Models based on artificial neural networks (ANN) are able to predict the outcome of several disorders. However, there is currently no predictive model for acute lower-gastrointestinal hemorrhage.
In this study, a research team from the United States investigated whether ANN models could predict clinical outcome in patients with acute lower-GI bleeding.
The team constructed ANN and multiple-logistic-regression (MLR) models using non-endoscopic data from patients admitted with acute lower-GI hemorrhage.
|The predictive accuracy of ANN was significantly better than that of BLEED.|
The team compared the performance of their model in classifying patients into high- and low-risk groups with that of the validated scoring system BLEED.
They measured recurrent bleeding, death, and therapeutic interventions for the control of hemorrhage.
The ANN models were trained with data from 120 patients admitted during a 12-month period. They were then validated using data from 70 patients who were admitted in the following 6 months.
The team then validated the ANN models externally. They also compared ANN and MLR in with 142 patients admitted to an independent institution.
The researchers found that the predictive accuracy of ANN was significantly better than that of BLEED, and was similar to MLR.
During the external validation ANN performed well.
Dr Ananya Das's team concluded, "ANN can accurately predict the outcome for patients presenting with acute lower-gastrointestinal hemorrhage".