and those residing in certain regions of the United States.
There is broad variation in the “intensity” of vascular care provided to Medicare patients in the year before amputation; in some regions, patients receive extensive invasive care, whereas in other regions, little vascular care is provided.
This variation is evident in traditional leg bypass surgery and in newer diagnostic and therapeutic endovascular procedures.
Patients who are treated only with newer, less-invasive therapeutic endovascular procedures are likely to undergo several procedures in the year before amputation.
We used the Medicare Physician/Supplier file and the Medicare Denominator file for these analyses, for the years between 2003 and 2006. The Physician/Supplier file contains all claims submitted by physicians for performance of procedures under the Medicare part B program, including Current Procedural Terminology codes,17 International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes, date of procedure, and the age, sex, and race of the beneficiary receiving the procedure. The denominator file contains information about eligibility by year for part B and information about age, sex, and race of eligible beneficiaries. We excluded patients <65 years. Records with missing values for hospital referral region (HRR), sex, age, and race strata were also removed from the analysis.
Establishing a Cohort of Patients With Severe PAD
In this project, we studied the use of lower extremity revascularization for CLI. Although distinct ICD-9 codes exist for claudication and CLI, many patients are assigned other nonspecific PAD codes. Preliminary work using a linked clinical-claims data set suggested that up to 40% of patients with PAD receive nonspecific PAD codes, making delineation of the extent of PAD difficult using administrative claims alone. Therefore, to establish a cohort of patients with severe CLI, wherein all patients have a similar extent of disease, we studied the vascular care provided in patients in the 12 months before major lower extremity amputation (below or above the knee). To create this cohort, we first identified all patients with PAD who underwent major lower extremity amputation (below or above the knee) using Medicare part B claims between 2003 and 2006. To ensure that these amputations were not the result of causes other than PAD, such as malignancy or trauma, we omitted any patients without ICD-9 diagnosis codes for PAD. After we identified a cohort of patients with diagnosis codes for lower extremity PAD (as shown in online-only Data Supplement Appendix I) who had undergone major amputation, we “looked back” 12 months before the date of amputation to determine if these patients had undergone lower extremity vascular procedures during that time period.
Unit of Analysis
Our unit of analysis was the patient, and we used indicator variables to study the first and any subsequent procedures each patient underwent in the year before amputation. Because patients may have been treated with >1 type (open or endovascular) of vascular procedure during the time interval, we categorized vascular procedures in 4 ways: (1) diagnostic endovascular intervention only, (2) therapeutic endovascular intervention, (3) open surgical treatment, and (4) both open and endovascular revascularization (Figure 1). We recorded patient characteristics, such as age, sex, race, and comorbidities (both individually and using the Charlson index18). Furthermore, while we collected the details of each individual Current Procedural Terminology code performed, we collapsed the open bypass surgery and endovascular procedures into inflow (above the inguinal ligament) and outflow (distal to the inguinal ligament) procedures for presentation.
Comparison of Vascular Procedure Rates Across Regions
Vascular procedure rates were defined as the number of patients undergoing vascular procedures(either bypass surgery or endovascular interventions), divided by the number of patients undergoing amputation secondary to PAD. To examine geographic variation in vascular procedure rates, we examined the rates of bypass surgery and endovascular interventions during this time period within each of the 306 HRRs in the United States.
HRRs, as described by the Dartmouth Atlas of Healthcare,19 represent distinct tertiary medical care markets (as defined by where cardiovascular and neurosurgical care is provided). Each HRR has at least 1 tertiary care center and several smaller centers. After defining crude rates of bypass surgery and endovascular intervention within each HRR, regions were aggregated into evenly sized quintiles, according to the use of vascular procedures, ranging from highest to lowest. HRR level rates of use for vascular procedures were calculated between 2003 and 2006.
We used t tests to compare rates between regions, and P<0.05 was considered significant. Univariate associations with P<0.20 were entered into a multivariable logistic regression model used to predict the likelihood of undergoing a vascular procedure in the year before amputation. Model performance was assessed using receiver operating characteristic curves. All analyses were performed using SAS (SAS Institute; Cary, NC), and STATA 10 (Stata Corporation, College Station, TX).
Characteristics of the Cohort of Patients Undergoing Amputation
Overall, we identified 20 464 Medicare patients who underwent major lower extremity amputation between 2003 and 2006. These amputations consisted of 42% below-knee amputations and 58% above-knee amputations. The 10 most common ICD-9 diagnostic codes for PAD in these patients are shown in Table 1. Gangrene was seen most commonly, followed by nonspecific PAD and ulcerations. No patients were included in the cohort based on diagnosis codes for non–limb-threatening diagnoses, such as claudication.
