Improvement of Dialysis Dosing Using Big Data Analytics

A Master of Science thesis in Biomedical Engineering by Syeda Leena Mumtaz entitled, “Improvement of Dialysis Dosing Using Big Data Analytics”, submitted in April 2021. Thesis advisor is Dr. Abdulrahim Shamayleh. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).

Abstract

Data is transforming the healthcare sector and making it more dependent on data science. Data science is becoming a critical tool that allows looking at the data generated from various sources, such as patient health records, diagnosis, treatment, smart devices, and wearables. Extracting insights from health data has the potential to transform the healthcare from traditional symptom-driven practice into a precision personalized medicine. The dialysis treatment generates a vast amount of data that can be utilized. Data of each dialysis patient constitutes over 100 parameters that must be regulated every dialysis session. Moreover, an individual dialysis dosing may depend upon complex linkage within multiple clinical and demographical parameters, early dialysis prescriptions, medications, or other health interventions. With dialysis complications, understanding the electrolyte parameters and predicting their outcome for each patient to deliver the optimal dialysis dosing is a challenge. This research approach is intended to improve dialysis dosing from the emerging data and the rising volume of dialysis patients, with the purpose of increasing patient’s quality of life and their welfare from the right dialysis treatment. Exploratory data analysis and data prediction approach were performed to provide insights on how to improve the patients’ dialysis dosing. Analysis of vital electrolytes displayed high variability amongst patients, which identified the needs to improve the dialysis dosing. Four data prediction models were used to predict patient electrolytes from various parameters. The models include decision tree, neural network, support vector machine, and linear regression. The results from the prediction identified that pre urea (BUN), anticoagulation, HBA1C, gender, and cumulative blood volume, had the most significant predictor weights. The important predictors interpreted that patient’s lifestyle and diet patterns are the major factors towards improper variability of the electrolytes.