An Integrated Framework Using Variable Encoding-TF-IDF-PCA-Classification for Predicting Adverse Event Action

Adverse events have a significant impact on both patients and healthcare institutions, leading to substantial financial losses, reputational harm, and operational disruptions. As an action plan, the first step to mitigate these events involves having a quality representative and a nurse spend hours daily determining which events require further action and which do not, after an adverse event is reported. This process is time-consuming, especially when the volume of adverse events is high. In large-scale organizations, the increased volume of reports can lead to inaccuracies and biases in decision-making, as human judgment is fallible. This research aims to develop a framework that integrates categorical encoding, Term Frequency-Inverse Document Frequency (TF-IDF), and Principal Component Analysis (PCA) with classification models to automatically predict whether an adverse event should be addressed with further action or dismissed to save time, effort, and money while achieving more accurate and unbiased adverse event classification.

Attendees will be able to apply the proposed methodology to other datasets and adapt it for different objectives, enabling them to customize it based on their specific needs.

Attendees will be able to distinguish between adverse events that require escalation and those that do not using machine learning models and decision-making criteria.

Attendees will be able to apply categorical encoding techniques, such as binary encoding, and incorporate oversampling strategies to balance healthcare datasets for more accurate machine learning predictions and explain the importance of using dimensionality reduction techniques and feature extraction as TF-IDF in improving predictive accuracy for adverse event classification.

Authors

A graduate research associate with a strong background in simulation for central fill pharmacy systems. Having gained a year of valuable experience in this area, I have developed a keen understanding of how to optimize healthcare processes for efficiency and accuracy. Currently, I am expanding my expertise as a Systems Engineering Graduate Research Associate at Memorial Sloan Kettering Cancer Center. Focusing on improving patient care and operational processes in one of the most renowned cancer treatment and research institutions.

SUNY Distinguished Professor and Chair, Systems Science and Industrial Engineering
Director, Watson Institute for Systems Excellence
Director, Healthcare Systems Engineering Center
Graduate Program Director, Executive Master of Science in Health Systems

Dr. Susan Lu has a  BS from Hebei University of Technology, a MS from Tianjin University, and a PhD from Texas Tech University. She is a professor at the Systems Science and Industrial Engineering Department in Binghamton University.