Patient characteristics of those in the cohort are shown in Table 2. Overall, patients were elderly (mean age, 78 years) and had a history of diabetes (49%), coronary artery disease (14%), and congestive heart failure (35%). Although patients undergoing revascularization were younger than those not undergoing revascularization (77 versus 79 years; P<0.001), patients undergoing revascularization had a slightly higher incidence of comorbidities, such as myocardial infarction, congestive heart failure, and renal failure.
Use of Vascular Procedures in the Year Before Amputation
Overall, of the 20 464 patients, we found that 54% (n=11 115) did not undergo any vascular procedure in the year before amputation, whereas 46% (n=9349) underwent a vascular procedure in the year before amputation (Figure 1). For example, a diagnostic endovascular intervention alone, such as “catheter placement into the aorta from the groin or arm,” was performed in 3791 patients (14.7% of the entire cohort). Similarly, therapeutic endovascular interventions, such as percutaneous femoral-popliteal angioplasty (n=2015 [7.8% of patients]) or open surgical procedures (such femoral-popliteal bypass; n=854 [3.3% of patients]), were performed on patients during the year before amputation.
Endovascular diagnostic procedures were used most commonly, followed by endovascular therapeutic and open surgical procedures. The 10 most commonly used endovascular and open lower extremity vascular procedures are listed in Table 3. For anatomic location in which revascularization procedures were performed, outflow arteries (such as the femoral-popliteal segment or the tibial arteries) were more commonly treated than inflow arteries (such as the iliac arteries). For example, endovascular interventions were more commonly used in outflow arteries than in inflow arteries (83% outflow arteries and 17% inflow arteries). This pattern was similar in open revascularization (93% outflow arteries and 7% inflow arteries).
Use of Multiple Revascularization Procedures in the Year Before Amputation
Many patients underwent >1 vascular procedure in the year before amputation. Across 9349 patients, 25 800 vascular procedures were performed. Although 14.3% of patients underwent 1 vascular procedure, 23.3% underwent 2 to 3 vascular procedures, and 8.4% underwent 3 vascular procedures. These findings remained similar, even when we analyzed our results exclusive of diagnostic angiography.
In Figure 2, we compared, by procedure type, the proportion of patients who received 1, 2, and ≥3 vascular procedures during the year before amputation. We found that patients undergoing therapeutic endovascular interventions were likely to undergo ≥3 procedures more commonly than patients treated with open surgery (68% versus 39%; P<0.001). Therefore, treatment with endovascular interventions made it more likely that a patient would experience a subsequent revascularization.
Overall, 653 patients (3.2%) underwent both types (therapeutic endovascular intervention and open surgical revascularization) in the year before amputation. Patients who underwent open surgery had a slightly higher risk of undergoing additional endovascular intervention (31%) compared with the risk of endovascular intervention patients undergoing additional open surgery (27%) (risk ratio, 1.15; 95% CI, 1.09–1.19; P=0.04).
Variation in the Use of Vascular Procedures by Patient Characteristics
We examined differences in age and race between patients selected to undergo vascular procedures and those who did not undergo a vascular procedure in the year before amputation. We found that older patients and black patients were less likely to undergo revascularization procedures in the year before amputation. For example, patients >90 years were less likely to undergo a vascular procedure than patients <70 years (21% versus 41%; P<0.001). Black patients were less likely to undergo any revascularization procedure in the year before amputation than white patients (32% versus 37%; P=0.001). These findings were similar in both open surgical and therapeutic endovascular interventions.
Variation in the Use of Vascular Procedures by HRR
To examine regional differences in practice patterns, we studied the use of invasive vascular procedures across the 306 HRRs in the United States. In the most “intensive” HRRs (eg, Elyria, OH; Munster, IN; and Santa Cruz, CA), between 71% and 80% of patients underwent a vascular procedure in the year before amputation. However, in the least aggressive regions (eg, Sayre, PA; Billings, MT; and Bryan, TX), <12% of patients underwent a vascular procedure in the year before amputation.
A national map of regional use rates (Figure 3) demonstrates that regions of high intensity of care (shown in dark red) are not concentrated in any 1 area of the United States. Rather, these regions are widely distributed across the United States. Furthermore, when we examined the specific use of open surgery or therapeutic endovascular interventions (Figure 4), we found broad variation in the use of all types of revascularization procedures, including diagnostic and therapeutic endovascular interventions and open surgical revascularization.
