Enhancing patient care with predictive analytics: Early identification of code blue events, fall detection, and IVF success prediction

Guide(s)

Kumar, U Dinesh

Department

Decision Sciences

Area

Decision Sciences

University

Indian Institute of Management Bangalore

Place

Bangalore

Publication Date

3-31-2025

Year Awarded

March 2025

Year Completed

March 2025

Year Registered

June 2019

Abstract

The healthcare sector is an essential and complex industry encompassing a variety of services and establishments dedicated to the preservation and enhancement of human health. This includes hospitals, pharmaceutical companies, medical device manufacturers, and health insurance providers. With the advancement of digitalization, the healthcare sector has witnessed a recent progression in utilizing analytics, Machine Learning (ML) algorithms, and Artificial Intelligence (AI) to innovate healthcare delivery. The complete extent of the utilization of AI in enhancing patient care remains largely unexplored. This thesis looks into three significant problems related to the healthcare sector. The first problem focuses on the prediction of Code Blue (Sudden Cardiac Arrest) events in hospital settings. The second problem deals with the detection of fall events for the elderly population, and the third problem investigates the prediction of successful pregnancy outcomes following In Vitro Fertilization (IVF) treatment. The first chapter provides an overview of the thesis, followed by Chapter 2, which offers an in-depth review of the literature relevant to the three healthcare problems addressed in this study. In Chapter 3 of the thesis, we address the first problem, with the objective to predict sudden cardiac arrest for in-hospital patients using Explainable Artificial Intelligence (XAI). Code Blue is an emergency alert code activated when any in-hospital patient faces cardiac or respiratory arrest. In a hospital, a cardiac arrest patient’s survival depends on the early identification of signs like unresponsiveness or pulselessness and the initiation of cardiopulmonary resuscitation (CPR) at the earliest. This research aims to build an interpretable machine-learning model for predicting Code Blue events before they can happen. This study uses the electronic medical record data of the patient for the last 24 hours and the doctor’s clinical notes. We use the techniques of Natural Language Processing (NLP) to extract features from the doctor’s clinical notes. The extracted features are combined with other electronic medical data, such as vital information, pre-existing diseases, and organ dysfunction, to build the final prediction model. The study establishes that clinical notes are a significant source of information on a patient's condition and have a substantial impact on their treatment. The enhanced ability to explain the model will help healthcare professionals better understand the early warning signals of Code Blue. The prompt alert from an early warning signal ensures that patients receive immediate care from medical practitioners, thereby substantially increasing the potential for saving lives.In Chapter 4, we predict fall events among elderly people using deep learning algorithms while incorporating explainable AI techniques to provide reasoning behind the model’s prediction. The World Health Organization (WHO) claims falls are the =second most common cause of unintentional injury worldwide. Although various ML techniques are used to predict fall events, these methods often lack the ability to explain the reason behind their prediction. We have employed a transformer with a multi-head attention mechanism to study the fall detection problem. The selection of multi-head attention comes from its fundamental concept of self-attention, which forms the basis of our research approach. Our findings have explored the complexities of fall detection utilizing various deep-learning models. Among all these models, the chosen strategy has shown superior performance across multiple metrics used for model evaluation. The practical implications of the research findings in this paper are the design and development of new products, especially for smartwatch makers who track users, especially high-risk groups such as senior citizens. In Chapter 5, the focus is on “In Virto Fertilization” (IVF) pregnancy and the pregnancy results following blastocyst transfer for couples undergoing IVF treatment. A blastocyst represents the initial phase of an embryo characterized by a cluster of dividing cells derived from fertilized eggs. Individuals with infertility frequently face significant societal expectations. Couples coping with infertility may find it challenging to make decisions due to multiple social and emotional stressors. Disparities in information, together with the effects of these stressors, can have a significant impact on fertility treatment decisions. In this study, we propose an effective system to help individuals make decisions based on information. We developed a model that uses a variety of clinical and embryological parameters to assist couples in making the best decisions for their specific circumstances. Couples seeking infertility treatment frequently experience difficulties in determining whether to use their own or donor gametes due to several age-related factors. This study focuses on how a woman's age affects her preference for using her own or donor gametes. Our analysis of clinical indicators reveals an association between the source used and pregnancy outcomes. A strong moderating effect of female age on the source used in treatment was discovered, giving patients a thorough explanation of why physicians advocate for a particular source. Furthermore, our evaluation of embryo quality yielded diverse and noteworthy results. Within single embryo transfers, we found that embryos of grade B and grade C quality had equal pregnancy rates. These findings enhance our understanding of the complexities involved in IVF success and have the potential to influence the treatment approaches. The findings of this dissertation possess a significant implication for the incorporation of data-driven models to enhance patient care, applicable in both hospital and clinical environments as well as in remote settings. These findings are expected to support the healthcare industry in utilizing data-driven models for judicious decision-making and personalized patient recommendations. Through the implementation of AI and ML algorithms to tackle healthcare problems, this research aspires to contribute to evidence-based decision-making. Our work contributes to the literature of critical events, emotional decision-making, information asymmetry, and focuses on the interpretability and explainability of the model. This approach aims to not only enhance patient care but also to improve the efficacy of healthcare systems, optimize the distribution of resources, and encourage innovation in treatment strategies.

Pagination

xvi, 138p.

Copyright

Indian Institute of Management Bangalore

Document Type

Dissertation

DAC Chairperson

Kumar, U Dinesh

DAC Members

Saranga, Haritha; Pareek, Bhuvanesh

Type of Degree

Ph.D.

Relation

DIS-IIMB-FPM-P25-01

This document is currently not available here.

Share

COinS