Impact of Renal Insufficiency on Short-Term Morbidity and Mortality after Lower Extremity Revascularization: Data from the Department of Veterans Affairs’ National Surgical Quality Improvement Program

Database

The VA NSQIP is an ongoing quality management initiative for surgical care. Since its inception on January 1, 1994, NSQIP has prospectively collected data on most major surgeries occurring at VA medical centers (VAMCs) across the country. A detailed account of NSQIP study design is provided elsewhere (25). At present, a total of 123 VAMCs participate in NSQIP. Patients are enrolled in NSQIP at the time of surgery. At this time, data on baseline clinical and demographic characteristics are obtained from the medical record, patient interview or from the surgeon caring for the patient. Pre-operative laboratory values are transmitted electronically from the VA’s decentralized hospital computer system (VistA) to the coordinating center at the Denver VA Medical Center and the University of Colorado Health Outcomes Program. Patients are then followed prospectively into surgery and for 30 d after surgery by surgical clinical nurse reviewers. Data on the surgery itself and on post-surgical complications occurring within 30 d are ascertained by these nurse reviewers using the medical record, surgeon interview and by patient follow-up (by letter or telephone call requesting follow-up data). All deaths occurring within 30 d post-operatively are verified against the VA Beneficiary Identification and Records Locator System (BIRLS) death records (26,27⇓). Patients are eligible for inclusion in NSQIP if their procedure is performed under general, spinal or epidural anesthesia.

Sample Selection

We searched the NSQIP database for the first occurrence of a lower-extremity surgical revascularization procedure for individual patients–as coded by Current Procedural Terminology (CPT) – from January 1, 1994 to September 30, 2001 (CPT lower extremity revascularization codes used in procedure code search: 35286, 35372, 35521, 35533, 35541, 35546, 35548, 35549, 35551, 35556, 35558, 35563, 35565, 35566, 35571, 35582, 35583, 35585, 35587, 35621, 35623, 35641, 35646, 35651, 35654, 35656, 35661, 35663, 35665, 35666, 35671). We included only the chronologically first procedure for each patient occurring during the study period to analyze all data at the individual patient level and to exclude as many secondary procedures as possible. Patients were included in the study if a CPT code for lower extremity revascularization appeared in the principal procedure code field listed in NSQIP. In this manner, we identified all patients who underwent at least one lower extremity revascularization between January 1, 1994 and September 30, 2001 recorded as a principal procedure in NSQIP.

Ascertainment of Postoperative Outcomes

Operative outcomes examined in this analysis were death and several other adverse clinical outcomes occurring within 30 d of the procedure. These included operative site complications (wound infection [superficial and deep], wound dehiscence, and graft failure requiring return to the operating room), cardiac complications (cardiac arrest requiring CPR, myocardial infarction), pulmonary complications (unplanned post-operative intubation for respiratory or cardiac failure and failure to wean from the ventilator after 48 h), infectious complications (sepsis, pneumonia) and other complications (stroke, coma >24 h, post-operative bleeding requiring >4 units of packed red blood cells or whole blood within the first 72 h after surgery, and acute renal failure requiring dialysis in a patient not previously on dialysis).

Determination of Renal Function

A single pre-operative serum creatinine measurement was available for most patients in the NSQIP database. In addition, nurse reviewers (who have access to the medical record) also recorded whether patients experienced pre-operative acute renal failure (defined as rapidly rising creatinine above 3mg/dl and oliguria). For the primary analysis, GFR was calculated for each patient using the abbreviated Modification of Diet in Renal Disease (MDRD) formula (28) which predicts GFR based on serum creatinine, albumin, age and race. This equation probably estimates GFR more accurately than serum creatinine alone because it takes into account racial and gender differences in creatinine production (28).

