The American Journal of Surgery (2015) 209, 219-229

Clinical Science

Patient-specific risk factors are predictive for postoperative adverse events in colorectal surgery: an American College of Surgeons National Surgical Quality Improvement Program–based analysis Adrian Y. Kohut, B.S., M.D., Candidate, Class of 2015, James J. Liu, B.A., M.D., Candidate, Class of 2015, David E. Stein, M.D., F.A.C.S., F.A.S.C.R.S., Richard Sensenig, M.S., M.S.E.E., Juan L. Poggio, M.D., M.S., F.A.C.S., F.A.S.C.R.S.* Division of Colorectal Surgery, Department of Surgery, Drexel University College of Medicine, 245 North 15th Street, MS 413, Philadelphia, PA 19102-1192, USA

KEYWORDS: Colorectal surgery; Colorectal resection; Postoperative complications; ACS NSQIP; Surgical site infection; Pay for performance

Abstract BACKGROUND: Pay-for-performance measures incorporate surgical site infection rates into reimbursement algorithms without accounting for patient-specific risk factors predictive for surgical site infections and other adverse postoperative outcomes. METHODS: Using American College of Surgeons National Surgical Quality Improvement Program data of 67,445 colorectal patients, multivariable logistic regression was performed to determine independent risk factors associated with various measures of adverse postoperative outcomes. RESULTS: Notable patient-specific factors included (number of models containing predictor variable; range of odds ratios [ORs] from all models): American Society of Anesthesiologists class 3, 4, or 5 (7 of 7 models; OR 1.25 to 1.74), open procedures (7 of 7 models; OR .51 to 4.37), increased body mass index (6 of 7 models; OR 1.15 to 2.19), history of COPD (6 of 7 models; OR 1.19 to 1.64), smoking (6 of 7 models; OR 1.15 to 1.61), wound class 3 or 4 (6 of 7 models; OR 1.22 to 1.56), sepsis (6 of 7

The authors declare no conflicts of interest. Disclosure information: All authors deny any disclosure of potential conflicts of interest, including financial interests, activities, relationships, and affiliations. The American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) and the hospitals participating in the ACS NSQIP are the source of the data used here; they have not verified and are not responsible for either the statistical validity of the data analysis or the conclusions derived by the authors. This study was presented as a podium presentation at the 2013 Annual Meeting of the Pennsylvania Society of Colon and Rectal Surgeons, The Union League, March 15, 2013, Philadelphia, PA. * Correspondence author. Tel.: 11-215-762-1750; fax: 11-215-762-8389. E-mail address: [email protected] Manuscript received March 12, 2014; revised manuscript July 22, 2014 0002-9610/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjsurg.2014.08.020

220

The American Journal of Surgery, Vol 209, No 2, February 2015 models; OR 1.14 to 1.89), corticosteroid administration (5 of 7 models; OR 1.11 to 2.24), and operation duration more than 3 hours (5 of 7 models; OR 1.41 to 1.76). CONCLUSIONS: These findings may be used to pre-emptively identify colorectal surgery patients at increased risk of experiencing adverse outcomes. Ó 2015 Elsevier Inc. All rights reserved.

Colorectal resection patients have higher rates of postoperative complications than all other surgical subpopulations.1–4 Surgical site infections (SSIs) are particularly prevalent and are associated with increased postoperative length of hospitalization and higher costs of care.3–13 As the surgical field moves toward pay-for-performance measures, SSI rates are being incorporated into reimbursement policies;4,5,14–18 however, pay-for-performance measures do not incorporate factors predictive of SSIs and other adverse postoperative outcomes into the provider reimbursement algorithms.5,17 Policies that fail to account for risk factors may result in negative consequences including decreased access to care for higher risk patients, discontinuation of payments to surgeons and institutions that provide care for higher risk patients, and increased racial and socioeconomic disparities in access to surgical care.5,18,19 Quality improvement programs have taken center stage in the surgical field as a means of quantifying patient outcomes.1,2,20–24 The American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) has been on the forefront of these efforts and has created a national tool consisting of a risk-adjusted evaluation of surgical outcomes.2,20–24 Institutions participating in the ACS NSQIP may use risk-adjusted data to construct predictive risk models capable of offering surgeons quantifiable and objective data for assessing the clinical and socioeconomic implications of adverse outcomes in colorectal surgery. Data-driven predictive modeling provides a means for rational review of pay-for-performance algorithms, which should lead to more equitable policies for health care providers, patients, and insurance companies alike. We undertook a large-scale population-based study to assess predictive factors associated with commonly occurring postoperative complications in colorectal resection patients. The purpose of this work was to both objectively and quantifiably delineate patient-specific independent risk factors associated with various measures of postoperative morbidity. The results provide a framework for preoperatively identifying patients at high risk of experiencing adverse postoperative outcomes and demonstrate the need to protect surgeons who choose to operate on high-risk patients.

