Global variation in anastomosis and end colostomy formation following left‐sided colorectal resection

Background End colostomy rates following colorectal resection vary across institutions in high‐income settings, being influenced by patient, disease, surgeon and system factors. This study aimed to assess global variation in end colostomy rates after left‐sided colorectal resection. Methods This study comprised an analysis of GlobalSurg‐1 and ‐2 international, prospective, observational cohort studies (2014, 2016), including consecutive adult patients undergoing elective or emergency left‐sided colorectal resection within discrete 2‐week windows. Countries were grouped into high‐, middle‐ and low‐income tertiles according to the United Nations Human Development Index (HDI). Factors associated with colostomy formation versus primary anastomosis were explored using a multilevel, multivariable logistic regression model. Results In total, 1635 patients from 242 hospitals in 57 countries undergoing left‐sided colorectal resection were included: 113 (6·9 per cent) from low‐HDI, 254 (15·5 per cent) from middle‐HDI and 1268 (77·6 per cent) from high‐HDI countries. There was a higher proportion of patients with perforated disease (57·5, 40·9 and 35·4 per cent; P < 0·001) and subsequent use of end colostomy (52·2, 24·8 and 18·9 per cent; P < 0·001) in low‐ compared with middle‐ and high‐HDI settings. The association with colostomy use in low‐HDI settings persisted (odds ratio (OR) 3·20, 95 per cent c.i. 1·35 to 7·57; P = 0·008) after risk adjustment for malignant disease (OR 2·34, 1·65 to 3·32; P < 0·001), emergency surgery (OR 4·08, 2·73 to 6·10; P < 0·001), time to operation at least 48 h (OR 1·99, 1·28 to 3·09; P = 0·002) and disease perforation (OR 4·00, 2·81 to 5·69; P < 0·001). Conclusion Global differences existed in the proportion of patients receiving end stomas after left‐sided colorectal resection based on income, which went beyond case mix alone.


Introduction
In 2015, the Lancet Commission on Global Surgery highlighted a substantial gap in access to safe and affordable surgical care across low-and middle-income countries (LMICs), raising the priority of surgery on the global health agenda 1 . Despite this, reporting of outcomes following abdominal surgery from many LMICs remains unstandardized and of mixed quality. Where high-quality evidence is available, a threefold higher risk of death in lowversus high-income settings has been observed 2 . However, other key outcomes from the surgical management of colorectal cancer or benign colorectal disease in LMICs have been particularly poorly profiled to date 3 .
End colostomy rates following colorectal cancer resection vary substantially between centres in high-income countries, ranging from 15 to 70 per cent 4 . This may reflect variations in case mix, as the decision to create an end colostomy rather than a primary restorative anastomosis is influenced by the urgency of presentation, the presence of operative field contamination, disease severity and stage, as well as functional status of the pelvic floor. For patients, quality of life with an end colostomy is influenced by multiple factors, including functional  status, social support, income level, education and availability of specialist services 5 . The care requirements of a stoma may present a different psychosocial and physiological burden for patients in LMICs compared with those in high-income settings. For example, geographical barriers and limited health resources are likely to raise treatment costs and reduce access to specialist equipment and services 6 , increasing the risk of catastrophic expenditure following colorectal surgery 7 . Examining international practice in stoma formation is therefore important in seeking to identify areas of variation and improve outcomes.
The primary aim of this study was to determine variation in rates of end colostomy formation following colorectal resection between low-, middle-and high-Human Development Index (HDI) strata, after adjusting for patient, disease and operative factors. Secondary aims were to report the mode of presentation, rate of laparoscopic surgery, and to determine any relationship between stoma formation and postoperative mortality in patients undergoing resections.

Protocol and network
This study was an exploratory subgroup analysis from two international, multicentre, prospective cohort studies conducted according to previously published protocols (NCT02179112, NCT02662231) 2,8 . These protocols were disseminated through social media, and national and international surgical and anaesthetic associations. Briefly, the model required small teams of local investigators to collect data on prospectively determined items, coordinated by regional and national lead investigators, across short time windows, with pooled analysis by a central steering committee.

Patients and settings
Any hospital providing both emergency surgery and elective colorectal surgical services was eligible to contribute patients to this study. Patients were included during at least one discrete 2-week period between 1 July 2014 and 31 December 2014 (GlobalSurg-1) and 4 January 2016 and 31 July 2016 (GlobalSurg-2). To maximize inclusiveness and minimize burden on resource-constrained clinicians, collaborators were permitted to collect data within any 2-week interval across this time window, so long as data collection was consecutive and case ascertainment was complete. Adult patients (aged over 16 years) undergoing elective (GlobalSurg-2) or emergency (GlobalSurg-1 and -2) left hemicolectomy, sigmoid colectomy or rectal resection were included. Emergency procedures were defined as unplanned operations occurring within 2 weeks of hospital admission, and included procedures for trauma and reoperation following surgical complications. Open, laparoscopic and laparoscopic converted to open procedures were all eligible. To reduce risk of bias based on case mix, only colorectal resections for a primary gastrointestinal indication were included. Patients were excluded if the primary indication for surgery was vascular, gynaecological, obstetric, urological or

Outcome measures
The primary outcome measure was the end colostomy formation rate, defined as formation of an end colostomy during the index procedure without restorative anastomosis. The secondary outcome measure was the postoperative mortality rate (death within 30 days of the index procedure).