Patients were grouped according to the level of renal function as follows: normal or mildly reduced renal function (estimated GFR≥60 cc/min/1.73m²), moderate renal insufficiency (estimated GFR 30 to 59 cc/min/1.73m²), severe renal insufficiency (estimated GFR <30cc/min/1.73m²) and dialysis-dependent renal failure. This classification is similar to the definition recommended by the National Kidney Foundation’s Dialysis Outcome Quality Initiative (NKF-DOQI) guidelines for chronic kidney disease (29) with one exception: To draw comparisons between dialysis patients and non-dialysis patients (a clinically relevant distinction), the small number of patients with an estimated GFR<15cc/min/1.73m² but not on dialysis were included in the severe renal insufficiency group (estimated GFR <30cc/min/1.73m²) and not included with dialysis patients in a “kidney failure” category as suggested by NKF-DOQI guidelines. We excluded from the analysis all patients who had pre-operative acute renal failure to include only patients likely to have chronic renal insufficiency. Because the MDRD formula has not been validated in a veteran population, as a sensitivity analysis we also measured the association of operative outcomes with level of renal function defined purely by serum creatinine levels. For this purpose we adopted the following creatinine cutoffs to correspond respectively with a GFR of <60cc/min/1.73m² and <30cc/min/1.73m² as reported by Couchoud et al. (30): 1.57 mg/dl and 2.19 mg/dl for men and 1.17 mg/dl and 1.77 mg/dl for women. As for the primary analysis, dialysis patients were included as a separate category.

Ascertainment of Patient Baseline Characteristics, Preoperative Condition, and Operative Characteristics

Patient characteristics included age, white race (versus non-white), sex, co-morbid conditions at the time of the procedure (history of diabetes, insulin use, chronic obstructive pulmonary disease (COPD), history of congestive heart failure (CHF) during the month before surgery, history of transient ischemic attack, history of stroke with neurologic deficit, current smoking, current alcohol use). To develop a complete picture of disease burden, we also included whether the patient had a do-not-resuscitate (DNR) order in the chart signed by an attending in the 30 d before surgery, history of >10% unintentional weight loss during the preceding 6 mo, and presence of dyspnea at rest.

We also compared the acute pre-operative condition of patients in each renal insufficiency group. Variables indicating patient condition at the time of procedure included pre-operative albumin, hematocrit and white blood cell count, whether the patient was admitted from home (versus transferred from nursing home, acute care hospital or other setting), the presence of wound infection, pre-operative sepsis, and length of pre-operative hospital stay.

Finally, we compared operative characteristics between patients with different levels of renal function. Revascularization procedures were classified as inflow procedures if they provided a blood supply from vessels above the inguinal ligament to the inguinal vessels (common femoral, superficial femoral or profunda femoral arteries) and outflow procedures if they provided a blood supply to or beyond the popliteal artery (CPT codes were used to categorize revascularization procedures in this way). Other operative characteristics included were type of graft material (prosthetic or composite versus autogenous vein), whether a lower extremity amputation was performed during the same operating session, whether the procedure was an emergency, wound classification (contaminated versus clean), operation time and type of anesthesia (general versus other).

Statistical Analysis

We compared patient characteristics, acute pre-operative condition and operative outcomes in each reduced renal function category to the reference category of patients with normal or mildly reduced renal function using a χ² test. For simplicity, continuous variables were presented and analyzed as categorical variables dichotomized either at the mean value for normally distributed variables (age), median value for non-normally distributed variables (pre-operative hospital stay and operation time) or at clinically meaningful cutoffs (e.g. albumin ≤3g/dl, hematocrit ≤30%, white blood cell count >12,000 cells/microliter) in descriptive analysis. We measured the association of level of renal function with each operative outcome using logistic regression analysis adjusted for the three groups of variables identified above (baseline characteristics, pre-operative condition and operative characteristics). To accommodate the potentially correlated nature of procedure data from individual surgical centers, all multivariable models were adjusted for a random center effect (31). For logistic regression analysis, continuous variables were not dichotomized as for descriptive analysis above. Primary analyses were conducted using estimated GFR categories as defined by the abbreviated MDRD equation. A sensitivity analysis was conducted using categories of renal function defined by serum creatinine cutoffs as described above.

Some variables were missing data. For most of these variables less than 1% of values were missing (race, history of TIA, dyspnea at rest (<2%), alcohol use, pre-operative hematocrit, history of weight loss, DNR order, pre-operative wound infection, white blood count, pre-operative hospital stay (<3%), and operation time). Several variables were missing larger amounts of data (i.e. pre-operative albumin (37%), setting from which patient was admitted (home versus other) (23%) and pre-operative sepsis (23%)). For continuous variables, a mean value was imputed for missing values with an additional variable (missing indicator variable) included in the model that was equal to one if the variable was originally missing and equal to zero otherwise. This allowed all observations to be used in estimating the effects of variables that were not missing, but including the missing indicator variable ensured that the particular values imputed (in this case the overall means), had no influence on the variables’ estimated effects. Missing values for categorical variables were included as a separate “missing” category. All analyses were completed using Stata Statistical Software (College Station, TX).

