UCLA

Chronic diseases are the leading causes of death and disability in the United States. More than 70% of deaths among Americans are caused by chronic diseases and more than 133 million Americans have at least one chronic disease. Due to the prevalence of chronic disease-related issues, it is prudent to seek out methodologies that would facilitate the prevention, monitoring, and feedback for patients with chronic diseases.

This dissertation describes WANDA (Weight and Activity with Other Vital Signs Monitoring System); a system that leverages sensor technologies and wireless communications to monitor the health-related measurements of patients with chronic diseases. The system was developed and validated in conjunction with the Computer Science Department, the School of Nursing and Ronald Regan Medical Center at the University of California, Los Angeles to enable real-time patient monitoring, user task optimization, missing data imputation and key clinical symptom prediction. The main contributions of designing and developing the WANDA system are 1) data abstraction and integration of the server side; 2) development of the smartphone and web applications; 3) data backup and recovery; 4) algorithm design and development of missing data imputation; 5) algorithm design and development of task optimization and early adaptive alarm; and 6) system deployment for clinical trials.

The WANDA system is a three-tier architecture consisting of wireless sensors, web servers, and back-end data analytics engines. The first tier comprises sensors that measure patients' vital signals and transmit data to the web server tier. The second tier consists of web servers that receive data from the first tier and maintains data integrity. The third tier is a back-end database server that performs data backup and recovery and various data analysis including dynamic task optimization, missing data imputation and adverse event prediction.

The WANDA dynamic task optimization function applies data analytics in real-time to discretize continuous features and apply data clustering and association rule mining techniques to manage a sliding window size dynamically and to prioritize required user tasks. The developed algorithm minimizes the number of daily action items required by patients using association rules that satisfy a minimum support, confidence, confirmation and conditional probability thresholds. Each of these tasks maximizes information gain, thereby improving the overall level of patient adherence and satisfaction. Experimental results from applying EM-based clustering and Apriori and confirmation-based rule mining algorithms show that the developed algorithm can reduce the number of user tasks by up to 76.19% with higher confidence levels.

Although missing data is highly undesirable as automated alarms may fail to notify healthcare professionals of potentially dangerous patient conditions, many studies reported high missing data rates in remote health monitoring. In this dissertation, I exploit machine learning techniques including projection adjustment by contribution estimation regression (PACE), Bayesian methods, and voting feature interval (VFI) algorithms to predict both non-binomial and binomial data. The experimental results show that the aforementioned algorithms are superior to other methods with high accuracy and recall. This approach also shows an improved ability to predict missing data when training on entire populations, as opposed to training unique classifiers for each individual.

The WANDA early adaptive alarm function discretize continuous features, applying Expectation Maximization (EM) clustering and association rule mining techniques for predicting future sensor readings and system non-use dynamically. The experiment results shows that developed algorithm can predict upto 27.08% of sensor readings and system non-use within the next three days.

Additionally, the results of performed clinical trials shows that patients monitored by WANDA are less likely to have readings fall outside a healthy raange and have higher adherence rate and more communications with health care professionals.