The details of the ACS NSQIP have been previously described.15
The ACS NSQIP provides risk-adjusted outcomes data to participating hospitals for the purpose of quality improvement. The program focuses on 30-day postoperative outcomes, including mortality and 21 categories of morbidity. Data collection at each participating institution is performed by a dedicated surgical clinical reviewer (SCR) with support and oversight from an ACS NSQIP Nurse Coordinator. The SCR, using medical chart extraction and other methods, collects detailed data on patient demographics, comorbidities, preoperative laboratory values, operative variables, and postoperative outcomes including 30-day complications, 30-day mortality, reoperation, and length of stay. Full descriptions of the qualifications, training, and auditing of data collection personnel, case inclusion criteria, sampling and data collection strategy, and variable and outcome definitions are available online from the ACS NSQIP website.19
In brief, the ACS NSQIP captures and reports 30-day morbidity and mortality outcomes for a sample of operations performed at participating institutions. In an attempt to limit bias in case selection, the NSQIP has developed a systematic sampling algorithm called the 8-day cycle. The number of patients included from each of the forty 8-day cycles per year is dependent upon the surgical volume of the participating institution. The SCR at a high-volume hospital will capture the first 40 operations in the 8-day cycle thereby including 1640 patients per high-volume center. All patients from each 8-day cycle are captured at low-volume hospitals, which have to include at least 900 patients per year. A full description of the temporal sampling procedure as well as case inclusion and exclusion criteria can be obtained in the ACS NSQIP data user guide.19
Participant use files (PUF) containing 240 HIPAA (Health Insurance Portability and Accountability Act) compliant variables for each patient included from each participating site are made available yearly by the ACS. The 2008 PUF contains 271,368 cases submitted from 211 participating institutions. The 2005 to 2006 and 2007 PUFs contain 152,490 and 211,407 cases submitted by 183 sites, respectively. These files are available free of cost to individuals from participating institutions. The human subjects committee at the University of Wisconsin reviewed the database and deemed it to be exempt from Institutional review board review.
All 635,265 patients in the NSQIP database between 2005 and 2008 including all undergoing emergency and nonemergency cardiac and noncardiac surgery were considered for inclusion into this study. Patients undergoing open colon surgery were selected using the principal Current Procedural Terminology (CPT) codes 44140, 44141, 44145, 44146, 44147, 44150, and 4416. Those patients undergoing laparoscopic colon surgery were selected using the CPT codes 44204, 44205, 44206, 44207, 44208, and 44210. From these groups, patients undergoing surgery for colon cancer were selected using International Classification of Disease, 9th edition codes 153, 153.1, 153.2, 153.4, 153.5, 153.6, 153.7, 153.8, and 153.9. Those who underwent emergency operations or classified as American Society of Anesthesiology (ASA) class 5 were excluded. Other exclusion criteria included ventilator dependence, preoperative sepsis, comatose, and age less than 65.