Study Population

A total of 18,939 patients underwent at least one lower extremity revascularization that was listed as a principal procedure in NSQIP during the study period. Of these, 110 had pre-operative acute renal failure and were excluded from the analysis. Of the remaining 18,829 eligible patients, 612 patients were excluded because they lacked data on renal function (in the form of either a pre-operative serum creatinine or an indication that they were on dialysis at the time of surgery). Our study population consisted of the 18,217 remaining patients who underwent at least one revascularization during the study period, had data on renal function, and did not have acute renal failure during the pre-operative period. In this group, 73% of patients had normal or mildly reduced renal function (mean creatinine 1.0 ± 0.2mg/dl), 23% had moderate renal insufficiency (mean creatinine 1.5 ± 0.3 mg/dl), 2% had severe renal insufficiency but were not on dialysis (mean creatinine 3.2 ± 1.4mg/dl) and 2% were dialysis dependent.

Patient Characteristics

Table 1 shows that most baseline characteristics differed substantially by level of renal function. Most notably, the percentage of patients with diabetes increased with advancing degrees of renal insufficiency. Table 2 shows that there were also differences in acute pre-operative condition across groups, with patients with renal insufficiency being less likely to be admitted from home and more likely to present with pre-operative wound infection, pre-operative sepsis, elevated white blood cell count, and to have longer pre-operative hospital stays compared to patients with normal or mildly reduced renal function. Table 3 shows that the types of procedures differed between groups. Patients with renal insufficiency were less likely to undergo inflow procedures and were also less likely to have a prosthetic (versus autogenous vein) graft than patients with normal renal function. Lower extremity amputation performed under the same anesthetic as the principal revascularization procedure was a more frequent occurrence in patients with severe or dialysis dependent renal insufficiency than in those with normal renal function or milder degrees of renal insufficiency. Patients with renal insufficiency were also more likely to undergo emergency procedures, more likely to have contaminated wounds, and less likely to undergo general anesthesia than patients with normal renal function.

Postoperative Complications

Table 4 shows the incidence of 30-d post-operative complications by degree of renal insufficiency. Death within 30 d, cardiac arrest, unplanned intubation, systemic sepsis, coma, and post-operative bleeding were more common in patients with renal insufficiency (even those with moderate renal dysfunction). In contrast, patients with renal insufficiency did not appear to have a higher rate of operative site complications such as wound infection, wound dehiscence, and graft failure requiring return to the operating room than those with normal renal function.

The incidence of most of the aforementioned complications was similar to or higher in patients with severe renal insufficiency than patients on dialysis. In addition, patients with moderate or severe renal insufficiency had a higher incidence of myocardial infarction, prolonged intubation and (by definition) post-operative acute renal failure requiring dialysis compared to patients on dialysis.

Table 5 and Figure 1 show adjusted associations of degree of renal insufficiency with post-operative complications. While we tested the association of all outcomes presented in Table 4 with level of renal function, only those associations that were statistically significantly associated with the outcome are presented. Even after adjustment for differences in baseline characteristics, death within 30-d, cardiac arrest, myocardial infarction, unplanned or prolonged intubation, and acute renal failure requiring dialysis were strongly associated with level of renal function. Dialysis patients had a higher incidence of death, cardiac arrest, sepsis and coma but not of other complications listed above than for patients with lesser degrees of renal insufficiency. In addition, we found no statistically significant adjusted (or unadjusted) associations for operative site complications such as wound infection or graft failure with renal insufficiency. The results of multivariate analysis using categories of renal function defined by serum creatinine alone did not differ substantially from the results presented for the primary analysis using estimated GFR.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Odds ratio plots for association of level of renal function with post-operative complications. All complications for which there was a statistically significant association with at least one renal insufficiency group are presented. The reference category for all plots is patients with a GFR >60 cc/min/1.73m².