Outcomes of obstructed abdominal wall hernia: results from the UK national small bowel obstruction audit

Background Abdominal wall hernia is a common surgical condition. Patients may present in an emergency with bowel obstruction, incarceration or strangulation. Small bowel obstruction (SBO) is a serious surgical condition associated with significant morbidity. The aim of this study was to describe current management and outcomes of patients with obstructed hernia in the UK as identified in the National Audit of Small Bowel Obstruction (NASBO). Methods NASBO collated data on adults treated for SBO at 131 UK hospitals between January and March 2017. Those with obstruction due to abdominal wall hernia were included in this study. Demographics, co‐morbidity, imaging, operative treatment, and in‐hospital outcomes were recorded. Modelling for factors associated with mortality and complications was undertaken using Cox proportional hazards and multivariable regression modelling. Results NASBO included 2341 patients, of whom 415 (17·7 per cent) had SBO due to hernia. Surgery was performed in 312 (75·2 per cent) of the 415 patients; small bowel resection was required in 198 (63·5 per cent) of these operations. Non‐operative management was reported in 35 (54 per cent) of 65 patients with a parastomal hernia and in 34 (32·1 per cent) of 106 patients with an incisional hernia. The in‐hospital mortality rate was 9·4 per cent (39 of 415), and was highest in patients with a groin hernia (11·1 per cent, 17 of 153). Complications were common, including lower respiratory tract infection in 16·3 per cent of patients with a groin hernia. Increased age was associated with an increased risk of death (hazard ratio 1·05, 95 per cent c.i. 1·01 to 1·10; P = 0·009) and complications (odds ratio 1·05, 95 per cent c.i. 1·02 to 1·09; P = 0·001). Conclusion NASBO has highlighted poor outcomes for patients with SBO due to hernia, highlighting the need for quality improvement initiatives in this group.


Introduction
Much of the literature addressing the treatment of small bowel obstruction (SBO) focuses on management of the most common cause, adhesions [1][2][3] . Emergency treatment of the obstructed abdominal wall hernia contains knowledge gaps that include the optimal surgical approach and type of repair 4,5 . Care for this group is particularly challenging as patients are often elderly with high levels of co-morbidity 6 .
The quality of emergency surgical care has been scrutinized through initiatives such as the American College of Surgeons' National Surgical Quality Improvement Program (NSQIP) 7 and the UK National Emergency Laparotomy Audit 8 , although the latter study did not include many emergency repairs that failed to meet criteria for inclusion. UK evidence also suggests that 'watchful waiting' as an alternative to elective hernia repair has increased the rate of emergency procedures 9 .
The aim of this study was to analyse cohort data from the UK National Audit of Small Bowel Obstruction (NASBO) to describe the current management and outcomes of patients with SBO due to abdominal wall hernia.

Methods
NASBO was a trainee-led and trainee-delivered multicentre prospective cohort study carried out between 16 January and 13 March 2017, with support from all major professional stakeholders. Any acute hospital in the UK that performed emergency general surgery was eligible to contribute data. Hospitals were recruited through the NASBO network, personal contacts and social media. This study is reported in line with STROBE 10 and Statistical Analyses and Methods in the Published Literature (SAMPL) 11 guidelines.

Approvals
At each site, local collaborators were responsible for registering the study and securing Caldicott Guardian permissions. As this study was considered clinical audit, national research ethical approval was deemed not to be required (NHS East Scotland Research and Ethics Committee, NR/1610AB10).

Inclusion and exclusion criteria
The study included non-pregnant patients aged 16 years and over who had a clinical suspicion of SBO. Clinical suspicion was either identified by a member of the surgical team or agreed in consultation with other specialties. Patients subsequently found to have pathologies other than SBO were excluded.

Data collection
Data were captured during the study period according to a prespecified protocol 12  Date of resumption of enteral nutrition was recorded by the local teams as the first date that oral intake was tolerated. Hernia types were broadly categorized as inguinal/femoral, primary midline, incisional and parastomal. The case report form and definitions of outcomes have been published previously 12 . Surgical interventions were classified as immediate operation (within 24 h of admission), delayed operation (24 h after admission) or non-operative. Components of emergency procedures were also recorded, including small bowel resection and stoma formation.