Procedural Terminology coding criteria. The following procedures were included in the study (numbers in parentheses are laparoscopic and open procedures, respectively): ileocolic resection (44,205; 44,160), partial colectomy (44,204; 44,140), Hartmann’s procedure (44,206; 44,143), low pelvic anastomosis (44,207 or 44,208; 44,145 or 44,146), total abdominal colectomy (44,210; 44,150), total proctocolectomy/ileal pouch anal anastomosis (44,211; 44,158), and total proctocolectomy/end ileostomy (44,212; 44,155). Laparoscopic partial colectomy (44,204) was used as the reference procedure by which all other procedures were compared against. Because patients’ presurgical comorbidity status varied greatly, we excluded emergency cases. To offset the limitations posed by missing data, we eliminated patients for which data were not available for at least 90% of variables in the net analysis sample. List-wise deletion was subsequently used for addressing residual missing data. After correction for missing data, the final analysis sample consisted of approximately 89% of the initial elective colorectal population.

Variable selection Prospective variables of interest in the study were selected based on clinical relevance and the literature.3–13 Definition criteria for all variables included in the study are found in the 2005 to 2010 American College of Surgeons User Guides for the Participant Use Data Files.25 Dependent outcome variables, ie, the adverse outcomes assessed by each predictive model include: 1. 2. 3. 4.

superficial SSI deep incisional SSI organ space SSI any SSI (includes presence of any of the 3 aforementioned SSIs) 5. wound disruption 6. return to the operating room 7. increased length of stay These outcome variables were evaluated against an array of predictor variables. All variables included in the study are contained in Table 1.

Methods Statistical analysis Patient selection/inclusion criteria We queried the 2005 to 2010 ACS NSQIP databases for elective colorectal surgery patients using primary Current

The queried colorectal surgery population of 67,445 patients was randomly sorted and divided into samples of 2 subpopulations:

A.Y. Kohut et al. Table 1

Adverse events in colorectal surgery

221

Model build population vs validation population

Variable Any SSI* Superficial SSI* Deep SSI* Organspace SSI* Wound disruption* Return to operating department* Increased length of stay*,† Laparoscopic ileocolic resection Laparoscopic partial colectomy Laparoscopic Hartmann’s procedure Laparoscopic low pelvic anastomosis Laparoscopic total abdominal colectomy Laparoscopic total proctocolectomy, ileal pouch anal anastomosis Laparoscopic total proctocolectomy, end ileostomy Open ileocolic resection Open partial colectomy Open Hartmann’s procedure Open low pelvic anastomosis Open total abdominal colectomy Open total proctocolectomy, ileal pouch anal anastomosis Open total proctocolectomy, end ileostomy Male sex Diabetes Smoking Alcohol History of COPD History of CHF History of myocardial infarction History of cardiac surgery High blood pressure History of PVD Dialysis History of cerebrovascular accident Disseminated cancer Steroid Weight loss Transfusion Chemotherapy Radiotherapy Wound class 3 or 4 ASA classification 3, 4, or 5 Operation duration .3 h Systemic sepsis‡ Ascites Dependent functional status Race, white Race, black Race, other Age 35–49 y 15–34 y 50–64 y 65–79 y 801 y

Model build population (n 5 44,963)

Validation population (n 5 22,482)

P

5,967 3,677 637 1,653 648 2,535 9,424 3,903 9,341 306 4,357 603 467

(13.3%) (8.2%) (1.4%) (3.7%) (1.4%) (5.6%) (21.0%) (8.7%) (20.8%) (.7%) (9.7%) (1.3%) (1.0%)

2,982 1,812 311 859 300 1,264 4,697 1,976 4,549 160 2,245 330 257

(13.3%) (8.1%) (1.4%) (3.8%) (1.3%) (5.6%) (20.9%) (8.8%) (20.2%) (.7%) (9.9%) (1.5%) (1.1%)