To allow for comparison across regions, we categorized all HRRs into 5 evenly sized groups, ranging from very low intensity (quintile 1) to very high intensity (quintile 5). Although overall rates of revascularization were low in the very-low-intensity group (32.6%), the rate of revascularization was significantly higher in the very-high-intensity group (58.4%) (P<0.001) (Figure 5A). Variation in procedural vascular care by intensity was again evident in diagnostic endovascular, therapeutic endovascular, and open surgical revascularization (Figure 5B).
Variation in intensity was not directly explained by differences in patient age or race. For example, the differences in revascularization across very-high- and very-low-intensity regions existed in elderly patients (32.3% in very-low-intensity regions and 57.9% in very-high-intensity regions) and in black patients (33.1% in very-low-intensity regions and 58.1% in very-high-intensity regions).
Multivariate Analysis to Predict the Likelihood of Undergoing Revascularization
To examine interactions between patient characteristics and regional patterns in use, we developed a multivariable logistic model to identify variables associated with the use of vascular procedures. Even when adjusting for the effect of age, sex, race, and comorbidities, patients in very-high-intensity regions were more than twice as likely to undergo revascularization in the year before amputation (odds ratio for very high intensity versus very low intensity, 2.4; 95% CI, 2.1–2.6; P<0.0001) (Table 4). Patient age and race also were independently associated with differences in the use of vascular procedures in the year before amputation. These differences were consistent across different definitions of procedural vascular care (endovascular diagnostic, endovascular therapeutic, or open procedure) or as any of its individual components, as shown in online-only Data Supplement Appendix II. Our model had moderate predictive ability, with an area under the curve of 0.64.
The treatment of lower extremity PAD and its consequences are among the most costly and morbid challenges faced by elderly Medicare patients. Therefore, efforts to limit major amputation secondary to PAD are a priority recognized by several societies and leaders in vascular care.7,20 In addition, although prior research suggests that broad expansion in the use of vascular care has occurred in recent years,11 our analyses demonstrate that aggressive vascular care for patients at risk for amputation has been unevenly applied across the United States. In many regions of the United States, most patients with severe PAD undergo amputation without even a diagnostic arteriogram performed in the year before amputation. However, in other regions, patients with a similar extent of PAD undergo a multitude of vascular procedures, especially therapeutic endovascular interventions. These decisions are unlikely to be driven solely by insurance access, because all patients in our analysis are >65 years and, thus, insured by Medicare. Therefore, if the variation in vascular care is not entirely explained by patient-level factors, this variation must be due to differences in regional practice patterns.
Physicians who care for patients with vascular disease will test that treatment decisions in patients with CLI can, at times, be straightforward. Patients with poor functional status, such as those living in a nursing home or those unable to ambulate, have had poor results in limb salvage attempts,21,22 and most agree that these patients should undergo primary amputation without attempts at limb salvage.23 Conversely, patients with good functional status, favorable anatomic characteristics, and few comorbidities have consistently good outcomes after lower extremity revascularization, in either an open or an endovascular fashion.24,25 However, decision making in lower extremity revascularization is often not so clear-cut, and physicians who care for patients with CLI have many treatment options. In settings in which clinical equipoise meets with multiple treatment options, the occurrence of health care variation has been well documented,26 and prior work has shown that the treatment of vascular disease is no exception.12,27 Accordingly, our present study indicates that, in some regions, patients are treated with intensive revascularization strategies, whereas in other regions, patients with a similar extent of CLI commonly undergo primary amputation.
Why does regional vascular practice vary, and what is the impact of this variation? Our future work aims to address these questions. First, in terms of variation, prior research in cardiovascular disease has demonstrated that the structural characteristics of hospitals (size, teaching status, and financial status) and surgeons (volume, specialty, and use of endovascular procedures) may explain this variation.28 In addition, although our current data set is limited in its ability to describe these variables, our future efforts will study the effect of these covariates on variation in the intensity of vascular care. Second, it remains uncertain if treatment intensity is related to outcome in vascular care (ie, “is more better”). Too little treatment intensity, we hypothesize, may be associated with inappropriately low rates of limb salvage. However, we also suspect that thresholds will exist at which treatment intensity reaches the “flat of the curve,” and more vascular procedures may not provide added value. Accordingly, our future work will study the relationships between intensity of vascular care and limb salvage.