Data validation
Accuracy of data submission was validated by local investigators who were independent from data collection teams. Selected fields from 25 per cent of all submitted patient records were sampled randomly at each site. Data were re-entered separately by the validator, who was blinded to the original data. Categorical variables were deemed to be accurate on exact match, and continuous variables when the value was within 0⋅5 units of the collected data. Accuracy is expressed as a percentage of correct fields of the total fields sampled.

Statistical analysis
After exclusion of ineligible patients, data were summarized using simple descriptive statistics for comparisons across specific hernia groups. Data are presented as numbers and percentages for categorical data and as mean(s.d.) values for normally distributed continuous data; where data were not Groin hernia Incisional hernia Midline hernia Parastomal hernia normally distributed, median (i.q.r.) values were used. χ 2 and Kruskal-Wallis tests were used to test for differences between categorical and continuous variables respectively.
To adjust for confounding factors, clinically plausible variables were entered into a Cox proportional hazards model, clustered by centre to adjust for hospital-level effects. Effect estimates are presented as hazard ratios (HRs) with corresponding 95 per cent confidence intervals. For the outcome of complications, a multilevel logistic regression model was used to estimate odds ratios (ORs) with 95 per cent confidence intervals. Briefly, patient risk factors were adjusted for at level 1 (fixed effect) and centre effects at level 2 (random effects). Model selection was guided by minimization of the Akaike information criterion. Models were examined for first-level interactions, and those found to be significant were retained in the model. Statistical significance was taken at the level of P ≤ 0⋅050 a priori.
All analyses were performed in R version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria) using the tidyverse and finalfit packages.

Results
NASBO collected data on 2604 patients from 131 hospitals during the study period (Fig. 1). After screening,
More than one-third of patients (149, 35⋅9 per cent) were at moderate to severe risk of malnutrition using the Nutritional Risk Index, which was highest in the groin and incisional hernia groups. Patients in the midline, incisional and parastomal hernia groups were significantly more likely to be overweight or obese as measured by BMI (77, 73⋅6 and 69 per cent respectively), in contrast with those in the groin hernia group, who tended to have normal BMI (31⋅4 per cent combined obese or overweight; 46⋅4 per cent normal BMI) ( Table 1).

Imaging
Some 91 patients (21⋅9 per cent) had plain abdominal radiography only, and the remainder had abdominal CT with or without previous plain radiography. Patients with obstructed parastomal or incisional hernia underwent diagnostic CT in 86 and 84⋅0 per cent of cases respectively, compared with 71⋅9 per cent of patients with groin hernia and 66 per cent of those with midline hernia. Thirty-five patients (11⋅1 per cent) had a subsequent water-soluble contrast study to assess whether the hernia was obstructed. Nine patients (5 groin, 1 incisional, 3 midline) had no imaging performed to support the clinical diagnosis of obstruction ( Table 2).

Operative interventions
Operative management was performed in 312 (75⋅2 per cent) of the 415 patients; Table S1 (supporting information) summarizes the surgical procedures carried out. Immediate operation was more common in the groin hernia group  at 75⋅2 per cent compared with midline (71 per cent), incisional (53⋅8 per cent) and parastomal (26 per cent) hernias (Table S1). In contrast, non-operative management was initiated in 54 per cent of patients with parastomal hernia and 32⋅1 per cent of those with incisional hernia. Overall, small bowel resection was common, being required in 198 (63⋅5 per cent) of the 312 patients, and stoma formation was performed in 55 (17⋅3 per cent). Values in parentheses are *percentages and †95 per cent confidence intervals, unless indicated otherwise; ‡values are mean(s.d.). CCI, Charlson Co-morbidity Index; WCC, white cell count; ED, emergency department; GP, general practitioner; AKI, acute kidney injury.
Multilevel regression analysis to identify factors associated with major complications (death, unplanned high-dependency unit/ICU admission, 30-day readmission) indicated in a univariable model that operative management, increasing age, increasing co-morbidity, CT within 24 h of admission or beyond 72 h after admission, and referral from other inpatient team were all significant predictors of complications, but only age persisted in the multivariable model (adjusted OR 1⋅05, 95 per cent c.i. 1⋅02 to 1⋅09; P = 0⋅001) ( Table 5).

