Coronary Assessment for Mortality Prediction (CAMP)

The study uses a retrospective observational study design.

Due to the nature of the subject of the study and its retrospective design no patient contact will be necessary and all imaging reviewed will have been acquired as part of their routine level of care.

Plan for the gathering of study data

A list of patients with the following surgical codes will be identified from the hospital surgical performance database to create retrospective dataset of 5 years' worth of cases (to December 2013). This date range may be extended to ensure an adequate sample size. :

Surgical codes (Healthcare Resource Groups (HRG) codes) to be included:

F31 Large Intestine - Complex Procedures F32 Large Intestine - Very Major Procedures F33 Large Intestine - Major Procedures with complications and comorbidities (w cc) F34 Large Intestine - Major Procedures without complications and comorbidities (w/o cc) F36 Large Intestinal Disorders >69 or w cc F37 Large Intestinal Disorders <70 w/o cc

F41 General Abdominal - Very Major or Major Procedures patients over 69 years of age (>69) or w cc F42 General Abdominal - Very Major or Major Procedures patients under 70 years of age (<70) w/o cc F46 General Abdominal Disorders >69 or w cc F47 General Abdominal Disorders <70 w/o cc

Cross-reference with Somerset Cancer Register and remove any who were not recognised as having bowel cancer.

This patient list will be cross-referenced with records of peri-operative mortality (within 90 days)

Propensity match each deceased patient to a patient who survived beyond 90 days. Our matching rules will be in descending order of importance :

• HRG code
• gender,
• cancer staging
• Age+/-5yrs.
• American Society of Anesthesiologists (ASA) Grade

Where there are multiple potential matches then this will be resolved by consultation between the research team members.

Pairs of CT scans (deceased plus matched control) will be procured from WebPACS (Web Picture Archival Communication System).

Scans will be anonymised and stored on a dedicated secure radiology research server.

Abbreviated Calcium scores will be generated for all patients in the trial by a trained radiologist using the calcium scoring method. Calcium burden will be recorded on a scale of 1-12 . The radiologist will be blinded to the group that the patient belongs to, to minimise bias.

Scans (deceased patients and matched controls) will be analysed in batches of 30 pairs. Radiologist (s) will be blinded to group allocation. Scans will not be "paired" at the time of reading.

Interim statistical analysis (of calcium scores) will be performed after each batch of scans.