Abstract
Background:
Post-operative infectious complications after repair of intestinal-cutaneous fistulas (ICF) represent a substantial burden and these outcomes vary widely in the literature. We aimed to evaluate the use of the modified frailty index-5 (mFI-5) to account for physiologic reserve to predict infectious complications in patients with ICF undergoing operative repair.
Methods:
We used the American College of Surgeon National Surgical Quality Improvement Program (ACS-NSQIP) 2006–2017 dataset to include patients who underwent ICF repair. The main outcome measure was 30-day infectious complications (surgical site infection [SSI], sepsis, pneumonia, and urinary tract infection [UTI]). The risk of 30-day post-operative infectious complications was assessed based on mFI-5 score. We performed multivariable logistic regression analyses to evaluate the association between infectious complications and mFI-5.
Results:
We identified 4,197 patients who underwent an ICF repair. The median age (interquartile range [IQR]) was 57 (46, 67) years, and the majority of patients were female (2,260; 53.9%); white (3,348; 79.8%); and 1,586 (38.3%) were obese. After adjustment for relevant confounders such as baseline patient characteristics, and operative details, mFI-5 was independently associated with infectious complications (odds ratio [OR], 2.00; 95% confidence interval [CI], 1.25–3.21), particularly SSI (OR, 2.16; 95% CI, 1.28–3.63) and pneumonia (OR, 5.31; 95% CI, 2.29–12.35), but not UTI or sepsis.
Conclusions:
We showed that the mFI-5 is a strong predictor of infectious complications after ICF repair. It can be utilized to account for physiologic reserve, therefore reducing the variability of outcomes reported for ICF repair.
Post-operative complications frequently burden patients who have intestinal-cutaneous fistulas (ICF) requiring surgical repair, with more than 80% experiencing at least one complication [1–3]. Infectious complications are a major contributor to these high morbidity rates, with 36% of patients developing an infectious complication [4]. Additionally, infectious complications in patients with ICF have been shown to predict worse outcomes including delayed spontaneous closure and death [5]. With the advent of total parenteral nutrition, sepsis is now considered the most common cause of death in patients with ICF [2,6]. Infectious complications can be more serious in patients with ICF with poor baseline physiologic and functional reserve [7–10]. However, there is a wide variation in the reported outcomes of infectious complications after ICF repair. The fact that physiologic reserve has not been accounted for in these studies could be a contributing factor to the wide variation in outcomes. Therefore, accounting for physiologic reserve in the operative repair of ICFs is needed, which is essential to reducing the variation in infectious outcomes in this population.
Frailty is well-known concept and is defined as age-related decline in functional reserve that renders patients susceptible to stressors [11]. Frailty has been associated with worse outcomes in both ambulatory and hospitalized patients [12–15]. Several indexes have been described in the literature to measure frailty. For example, the modified frailty index 11-item (mFI-11) is an index derived from the American College of Surgeon National Surgical Quality Improvement Program (ACS-NSQIP) database and modified from the Canada Study of Health and Aging frailty index (CSHA-FI) [16,17]. Both of these indexes are robust predictors of morbidity and mortality after different surgical operations [18–21]. Because some variables have been dropped from the ACS-NSQIP database, a modified version of mFI-11 has been described recently based on five parameters that remained in the ACS-NSQIP database [16]. Evidence regarding the predictive capacity of the mFI-5 is increasing [22,23]. However, there are no published studies evaluating the use of mFI-5 to predict infectious complications after ICF surgical repair.
Our study hypothesis is that the mFI-5 is a robust tool to adjust for physiologic reserve in patients undergoing operative repair of ICFs and adequately predicts post-operative infectious complications. We sought to evaluate the ability of the mFI-5 index in predicting post-operative infectious complications in patients with ICF after operative repair.