.980 .591 .678 .366 .252 .915 .833 .633 .102 .657 .217 .173 .235

223 4,874 10,721 1,839 5,985 1,247 449

(.5%) (10.8%) (23.8%) (4.1%) (13.3% (2.7%) (.9%)

95 2,415 5,420 925 2,964 617 208

(.4%) (10.7%) (24.1%) (4.1%) (13.2) (2.7%) (.9%)

.154 .693 .439 .902 .639 .823 .416

648 21,291 6,441 8,074 1,599 2,399 435 260 2,519 22,794 611 334 993 1,888 2,993 2,429 334 533 1,116 7,591 21,061 15,408 2,312 500 2,569 38,861 4,418 1,684

(1.4%) (47.4%) (14.3%) (18.0%) (3.6%) (5.3%) (1.0%) (.6%) (5.6%) (50.7%) (1.4%) (.7%) (2.2%) (4.2%) (6.7%) (5.4%) (.7%) (1.2%) (2.5%) (16.9%) (46.8%) (34.3%) (5.1%) (1.1%) (5.7%) (86.4%) (9.8%) (3.7%)

321 10,768 3,245 4,057 798 1,185 192 125 1,246 11,247 299 142 481 914 1,453 1,161 165 302 552 3,853 10,471 7,690 1,194 227 1,279 19,518 2,109 855

(1.4%) (47.9%) (14.4%) (18.0%) (3.5%) (5.3%) (.9%) (.6%) (5.5%) (50.0%) (1.3%) (.6%) (2.1%) (4.1%) (6.5%) (5.2%) (.7%) (1.3%) (2.5%) (17.1%) (46.6%) (34.2%) (5.3%) (1.0%) (5.7%) (86.8%) (9.4%) (3.8%)

.918 .178 .727 .774 .947 .703 .126 .746 .749 .101 .750 .107 .558 .425 .324 .191 .886 .097 .875 .396 .524 .877 .348 .236 .916 .162 .062 .748

6,754 2,471 15,188 14,458 6,092

(15%) (5.5%) (33.8%) (32.2%) (13.5%)

3,449 1,247 7,675 7,087 3,024

(15.3%) (5.5%) (34.1%) (31.5%) (13.5%)

.274 .789 .352 .093 .720

(continued on next page)

222 Table 1

The American Journal of Surgery, Vol 209, No 2, February 2015 (continued )

Variable

Model build population (n 5 44,963)

BMI ,26 26–29 30–35 .35

18,665 11,735 9,590 4,973

(41.5%) (26.1%) (21.3%) (11.1%)

Validation population (n 5 22,482) 9,349 5,811 4,812 2,510

(41.6%) (25.8%) (21.4%) (11.2%)

P .862 .485 .834 .697

Proportion of colorectal patients presenting with each variable in both the model build population and the validation population. P values derived from a 2-sample Z-test of proportions assessment for all variables in the study. Insignificant P suggests no significant differences in the proportions of patients presenting with each variable in the Model Build population vs the validation population. *Signifies outcome variable. † Defined as postoperative hospitalization length within the upper quartile of the analysis sample (9 days). ‡ NSQIP subcategorizes the systemic sepsis variable as the presence of Systemic Inflammatory Response Syndrome, sepsis, or severe sepsis/septic shock at any time during the 48 hours preceding surgery. Within the patients included in our study, there were none classified as having severe sepsis or septic shock.

1. a model build sample (n 5 44,963; 66.7%) used to generate all models and 2. a validation sample (n 5 22,482; 33.3%) fitted for all models generated by the model build sample as a means of internal validation. All predictor variables included in the study were individually analyzed against each of the 7 outcome variables via univariate logistic regression. Saturated models were constructed for each dependent outcome variable using all predictor variables, yielding P , .15 in the univariate analyses. The saturated models were then reduced using a backward stepwise Akaike information criterion–based methodology.26 All variables that remained following the Akaike information criterion–based reduction were then used to construct the final adjusted multivariable models. Patient-specific predictive probabilities, which were used to measure model performances via the receiver operating characteristics (ROC) curve–derived c statistic and the Hosmer–Lemeshow goodness-of-fit test, were calculated from the final, adjusted multivariate logistic regression models in the form of the logit transformation: Predictive probabilityðoutcome occurringÞ   5 ea1b1 X11b2 X21.1bi Xi 11ea1b1 X11b2 X21.1bi Xi ; where a corresponds to the b coefficient of the model intercept and b corresponds to the b coefficient of individual variables of interest. Model performances of both the model build sample and the validation sample were assessed by measures of both discrimination and calibration using the ROC curve–derived c statistic and the Hosmer–Lemeshow goodness-of-fit test, respectively. The interpretation of such analyses has been widely discussed.2,27–32 Lastly, the model build sample and validation sample were compared for demographic differences by performing a 2-sample Z test of proportion assessments for all variables in the study. All statistical analyses were performed using the R software

