Participant: PROMISE AGEP Research Symposium
Department: Computer Science
Institution: University of Maryland, Baltimore County (UMBC)
2016 RESEARCH SYMPOSIUM ABSTRACT
Diabetes Comorbidities Analysis
Various studies have shown that the relationship between diabetes and other diseases impacts the clinical outcomes in a patient. However, more studies are needed to understand how the temporal occurrence of clinical events affect this relationship. In this research, we identify the common comorbidities of diabetes and investigate the timing of the associated clinical events affect the patient outcome. Understanding this relationship can help improve treatment practices. We use ICU Medical Information Mart for Intensive Care III (MIMIC-III) data comprising of over 58,000 hospital admissions in critical care units of the Beth Israel Deaconess Medical Center. We use the data from MIMIC-III to train a learning model that predicts potential clinical events in patients with diabetes. The goal is to incorporated this model to a clinical decision support system that can be used by clinicians at the point to care.
2015 RESEARCH SYMPOSIUM ABSTRACT
Predictive Modeling with Patient Reported Data
Machine learning techniques have been utilized in clinical practice to create predictive models for diseases with impressive success. In personalized medicine, these models have helped identify patients with various risk factors using both genomic and clinical data. However, these predictive models do not contain information extracted from patient reported outcomes. Part of the lack of patient reported data is the difficulty involved in capturing, processing, and extracting information from patient reported sources. Research has shown that patient-reported data contains useful information that can be used to create a safety risk profile for each individual patient. The purpose of my research is twofold. First, I create an ensemble machine learning algorithm that incorporates patient-reported data as an input. Second, I use this algorithm to predict safety events with improved accuracy over existing systems. A safety event is a situation where best or expected practice does not occur. Predicting safety events can help avoid them, therefore improving patient outcomes and experience cost savings. As a proof of concept, I investigate the effectiveness of using the algorithm in predicting safe antibiotics prescriptions for hospitalized patients, as well as predicting central line infections.
2014 RESEARCH SYMPOSIUM ABSTRACT
Electronic Health Records Impact on Safety and Efficacy
The use of information technology in healthcare provision has increased significantly over the last decade. New regulations require healthcare providers utilize certain levels of technology. Technology in healthcare has been used in various ways including clinical decision support, prescribing of medication, validation of clinical data, genome mapping, etc. Several concerns such as privacy of patient data, confidentiality, clinical workflow interruption, and data ownership have been raised in various implementations of technology in healthcare. In particular, electronic health records (EHRs) have been adapted by many providers with the aim of improving the quality of care. However, not much research has been done in determining the impact of EHRs on clinical outcomes. This paper does a survey of clinical institutions that have adapted EHRs and the implications on safety and efficacy. A distinction is made between the direct and indirect impacts, and a connection is made between this impact and observed clinical outcomes.
Isaac Mativo earned his Bachelor of Science degree in Computer Science from University of Maryland Baltimore County in 2000. He received his Master of Science degree in Computer Science from University of Maryland Baltimore County in 2011. In 2012, he joined the doctoral program in Computer Science in at the University of Maryland Baltimore County.
Mr. Mativo is a senior software engineer at Leidos Biomedical Research Inc. where he works on precision medicine in cancer research. Previous employers he has worked for include Computer Sciences Corporation, CareFirst BlueCross BlueShield and the University of Maryland School of Medicine.
Mr. Mativo has been a recipient of several academic and professional honors and awards including a certificate of outstanding achievement at Computer Sciences Corporation. Mr. Mativo is a member of the American Medical Informatics Association (AMIA), Upsilon Pi Epsilon Computer Science Honor Society, Golden Key Honor Society, and University of Maryland Baltimore County Honors College.
Mr. Mativo’s research area is in the intersection of machine learning and clinical informatics. He is interested in using machine learning techniques to create predictive models that improve clinical outcomes. His goal is to have computationally enabled clinical systems that process data from disparate sources and avail them to clinicians at the point of care. Mr. Mativo lives in Columbia Maryland with his wife and three children. He is an avid marathoner and an active volunteer in his community.
GENERAL SUMMARY OF GRADUATE RESEARCH
My master’s paper was on the technology used in personalized medicine, focusing on electronic medical records (EMRs). In that work, I identified the role EMRs play in healthcare provision as well as the barriers that exist for their adaption. I also assessed their actual usefulness and design considerations. I then identified key elements that needed to be considered in a successful EMR design and implementation. I wrote a paper on personal health records (PHR) where I surveyed several implementations. I analyzed the integration of PHRs and EMRs and identified technological challenges including secure messaging, authentication, and privacy.
My PhD research is in clinical predictive modeling. I use a machine learning ensemble algorithm to come up with results that can be used in a clinical decision support system. Part of my research involves feature extraction from patient reported data. My current domain in patient safety is in central line infections and antibiotic selection in septic patients.
SELECTED LIST OF PRESENTATIONS AND PUBLICATIONS
- Clinical Predictive Modeling with Patient Reported Data, Mid-Atlantic Student Colloquium on Speech, Language and Learning (MASC-SLL) January 2015, Johns Hopkins University
- Predictive Modeling for Clinical Decision Support Systems, Ebiquity Lab, Computer Science Department, University of Maryland Baltimore County, October 2014
- Policy, Research, and Technology Challengers in Healthcare, American Medical Informatics Association (AMIA) Policy Invitational Meeting, September 2014, Washington DC
- Impact of Electronic Medical Records on Safety and Efficacy, PROMISE AGEP Research Symposium, February 2014, University of Maryland College Park
Disclaimer: Information on this page has been provided by and is owned by the student presenter.