In 1994, Dr. William Kissick introduced the iron triangle of health care (cost, quality, and access) in his book, “Medicine's Dilemmas: Infinite Needs Versus Finite Resources”. In recent years, the U.S. healthcare cost is becoming higher and higher, and if this trend continues, it will eventually make the healthcare unaffordable, hard to access, and with low quality. The central goal of my research is to improve the healthcare quality and access, but reduce its cost. To reach this goal, my approach is to
My research approach is consistent with the ideas promoted by evidence-based practice and the precision medicine initiative, which require healthcare providers to collect patient data from various sources (e.g., lifestyle, environment, genomics, and medical records) and integrate them with solid evidence from the literature to come up with personalized treatment plans.
In my health IT system development and evaluation research, I have created and evaluated stand-alone software programs, mobile apps, and web-based portals so that we can collect valuable data from different sources, integrate them together, and deliver the clinically relevant information to patients and healthcare providers in an easily accessible approach. I also designed new questionnaires to evaluate usability, accessibility, and security of mobile health apps.
Currently existing well-being assessment scales often only evaluate a person's well-being in one particular aspect, such as physical health, socioeconomic status, behavioral well-being. Plus, most of the existing well-being scales were designed either for the general population, or patients with one specific disease. In this project, we collaborated with a free health center (formerly FOCUS Pittsburgh, now Neighborhood Resilience Project - NRP) in the Hill District in Pittsburgh and developed a comprehensive well-being assessment system, which assess a person's well-being in five different aspects: physical health, behavioral health, relational health, socioeconomic status, and spiritual health. The targeted population is people living in medically underserved communities. The result of this assessment can provide a comprehensive picture about the person's well-being and can be used to guide the design of personalized intervention for that person. This system includes a newly designed and validated comprehensive well-being assessment scale, multiple mobile apps for data collection, a custom-built Electronic Health Record (EHR) system for managing the collected patient data, a data integration and analytics program for analyzing the collected data, and a web-based portal for presenting the summary of well-being assessment results in an easily accessible and graphical manner.
In current language rehabilitation programs, PwCDs use dedicated Augmentative and Alternative Communication (AAC) devices to create sentences. The language performance data are stored on the device and they can only be downloaded by Speech Language Pathologists (SLPs) when PwCDs visit SLPs' office. The SLPs need to use a specific software program to analyze the downloaded data to understand the data collected by the device. Therefore, the data analysis is always done after the clients’ visit, if it is performed at all. We created a mobile app (EuTalk) to facilitate communication between PwCDs and others. This app can collect clinically relevant data items and forward the data to a secure database. We also built a web-based portal to access the database and present the analyzed data in various methods, such as diagrams, tables, and reports. This system makes it possible for SLPs to track their patients' situation in real time and adjust their treatment approach accordingly. This work was funded in part by grants from the National Institutes of Health and the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR).
In recent years, clinical research projects have often been conducted by multiple study sites and various types of research data have been collected from multiple sources, managing these complex clinical data sets has become a big challenge for data scientists, especially if clinical researchers frequently change their plans for data processing and data analysis procedures. In this project, we created an elegant web-based research data preprocessing and retrieval system, in which clinical researchers can easily change the rules for data preprocessing without the need of understanding the data structures used to store the raw data and click a few buttons to retrieve the processed data set. The obtained data set is ready to be loaded into a statistical analysis package (e.g., SPSS, SAS, and R) for desired statistical analysis or modeling. This system has been implemented for a complex data set in a collaborative project with researchers in the Department of Communication Science and Disorders (CSD) at the University of Pittsburgh and a usability study on this web-based system has been performed. This project was funded in part by a grant from the Veterans Affairs (VA).
The improvement in DNA sequencing technologies makes it affordable for patients to perform whole genome sequencing. The research on the genetic association to disease also produced a large amount of results. However, both genomic data analysis and research data integration are challenging for most of the current healthcare providers in the workforce. To make it convenient for healthcare providers to utilize patient genomic information in their clinical practice, we created a patient genomic information analysis and management system. We first collected data from multiple sources, such as research papers and molecular biology databases, and integrated the collected data into one database. We then built a genetic variation analysis program to analyze patient's genetic variations against disease or medication associated variations. We constructed a graphical user interface for healthcare providers so that they can easily query patient's genomic information and view patient-specific genetic analysis reports. This project was supported by a fund from the Jordan government.
Usability of a software program is critically important to user engagement. In the past, we used existing usability questionnaires (e.g., Post-Study System Usability Questionnaire -PSSUQ, System Usability Scale - SUS) to evaluate the usability of the mobile health apps we created. However, PSSUQ and SUS were designed for general software systems. They cannot capture the usability issues unique to mobile health apps, such as small form factor of mobile devices and user’s concern about privacy. We conducted a systematic review on those existing usability questionnaires and designed a new usability questionnaire (mHealth App Usability Questionnaire - MAUQ) in order to evaluate the usability of mobile health apps. The results from MAUQ are consistent with PSSUQ and SUS on general system usability. MAUQ can also identify unique usability issues in mobile health apps. This work was funded in part by a grant from the NIDILRR.
