Elise H. Lawson, MD, MSHS, discusses Risk Factors for Superficial vs Deep/Organ-Space Surgical Site Infections: Implications for Quality Improvement Initiatives.
eTable. Risk-adjusted odds ratios for SSI among patients undergoing colectomy procedures in 2011.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Lawson EH, Hall BL, Ko CY. Risk Factors for Superficial vs Deep/Organ-Space Surgical Site Infections: Implications for Quality Improvement Initiatives. JAMA Surg. 2013;148(9):849–858. doi:10.1001/jamasurg.2013.2925
Surgical site infections (SSIs) are the focus of numerous quality improvement initiatives because they are a common and costly cause of potentially preventable patient morbidity. Superficial and deep/organ-space SSIs differ in terms of anatomical location and clinical severity.
To identify risk factors that are uniquely predictive of superficial vs deep/organ-space SSIs occurring after colectomy procedures.
Retrospective cohort study.
American College of Surgeons National Surgical Quality Improvement Program.
Patients undergoing colectomy procedures in 2011 were identified by Current Procedural Terminology codes.
Main Outcomes and Measures
We compared rates of superficial SSI and deep/organ-space SSI associated with perioperative variables of interest: demographics; preoperative clinical severity, risk factors, and comorbidities and variables related to the hospitalization or procedure. Hierarchical multivariable logistic regression models were developed to identify risk-adjusted predictors of each SSI type.
Among 27 011 patients identified from 305 hospitals, 6.2% developed a superficial SSI and 4.7% developed a deep/organ-space SSI. Risk factors common to the occurrence of both SSI types were identified: open surgery (vs laparoscopic) and current smoker. Risk factors with differential effects on each SSI type included specific postoperative diagnoses, disseminated cancer, and irradiation therapy, which were all associated with increased odds of deep/organ-space SSI only. The graded relationship between increasing body mass index and SSI occurrence appeared to be stronger for superficial SSI.
Conclusions and Relevance
Risk factors for superficial SSI and deep/organ-space SSI vary in terms of magnitude and significance, suggesting that these SSI types are somewhat different disease processes. Groups interested in preventing SSIs might improve success by considering these SSI types independently for root-cause analyses and development of best practices and interventions.
Surgical site infections (SSIs) are common and costly postoperative complications.1-7 This persistent patient morbidity is particularly common for colorectal procedures, with reported rates ranging from 5% to 26%.8-11 The additional cost attributed to SSI is estimated to be between $10 4436 and $25 546 per infection7 and patients with an SSI have been shown to have a longer length of hospital stay, reduced health-related quality of life, and higher mortality.1-5
The Centers for Disease Control and Prevention and the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) each divide SSI into 3 distinct types based on the anatomic level of the infection: superficial incisional, deep incisional, and organ space. Superficial SSIs are more prevalent than deep/organ-space SSIs but generally require less aggressive interventions. In contrast, deep/organ-space SSIs often require invasive management with intravenous antibiotics, percutaneous drainage procedures, and reoperations. Despite these clinical differences, research studies and quality improvement initiatives often group these SSI types together to increase the overall event rate, with the hope that this will increase the likelihood of detecting a significant change in performance after an intervention. However, this aggregation of potentially different disease processes could actually have negative implications for quality measurement or improvement if the clinical processes of care that lead to superficial SSI are different than the processes that lead to deep or organ-space SSI.
The objective of this study was to characterize and compare superficial and deep/organ-space SSIs occurring after colectomy procedures. Our goal was to elucidate whether quality improvement initiatives should focus on the overall postoperative SSI rate or whether they would be better served by considering each type of SSI as an independent postoperative complication. Our hypothesis was that superficial and deep/organ-space SSIs represent different disease processes and as such will have differing perioperative profiles in terms of risk factors and patient populations affected.
