Predicting outcome following colorectal cancer surgery using a colorectal biochemical and haematological outcome model (Colorectal BHOM)

Abstract

The Colorectal Biochemistry and Haematology Outcome mortality model suggests good discrimination (c-index > 0.8) and uses only a minimal number of variables. However, it needs to be tested on independent datasets in different geographical locations.

Model for risk adjustment of postoperative mortality in patients with colorectal cancer

“…Seven of the models were developed for patients with colorectal cancer, and the largest of these included 7374 patients. Three models were explicitly intended for case‐mix adjustment, although two of these included predictors that can be influenced by the provider, such as surgical urgency and procedure, and the other model included only 55 deaths. Ten of the models included data items not routinely collected in national databases, such as serum measurements, lifestyle factors including smoking and alcohol consumption, and clinical observations including blood pressure and body mass index.…”

“…Fifteen prognostic models for short‐term mortality after colorectal surgery were identified in English‐language journals, the largest of which included 975 825 patients ( Table S1 , supporting information). Seven of the models were developed for patients with colorectal cancer, and the largest of these included 7374 patients.…”

“…Three of these nine models were validated in a different health system and their discriminatory ability in these external data sets was often substantially lower than in the original data sets (0·90 reduced to between 0·69 and 0·78; 0·78 reduced to between 0·70 and 0·73; 0·8 maintained at 0·81). Four of the other six models are likely to suffer from a large amount of overfitting as the models were developed on fewer than 300 deaths. Although these models appear to show very good discrimination, with internal C‐indices ranging from 0·81 to 0·93, it is likely that their discrimination would be considerably lower in other settings.…”

“…The largest study to date – including 7374 patients of whom 553 died – considered only patients who died in hospital, thereby not counting those who died as a result of complications after discharge. In addition, most risk models were developed using suboptimal modelling approaches (for example selecting risk factors based on significance testing, categorizing continuous risk factors, ignoring potential interactions between risk factors, or having small sample size).…”

Efficacy of VBHOM to Predict Outcome Following Major Lower Limb Amputation

“…17 Using a similar multivariate approach serum albumin has recently been shown to be independently predictive of short-term mortality following colorectal surgery. 18 Thus, patients with low albumin levels are most likely to be undernourished, have chronic medical illness, and lack adequate energy reserves to recover appropriately following major surgical stress.…”