environment for statistical computing and graphics, version 2.15.1.33 This study was reviewed by the Drexel University College of Medicine Institutional Review Board and was deemed appropriate for exemption from institutional review board oversight as no personal identifiers were used.

Results Patient characteristics A total of 89,098 patients who had undergone colorectal surgery were identified from the 2005 to 2010 ACS NSQIP databases. A total of 13,938 emergency cases (15.6%) were excluded, leaving an elective colorectal population of 75,160 patients. We performed list-wise deletion to account for missing data, arriving at a final analysis sample of 67,445 patients (89.7% of the initial elective population). As demonstrated in Table 1, no significant variances in the rates of occurrence of variables in the model build sample vs the validation sample were identified. SSIs were found in 13.3% of patients in both populations, with superficial SSIs occurring most commonly: 8.2% (model build) and 8.1% (validation). Approximately one-fifth of all patients required increased lengths of stays (z9 days). Partial colectomy was the most frequently performed procedure in both populations. Approximately 52% of all patients were women, and whites were the overwhelmingly predominant race (.86%). More than 79% of all patients were of 50 years or older, and 46% of all patients were identified as American Society of Anesthesiologists class 3, 4, or 5. The rates of occurrence of SSIs in the sample population were any SSI (13.3%), superficial SSI (8.1%), deep incisional SSI (1.4%), and organ/space SSI (3.7%). Rates of occurrence for the additional outcomes assessed in this study were wound disruption (1.4%), return to operating room (5.6%), and increased length of postoperative hospitalization (20.9%). Given the large sample size, these rates

A.Y. Kohut et al.

Adverse events in colorectal surgery

223

Figure 1 (A) Any SSI: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with any SSI. (B) Superficial SSI: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with superficial SSI. (C) Deepincisional SSI: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with deepincisional SSI. (D) Organspace SSI: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with organspace SSI.

likely represent accurate estimates of rates of occurrences in the colorectal population.

regression models pertaining to SSIs are detailed in Fig. 1A–D.

Surgical site infections

Other adverse outcomes

Odds ratios, 95% confidence intervals, and significance data derived from final, multivariate (adjusted) logistic

All final, multivariate (adjusted) logistic regression models pertaining to wound disruption return to operating

224

The American Journal of Surgery, Vol 209, No 2, February 2015

Figure 1

room, and increased length of stay are detailed in Fig. 2A–C.

Discrimination and calibration The c statistic was similar in all SSI models: any SSI, superficial SSI, deep incisional SSI, and organ/space SSI yielded areas under the ROC curve (c statistic) of .640, .630, .679, and .665, respectively (model build population). Greater discriminatory power was found in the remaining

(Continued)

models, with wound disruption, return to operating room, and increased postoperative length of stay yielding area under the ROC curve values of .811, .779, and .796, respectively. When the validation population was fitted to each model, there was no suggestion of loss of discrimination in any model (Table 2). The Hosmer–Lemeshow goodness-offit test indicated no significant P values, among all SSI models, suggesting that the fit was adequate in all SSI models. Significant P values were found in wound disruption (P 5 .042), return to operating room (P 5 .002), and increased length of postoperative hospitalization models

A.Y. Kohut et al.

Adverse events in colorectal surgery

225

Figure 2 (A) Wound disruption: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with wound disruption. (B) Return to operating room: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with return to OR. (C) Increased length of stay: odds ratio and corresponding 95% confidence interval for each predictor variable found to be associated with increased length of stay.

226

The American Journal of Surgery, Vol 209, No 2, February 2015

Figure 2

Table 2

(Continued)

Model performance statistics

Model performance

C statistic

Hosmer–Lemeshow P value

Any SSI (model build) Any SSI (validation) Superficial SSI (model build) Superficial SSI (validation) Deep SSI (model build) Deep SSI (validation) Organspace SSI (model build) Organspace SSI (validation) Wound disruption (model build) Wound disruption (validation) Return to operating room (model build) Return to operating room (validation) Increased length of stay (model build) Increased length of stay (validation)

.64 .64 .63 .616 .679 .674 .665 .665 .811 .807 .779 .775 .796 .803

.445 .881 .143 .401 .874 .570 .802 .644 .042 .056 .002 .033 ,.001 ,.001

C statistic quantifies the discriminatory power of each model and is derived from the area under the receiver operator curve. Hosmer–Lemeshow P value quantifies the calibration of each mode where insignificant P values insinuate superior calibration.

