MM O'Brien, Denver VAMC/University of Colorado Health Sciences Center; R Gonzales, University of Colorado Health Sciences Center; AL Shroyer, Denver VAMC/University of Colorado Health Sciences Center; G Grunwald, Denver VAMC/University of Colorado Health Sciences Center; SF Khuri, West Roxbury VAMC/VA Boston Health Care System; J Daley, West Roxbury VAMC/Massachusetts General Hospital; WG Henderson, Edward Hines Jr. Hospital; R Anderson, University of Colorado Health Sciences Center

Objectives: To determine if a mildly elevated serum creatinine concentration is associated with risk-adjusted morbidity and mortality after general surgery, and to approximate a threshold of serum creatinine concentration which predicts adverse outcomes.

Methods: Using a pre-existing Veterans Affairs (VA) database (National Surgical Quality Improvement Program or NSQIP), we performed univariate and multivariate analyses of preoperative risk variables in 49,081 general surgery cases done between 10/1/96 and 9/30/98 at any one of 100 VA medical centers. Cases were categorized according to the preoperative (baseline) serum creatinine: (1) less than 1.5 mg/dL (normal value for most hospital laboratories), (2) 1.5 to 3.0 mg/dL (considered by many to be mild to moderate renal failure, usually asymptomatic), and (3) greater than 3.0 mg/dL (considered more indicative of more advanced renal failure). Two outcomes were analyzed: all-cause 30-day mortality and 30-day morbidity after general surgery. Risk models were built on learning datasets using logistic regression and disallowing serum creatinine to enter the models. The risk models were then applied to the test datasets, calculating predicted risk of death and morbidity. The three-level serum creatinine variable was subsequently added to each of the newly created risk models, using the logistic transformation of the previously derived mortality/morbidity risk estimate as an offset so that the initial risk parameters were not re-estimated. Comparisons were made of the predictive power of the models before and after serum creatinine, using the c-index, Hosmer-Lemeshow goodness-of-fit statistic, and the likelihood ratio test. Expanding the initial three-level serum creatinine variable to a five-level variable, contrast statements were then used in logistic regression so that multiple groupings of serum creatinine could sufficiently be compared to each other, and the threshold of statistical significance could be identified. Since renal failure results in physiologic abnormalities that could potentially predispose patients to such postoperative complications as bleeding, infections, cardiac and respiratory complications, further analyses were done to predict these individual morbidities.

Results: Unadjusted 30-day mortality, unadjusted 30-day morbidity and several pulmonary, cardiac, neurologic, infectious and hemorrhagic morbidities were significantly higher (p<0.01) in patients with a serum creatinine between 1.5 and 3.0 mg/dL compared to patients with a serum creatinine less than 1.5 mg/dL. With multivariable analysis, a serum creatinine between 1.5 and 3.0 mg/dL (compared to serum creatinine <1.5 mg/dL) was associated with an adjusted odds ratio of 1.44 for mortality and 1.23 for morbidity (respective 95% confidence intervals of 1.219 to 1.711 and 1.105 to 1.373). We found the threshold of serum creatinine concentration for 30-day mortality to be approximately 2.0 mg/dL, and for 30-day morbidity, approximately 1.5 mg/dL.

Conclusions: Mild renal insufficiency is significantly associated with risk-adjusted 30-day mortality and morbidity after general surgery. A preoperative serum creatinine of 1.5 mg/dL or higher, may significantly increase the risk of mortality and/or morbidity after general surgery.

Impact: Our study identifies a potentially modifiable risk factor for adverse outcomes after general surgery. A preoperative serum creatinine greater than 1.5 mg/dL may necessitate enhanced clinical evaluation and surveillance, and perhaps specific prophylactic maneuvers to potentially improve outcome.