Introduction

Management of acute diverticulitis remains controversial, influenced by training, geography, and available resources. Current dogmatic practice parameters advocate for antibiotic therapy in uncomplicated disease, despite a dearth of evidence for causality and recent high-level data suggesting antibiotic therapy neither lessens severity nor prevents recurrence [1–3]. A large majority of uncomplicated disease is managed in the outpatient setting [4, 5]. For patients failing outpatient management, fluid resuscitation, bowel rest with gradual resumption of a low residue diet upon resolution of abdominal pain, and intravenous antibiotics remain the mainstay. Urgent or emergent surgical management is often reserved for complications of acute diverticulitis, such as perforation with generalized peritonitis, abscess, obstruction, or fistula [1, 6]. Though many studies advocate for the relative safety of resection with primary anastomosis, Hartmann’s procedure (HP) persists as the de facto standard of care when urgent surgical intervention is felt to be unavoidable [6–8]. Still, a data-driven treatment algorithm for the emergent or urgent operative management of diverticulitis has yet to be formulated.

While the overall prevalence of diverticulosis is estimated at 2–10 %, among adults aged 50 and older, the burden of disease approaches 50–80 % [2, 9–12]. Also, strongly associated with advancing age, approximately 10–25 % of this population goes on to develop complicated diverticulitis [1, 2, 9, 10, 13, 14]. Though increasingly managed in the outpatient setting, acute diverticulitis accounts for approximately 300,000 annual hospital admissions, equating to 1.5 million inpatient hospital days and upwards of $2.5 billion in treatment costs annually in the United States alone [1, 15–18]. Importantly, there remains a paucity of high-level data regarding the optimal timing of operations in the emergent or urgent setting. Practice parameters published by the American Society of Colon and Rectal Surgeons (ASCRS) advocate for intervention without delay for patients with a toxic, acutely ill clinical presentation; however, there is no best-practice recommendation for operative management of patients who fail on the initial conservative treatment and subsequently require emergent intervention [1]. We hypothesize that patients requiring subsequent emergent surgical intervention for acute diverticulitis, following more than a 24-h delay from admission, experience worsened outcomes. Here, we examine patient characteristics and outcomes as they relate to operative delays for acute diverticulitis in the emergent setting.

Methods

The American College of Surgeons’ National Surgical Quality Improvement Program (ACS-NSQIP) database from 2005 to 2012 was the data source for this study. This is a national database with data entered by trained clinical reviewers. It includes pre-operative risk factors, lab values, intraoperative variables, and 30-day post-operative morbidity and mortality data [19, 20]. The ACS-NSQIP administration periodically audits the data to ensure high reliability. Acute diverticulitis patients were initially identified by PODIAG (post-operative diagnosis) ICD-9 codes 562.11 (diverticulitis of colon without mention of hemorrhage) and 562.13 (diverticulitis of colon with hemorrhage). Subsequent exclusion criteria included non-emergent case (EMERGNCY variable = “YES”), previous operation within last 30 days, and time from admission to operative intervention greater than 14 days. EMERGNCY includes patients identified preoperatively as “case usually performed within a short interval of time between patient diagnosis or the onset of related preoperative symptomatology. It is implied that the patient’s well-being and outcome is potentially threatened by unnecessary delay and the patient’s status could deteriorate unpredictably or rapidly.” Here, we utilized this variable to ensure evaluation of only those patients requiring emergent intervention for clinical deterioration related to sepsis. To reduce potential bias from inappropriately identified “emergent” cases, all patients with PRSEPIS variable equal to “NULL” or “None” were excluded from the analysis. PRSEPSIS includes patients identified preoperatively with “SIRS, Sepsis, or Septic Shock” according to ACS-NSQIP User Guide.