Quiz Ref IDThe ACS-NSQIP is a validated, institution-based, multispecialty surgical registry of patient risk factors and 30-day postoperative outcomes. Hospital participation in ACS-NSQIP is voluntary but requires employment of a dedicated data abstractor who is trained to use strict variable definitions and collection methods.12,13 As previously described, the sampling strategy consists of collecting data for the first 40 cases performed within consecutive 8-day cycles. General surgery, vascular surgery, and specific surgical subspecialties are included in the sample. Data are abstracted from medical records and direct communication with the patient. Data collected include demographics, preoperative risk factors, operative information, and postoperative outcomes within 30 days of the index operation. Hospitals are audited to ensure standardized data collection, with audit results in the past demonstrating substantial or almost perfect agreement on the coding of most variables.14 Information in the database is deidentified. Participating hospitals receive semiannual reports with risk-adjusted outcomes from ACS-NSQIP that allow them to benchmark their performance with national averages.12,13
Patients who underwent a colectomy procedure in 2011 were identified by Current Procedural Terminology (CPT) code.15 We chose to focus on colectomy procedures because they are frequently performed and have relatively high reported rates of SSI compared with other common procedures, resulting in a high societal burden.16 The CPT codes included in this study are listed in Table 1 and are consistent with the codes used by the ACS-NSQIP for quality measure modeling. This group of CPT codes was also used in the development of the ACS Colectomy Composite Outcome Quality Measure, which was approved by the National Quality Forum for use as a quality measure.17 Each case was assigned as either a laparoscopic or open procedure based on CPT code and to 1 of 10 postoperative diagnosis groups (benign neoplasm, malignant neoplasm, diverticulitis, ulcerative colitis/Crohn disease, obstruction/perforation, vascular insufficiency, fistula, infectious colitis/peritonitis, hemorrhage, and other) based on International Classification of Diseases, Ninth Revision code.
Quiz Ref IDThe primary outcomes of interest were 30-day postoperative (1) superficial SSI; (2) deep and/or organ-space SSI; and (3) any occurrence of SSI (superficial and/or deep and/or organ space). The ACS-NSQIP defines superficial SSI as infections involving only the skin or subcutaneous tissue of the incision. Infections that extend into the fascial and muscle layers are thus not considered to be superficial. Deep SSI is defined as infections involving the deep soft tissues (eg, fascial and muscle layers) of the incision, while organ-space SSI includes infections involving any part of the anatomy that was opened or manipulated during an operation (other than the incision). Each SSI type has specific inclusion and exclusion criteria, which are detailed in the ACS-NSQIP operations manual. We chose to group deep and organ-space infections together because infections that originate as organ space and drain through the incision are by protocol labeled as deep SSI. This classification is consistent with Centers for Disease Control and Prevention guidelines.18 The ACS-NSQIP has implemented training and audit procedures for data abstractors to ensure consistent collection of robust data using these strict definitions.14 For this study, patients with both a superficial SSI and a deep/organ-space SSI were included only in the deep/organ-space SSI group (n = 79 patients, 0.29% of study population).
All data preparation and analyses were performed using SAS version 9.2 software (SAS Institute Inc). Missing values for body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), American Society of Anesthesiologists class, and functional status were imputed using the hot-deck method. Superficial and deep/organ-space SSI rates for patients with various variables of interest were calculated and compared. χ2 Tests and t tests were performed for categorical and quantitative variables, respectively.
We used SAS PROC GLIMMIX to develop hierarchical multivariable logistic models. This approach accounts for clustering of patients within hospitals by allowing each hospital to have a different random intercept and incorporates the empirical Bayes method. Models were developed for the following outcomes: (1) any SSI vs no SSI; (2a) superficial SSI vs no superficial SSI; (2b) deep/organ-space SSI vs no deep/organ-space SSI; (3a) superficial SSI vs no SSI (patients with deep/organ-space SSI excluded); (3b) deep/organ-space SSI vs no SSI (patients with superficial SSI excluded); and (4) deep/organ-space SSI vs superficial SSI (patients with no SSI excluded). The aim of model 1 was to identify risk factors with a common impact for superficial and deep/organ-space SSI, while the aim for models 2a and 2b was to identify risk factors that are uniquely predictive of superficial or deep/organ-space SSI. As a form of sensitivity analysis, models 3a and 3b compared patients with 1 SSI type vs patients with no SSI, excluding patients with the other SSI type. The results of these models were very similar to models 2a and 2b (eTable in the Supplement). Finally, the aim of model 4 was to determine if there are significant predictors for whether a patient will develop a superficial vs deep/organ-space SSI, conditional on an SSI of some type occurring (ie, patients with no SSI occurrence were excluded).