Patients were dichotomized throughout our study into open or laparoscopic resection based on CPT codes. Outcome variables included 30-day mortality and 30-day complications (any—yes/no; and by category). Control variables include demographics (gender, race, age), as well as preoperative health status and comorbidities, perioperative variables, and preoperative laboratory values. Age was treated as both continuous and categorical variables (1, 65–75 years; 2, 75–85 years; and 3, 85+ years). Preoperative comorbidities included body mass index (BMI) (continuous—calculated from height and weight; categorical—underweight [BMI <18.5], normal [BMI = 18.5–24.9 kg/m2
], overweight [BMI = 25–29.9 kg/m2
], obese [BMI = 30–49 kg/m2
], and superobese [BMI >50 kg/m2
diabetes mellitus (no/oral hypoglycemic agents or insulin), history of smoking in the year before surgery (no/yes), intake of more than 2 alcoholic drinks in the 2 weeks before surgery (no/yes), functional category (independent/totally or partially dependent), history of chronic obstructive pulmonary disease (COPD) (no/yes), current pneumonia (no/yes), presence of difficult or labored breathing (dyspnea no/yes), history of coronary artery disease (no/yes), history of peripheral vascular disease (no/yes), history of neurologic disease (no/yes), currently on dialysis (no/yes), use of steroids for chronic condition (no/yes), >10% weight loss in the 6 months prior to surgery (no/yes), ascites (no/yes), history of esophageal varices documented within 6 months of surgery (no/yes), history of congestive heart failure within 30 days of surgery (no/yes), history of bleeding disorders (no/yes), history of chemotherapy exposure in ≤30 days prior to surgery (no/yes), and presence of open wound/wound infection preoperatively (no/yes). Other control variables include the ASA category (1: no/mild; 2: severe; 3: life threatening) as well as intraoperative variables including surgical wound category (1: clean/contaminated; 2: contaminated; 3: dirty/infected), intraoperative blood transfusion (1: none; 2: 1–2; 3: ≥3), and length of operation (1: <4 hours, 2: 4–6 hours; and 3: >6 hours). Preoperative laboratory variables included creatinine (low, <1.2 mg/dL; high, >1.2 mg/dL; or missing), serum sodium (low, <135 mmol/L; normal, 135–145 mmol/L; high, >145 mmol/L; or missing), serum alkaline phosphatase (low, <125 U/L; high, >125 U/L; or missing), white blood cell count (WBC low, <4.5 K/μL; normal, 4.5–11 K/μL; high, >11 K/μL; and missing), hematocrit (low, <38%; high, >38%; and missing), blood urea nitrogen (BUN low, <40 mg/dL; high, >40 mg/dL; or missing), serum glutamic oxaloacetic transaminase (low, 40 U/L; high, >40 U/L; or missing), albumin (low, <2.4 g/dL; normal, 2.5–3.39 mg/dL; high, >3.4 mg/dL; or missing) and platelets (low, <50 K/μL; normal, 50–400 K/μL; high, >400 K/μL; or missing).
Outcomes of interest were complications occurring within 30 days of the procedure and death occurring within 30 days of the procedure. All surgical outcomes and potential control variables were compared between the surgical approach groups using χ2 tests for categorical variables and Wilcoxon rank sum tests for continuous variables. All variables that were found to be associated with surgical outcomes at a significance level of 0.10 in these univariate analyses were included in the multivariable model. Preoperative albumin and serum glutamic oxaloacetic transaminase categories were excluded from the final model to minimize the potential effect of multicollinearity.
To partially adjust for the nonrandom assignment of surgical approach, a propensity score model was constructed based on multivariate logistic regression using all the preoperative patient characteristics, comorbidities, and laboratory values. Propensity scores were divided into quintiles so that patients in the laparoscopic and open surgical approach groups had a similar distribution of preoperative baseline risk factors within each quintile.18,21
The propensity score variable was included in the final multivariable model (along with control variables that remained significant at P
= 0.1) to examine the association between the surgical approach and morbidity and mortality as has been previously used by Bilimoria et al.18
Interaction terms involving the propensity score quintiles and other factors did not contribute to the overall quality of the multivariable model.
The NSQIP database includes a variable that estimates risk for either morbidity or mortality for each patient (probmorb and probmort, respectively). These variables were not included in the multivariable analyses. However, unadjusted subgroup analyses were performed to assess whether laparoscopy had a significant association with postoperative complications within 3 probability of morbidity groups: ≤0.2, 0.2–0.4, and ≥0.4; 3 age groups: 65–75, 75–85, and ≥85; and 3 ASA categories: no/mild illness (ASA 1–2), severe illness (ASA 3), and life-threatening illness (ASA 4) using a χ2 test. Statistical significance was considered with P < 0.05. All analyses were performed using SAS 9.1.3 for Windows (SAS Institute; Cary, NC).