We grouped patients undergoing emergent surgical intervention based upon time from admission into the following cohorts: <24 h from admission, 24–72 h from admission, 3–7 days from admission, and >7 days from admission. We hypothesized that emergent cases within 24 h were not candidates for conservative management, while patients undergoing emergent operation after 24 h from admission were initially deemed appropriate for non-operative therapy. Baseline demographics included age, body mass index (BMI), gender, presence of comorbid disease, history of active smoking, work relative value units (wRVU), and chronic steroid use. Comorbidities were grouped by organ system as reported in prior studies utilizing ACS-NSQIP data [21, 22]. American Society of Anesthesiologist (ASA) classification of patient’s physical condition was analyzed in two ways, by grouping ASA3/4 and ASA4 alone. Surgical interventions were identified using common procedure terminology (CPT) and classified into Hartmann-type procedures, primary anastomosis, primary anastomosis with diversion, diversion alone, and drainage alone.

Primary outcomes included 30-day mortality and 30-day overall morbidity. Overall morbidity consisted of all minor (surgical site infection, pneumonia, deep venous thrombosis or thrombophlebitis, unplanned reintubation, urinary tract infection, peripheral nerve injury) and serious morbidities (organ space infection, sepsis, septic shock, wound disruption, pulmonary embolism, ventilator dependence greater than 48 h after surgery, progressive renal insufficiency, acute renal failure, stroke, coma, cardiac arrest requiring CPR, myocardial infarction, blood transfusion for bleeding). Additional outcomes included operative time, post-operative length of hospital stay, serious morbidity, minor morbidity, septic complications, reoperation, and readmission.

Approval for this study was obtained through an agreement with ACS-NSQIP and exempt status from the East Carolina University Institutional Review Board. Data analysis was performed using SPSS for Windows version 22 (IBM, Somers, NY). Univariate analysis was performed using Chi-square for nominal and ordinal variables with frequencies reported. Continuous variables were analyzed with independent sample t test or one-way ANOVA with standard deviation. Multivariable logistic regression included all baseline variables with clinical and statistical significance on univariate analysis entered into the model. Hosmer and Lemeshow test was used to assess for goodness of fit, remaining non-statistically significant for both models. Odds ratios (OR) with 95 % confidence intervals (CI) were reported. A P value <0.05 was considered statistically significant.

Results

During the 7-year study period, 2119 patients with preoperative sepsis who underwent emergent surgical intervention for acute diverticulitis were captured in ACS-NSQIP. Of those, 1212 (57.2 %) underwent operation within 24 h of presentation, 558 (26.3 %) between days 1–3, 273 (12.9 %) between days 3–7, and 76 (3.6 %) greater than 7 days after presentation. Unadjusted demographic information is presented in Table 1. Significant differences were identified in age, wRVU, diabetes, ASA 3/4, and ASA 4. Rates of vascular, cardiac, pulmonary, neurologic, and hepatic comorbidities significantly increased with delay in surgical treatment. Hartmann’s procedure represented the overwhelming majority of procedures performed in an emergent setting, followed distantly by resection with primary anastomosis. Fewer than 7 % of patients underwent primary resection with diversion, while diversion alone occurred in almost 7 % of operative candidates.

Table 1 Preoperative characteristics of patients treated with emergent surgery for acute diverticulitis
Full size table

A total of 1212 patients underwent surgery within 24 h of admission, yielding an overall morbidity rate of 38 % (Table 2). While the number of patients undergoing delayed operative intervention decreased with longer time from admission, overall morbidity significantly and steadily increased to a rate of 61.8 % when performed greater than 7 days from presentation. Similarly, the mortality rate was lowest in patients undergoing prompt operative intervention (5.4 %) but steadily increased with surgical delay to 11.8 % after 7 days. While post-operative renal and neurologic complications did not significantly increase with delay in definitive surgical management, cardiovascular and pulmonary complications did. Rates of cardiovascular complication more than doubled, from 10.3 % with early surgery to 25 % when performed after 7 days. Further, pulmonary complications more than doubled from 15.8 to 39.5 % in the same time frame. Surgical site infections and rates of reoperation were not different based on timing of surgical resection. However, septic complications significantly increased with delay in surgical treatment. In unadjusted analysis, post-operative length of stay significantly increased from 10.72 ± 9.35 days when operative intervention was performed within 24 h, to 22.72 ± 12.06 days in those receiving an operation more than 7 days following admission.