There are a huge number of mobile health apps in the market, however, most of them were created without persons with disabilities’ needs in mind, even though more than 12% of the U.S. population has one or multiple types of disabilities. In this project, we designed and implemented several new mobile app accessibility features in a mobile app so that persons with various types of disabilities, such as vision impairment and dexterity impairment, can conveniently use this mobile apps to manage their own health. Our approach is highly generic and therefore other mobile app developers will be able to adopt our approach to improve the accessibility of their mobile apps, which will make a large number of mobile health apps accessible to persons with disabilities. This work was funded in part by a grant from the NIDILRR.
In healthcare, information security is extremely important because patient health information is highly sensitive. One single security breach in a healthcare organization can cost millions of dollars and also the reputation of the organization, which can be devastating to the organization. Therefore, it is highly desired to make health IT systems secure. We conducted multiple security related projects, for instance, identifying and evaluating security features in mobile health apps, collecting user's perspectives on security features in mobile apps, creating mobile apps for providing security education to app users, determining user preferred privacy protection methods in mobile health apps, and conducting systematic reviews on security in telehealth systems, all with the purpose for assessing and enhancing security of mobile health apps and telehealth systems. This work was funded in part by a grant from the NIDILRR.
In my data analysis and computational modeling research, I designed methods and algorithms to understand large-scale biological or medical data sets. I also created computational models using various approaches, especially statistical modeling and agent-based modeling, to understand the mechanism of disease and determine the consequence of therapeutic approaches in silico.
Researchers use animal models to understand the mechanism of disease and evaluate the effectiveness of medication. However, in many cases, a very effective medication in animal models does not work well on human beings. One reason behind this is the difference of the genetic makeup between animals and human beings. In this project, we performed gene annotation and analyzed orthologous genes in different species with the purpose of understanding their differences in gene structure. More specifically, I have worked on 1) designing and developing algorithms and tools for DNA sequence analysis; 2) creating methods for annotating genomes; 3) investigating evolution of gene intron-exon structure in multiple species and creating a database to document all the reported intron polymorphism cases in the literature; and 4) building algorithms for analyzing patient genomic data.
Because of the wide adoption of Electronic Health Record (EHR) systems, it becomes convenient for us to access detailed patient data. My collaborators in the Department of Critical Care Medicine at Pitt accumulated a huge database with itemized data for more than 54,000 patients who were admitted to one of eight ICUs in the University of Pittsburgh Medical Center (UPMC) in a 8-year period. We performed statistical analysis and modeling on this database and answered critical research questions related to acute kidney injury (AKI), a complication with high mortality rate for older patients. More specifically, we investigated the cost of treating AKI and its relationship with different situations such as community acquired AKI, hospital acquired AKI, multiple risk factors, and the timing of detecting AKI severity stages based on the data collected in the database.
I have worked with collaborators in the Department of Communication Science and Disorders at Pitt on a statistical analysis and modeling project for understanding the language comprehension process in the brain of patients with aphasia. Structural equation modeling was used to determine the correct model describing the relationship among conflict resolution, short-term memory, and language processing. We also used statistical analyses to determine whether a shortened language comprehensive test (28 subtests in total and they were repeated 10 times for reliability purpose) is equally good as the full test. Our results indicated that they can safely remove 10 subtests and reduce the number of iterations to 3 times, but still obtain a comparable outcome. This result alone can significantly reduce the time of the test.
Computational modeling is one important approach for understanding dynamic and complex systems. I have created computer models in multiple fields using various approaches. For instance, I created differential equation models to simulate the interactions among quarks in hadron. I built statistical models to identify patterns in cross-species sequence alignment and the trend of AKI development. I used agent-based modeling technique to investigate various medical problems, such as the formation of coronary artery restenosis following balloon angioplasty procedure and bare metal stent implantation; the development of pressure ulcer following spinal cord injury; the development of osteoporosis. These computational models can be used to both perform patient-specific simulations and conduct large-scale simulations on a chosen population. The later could be considered as in-silico clinical trials. This project was supported in part by a grant from the National Science Foundation.
Sometimes, we need to deliver intervention to a large number of people in one community, however, because of restrictions in financial and human resources, it is not practical to deliver the intervention to everyone in the community simultaneously. In this case, we need to determine who should receive the intervention first so that these people can provide positive influence to others in the same community. For that purpose, we collected social network information via a brief survey, performed social network analyses, and identified the top influential persons in a community. These opinion leaders are also the ones who are typically open to innovative ideas and willing to apply the new intervention on themselves. This approach can be applied in many fields, especially behavioral change interventions such as smoking cessation and weight loss.
One of my major projects in the last two years is the statistical analysis and machine learning on a multimodal data set collected from 1000 patients with chronic low back pain. In this project, we are collecting patients' electronic health records, wearable sensor data, mobile health app data, answers to a number of behavioral and performance questionnaires, genomic data, and proteomic data. My responsibility in this project is to conduct statistical analysis and machine learning to identify key factors which can be used to determine the best treatment pathway for individual patients/patient groups. This project is currently ongoing. It is supported in part by a grant from the NIH HEAL Initiative.