Model-building on survivability of upper gastrointestinal bleed patient’s

• Main Authors: Pillay, Khuneswari Gopal, Mohd Padzil, Siti Aisyah
• Format: Conference or Workshop Item
• Language: English
• Published: 2018
• Subjects: T Technology (General)
• Online Access: http://eprints.uthm.edu.my/7068/1/P10449_283d7b7d1284090cbc276e23a6d8a38e.pdf; http://eprints.uthm.edu.my/7068/; https://doi.org/10.1088/1742-6596/1132/1/012067
• Institution: Universiti Tun Hussein Onn Malaysia

Statistical modelling by using Multiple Binary Logit (MBL) or Logistic Regression (LR) on a medical data has been a common practice by the researchers. However, there are no agreed guidelines for how best to carry out model-building using MBL to obtain the best model. This research will propose an appropriate guideline for beginners and demonstrate the flow of model-building process using Rockall score data as well as highlighting the significant factors of survivability of Upper Gastrointestinal bleed (UGIB) patients in Sabah. Rockall score is a scoring system used in identifying the risk of survivability for patients with UGIB. The patient’s data were retrieved from Hospital Queen Elizabeth in Sabah. Seven categorical variables related to the Rockall scoring system were studied and the steps to obtain best model using MBL were illustrated in four phases. The phases include all possible models, selected models, best model and goodness-of-fit test. All possible models were considered without interaction variables. A progressive elimination (one by one, least significant first) of the insignificant variables is carried out to a set of selected models (with significant variables). Model selection criteria AIC, corrected AIC (AICc) and BIC were used to single out the best model among the selected models. Pearson chi-square test and deviance chi-square test were carried out to ensure the best model validity and appropriateness. The results showed that the factors affecting the survivability of UGIB patients in Sabah are shock score, comorbidity and rebleed. In conclusion, the study showed that the Rockall scoring system had satisfactory validity for the prediction of shock score, comorbidity and rebleeding in patients with UGIB. There was a negative relationship between the clinical Rockall scores and patient outcomes in terms of shock score and comorbidity.