Binary Logistic Regression Model to Predict the Need of Intensive Care Unit (ICU) Support of Sepsis Patients during Hospital Stay
Keywords:Binary logistic regression model; Sepsis, Severe sepsis; Intensive care support; Hospital mortality
Hospital mortality in ICU patients of Bangladesh suffering from severe sepsis, which is significantly higher than countries reported. Binary logistic regression model of important clinical variable will use for this study to predict the category of severe sepsis, fatal outcome and need of Intensive Care Support (ICU) during hospital stay. The aim of this study was to achieve accurate prediction of the category of severe sepsis and the need of ICU of sepsis and severe sepsis patients by using binary logistic regression model. This study was carried out in the Department of Medicine, Dhaka Medical College Hospital (DMCH) during January 2015 to December 2015.We select 100 patients from them 35 were sepsis and 65 were severe sepsis. Data was collected in a pre-designed proforma. All data compiled together and statistical analyses was carried out by using IBM SPSS Statistics 22.0 & MS-Excel 2016. Total 100 patients from both the groups (sepsis 35 and severe sepsis 65) completed the study. The aim of the logistic regression analysis models was to predict outcome variables like improvement during hospital stay, need of ICU support, fatal outcome and fall in the category of severe sepsis. In our study, we found 55 of the 65 cases (84.6 %) are classified correctly as severe sepsis, 22 of the 35 cases (62.9 %) were not categorized as severe, 4 of the 14 cases (28.6 %) who had fatal outcome during hospital stay are classified correctly. Overall, 77% of the cases are classified correctly 83 of the 86 cases (96.5 %) who did not have fatal outcome during hospital stay. Overall, 87 % of the cases are classified correctly. 74 of the 79 cases (93.7 %) who required Intensive Care Unit (ICU) support during hospital stay are classified correctly. 5 of the 21 cases (23.8 %) who did not require ICU support during hospital stay are classified correctly. Overall, 79 % of the cases are classified correctly. From this study, it is concluded that sepsis & severe sepsis is very common in a hospital of Dhaka at a tertiary level. We found by applying binary logistic regression model there 77% severe sepsis patients, 87% fatal outcome and 93.7% patients need intensive care unit (ICU) support. These results suggest that binary logistic regression model can help to predict need for ICU support and eventually help to measure the outcome during hospital stay.
J Shaheed Suhrawardy Med Coll 2021; 13(2): 94-99