Methods
Study population, setting, and design
We queried the ACS-NSQIP database to include adult patients (≥18 years old) who had surgical repair of ICF between 2006 and 2017. We defined ICF as an abnormal communication between the skin and the small or large intestine. We used the Current Procedural Terminology (CPT) code 44640 to identify ICF repair (including patients undergoing ICF repair as a primary or secondary procedure). Furthermore, we used the following CPT codes to identify small bowel resection (44110, 44120, 44121, 44125, 44130, 44202, 44203); ileostomy/jejunostomy (44310, 44312, 44314); enterorrhaphy (44602, 44603); colectomy (44139, 44140, 44141, 44143: 44144: 44145, 44146, 44147, 44150, 44155, 44160, 44204, 44205, 44206, 44207, 44208, 44210, 44213, 44227); colostomy (44345, 44346); intra-abdominal abscess drainage (49020, 49021, 49060); repair of ventral hernia (49560, 49561, 49565, 49566, 49568, 49654, 49655); myocutaneous/fasciocutaneous flaps (15734); and abdominal wall debridement (11005, 11008).
Exclusion criteria were patients with any other type of fistula, including colovesical fistula, vesicovaginal fistula, rectovaginal fistula, and perianal fistula. Patients were also excluded if they had other procedures along with ICF repair to limit post-operative complications from non-ICF surgery. These operations were any gastric procedure except for gastrostomy tube; repair or resection of duodenum; any hepatobiliary procedure except for cholecystectomy and liver biopsy; any pancreatic procedure; any repair of the bladder; excision of any tumor; any procedure of the anus; resection of rectum; creation, resection, or revision of ileoanal reservoir.
Study variables and outcomes
The mFI-5 was based on the following conditions functional dependence at baseline, chronic obstructive pulmonary disease (COPD), hypertension (HTN), congestive heart failure (CHF), and diabetes mellitus. Functional dependence is defined by ACS-NSQIP as a “patient's partial or total assistance from another person to perform activities of daily living 30 days prior to surgery, including but not limited to dressing, bathing, toileting, eating, and ambulating” [24]. The mFI-5 was estimated by multiplying number of variables present by 0.2, where each component of the mFI-5 represents a value of 0.2. The maximum mFI-5 score is 1 (0.2 multiplied by 5). If a patient had a missing variable of the mFI-5, this variable is not included in the denominator.
The study outcome measure was 30-day infectious complications, defined as the occurrence of any of the following complications: SSI infection (including superficial, deep, and organ/space SSI as well as wound dehiscence), sepsis, UTI, and pneumonia.
Statistical analysis
Univariable analysis and multivariable logistic regression analyses were performed to evaluate predictors of post-operative infectious complications. For univariable analysis, χ2 test and Fisher exact tests were performed to compare categorical variables as appropriate. For continuous variables, t-test or Wilcoxon rank sum tests were used as appropriate to compare patient mean difference or mean ranks, respectively. For multivariable logistic regression analyses, covariates were selected if their p value was below 0.2 in the univariable analysis or deemed clinically significant. Statistical analyses were conducted using StataCorp 2017 (Stata Statistical Software: Release 15; College Station, TX: StataCorp LLC).
Results
Patient characteristics
We included a total of 4,197 patients who had an ICF repair. A total of 1,014 (24.2%) had an operation of the small intestine only and 768 (18.3%) of the large intestine only (Fig. 1). The median age (interquartile range [IQR]) was 57 (46, 67) years; more than half of the patients were females (2,260; 53.9%) and white (3,348; 79.8%), and 1,586 (38.3%) of the cohort were obese (Table 1). Higher mFI-5 was associated with advanced age. For example, the median (IQR) age of patients with a mFI-5 of 0 was 50 (38, 61) years versus 67 (58, 76) years in those with a mFI-5 of 1.0 (p < 0.001; Table 2).
Data cleaning flowchart.