A.Y. Kohut et al.

Adverse events in colorectal surgery

(P , .001). The validation population yielded Hosmer–Lemeshow goodness-of-fit test results that were similar to the model build population in all models (Table 2).

Comments The results of this study demonstrate that quantifiable objective determinants of patient-specific risks factors (see Fig. 1A–D and Fig. 2A–C) are associated with postoperative complications frequently occurring in patients undergoing elective colon and rectal surgery. We have used the power of an immense sample size (n 5 67,445 patients) to both derive and validate risk models individualized for each specific outcome variable in question. Thus, clinicians may use our findings as a tool to pre-emptively identify colorectal surgery patients at increased risk of experiencing specific postoperative, adverse events. Additionally, the descriptive statistics offered in Table 1 allow for an indepth depiction of the demographic and clinical intricacies present in the colorectal resection population as a whole. Although the main focus of this study was to isolate predictive factors associated with individual outcome variables, various overarching themes coincide. Certain risk factors were positively associated with postoperative complications in multiple models. Notable examples (see Table 3) include American Society of Anesthesiologists Table 3

Significantly occurring predictor variables

Predictor variable Open partial colectomy Open total abdominal colectomy ASA classification 3, 4, or 5 Open low pelvic anastomosis BMI . 35 History of COPD Wound class 3 or 4 Systemic sepsis Smoking Open ileocolic resection Open Hartmann’s procedure Open total proctocolectomy/ end ileostomy Steroid administration Operation duration . 3 h Laparoscopic proctocolectomy, end ileostomy Open total proctocolectomy, ileal pouch anal anastomosis BMI 30–35 26–29

No. of models containing predictor variable 7 7 7 6 6 6 6 6 6 5 5 5 5 5 4 4

4 4

Risk factors found to be positively associated with postoperative complications in multiple predictive models. There were a total of seven models in study.

227 class 3, 4, or 5 (in 7 of 7 models), body mass index (BMI) greater than 35 (in 6 of 7 models), history of chronic obstructive pulmonary disease (COPD) (in 6 of 7 models), wound class 3 or 4 (in 6 of 7 models), systemic sepsis (in 6 of 7 models), smoking (in 6 of 7 models), operation duration more than 3 hours (in 5 of 7 models), and steroid administration (in 5 of 7 models). Three significant variables of particular interest were smoking, increasing BMI, and history of COPD. Our findings indicated that smoking was significantly associated with SSIs, wound disruptions, and return to the operating room. Smoking has previously been suggested as a risk factor for surgical complications.35,36 Possible mechanisms of action include the undermining of wound healing and tissue ischemia.34,35 Patients with COPD may have similarly impaired wound healing via tissue ischemia and altered immunologic response and repair mechanisms.36,37 Furthermore, increasing thickness of subcutaneous fat correlates with wound infection,38 which corresponds to increasing BMI as a consistent predictor as seen in the models for any SSI, superficial SSI, deep SSI, wound disruption, and return to operating room. Thus, our study provides further clinical and economic incentive to promote nationwide measures for smoking cessation and healthy body fat composition levels. Widespread improvement in these 2 areas alone would result in decreased costly adverse outcomes in colorectal resection patients. One of the most consistent predictive factors was the open approach to various procedures. However, a few laparoscopic approaches also proved frequently positively predictive, including laparoscopic proctocolectomy/end ileostomy (in 4 of 7 models) and laparoscopic total proctocolectomy/ileal pouch anal anastomosis (in 3 of 7 models). Both these laparoscopic procedures are predominantly indicated in ulcerative colitis patients. We hypothesize that ulcerative colitis patients present with a unique set of comorbidities, which may cause a confounding effect on other risk factors. Thus, we recommend that clinicians remain cognizant of the high-risk laparoscopic procedures noted in Figs. 1A–D and 2A–C. Various limitations to our study should be noted. Although the ACS NSQIP produces the largest, most comprehensive, and risk-adjusted database of surgical outcome data in the world, it reports only a percentage of randomly selected cases occurring at participating hospitals. Thus, our data were limited to hospitals participating in the ACS NSQIP program. Also, select variables of interest such as serum albumin were excluded from study because of the sizable proportion of patients lacking data. Future studies may expand on such limitations using advanced multiple imputation methods; however, such methods allow for the possibility of inferential bias. Additionally, data for anastomotic leakage and antibiotic prophylaxis were not available, thus precluding us from studying their effects in this study. Lastly, we do not address the NSQIP-limited information on surgeon reasoning in laparoscopic vs open approaches. The subjective nature of this determination limited us from being