Covariates included in these models included procedure type, postoperative diagnosis group, demographics (age, sex, race, Hispanic ethnicity, and admission source), and preoperative health status, risk factors, and comorbidities (BMI, functional status prior to surgery, current smoker, alcohol use, renal failure, diabetes, dyspnea, chronic obstructive pulmonary disease, pneumonia, steroid use, bleeding disorder, congestive heart failure, hypertension requiring medication, myocardial infarction in previous 6 months, disseminated cancer, >10% weight loss in previous 6 months, recent chemotherapy, recent irradiation therapy, preoperative sepsis, preoperative blood transfusion, American Society of Anesthesiologists class, wound class, and emergency procedure). Laboratory values were not included as covariates because of frequent missing values.
As a form of sensitivity analysis, an indicator variable was created for colectomy CPT codes that include the descriptor “low pelvic anastomosis” (Table 1). When included in the earlier-described models, this variable did not reach statistical significance and did not substantially change the results of the models. Additionally, we created an indicator variable for the interaction between the variables for “low pelvic anastomosis” and “irradiation”; however, this interaction term did not reach statistical significance when included in the models nor substantially change the results of the models. Finally, we repeated our analyses after excluding patients with a “low pelvic anastomosis” CPT code; however, this did not substantially change the conclusions of our study.
Operation duration was converted into a categorical variable by ranking operation durations and splitting them into equal deciles. The relationship between this intermediate outcome and SSI was evaluated by adding the categorical operation time variable into the earlier-described models for any SSI (model 1), superficial SSI (model 2a), and deep/organ-space SSI (model 2b), using the lowest operation time decile (<77 minutes) as the reference category
We identified 27 011 colectomy cases performed in 2011 and entered into the ACS-NSQIP database from 305 hospitals. The mean (SD) number of cases per hospital was 89 (71) and the median was 70. Overall, 1672 patients developed a superficial SSI (6.2% of study population) and 1271 developed a deep/organ-space SSI (4.7% of study population). The median number of days until SSI occurrence was 9 (mean [SD] 11.2 [6.3]) for superficial and 10 (mean [SD] 12.2 [6.7]) for deep/organ-space SSI.
Table 2 displays 30-day postoperative SSI rates for colectomy patients with specific demographic and preoperative risk factor variables. Compared with open procedures, laparoscopic colectomy was associated with significantly lower rates of superficial SSI (7.12% vs 4.88%; P < .05) and deep/organ-space SSI (5.66% vs 3.36%; P < .05). The postoperative diagnoses with the highest associated superficial SSI rates were hemorrhage (9.56%), fistula (7.94%), and obstruction/perforation (6.76%). Infectious colitis/peritonitis and vascular insufficiency had the lowest associated superficial SSI rates (3.15% and 3.06%, respectively). In contrast, the postoperative diagnoses with the highest associated deep/organ-space SSI rates were fistula (7.94%), ulcerative colitis/Crohn disease (7.66%), and obstruction/perforation (5.59%). Benign neoplasm and infectious colitis/peritonitis had the lowest associated rates of deep/organ-space SSI (3.21% and 3.65%, respectively).
There was a significant difference in both superficial SSI and deep/organ-space SSI rates by age group, with older age groups tending to have lower associated SSI rates (P < .05). Similarly, when comparing BMI categories, patients with a normal BMI (18.5-24.9) had significantly lower superficial and deep/organ-space SSI rates (4.91% and 4.31%, respectively) than patients in higher BMI categories. Of patients with class II and III obesity (BMI 35-39.9 and ≥40, respectively), 9.34% and 8.98% developed a superficial SSI and 5.27% and 6.34% developed a deep/organ-space SSI, respectively. Notably, patients with a totally dependent functional status prior to surgery had significantly lower SSI rates than patients who were partially dependent or independent prior to surgery. Patients who were smokers within the year prior to surgery had significantly higher rates of superficial and deep/organ-space SSI (7.36% and 5.72%, respectively).