Table 2 –Outcomes of patients treated with emergent surgery for acute diverticulitis
Full size table

Multivariable analysis models were constructed to evaluate associations with 30-day overall morbidity and mortality (Table 3). Factors independently associated with worsened mortality in this analysis included age (OR 1.06, 95 % CI 1.04–1.08), ASA3/4 (OR 15.16, 95 % CI 3.62–63.50), pulmonary (OR 2.61, 95 % CI 1.78–3.81), neurologic (OR 2.32, 95 % CI 1.55–3.48), and hepatic comorbidity (OR 1.72, 95 % CI 1.72–5.30). (Table 3) While type of procedure was not significantly associated with worsened mortality as a whole, diversion alone in this patient population was associated with higher mortality (OR 3.58, 95 % CI 1.40–9.17). For overall morbidity, pulmonary (OR 2.12, 95 % CI 1.62–2.78), hepatic (OR 1.75, 95 % CI 1.14–2.69), and diabetes complications (OR 1.41, 95 % CI 1.08–1.85) were significantly associated with worsened outcomes on multivariate analysis. Work relative value units demonstrated a minimally protective association in this model (OR 0.96, 95 % CI 0.93–0.99), while ASA3/4 remained associated with worsened overall morbidity (OR 1.92, 95 % 1.53–2.42). Procedure type was again not associated with worsened overall morbidity as a whole, but individually, drainage alone demonstrated protection (OR 0.32, 95 % CI 0.10–0.98). Delay in surgical intervention was independently associated with higher overall morbidity, specifically at 3–7 days (OR 1.83, 95 % CI 1.38–2.43) and >7 days post-admission (OR 1.74, 95 % CI 1.05–2.88).

Table 3 Multivariate analysis evaluating 30-day morbidity and mortality following emergent procedure for acute diverticulitis
Full size table

To better understand the pre-operative characteristics of patients who died within each delay period, univariate analysis was performed comparing survivors to non-survivors (Supplemental Tables 1–4). Factors that remained significantly different across all delay groups included age, pulmonary comorbidity, ASA3/4, and ASA4. Rates of chronic steroid use were higher in deceased patients across all groups, saving those undergoing operative intervention more than 7 days after admission.

Discussion

Septic patients with peritonitis from acute diverticulitis require expeditious evaluation and emergent surgical management. Unfortunately, the management of acute diverticulitis without overt signs of severe sepsis or peritonitis is less clear. The present study utilizes ACS-NSQIP to assess patients initially deemed appropriate for non-operative management of acute diverticulitis, who subsequently require emergent surgical intervention. Our findings indicate that those with delay in surgical treatment are older, more comorbid, have higher ASA, and are treated with Hartmann’s procedure more than 65 % of the time. Interestingly, while surgical delay resulted in significant unadjusted differences for most outcome measures, multivariable analysis only demonstrated an association between operative delay and worsened 30-day overall morbidity. Specifically, patients with surgical delay greater than 3 days following admission were more likely to suffer worsened 30-day overall morbidity. In contrast, delayed emergent surgical management for acute diverticulitis was not associated with worsened 30-day mortality. Independent of surgical timing, ASA level greater than 3 was highly associated with mortality. As such, surgeons should consider aggressive watchful waiting in these patients for upwards of 72 h, trading increased 30-day morbidity for elevated operative mortality.

