Olawale Ayilara, a second-year PhD student in the Max Rady College of Medicine’s department of community health sciences, in Montreal during his VADA work placement.
Program enriches health data analysis and techniques
The Visual and Automated Disease Analytics (VADA) program, which trains graduate students to translate complex health data to monitor chronic diseases and quickly detect and prevent the outbreak of infectious diseases, will graduate its first PhD students this summer.
The program, a collaboration between the University of Manitoba and University of Victoria, is completed concurrently by graduate students while working toward their graduate degree. It can be completed in one year at the master’s level, or two years at the PhD level. Two master’s students from U of M graduated last year.
Olawale Ayilara, a second-year PhD student in the Max Rady College of Medicine’s department of community health sciences, is among three U of M PhD students who will complete the program in August 2019. Last summer he interned at the Human Computer Interaction (HCI) Lab at the U of M’s department of computer science, and with the Canadian Network of Observational Drug Effect Studies (CNODES), a national research network that conducts drug safety and effectiveness research using population-based administrative health data, at McGill University in Montreal.
Why did you choose to apply to the VADA program?
When I was putting my PhD application together, I heard about the program from my advisor, Dr. Lisa Lix, [professor and Canada Research Chair in methods for electronic health data quality.] It trains graduate students on the development and use of automated techniques and tools to collect, analyze and visualize chronic and infectious disease data. The program’s objective aligns with my desire to put knowledge acquired into practice in the health sector by addressing problems that have real-world relevance.
What is the focus of your research within the program?
Patient Reported Outcomes (PROs) are appraisals from patients about their quality of life, which includes self-perceived physical, mental and social health. The way that people interpret and respond to PROs may change when they have a medical procedure or major health event, such as surgery or a new chronic disease diagnosis. This phenomenon is known as response shift. If unaccounted for in the analysis of PROs, response shift can lead to an under/over estimation of change in PRO scores over time, which can mask intervention effects.
My research will develop and apply new response shift methods. My research focuses on longitudinal item response models and latent variable mixture models. I will develop new statistical methods that will benefit health-care providers, as they will be able to better interpret PROs and develop routinized software tools to analyze these data.
How do latent variable mixture models and longitudinal item response models shape your work?
A latent variable model is a statistical model that relates a set of observed variables to an unobserved variable, such as depression, anxiety, mental or physical health that cannot be directly measured. We can measure these latent traits by giving respondents a set of questions (called indicators) that are related to these traits.
Longitudinal item response models are one kind of latent variable model, but we can incorporate the variability within individual responses across time.
What did you find most rewarding about participating in the VADA program?
The VADA program training helped broaden my horizons in analyzing complex health data and advanced my techniques in data visualization. I have also developed the skills to conduct collaborative and multidisciplinary research. By taking part in a course on the foundations of disease analytics, I became familiar with new computational software and methods, and developed my professional skills in networking and working with complex datasets.
How did the internship enhance your skills?
In the HCI lab under the supervision of Dr. Pourang Irani, I developed skills on visualization of PROs to understand missing data patterns and selection of features that are predictive of the outcome of interest or the missing data. At CNODES, I developed a training manual on how to use simulated datasets in studies about the safety of prescription medication.
What are your career plans going forward?
I hope to complete my PhD program in 2020 or 2021. My primary career goals are to conduct innovative biostatistical research and train students in biostatistical methods. I am also very interested in working as a member of multidisciplinary and collaborative research teams, bringing my skills in biostatistics to solve real-world clinical and population health problems.