Variables with significantly different associated superficial SSI rates but no difference in deep/organ-space SSI rates included renal failure (2.9%), insulin-dependent diabetes (8.11%, compared with 6.06% for nondiabetic patients and 6.30% for diabetic patients taking oral medications), chronic obstructive pulmonary disease (7.7%), and hypertension requiring medication (6.48%). Conversely, variables with significantly different associated deep/organ-space SSI rates but no difference in superficial SSI rate included male sex (5.21%), recent steroid use (6.86%), disseminated cancer (8.26%), more than 10% weight loss in prior 6 months (6.21%), recent irradiation therapy (7.89%), and preoperative blood transfusion (5.94%).
The mean operation time for the study population was 163.5 minutes (Table 3). Patients with a deep/organ-space SSI had the highest mean operation time (199.7 minutes), followed by patients with a superficial SSI (179.5 minutes). Patients with no SSI had a mean operation time of 160.5 minutes. Similarly, the median postoperative length of stay was 6 days for all colectomy patients as a group but 7 days for patients with a superficial SSI and 11 days for patients with a deep/organ-space SSI (Table 3).
Risk-adjusted odds ratios (ORs) for models 1 (any SSI), 2a (superficial SSI), and 2b (deep/organ-space SSI) are displayed in Table 4. Congestive heart failure and preoperative blood transfusion were the only variables that reached statistical significance in the any SSI model but not in either the superficial or deep/organ-space SSI models. Laparoscopic procedures were associated with a significantly lower risk-adjusted odds of both superficial SSI and deep/organ-space SSI (OR, 0.599 and 0.597; both P < .001) compared with open procedures, and this effect was also observed in the any SSI model (OR, 0.576; P < .001). Similarly, smokers had a significantly higher risk-adjusted odds of both superficial and deep/organ-space SSI (OR, 1.231 and 1.194; P = .002 and .02, respectively) and this effect was preserved in the any SSI model (OR, 1.229; P < .001). Older age categories were associated with significantly lower risk-adjusted odds of superficial, deep/organ-space, and any SSI compared with patients younger than 55 years. Finally, compared with wound class II (clean-contaminated wound), wound class III (contaminated wound) was associated with significantly higher risk-adjusted odds of superficial, deep/organ-space, and any SSI (OR, 1.195, 1.272, and 1.251; P = .02, .005, and <.001, respectively), while wound class IV (dirty-infected wound) was associated with significantly lower risk-adjusted odds of these SSI types (OR, 0.704, 0.682, and 0.672; P = .001, .001, and <.001, respectively).
Using benign neoplasm as the base category for postoperative diagnosis group, only vascular insufficiency had a significant (though reduced) risk-adjusted OR for superficial SSI (0.491; P = .007), while ulcerative colitis/Crohn disease, obstruction/perforation, fistula, and “other diagnosis” were all associated with significantly higher risk-adjusted odds of deep/organ-space SSI (OR, 1.752, 1.440, 2.066, and 1.370; P < .001, P = .02, P < .001, and P = .04, respectively). These observations were attenuated in the any SSI model, with only obstruction/perforation and fistula having significantly different (higher) odds of SSI compared with benign neoplasm (OR, 1.248 and 1.497; P = ..03 and .003, respectively).
For superficial SSI, all increased BMI categories had an increased risk-adjusted odds of SSI compared with patients with normal BMI, while underweight patients had a significantly decreased odds of superficial SSI. For deep/organ-space SSI, only obesity class II and III were associated with significantly greater odds of SSI compared with patients with normal BMI, and the ORs for SSI were of lesser magnitude than those observed in the superficial SSI model. In the any SSI model, all elevated BMI categories had a significantly higher risk-adjusted odds of SSI compared with patients with a normal BMI. The ORs observed in the any SSI model were of greater magnitude than those observed in the deep/organ-space SSI model and of lesser magnitude than those observed in the superficial SSI model.
Disseminated cancer and recent irradiation therapy were associated with significantly increased risk-adjusted odds of deep/organ-space SSI (OR, 1.664 and 2.514; P < .001 and P = .003), but not superficial SSI. These relationships were attenuated in the any SSI model, with disseminated cancer associated with a lower odds of any SSI compared with deep/organ-space SSI (OR, 1.275; P = .004) and the OR for irradiation therapy no longer reaching significance. Similarly, chronic obstructive pulmonary disease was associated with a significantly increased risk-adjusted odds of superficial SSI (OR, 1.310; P = .01) but not deep/organ-space SSI or any SSI.
