Study design

A multicentre prospective cohort study disseminated through collaborative medical student networks (figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

STARSurg ‘Mini-Team” structure, roles and responsibilities.

The STARSurg network

Student Audit and Research in Surgery (STARSurg) is the UK's national medical student research collaborative, run by the Surgical Research Gateway (SURG) Foundation (registered charity number 1 159 898). STARSurg is co-ordinated by a team of medical students and surgeons. The collaborative model and the educational benefits to participating students have previously been described elsewhere.11 ,12

Study setting

Hospitals in the UK and the Republic of Ireland that offer elective or emergency gastrointestinal or liver surgery will be invited to participate. Each centre may contribute up to four, 2-week sets of consecutive patient data.

Inclusion criteria

• Adult patients (age ≥18 years) undergoing any gastrointestinal resection, liver resection or reversal of ileostomy or colostomy;

• Elective and emergency patients may be included;

• Open, laparoscopic, laparoscopic-assisted, laparoscopic-converted and robotic procedures may be included.

Exclusion criteria

• No baseline preoperative creatinine available;

• Previous kidney transplant;

• Renal replacement therapy (any form of dialysis) in the 90 days preceding operation;

• Appendicectomy;

• Transplant surgery;

• Gynaecological primary indication;

• Urological primary indication;

• Vascular primary indication.

Patient identification and data collection

Operating lists and theatre computer systems will be screened daily to identify eligible patients. The preoperative and operative data fields will be completed at the earliest opportunity following surgery. Collaborators will actively follow-up patients in the postoperative period, completing 30-day and 1 year follow-up promptly.

Primary outcome

The primary outcome measure will be the incidence of AKI within 7 days of the indexed operative procedure. This period of 7 days has been selected to only capture AKI directly attributable to insults sustained during the perioperative period.

Collaborators will submit raw creatinine values to the study database. AKI will be identified using an adaptation of the National Algorithm for Detecting Acute Kidney Injury classification system (figure 2), which is based on Kidney Disease Improving Global Outcomes (KDIGO) AKI guidelines.13 It was introduced by NHS England in June 2014 in a Patient Safety Alert requiring standardisation of the identification of AKI across the UK.14 If serum creatinine is not available over the past 7 days, the National Algorithm defines baseline creatinine as the median of all measured creatinine levels over the past year. It would not be practical for study collaborators to manually derive this value; therefore the algorithm has been adapted to require the most recent preoperative serum creatinine measurement. To identify emergency patients who had already developed AKI at admission, both the first creatinine measured on admission and the most recent preadmission creatinine will be required for emergency patients. A full description required creatinine data points is available in the online supplementary file (data points 11 and 29). In this observational study no extra blood tests will be performed, other than those ordered by the clinical team.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Adaptation of the national algorithm for detecting acute kidney injury.14 RV, reference value; OAKS, Outcomes After Kidney injury in Surgery.

Secondary outcomes

Secondary outcome measures will include:

• Incidence of AKI within 48 h of surgery

• Persistent renal dysfunction at discharge

• Persistent renal dysfunction at 12 months

• Regular renal replacement therapy (any form of dialysis) at 12 months

• 30-day major complication rate (Clavien-Dindo grades III-V)

• 30-day minor complication rate (Clavien-Dindo grades I-II)

• Mortality at 12 months

Persistent renal dysfunction will be defined as an increase from baseline of chronic kidney disease (CKD) stage, as defined by the Kidney Disease Outcomes Quality Initiative.15 In addition to serum creatinine, all other variables required to calculate estimated glomerular filtration rate (eGFR) using the Chronic Kidney Disease Epidemiology Collaboration formula will be collected.16 For the 30-day outcome, the single eGFR value nearest to postoperative day 30 will be used. For the 1 year outcome, blood tests taken 9–15 months postoperatively will be considered. Two separate eGFR values recorded at least 3 months apart will be required to demonstrate a persistent increase in CKD stage.

Complications will be recorded using the Clavien-Dindo classification, a standardised, internationally validated scoring system for postoperative complications.17 The highest grade complication (grade I-V) experienced by each patient will be recorded.

Factors associated with AKI

A key aim of OAKS is to identify perioperative factors that may be associated with subsequent development of AKI. To record this, data will be collected on patient-related, disease-related and pharmaceutical variables (table 1). Detailed description of all data fields is available in the online supplementary file.

Centre questionnaire

Supervising consultants will be asked to complete a short centre questionnaire about the resources available at their centre (table 2).

Quality assurance

Each local team of medical students will be supervised by a qualified doctor (ranging from foundation year to registrar grade) and a consultant surgeon.

A detailed protocol for collaborators describing how to register and run the study will be made available online. It will provide a detailed data dictionary, along with guidance on potential data sources for collaborators to use when collecting the data. The protocol will be presented at collaborator meetings across the UK, enabling participants to clarify any uncertainties in person with the steering committee.

Collaborators will be asked to complete mandatory online e-learning modules covering study inclusion criteria, AKI, the Clavien-Dindo classification system and data governance protocols. This will ensure collaborators are able to safely collect accurate data.

To identify any local issues regarding data collection, all participating centres will be asked to pilot patient identification and case record completion for 1 day in the week leading up to the main study starting date. The steering committee will offer support throughout this period to resolve any problems.

Validation

Only data sets with >95% data completeness will be accepted. Final year medical students or qualified doctors not involved in initial data collection will act as independent assessors, reviewing data collected at a local centre. Overall independent assessors will validate a minimum of 5% of patient records, with a target of >95% case ascertainment and >98% data accuracy.

Data management

Data will be collected and stored online through a secure server running the Research Electronic Data Capture (REDCap)18 web application, hosted at the University of Edinburgh. REDCap is widely used internationally by academic organisations to store research databases. Collaborators will be given secure login details including a password for the REDCap project server. All transmission and storage of web-based information by this system is encrypted. Any patient identifiable information will not be available for data-analysis and will be automatically stripped from the database when exported from REDCap.

Anticipated minimum recruitment

Based on our previous collaborative studies,19 an average centre performs approximately 20 operations meeting the study inclusion criteria in a 14-day period. At most centres in our previous study three data collection teams participated at each centre, collecting data over consecutive 14-day periods. Previously at least four centres participated at each of the 38 medical schools involved. Overall, we estimate that a minimum of 6840 patients will be recruited. The observational nature of the study's primary outcome means a power calculation is not required.

Statistical analysis

Differences between demographic groups will be tested with the χ² test for categorical data, with differences in continuous data analysed using Student t test. Two-sided statistical significance will be defined at the level of p<0.05. Multivariable binary logistic regression will be used to provide adjusted estimates of effects. Variables to be entered to the model will be those that occur prior to the operation. Initially all these variables will be considered. Variables to remain in the final model will be selected by best using Akaike Information Criterion (AIC). Effect estimates will be presented as ORs and bootstrapped 95% CIs and p values. All relevant second order interactions will be examined. We will explore the possibility of creating a risk prediction model. Such a model would be derived from one half of the data set and validated in the other half. Data handling and analysis will be performed in REDCap and R (R Foundation Statistical Programming, Vienna, Austria).