Our study has limitations. First, many will argue that administrative data lack adequate anatomic, hemodynamic, and physiological information to study patients with CLI.29 However, our combining a known event (major amputation) and established diagnosis codes for PAD makes it unlikely that these amputations occurred in the absence of significant vascular disease, and our prior work has validated this approach in similar clinical-claims data sets.27,30 Second, the sidedness of vascular procedures is not recorded in claims data, and we cannot be sure that vascular interventions described in our Current Procedural Terminology codes were not contralateral to the limb that was eventually amputated. However, in our regional data set in the Vascular Study Group of New England (VSGNE), we found that among >4000 patients with CLI, this event occurred in <7% of all amputations and was usually bilateral (not contralateral). Third, although we studied vascular care in the 12 months before amputation, other studies have used longer intervals, without significant differences in the direction or size of their effect.16 Finally, our model had an area under the curve of 0.64, which indicates that our model has a “moderate” ability to discriminate between those likely to undergo revascularization and those who are unlikely to undergo revascularization. Achieving better model discrimination will likely require more granular detail, because of provider- and hospital-level differences in care. Our future work will aim to use clinical and claims data sets to explore these determinants of treatment intensity for procedural vascular care and analytic strategies, such as instrumental variables, to deal with unmeasured confounding. Especially in observational analyses, such as the work described herein, the role of unmeasured confounding must be acknowledged and considered, especially when determining the impact of our findings on future health policy decisions.
In conclusion, in many regions of the United States, most Medicare patients with CLI undergo amputation, with little procedural vascular care in the year before their amputation, whereas in other regions, patients with a similar extent of PAD undergo a variety of diagnostic, endovascular, and open interventions. Furthermore, these treatment decisions are not driven solely by patient characteristics. Rather, these differences in treatment appear to be explained, at least in part, by regional differences in the intensity of vascular care. Our future work aims to explore the determinants of intensity of vascular care and to characterize the effectiveness of intensive and nonintensive treatment strategies. It is our ultimate goal to use this information to design algorithms that allow delivery of the most effective vascular care at the lowest treatment intensity for patients with CLI.
Sources of Funding
Dr Goodney was supported by a Career Development Award from the National Heart, Lung, and Blood Institute (1K08HL05676–01) and a Society for Vascular Surgery Foundation/American College of Surgeons Supplemental Funding Award.
Presented at the Society for Vascular Surgery's Annual Meeting, Chicago, IL, June 15, 2011.
The online-only Data Supplement is available with this article at http://circoutcomes.ahajournals.org/lookup/suppl/doi:10.1161/CIRCOUTCOMES.111.962233/-/DC1.
- Received June 10, 2011.
- Accepted October 13, 2011.
- © 2012 American Heart Association, Inc.
- Selvin E,
- Erlinger TP
- McDermott MM
- Leng GC,
- Lee AJ,
- Fowkes FG,
- Whiteman M,
- Dunbar J,
- Housley E,
- Ruckley CV
- Peacock JM,
- Keo HH,
- Yu X,
- Oldeberg N,
- Duval S,
- Henry TD,
- Jaff MR,
- Baumgartner I,
- Hirsch AT
- White CJ,
- Gray WA
Dartmouth Atlas Cardiovascular and Thoracic Health Care. Boston, MA: Manning Selvage & Lee; 1998.
CPT schedule. Available at: https://catalog.ama-assn.org/Catalog/cpt/cpt_search.jsp. Accessed September 9, 2011.
Dartmouth Atlas of Healthcare. Available at: http://www.dartmouthatlas.org. Accessed October 1, 2007.
- Goodney PP,
- Nolan BW,
- Schanzer A,
- Eldrup-Jorgensen J,
- Bertges DJ,
- Stanley AC,
- Stone DH,
- Walsh DB,
- Powell RJ,
- Likosky DS,
- Cronenwett JL
- Taylor SM,
- Cull DL,
- Kalbaugh CA,
- Cass AL,
- Harmon SA,
- Langan EM III.,
- Youkey JR
- Goodney PP,
- Schanzer A,
- Demartino RR,
- Nolan BW,
- Hevelone ND,
- Conte MS,
- Powell RJ,
- Cronenwett JL
- Schanzer A,
- Goodney PP,
- Li Y,
- Eslami M,
- Cronenwett J,
- Messina L,
- Conte MS
- Goodney PP,
- Travis LL,
- Malenka D,
- Bronner KK,
- Lucas FL,
- Cronenwett JL,
- Goodman DC,
- Fisher ES