Publications by authors named "Fadi Islim"
Article Synopsis
- Using proper imputation methods is crucial in tree-based predictive analytics for healthcare, as learning missingness patterns can drastically reduce performance in different datasets.
- A novel simulation generated synthetic electronic health records to highlight this issue and demonstrated that imputation methods like randomized and Bayesian regression can alleviate the problem.
- The PICTURE system, developed using extensive patient data, showed promising results in predicting patient deterioration compared to existing early warning systems, achieving an AUROC of 0.83.
Introduction As the electronic health record becomes more sophisticated, commensurate advances in cost accounting have risen as a top priority for hospital leaders. This study explored: 1) the average time to complete common nursing tasks documented in the electronic health record, 2) nursing-related tasks that remain undocumented, 3) the association between observation data and actual nursing documentation, and 4) considerations for model development and report design to be used for activity based cost accounting in nursing. Methods This was an observational study completed on acute care inpatient nursing units at a large academic medical center.
View Article and Find Full Text PDF