In the mid-nineteenth century, diverticulosis was primarily encountered at autopsy. At the turn of the twentieth century, the highest reported average adult life expectancy in developed nations was still less than 60 years, while the average adult life expectancy of developed nations is now greater than 80 years [23]. In his 1969 review of 521 patients, Parks identified a mean age of presentation for symptoms of diverticulitis of 61.8 years, with over 92 % of patients in the series over the age of 50 [24]. Review of over 275,000 patients identified the annual average age at time of admission for management of acute diverticulitis between 61.8 and 64.6 years, while a recent large-scale retrospective evaluation of over 13,000 patients in the NSQIP database identified the average age of patients undergoing resection for diverticulitis as 59 years [17, 25]. Advancing age has been identified as an independent risk factor for mortality in emergent operations for acute diverticulitis [26], and our study was no different. In fact, analysis by time of delay on survivors showed a significantly higher age for non-survivors. Interestingly, our findings suggest a correlation between delay in surgical management and increasing age. We hypothesize that surgeons are less apt to emergently intervene on elderly patients with increased rates of comorbidity. Given no association between delayed surgical intervention and 30-day mortality, it is rational for surgeons to consider a course of non-operative management. This is especially true in patients with ASA3/4, where risk of 30-day mortality is significantly higher. However, discussions between surgeon, patient, and family about appropriate endpoints and triggers for intervention are necessary, along with explanation of increased risks of overall morbidity and prolonged hospital stay.

Urgent operation for failure of initial medical management is a grade 1B recommendation by the American Society of Colon and Rectal Surgeons (ASCRS) [1, 6]. However, timing of operation is undefined, with the decision based upon clinical judgment. Previous studies have examined outcomes of emergent operative intervention for acute diverticulitis among several subgroups of patients including the morbidly obese, immunosuppressed, and those with end-stage renal disease. While it appears that operative delay does not significantly impact outcomes for those with end-stage renal disease, there is evidence to suggest early surgical management is associated with significantly greater risk for those who are immunosuppressed [27, 28]. Additionally, morbid obesity portends an increased risk for presentation with acute, complicated diverticulitis, but the impact of delay to surgical intervention remains unclear [29]. In our study, we show increased presence of comorbidities among patients receiving delayed operative intervention. Specifically, vascular, cardiac, pulmonary, neurologic, hepatic, and renal comorbidities significantly increased on univariate analysis of the delayed groups. On multivariate analysis, pulmonary and hepatic comorbidities remained associated with increased 30-day morbidity and 30-day mortality. During subset analysis of deceased patients compared with survivors, the presence of pulmonary comorbidity was significantly higher in patients who died, regardless of timing of surgery. These findings suggest a more aggressive surgical approach is indicated for acute diverticulitis patients with pulmonary comorbidity who present with underlying signs of sepsis and no overt peritonitis.

Previous studies have demonstrated an association between higher ASA classification and worsened outcomes [30, 31]. Within our univariate and multivariate analyses, this pattern also emerged. With delay in surgical treatment, those undergoing surgery increasingly fall into the ASA3-4 classification. Possible theories for this include high ASA patients being offered conservative management at admission due to concern over operative complications, low ASA patients becoming high ASA patients as delay in surgical management produces worsening sepsis, and a combination of these two phenomena. While ASA3-4 classification demonstrates an odds ratio of 1.92 (95 % CI 1.53–2.42) for 30-day overall morbidity, the odds ratio for 30-day mortality jumps to 15.16 (95 % CI 3.62–63.5). Further, patients experiencing mortality within all delay groups almost exclusively are categorized as ASA3 or ASA4. Interestingly, only 1.9 % of the studied cohort had a “Do Not Resuscitate” (DNR) order recorded. It was present in 1.3 % of the early cohort, increasing to 3.9 % of the cohort undergoing surgery more than 7 days from admission. There were no differences in presence of DNR among deceased patients based upon the delay in surgery. In the context of this study, with a 30-day mortality rate ranging from 5.4 % in the earliest treatment group to 11.8 % in the latest treatment group, it behooves surgeons to have a frank discussion with highly-comorbid patients about the risks involved with both operative and non-operative management.