Model 4 identified significant predictors of whether a patient would develop a superficial vs deep/organ-space SSI, conditional on an SSI of some type occurring (ie, patients with no SSI occurrence excluded). For postoperative diagnosis, ulcerative colitis/Crohn disease, vascular insufficiency, and fistula were all associated with significantly increased odds of deep/organ-space SSI compared with the benign neoplasm category (OR, 1.560, 2.228, and 1.748; P = .049, .03, and .03, respectively). Other variables associated with a significantly increased odds of deep/organ-space SSI include male sex (OR, 1.243), disseminated cancer (OR, 1.771), recent irradiation therapy (OR, 3.250), and preoperative septic shock (OR, 2.981). Variables associated with significantly decreased odds of deep/organ-space SSI (meaning increased odds of superficial SSI) included obesity categories I and II (OR, 0.774 and 0.707, respectively) and chronic obstructive pulmonary disease (OR, 0.666).
There was a graded relationship between operation time decile and SSI for each SSI type, with higher deciles having higher risk-adjusted odds of SSI (Table 5). The raw rates of superficial and deep/organ-space SSI increased from 4.4% and 3.1% in the lowest operation time decile (<77 minutes) to 8.5% (P < .001) and 9.5% (P < .001) in the highest decile (>271 minutes), respectively. In the longest 3 deciles of duration, the increased odds of deep/organ-space SSI were of substantially greater magnitude than the increased odds of superficial SSI.
Reducing the occurrence of SSIs is the focus of numerous quality improvement initiatives because SSIs are a common and costly cause of potentially preventable patient morbidity. Many studies have investigated the causes and risk factors for SSI. However, to our knowledge, ours is the first to use a large national clinical database to compare patient risk factors associated with the occurrence of a superficial SSI vs a deep/organ-space SSI. Quiz Ref IDOn univariate and multivariable analysis, we found risk factors common to the occurrence of both SSI types, such as an open surgical approach (as opposed to laparoscopic) and current smoker status.Quiz Ref IDHowever, we also identified risk factors with differential effects on each SSI type. For example, the postoperative diagnoses ulcerative colitis/Crohn disease, obstruction/perforation, and fistula were each associated with a significantly increased risk-adjusted odds of developing a deep/organ-space SSI (relative to the diagnosis benign neoplasm) but were not associated with an increased odds of superficial SSI. Similarly, disseminated cancer and recent irradiation therapy increased the risk-adjusted odds of deep/organ-space SSI, but not of superficial SSI. While elevated BMI increased the odds of each SSI type, the graded relationship between increasing BMI category and SSI occurrence appeared to be stronger for superficial SSI.
Quiz Ref IDIn addition, we observed a positive graded relationship between increasing operation duration and odds of superficial or deep/organ-space SSI occurrence, even after risk adjusting with 30 other patient factors. In the longest duration deciles, the observed effect of duration was more pronounced for deep/organ-space SSIs than for superficial SSIs. The interpretation of the operation duration effect is challenging. A long operation could conceivably increase a patient’s risk of developing a postoperative SSI. However, a lengthy operation could also be the result of a predisposing confounding factor or an intraoperative complication, each of which could increase a patient’s risk of SSI. In the latter case, operation duration would merely be an intervening intermediate outcome. For the purpose of this study, the relationships between operation duration and superficial vs deep/organ-space SSI differ.
Most studies on colorectal SSI aggregate superficial, deep, and organ-space SSI together as 1 group when examining the effect of different risk factors. An exception to this is a single-institution study that compared risk factors for incisional vs organ-space infections in 428 patients undergoing colorectal procedures.19 On multivariable analysis, Blumetti et al found that BMI and creation/revision/reversal of an ostomy were independently associated with incisional SSI, while perioperative transfusion and previous abdominal surgery were independently associated with organ-space SSI. Our study builds on these findings by examining a greater number of risk factors and including data from 305 hospitals nationwide.