Other included explanatory variables
Data variables were designed to be assessed objectively, standardizable and internationally relevant. Variables deemed candidates in the causal pathway for stoma formation were indication for surgery, urgency of surgery (elective/planned or emergency/unplanned (within 2 weeks of hospital admission)) and colonic or rectal perforation noted at the time of surgery. Variables deemed to be confounders associated with both the causal pathway and outcome measures included age, sex, ASA fitness classification, smoking status, use of the WHO Surgical Safety checklist 10 , and use of laparoscopic surgery.

Data capture and validation
Study data were collected and managed using RED-Cap (Research Electronic Data Capture) tools hosted at the University of Edinburgh (https://www.projectredcap.org/). REDCap is a secure, web-based application designed to support data capture for research studies, providing: an intuitive interface for validated data entry; audit trails for tracking data manipulation and export procedures; automated export procedures for seamless data  downloads to common statistical packages; and procedures for importing data from external sources. In both studies, a local lead investigator was responsible for overall quality assurance, case ascertainment and data accuracy at each centre. Where missing data were identified, the lead investigator was contacted and asked to ensure completeness. Records from centres that had an overall data completion rate of less than 95 per cent were removed from this analysis.

Statistical analysis
Variation across different international health settings was assessed by stratifying participating centres by country into tertiles according to HDI. This is a composite statistic of life expectancy, education and income indices published by the United Nations (http://hdr.undp.org/ en/content/human-development-index-hdi). Differences between HDI tertiles were tested with the Pearson χ 2 test and Kruskal-Wallis test for categorical and continuous variables respectively. Descriptive percentages are listed as low HDI versus middle HDI versus high HDI throughout for consistency. To account for case mix, mixed-effects, hierarchical multilevel logistic regression models were constructed. Patients nested within countries were considered by a random-effects model. Patient-, disease-and operation-specific variables considered a priori to be candidates in the causal pathway, or confounders to the included outcomes, were included and treated as fixed effects. Model residuals were checked at both levels, checking for first-order interactions; these were included in final models if found to be influential. Final model selection was by minimizing the widely applicable information criterion (variables considered to be marginal candidates in the causal pathway, and that reduced the goodness of the model fit were removed). Any variables with an incident rate below 1 per cent were not taken forwards into the multivariable models. Model discriminative ability was determined using the C-statistic (area under the receiver operator curve, AUC). Coefficients generated were presented as odds ratios (ORs) with 95 per cent confidence intervals. All analyses were performed using the R version 3. resection were included in this study (Fig. 1); 113 patients (6⋅9 per cent) were from low-HDI, 254 (15⋅5 per cent) from middle-HDI and 1268 (77⋅6 per cent) from high-HDI countries. Patients from low-and middle-HDI settings were significantly younger, more frequently men, lower risk (ASA grade below III) and less likely to smoke than those in high-HDI settings ( Table 1). Patients were more likely to present as an emergency in low-HDI settings (low, 75⋅2 per cent; middle, 44⋅9 per cent; high, 45⋅5 per cent; P < 0⋅001) (Fig. 2) and more likely to have perforated disease at presentation (57⋅5, 40⋅9 and 35⋅4 per cent respectively; P < 0⋅001). Disease profiles in patients from low-HDI settings were different from those in middle-and high-HDI settings (Fig. 3). Fewer procedures were performed for malignancy (47⋅8, 58⋅3 and 64⋅3 per cent respectively; P = 0⋅001), diverticulitis (1⋅7, 4⋅3 and 14⋅2 per cent; P < 0⋅001) and inflammatory bowel disease (0, 1⋅6 and 1⋅4 per cent; P = 0⋅007), but a greater proportion of procedures were for volvulus (21⋅2, 7⋅5 and 2⋅4 per cent; P < 0⋅001) and trauma (9⋅7, 8⋅3 and 0⋅8 per cent respectively; P < 0⋅001). An overall delay from presentation to surgery of at least 48 h was more common in both low-and middle-HDI than high-HDI countries (33⋅6, 42⋅1 and 29⋅0 per cent; P < 0⋅001). A WHO checklist was used in only 40⋅7 per cent of operations in low-HDI countries compared with 72⋅4 and 84⋅1 per cent in middle-and high-HDI countries respectively. Half as many patients in middle-HDI countries had a planned laparoscopic operation than in high-HDI countries (15⋅4 versus 29⋅7 per cent; P < 0⋅001). Only one patient from a low-HDI country had laparoscopic surgery (this was subsequently excluded from the mixed-effects models).