Univariate Analysis of Potential Predictors of Post-Operative Infectious Complications after Intestinal-Cutaneous Fistula Surgical Repair
IQR = interquartile range; BMI = body mass index; ASA = American Society of Anesthesiologists; DM = diabetes mellitus; HTN = hypertension; COPD = chronic obstructive pulmonary disease; CHF = congestive heart failure; AKI = acute kidney injury.
Univariate Analysis of Demographics and Comorbidities across Different Modified Frailty Index Scores in Intestinal-Cutaneous Fistula Surgical Repair Cohort
mFI-5 = modified frailty index-5; IQR = interquartile range; DM = diabetes mellitus; HTN = hypertension; COPD = chronic pulmonary disease; CHF = congestive heart failure
Outcome and mFI-5
Approximately one-third of the patients (n = 1,416; 33.7%) developed infectious complications. A higher mFI-5 was associated with a higher risk of infectious complications (p < 0.001; Table 2 and Fig.2). For example, infectious complications were present in 609 patients (30.5%) with a mFI-5 of 0 versus 236 patients (38.3%) with mFI-5 of 0.4. All patients with a mFI-5 of 1.0 developed post-operative infectious complications (Table 2).
Association between the modified frailty index-5 score and post-operative infectious complications in intestinal-cutaneous fistula repair cohort.
Predictors of infectious complications
Patients with infectious complications were more likely to be obese, have high American Society of Anesthesiologists (ASA) classes (III–V) compared with patients who did not develop such complications (Table 1). They were also more likely to have pre-existing comorbidities such as diabetes mellitus, severe COPD, functional and ventilator dependence, bleeding disorders, chronic kidney disease, pre-operative wound infection/sepsis, pre-operative anemia, hypoalbuminemia, and a higher median mFI-5. Additionally, the infectious complications cohort was more likely to have a contaminated or dirty/infected surgical wound compared with the cohort that did not develop post-operative infectious complications. Finally, they were more likely to spend more time in the operating room, and to receive blood transfusions in the peri-operative period.
On multivariable analysis, the mFI-5 was independently predictive of post-operative infectious complications (OR, 2.00; 95% CI, 1.25–3.21; p = 0.004). Other independent predictors of infectious complications were pre-operative hypoalbuminemia, bleeding disorders, and operative time (Table 3).
Multivariable Logistic Regression Analysis for the Predictors of Post-Operative Infectious Complications after Intestinal-Cutaneous Fistula Repair
OR = odds ratio; CI = confidence interval; BMI = body mass index; ASA: American Society of Anesthesiologists; IQR = interquartile range; mFI-5 = modified frailty index-5.
In a subanalysis of the predictive power of the mFI-5 and individual infectious complications, the mFI-5 was a robust predictor of SSI (OR, 2.16; 95% CI, 1.28–3.63; p = 0.004) and pneumonia (OR, 5.31; 95% CI, 2.29–12.35; p < 0.001), but not UTI or sepsis (p > 0.05; Table 4).
Multivariable Logistic Regression Analyses for Different Post-Operative Infectious Complications with Modified Frailty Index-5 as One of the Independent Covariates
OR = odds ratio; CI = confidence interval.
Discussion
Infectious complications are common and have substantial impact on clinical outcomes after surgical management of ICFs. In this study, the mFI-5, a measure of physiologic reserve, was an independent predictor for post-operative infections after ICF surgical repair. More specifically, mFI-5 was independently predictive of SSI and pneumonia but not with post-operative sepsis. The combined effect of the four comorbidities and functional dependence represented by mFI-5 provided additional stratification of infectious complications that is not covered by other patient-related factors. This finding supports the idea that mFI-5 provides a unique perspective on outcomes after ICF repair that is introduced by frailty. Although there is a higher risk of infectious complications for increases in mFI-5 across its spectrum, there appears to be a breaking point in this relation at mFI-5 = 0.6 and infectious complication risk increases sharply past this point. Other independent predictors of infectious complications in our analysis were race (African American and American Indian/Alaska Native), history of bleeding disorder, hypoalbuminemia before surgery, and operation time.