Quality improvement initiatives aimed at reducing SSI rates are often hindered by limited or even conflicting evidence for proposed interventions to reduce SSI.20 For example, the debate regarding the risk-benefit balance of mechanical and oral antibiotic bowel preparation prior to colorectal procedures is ongoing, with conflicting data supporting each side.21,22 Even interventions based on published evidence can produce disappointing results. For example, the Surgical Care Improvement Project is a national initiative aimed at reducing postoperative SSIs by focusing on a series of evidence-based processes of care such as perioperative prophylactic antibiotic administration, skin (hair) clipping, and normothermia. However, despite evidence supporting the importance of these processes, high compliance is only weakly linked to improved outcomes, at best.23-25
Two approaches for successfully reducing postoperative SSIs have recently been described: Lean Six Sigma and implementation of a surgical Comprehensive Unit-based Safety Program (CUSP). Lean Six Sigma is a quality improvement method that uses a multidisciplinary team and a systematic data-driven approach to identify problems in care and develop targeted solutions. This approach was used in a recent multi-institutional collaborative project aimed at reducing the occurrence of SSIs after colorectal procedures. Through this project, which was co-led by the Joint Commission Center for Transforming Healthcare and the ACS, participating hospitals reduced their combined superficial SSI rate by 45% (P < .001).26 Though the project aimed to reduce both superficial and deep/organ-space SSIs, the collaborative found that the factors identified as contributing to the occurrence of superficial SSI differed from those that contributed to deep/organ-space SSI and, as such, different targeted solutions were needed for each SSI type. The Mayo Clinic Rochester Methodist Hospital, a participant in the collaborative, recently published a report detailing the successful experience of developing and implementing a colorectal SSI reduction bundle using this approach.27
The CUSP approach to quality improvement focuses on engaging an interdisciplinary team to implement a safety culture, which includes principles of safe design such as standardization, independent checks for important processes, and learning from mistakes. A team at Johns Hopkins used this approach to develop and implement a bundle of interventions that successfully reduced the colorectal SSI rate by 33.3% (P < .05).28 Because of their success, and the success of the CUSP approach in other clinical areas, the program is currently being expanded to colon surgery programs at more than 100 US hospitals.
Our study has limitations. First, superficial SSIs may be undercoded if hospitals record only the more severe SSI type in patients who develop multiple SSIs. In our data set, 0.29% of the study population had both SSI types recorded and we chose to count these patients only in the deep/organ-space SSI group. Second, we grouped deep and organ-space SSIs because we believe they represent intra-abdominal disease processes with similar clinical implications. Conceivably, the 3 types of SSI might all have very different clinical ramifications and combining any of them might not accurately reflect the complex etiology of SSI. Third, wounds that were left open at the time of surgery are generally not considered eligible for developing a superficial SSI, a practice that could interfere with these analyses. Fourth, ACS-NSQIP does not currently capture procedure-specific preoperative risk factors and postoperative outcomes such as anastomotic leak, which could be relevant to this study. Finally, ACS-NSQIP hospitals are predominantly medium and large academic medical centers; thus, our findings may not be universally generalizable.
In conclusion, we found that risk factors for superficial SSI and deep/organ-space SSI vary in terms of magnitude and significance, suggesting that these SSI types are somewhat different disease processes. Hospitals interested in reducing their SSI rates for patients undergoing colectomy procedures may thus benefit from investigating these SSI types independently when doing root cause analyses and should consider interventions targeted specifically at superficial or deep/organ-space SSI.
Corresponding Author: Elise H. Lawson, MD, MSHS, David Geffen School of Medicine at UCLA, Department of Surgery, 10833 Le Conte Ave, 72-215 CHS, Los Angeles, CA 90095 (email@example.com)
Accepted for Publication: April 19, 2013.
Published Online: July 17, 2013. doi:10.1001/jamasurg.2013.2925.
Author Contributions: Dr Lawson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition of data: Ko.
Analysis and interpretation of data: All authors.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: Hall, Ko.
Statistical analysis: Lawson, Hall.
Study supervision: Hall, Ko.
Conflict of Interest Disclosures: None reported.
Funding/Support: Dr Lawson’s time was supported by VA Health Services Research and Development program grant RWJ 65-020 and the ACS through the Robert Wood Johnson Foundation Clinical Scholars Program. Dr Hall is a consultant for the ACS and Dr Ko is staff at the ACS.
Previous Presentation: This study was presented at the 84th Annual Meeting of the Pacific Coast Surgical Association; February 17, 2013; Kauai, Hawaii, and is published after peer review and revision.