Discussion
This study demonstrated that end stoma rates in low-HDI countries were twice those in middle-and three times those in high-HDI countries. As each of the HDI strata included multiple hospitals of different size and nature, it suggests that variation based on income per capita may be more important than variation within countries. The difference between groups is partly explained by differences in case mix, with greater emergency presentation of both malignant and non-malignant conditions in low-HDI settings. This association persisted despite adjustment, suggesting that other factors may contribute to this variation. Patients in LMICs were more likely to present as emergencies and to have perforated disease than patients in high-HDI settings. In part, this reflects differences in the overall disease burden, with trauma and volvulus being more common in LMICs. However, the increased frequency of emergency procedures for malignancy in LMICs may reflect barriers to accessing care and treatment for non-communicable disease in LMICs 1,3 . These may include limited implementation of screening programmes, inefficient referral pathways, the relatively high cost of investigations such as endoscopy 3,11,12 , as well as some patients having limited access to health education or a preference to seek care from traditional healers 13 -16 . The greater burden of emergency surgery suggests that patients in LMICs may be more likely to delay a decision to seek healthcare until they have deteriorated with complicated, advanced disease. Because significant populations live more than a 2-h drive from the nearest hospital 17,18 , patients' conditions may deteriorate further owing to delays while identifying affordable and efficient means of transport 19,20 . In LMICs, once patients reach hospital, delayed and lack of appropriate investigations, staff shortages, erratic electric and water supplies, and insufficient funds to pay for care can limit and further delay surgery 21 . In the present study, patients in LMICs were more likely to experience significant in-hospital delays. Consistent with previous studies 22,23 , this was associated with end stoma formation. It should be noted in the present data, however, that in-hospital delay (48 h or more) was not associated with an increased risk of death in the mixed-effects model. This may reflect appropriate delay of surgical intervention (such as for preoperative optimization of an obstructing cancer) and appropriate rationalization of resources (the most unwell patients were prioritized for early access to  24 , all contribute to patients in LMICs being more likely to present acutely unwell with complicated disease that makes primary restorative surgery challenging, and influencing the decision whether primary anastomosis or end colostomy is appropriate 25 .
Differences in training and provision of specialist colorectal surgery, and lack of available or affordable equipment for technically difficult anastomoses, could also affect stoma rates. With fewer patients presenting with operable colorectal cancer in many low-HDI countries 3,12 and fewer formal training opportunities, access to subspecialist colorectal services is limited 3,26,27 . High baseline mortality rates 2 , inadequate provision of critical care support 28,29 and insufficient medicolegal protection 30 may also promote risk-averse practices. Stapling devices may be unaffordable for both patient and provider in many LMICs, meaning that only selected patients have access to these techniques 31 . Similarly, although laparoscopic colorectal resection was performed in middle-HDI settings, it was uncommon. Lack of affordable laparoscopic equipment, variable provision of training and hospital-level difficulties, such as a reliable electrical supply, remain barriers to minimal access surgery in LMIC settings 32 , despite potential for patient benefit 33,34 .
The high mortality rate for both elective and emergency surgery reported in this study supports previous findings that patients have a higher risk of death following surgery in low-HDI settings which cannot be accounted for by case mix alone 2,35 . The present analysis showed that patients undergoing end stoma formation were at increased risk of death. Despite adjustment, this finding could represent a surrogate marker of disease severity where the highest-risk patients are being selected to receive a stoma. In the present study, it was not possible to measure physiological markers of disease severity beyond ASA classification (such as hypotension, tachycardia, high lactate level or an end-organ perfusion deficit) that could influence surgical decision-making and outcomes.
This study has important limitations that could affect its generalizability. As it included a relatively low mean number of patients per centre in a 'snapshot' methodology, no analysis was performed at a per-centre or per-country level. Although only one-quarter of patients in the data set were from LMICs, sites across 30 countries contributed data, bolstering external generalizability across LMIC settings. Data were collected across all HDI tertiles in both emergency (GlobalSurg-1 and -2) and elective (GlobalSurg-2) settings, and are relevant to both planned and unplanned left-sided colorectal resections, but numbers in some groups (such as elective operations for cancer in low-HDI settings) were small. Further validation of these findings is therefore required in future work. Although there were no centre-level exclusion criteria for case volume or infrastructure, a sampling bias is likely to exist, wherein the best resourced and/or academically affiliated centres within LMICs were more likely to access the study protocol and provide patient data than those in remote and rural settings. This may have led to an underestimate of the true rate of end stoma formation within LMICs.
Reported end colostomy rates have varied from 0 to as high as 74 per cent 25,36 -39 in groups including emergency surgery 39 , late presentations of cancer 25 , complications of infectious disease 38 and traumatic injury 36 . The collaborative methodology in the present study enabled clinicians to enter data into a secure online platform contemporaneously alongside their clinical practice, in accordance with a prespecified protocol. This led to high levels of data accuracy and completeness 40 and has provided the basis on which further studies can be developed to examine other factors that influence outcomes in different settings.