228 able to quantify it in a manner appropriate for inclusion in our risk models. Among inferences that we draw from the data presented in this report is that the attempt to withhold surgeon reimbursement based on occurrence of adverse outcomes, without taking into account patient-specific risk factors, may lead to discontinued payments to surgeons who choose to regularly provide care for high-risk populations.5 Such a scenario would only further augment disparities in access to surgical care for high-risk population groups such as low-income and minority patients and, thus, inherently undermine an institution’s future capacity for treating high-risk patients. Although we encourage all attempts to both monitor and combat SSIs on a national level, we believe that patient-specific risk factors such as those described in this study should be considered so that current and upcoming pay-for-performance policies can be made more equitable for surgeons, patients, and insurance providers alike.

Acknowledgment Author contributionsdStudy conception and design: J.L.P., A.Y.K., J.J.L.; acquisition of data: J.L.P., A.Y.K., J.J.L.; analysis and interpretation of data: J.L.P., A.Y.K., J.J.L.; drafting of manuscript: J.L.P., A.Y.K., D.E.S., and R.S.; critical revision: J.L.P., A.Y.K., D.E.S., and R.S.

References 1. Schilling PL, Dimick JB, Birkmeyer JD. Prioritizing quality improvement in general surgery. J Am Coll Surg 2008;207: 698–704. 2. Cohen ME, Bilimoria KY, Ko CY, et al. Development of an American College of Surgeons National Surgery Quality Improvement Program: morbidity and mortality risk calculator for colorectal surgery. J Am Coll Surg 2009;208:1009–16. 3. Wick EC, Vogel JD, Church JM, et al. Surgical site infections in a ‘‘high outlier’’ institution: are colorectal surgeons to blame? Dis Colon Rectum 2009;52:374–9. 4. Young H, Knepper B, Moore EE, et al. Surgical site infection after colon surgery: National Healthcare Safety Network risk factors and modeled rates compared with published risk factors and rates. J Am Coll Surg 2012;214:852–9. 5. Wick EC, Hirose K, Shore AD, et al. Surgical site infections and cost in obese patients undergoing colorectal surgery. Arch Surg 2011;146: 1068–72. 6. Kao LS, Ghaferi AA, Ko CY, et al. Reliability of superficial surgical site infections as a hospital quality measure. J Am Coll Surg 2011; 213:231–5. 7. Hu¨bner M, Diana M, Zanetti G, et al. Surgical site infections in colon surgery: the patient, the procedure, the hospital, and the surgeon. Arch Surg 2011;146:1240–5. 8. Imai E, Ueda M, Kanao K, et al. Surgical site infection risk factors identified by multivariate analysis for patient undergoing laparoscopic, open colon, and gastric surgery. Am J Infect Control 2008; 36:727–31. 9. Serra-Aracil X, Garcı´a-Domingo MI, Pare´s D, et al. Surgical site infection in elective operations for colorectal cancer after the application of preventive measures. Arch Surg 2011;146:606–12.