Historically, sigmoidectomy with end colostomy was the procedure of choice for emergent and urgent management of acute diverticulitis [2, 17, 32, 33]. This study confirms the continued dominance of HP in this setting, with significantly less use of primary resection, with or without diversion. More recently, several studies have shown the safety of resection and primary anastomosis in the acute setting, and some groups strongly advocate for this to reduce complications of stomas and increase functional capacity in elderly patients unlikely to receive future stoma closure [25, 34]. Possible reasons for these findings include surgeon comfort, bias within the USA, and delay of adoption of published findings within clinical practice. To date, no prospective, randomized studies comparing procedure type, or laparoscopic vs open approaches, have been performed. The heterogeneity of patient populations, practice settings, and other confounding factors has impeded the development of a clear recommendation for, or against, HP in the non-elective setting [35]. Aside from the rare and outdated usage of diversion alone for acute diverticulitis, our current results show no increased risk of 30-day mortality or 30-day morbidity associated with the procedure performed. As such, surgeons should consider the possibility of primary anastomosis in this setting, with or without proximal diversion.

Surgical outcomes are coming under increasing scrutiny as the United States healthcare system shifts from a Fee-For-Service model to Accountable Care Organizations promoting value-based care. While various studies question the effectiveness of current metrics such as rates of readmission and hospital length of stay [36, 37], there remain clear financial and patient satisfaction benefits with shorter length of stay. Here, we show that delay in timing of surgical intervention leads to significantly increased post-operative hospital length of stay that more than doubles to 22 days in the group undergoing surgery more than 7 days following admission. Given the increased rate of overall morbidity seen as treatment delay increased, this result is expected. Combined with prior study findings, perhaps a more aggressive surgical approach in patients previously deemed “too sick” for surgery would produce fewer complications, less mortality, and shorter post-operative length of stay. As stated previously, this must be balanced with the possibility of increased morbidity and mortality for high-risk patients undergoing surgical intervention. Unfortunately, our analysis does not allow for an answer to this question due to limitations with the ACS-NSQIP dataset.

Due to study design, our retrospective review of prospectively collected data in the ACS-NSQIP participant use file is limited by the possibility of confounding, selection, information, and misclassification bias. Appropriate corrections were made during statistical analysis to reduce effects of these biases but cannot account for all biases. Unfortunately, this dataset only allows for a single post-operative diagnosis code. Therefore, secondary diagnoses such as 567.22 (peritoneal abscess) or 569.5 (abscess of intestine) cannot be queried. Further, there is no radiographic data to determine presence or absence of perforation, fluid collection, phlegmon, or other complicating factors. Other limitations include heterogeneous patient factors including procedures performed, differences in hospitals, surgical expertise, and lack of randomization. The lack of longer-term outcomes is a significant limitation of this study, precluding our ability to link post-operative complications with long-term outcomes. Lastly, 30-day surgical outcomes are influenced by factors not accounted for in this study including medical staff, patient compliance, and many others not included in the NSQIP dataset. Further, these data cannot account for septic patients successfully treated non-operatively. Because these data are unknown, our conclusions can only be applied towards those patients who ultimately undergo surgical intervention in this setting. These limitations should be taken into account when reviewing our data. Despite these limitations, the data analyzed are considered the most complete multi-institutional clinical data available, with the added advantage of impartial clinical reviewers to help reduce error and bias.

Conclusion

High risks of morbidity and mortality accompany emergent or urgent operation for acute diverticulitis. There are many patient factors including advanced age and major comorbidities that impact the clinical decision to abort medical management and proceed to an emergent or urgent operation. When surgical management is deemed necessary, the present study demonstrates favorable outcomes for earlier intervention. Facing an aging population living with chronic disease, further investigation is warranted to detail whether delay to operation is an independent risk factor and if so, whether certain subgroups may benefit from an early operation.