Post-operative infections are among the most commonly encountered complications and sepsis is the most common cause of death during both the pre-operative and post-operative period [1,2]. However, similar to other post-operative complications, reports of post-operative infections vary widely between studies [2,8,25]. This variation may be caused by differences in fistula characteristics, patient-related factors such as physiologic reserve, or management strategies. The relation between mFI-5 and post-operative outcomes could provide a method to account for baseline physiologic reserve in patients undergoing ICF repair and standardize the reporting of outcomes across studies. A major advantage of the mFI-5 is that it uses easily accessible variables for calculation that makes it applicable to most medical databases.
Higher mFI was associated with a higher risk of pneumonia in patients who had emergency general surgery, as well as head and neck surgery [26,27]. Similarly, mFI-11 was an independent predictor of re-operation because of SSI after degenerative spine disease surgery [28]. The finding that mFI-5 predicts infectious complications, specifically pneumonia and SSI, has been shown in several unrelated procedures and this clearly suggests that frailty measured by mFI-5 represents systemic alterations rather than procedure-specific factors.
Other indices of frailty such as mFI-11, CSHA-FI, Charlson comorbidity index, Fried index, and FRAIL index have been evaluated in the past and shown to be reliable predictors of worse post-operative outcomes [29–34]. Although some of these indices are quite robust and may have the advantage of measuring functional limitation, allowing for a more accurate representation of a patient's physiologic reserve, this also makes them unpractical tools for studying outcomes of ICFs. Most of the requested variables are frequently not included in most patient charts and large datasets.
Our study is not without important limitations. The ACS-NSQIP dataset is not specific to ICFs, therefore it is possible that our analysis missed certain factors associated with outcomes after ICF repair, such as fistula characteristics. Our analysis lacks fistula-related variables that are known to affect outcomes after repair such as etiology of fistula, complexity of fistula (enterocutaneous vs. enteroatmospheric fistula, undrained collections, infected mesh, etc.), level of fistula (e.g., gastric, duodenal/jejunal/ileal, colonic), fistula output, and time between fistula formation and repair [25]. Therefore, mFI-5 should be interpreted as one of the elements in the multifactorial relation between pre-operative status and post-repair outcomes rather than a definitive predictor. Additionally, the mFI-5 relies on binary predictors (e.g., diabetes, present or absent) and therefore it does not account for disease severity or control. The ACS-NSQIP database only tracks 30-day surgical outcomes, therefore, this study could not compare long-term outcomes after discharge.
Data contribution to this database occurs on a voluntary basis and contributing centers may disproportionately represent large academic institutions [35]. Although we have adjusted for significant clinical confounders in our multivariable logistic regression analysis, no risk adjustment method is perfect and there may be confounders that were not accounted for in this study such as intra-operative adverse events that may have led to high risk of infectious complications.
Despite all of these limitations, the strengths of this study include the nationwide representation of the ACS-NSQIP database increasing the external validity of these results. The ACS-NSQIP database has been shown to have robust data quality leading to reliable and reproducible studies published using this database [36,37]. Last, our study is the first study to assess the association between the mFI-5 and post-operative infectious outcomes in patients with ICF undergoing surgical repair.
Conclusions
This study demonstrates that frailty, when measured using the mFI-5, independently predicts post-operative infections after surgical repair of ICF. More specifically, higher pre-operative mFI-5 scores were associated with an increased risk of pneumonia and SSI after surgical ICF repair. These findings are important for risk stratification in patients undergoing ICF repair surgery and for standardization in reporting outcomes after ICF repair.
Footnotes
Acknowledgments
Our Institutional Review Board (IRB) approved the study. The Data Use Agreement (DUA) was signed as per the ACS-NSQIP's requirements.
Funding Information
This study received no funding.
Author Disclosure Statement
No competing financial interests exist