The American Journal of Surgery, Vol 209, No 2, February 2015 10. de Campos-Lobato LF, Wells B, Wick E, et al. Predicting organ space surgical site infection with a nomogram. J Gastrointest Surg 2009;13: 1986–92. 11. Gervaz P, Bandiera-Clerc C, Buchs NC, et al. Scoring system to predict the risk of surgical-site infection after colorectal resection. Br J Surg 2012;99:589–95. 12. Ho VP, Stein SL, Trencheva K, et al. Differing risk factors for incisional and organ/space surgical site infections following abdominal colorectal surgery. Dis Colon Rectum 2011;54:818–25. 13. Kiran RP, El-Gazzaz GH, Vogel JD, et al. Laparoscopic approach significantly reduces surgical site infections after colorectal surgery: data from national surgical quality improvement program. J Am Coll Surg 2010;211:232–8. 14. Wick EC, Gibbs L, Indorf LA, et al. Implementation of quality measures to reduce surgical site infection in colorectal patients. Dis Colon Rectum 2008;51:1004–9. 15. Mehrotra A, Damberg CL, Sorbero ME, et al. Pay for performance in the hospital setting: what is the state of the evidence? Am J Med Qual 2009;24:19–28. 16. Lindenauer PK, Remus D, Roman S, et al. Public reporting and pay for performance in hospital quality improvement. N Engl J Med 2007;356: 486–96. 17. Jancin B. Medicare may drop payments for SSI. General Surgery News 2007;3:1. 18. Lawson EH, Ko CY. Preventing unintended consequences of quality measurement: Comment on ‘‘Surgical site infections and cost in obese patients undergoing colorectal surgery.’’ Arch Surg 2011;146:1072–3. 19. Casalino LP, Elster A, Eisenberg A, et al. Will pay-for-performance and quality reporting affect health care disparities? Health Aff (Millwood) 2007;26:w405–14. 20. Fazio VW, Tekkis PP, Remzi F, et al. Assessment of operative risk in colorectal cancer surgery: the Cleveland Clinic Foundation colorectal cancer model. Dis Colon Rectum 2004;47:2015–24. 21. Khuri SF, Henderson WG, Daley J, et al, Principal Investigators of the Patient Safety in Surgery Study. Successful implementation of the department of Veterans Affairs’ national surgical quality improvement program in the private sector: the patient Safety in surgery study. Ann Surg 2008;248:329–36. 22. Khuri SF, Daley J, Henderson W, et al. The Department of Veterans Affairs’ NSQIP: the first national, validated, outcome-based, riskadjusted, and peer-controlled program for the measurement and enhancement of the quality of surgical care. National VA Surgical Quality Improvement Program. Ann Surg 1998;228:491–507. 23. Hall BL, Hamilton BH, Richards K, et al. Does surgical quality improve in the American College of Surgeons National Surgical Quality Improvement Program: an evaluation of all participating hospitals. Ann Surg 2009;250:363–76. 24. Dimick JB, Weeks WB, Karia RJ, et al. Who pays for poor surgical quality? Building a business case for quality improvement. J Am Coll Surg 2006;202:933–7. 25. American College of Surgeons. ACS NSQIP User Guide for the Participant Use Data File 2005–2010. Available at, http://site.acsnsqip.org/ participant–use–data-file/. Accessed July 22, 2012. 26. Akaike H. A new look at the statistical model identification. IEEE Trans Automatic Control 1974;19:716–23. 27. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143: 29–36. 28. Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. Wiley Series in Probability and Statistics. New York: Wiley; 2000. 29. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology 2010;21:128–38. 30. Kramer AA, Zimmerman JE. Assessing the calibration of mortality benchmarks in critical care: the Hosmer-Lemeshow test revisited. Crit Care Med 2007;35:2052–6. 31. Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007;115:928–35.

A.Y. Kohut et al.

Adverse events in colorectal surgery

32. Merkow RP, Hall BL, Cohen ME, et al. Relevance of the c-statistic when evaluating risk-adjustment models in surgery. J Am Coll Surg 2012;214:822–30. 33. R Development Core Team. A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2012. 34. Hawn MT, Houston TK, Campagna EJ, et al. The attributable risk of smoking on surgical complications. Ann Surg 2011;254:914–20.

229 35. Silverstein P. Smoking and wound healing. Am J Med 1992;93:22S–4S. 36. Fredholm BB. Adenosine, an endogenous distress signal, modulates tissue damage and repair. Cell Death Differ 2007;14:1315–23. 37. De Boer WI. Cytokines and therapy in COPD: a promising combination? Chest 2002;121(5 Suppl):209S–18S. 38. Fujii T, Tsutsumi S, Matsumoto A, et al. Thickness of subcutaneous fat as a strong risk factor for wound infections in elective colorectal surgery: impact of prediction using preoperative CT. Dig Surg 2010;27:331–5.

Patient-specific risk factors are predictive for postoperative adverse events in colorectal surgery: an American College of Surgeons National Surgical Quality Improvement Program-based analysis.

Pay-for-performance measures incorporate surgical site infection rates into reimbursement algorithms without accounting for patient-specific risk fact...
2MB Sizes 0 Downloads 4 Views

Recommend Documents