Discussion
This study provides contemporary evidence of high rates of mortality and morbidity in patients with SBO due to obstructing hernia. Although the morbidity of obstructed hernia has been well documented 5 , this study demonstrates that it still persists.
In patients undergoing conservative treatment of groin hernia, the reported annual risk of strangulation or incarceration ranges from 0⋅4 to 2⋅7 per cent 14 . A recent study 15 of emergency repaired midline hernia using the US NSQIP database showed the mortality rate to be 1⋅3 per cent, and a 2010 UK-based single-centre study 16 found a 30-day mortality rate after emergency groin hernia repair of 2⋅5 per cent. A single-centre observational study 17 of incarcerated hernia found morbidity comparable to that in the present study. These studies included a mix of obstructed and strangulated hernia, but the difference in mortality between these and the present study is stark. The finding that bowel resection in the present study might be associated with reduced mortality, as seen in Table 4, contrasts with other studies 18 , suggesting that prompt surgery for clinically apparent bowel ischaemia improved outcomes. Equally, this difference may reflect selection bias, whereby fitter patients underwent resection and less fit patients were managed conservatively, with either hernia reduction alone or supportive conservative management. The association between delayed operation and lower in-hospital mortality may reflect the predominantly incisional or parastomal hernia groups treated in this way. It might also reflect a group in which additional resuscitation measures had been undertaken, mitigating the impact of surgery.
The characteristics of the population studied may go some way to explaining the outcomes demonstrated here. Acute kidney injury was independently associated with poor outcomes, including death, in the present study, in line with other emergency surgeries 19 . Malnutrition is also associated with poor surgical outcomes. This study showed high rates of malnutrition with limited interventions to address this.
Beyond patient characteristics are those factors related to the organization of hospital services. Outcomes were worse in terms of complications for the patients referred from inpatient teams in this study, potentially reflecting delay where patients were admitted under medical specialties with 'coffee ground vomit' or 'gastroenteritis', whereas the correct diagnosis was obstructed hernia. Time to surgery may be one element of management that might be improved for this particular set of patients. However, it was noted that delayed surgery was associated with reduced mortality. This may reflect that the group who had delayed surgery were more likely to have incisional rather than groin hernias, the latter carrying a greater risk of strangulation.
Bowel resection rates were high in this study at 63⋅5 per cent, compared with 5 per cent in adhesive SBO 3 . Delay potentially increases the need for small bowel resection, which is associated with worse clinical outcomes including mortality 20 . A threshold of a 3-day trial of non-operative management has been demonstrated as the cut-off point for worsening outcomes in adhesive obstruction 1 . It is possible that a similar temporal cut-off exists for patients with obstructed hernia, albeit at a much earlier time point.
There were relatively high rates of unplanned critical care use (approximately 10 per cent) in the groin hernia population. Despite high levels of co-morbidity, patients were managed at ward level until deterioration requiring increased support. In the UK, some of this patient group fell outside the National Emergency Laparotomy Audit criteria 7 , and were not assessed routinely in the same way for mortality risk and frailty. As a result, acuity of the situation and the need for critical care may not have been recognized for the high-risk patient.
This study includes a subgroup of the whole patient population treated for SBO. As a result, it is missing some information that may be of interest to clinicians, specifically type of repair, use of mesh, and type of anaesthesia. As an observational study, no inference can be drawn on causality.
Guidelines for the management of hernia in the acute setting are limited, reflecting a paucity of high-quality evidence 5 . Work in emergency laparotomy 8 , a condition with similar rates of mortality and morbidity, has demonstrated improvements in the UK. Principles learned from that study might be translated into emergency hernia surgery. For example, comprehensive preoperative risk assessment could inform preoperative and postoperative intervention. Mortality risk and frailty scoring could be undertaken 8,21 . In turn, this information might influence case selection to avoid operating where futility is anticipated. Other researchers 9,22 have noted an increase in emergency hernia repair following the institution of watchful waiting policies, and the data presented here might shift the health economic balance of this policy to favour early repair in selected patients in the elective setting.
The findings of this study suggest that strategies should be developed to aid early diagnosis of hernia, particularly in patients admitted to non-surgical wards, and that similar standards of care as offered to patients having emergency laparotomy (early imaging to aid diagnosis, senior surgeon, routine postoperative high-dependency unit admission) should be instituted. To address these issues, additional data specific to emergency hernia repair, rather than only bowel obstruction, would be helpful. Prospective studies might investigate strategies to improve perioperative outcomes including preoperative resuscitation or the use of high-dependency facilities. Decision-making models to identify the optimal timing of repair for pre-existing hernias might help surgeons to balance the risk of obstruction against the risks of repair in elective and emergency settings. Poor outcomes in patients with SBO due to abdominal wall hernia highlight the need for quality improvement initiatives in these patients.