Nathaniel Osgood - Professor
Office : 254.4 Thorvaldson (Computational Epidemiology and Public Health Informatics Lab)
Phone: (306) 966 6102
Fax: (306) 966 4884 Email: osgood 'at' cs.usask.ca
BS EECS (MIT), MS EECS (MIT), PhD CS (MIT)
My research is focused on providing cross-linked system simulation, mobile data collection, and theory-informed machine learning/artificial intelligence tools to inform decision making in health. Such tools can, for example, aid public health decision makers in putting into place cost-effective preventive policies, design more effective screening or treatment strategies for an illness, help support epidemiological models that learn from incoming evidence and are kept perpetually up-to-date with the latest evidence so as to provide for more reliable policy planning. Such tools can further enable insight into the causes underlying changes in the number of cases of a disease reported, and react more quickly to an outbreak of infectious disease when it occurs.
Tools of choice include supplementing system science dynamic models (particularly Agent-Based models, System Dynamics, and discrete event simulation) with particle filtering and particle MCMC with such system science models, systems for visualizing state space reconstruction and for Convergent Cross Mapping (CCM), advancing machine learning and dynamic modeling toolsets, GPU-based computational statistics algorithms (PMCMC and Particle Filtering) and CCM. All such tools are applied within the health sphere, as this is our elected point of focus and dedication. We further make extensive investment in machine learning to sharpen and broaden the health surveillance, including using our Ethica epidemiological smartphone and wearable-based data collection system, time series of search volumes, time series of machine-learning-classifed Twitter messages, web-scraped data, and other mechanisms. For example, as part of our strategy of using twitter for health surveillance in Canada, we have amassed more than 100M tweets. The types of data we obtain through surveillance data are classified using machine learning tools (including more traditional tools through to deep learning) to flag tweets of relevance and inform our particle-filtered and PMCMC-regrounded models.
Our work includes both application and methodological components.
On the application side, our research involves collaborating with cross-disciplinary teams to create tools to inform the design of health interventions that are high leverage, robust, and cost-effective. Such applications are almost always pursued in close collaboration with broader teams, frequently including those with close clinical familiarity of the diseases and/or pathogens involved (particularly doctors and nurses), epidemiologists, biostatisticians, public health nurses, and researchers or others involved in data collection and surveillance.
Please see our applications page for more information on this work.
Data sources that we tap in this area include smartphone and wearable based data collection (particularly via the Ethica epidemiological data collection system, which emerged from our iEpi project), social media and search mining.
To help make sense of such data and make it actionable at policy, health services and clinical levels, we use a variety of tools, including Agent-Based modeling and ODE (System Dynamics, particularly when leveraged by Particle Filtering) and Discrete Event modeling & hybrids, MCMC, and PMCMC, deep learning for symptom and health-behaviour recognition, additional types of Machine Learning tools (e.g., HMMs and Bayesian Networks), CCM and State-Space Reconstruction. Much recent work has focused on fully homomorphic encryption and privacy-preserving technologies for health and social data. Where they fill a key gap, we also develop apps for smartphone and web platforms.
Some may be interested in a list of selected elements of my dynamic models (Agent-Based models, System Dynamics models, Hybrid Models, etc.) in Health.
Our methodological work reflects the fact that practitioners of dynamic modelling for public health currently face many hurdles and challenges in building, validating, executing, modifying and understanding their models.
Please see our methodological work page for additional information.
Our most recent contribution here comes in terms of the Ethica epidemiological data collection system, which provides a powerful platform for quickly defining , refining, running and monitoring, and analyzing and reporting on epidemiological data. This platform can tap into the exceptional power of smartphones as a health/mHealth data collection platform, but also leverages wearable devices and web-based data collection from consenting respondents. All of this work is undertaken in the context of strong privacy guarantees, including the capacity to pause data collection at any point, and (for many studies) the capacity for a respondent to retroactively delete previously collected data.
For information on earlier versions of our smartphone-based Ethica/iEpi epidemiological monitoring system, please see information on the Ethica epidemiological data collection system used globally and my iEpi page.
Registration and information for our coming 2019 bootcamp on Combining Data Science and Systems Science (Big Data and Dynamic Modeling) for Health 2019 (Jul 29-Aug 2, 2019)
Click here to see videos from our 2018 bootcamp on combining Data Science and System Science.
Please see my publications page and Curriculum Vitae for lists of some recent publications.
You can find my recent free library of public health agent-based and hybrid model at these materials from my August 2018 bootcamp. Look in the "Example Models" folder.
For added information on the below, please write firstname.lastname@example.org
2019 Bootcamp & Incubator on Understanding Health Behavior using Smartphones and Wearables
June 24-26, 2019
This tutorial introduces researchers & practitioners to tools, practical skills and the conceptual background required to define,deploy, monitor & analyze mHealth studies with no programming, and assists participants in getting started in applying such techniques to studies and applications of their specific interest. This tutorial will include hands-on work with the popular Ethica Health data collection system, which has been used by researchers across 4 continents (between it and predecessors, in approximately 100 studies).
Incubator participants will create first versions of mHealth studies in their areas of specific interest, including to build & test out custom study designs, survey instruments (sensor-triggered EMAs, image/audio&video reporting questions), Bluetooth Beacons to detect proximity to actors & resources, wearable technologies, crowdsourcing mechanisms, and sensor-triggered data collection mechanisms, and study data analysis via Apache Kibana.
Combining Data Science and Systems Science (Big Data and Dynamic Modeling) for Health
Jul 29-Aug 2, 2019
Will provide a particularly notable coverage of Particle Filtering and Particle MCMC methods for combining dynamic modeling with incoming data (especially streaming Big Data complementing traditional data), and diversity of case studies. Additional hands-on exercises, templates provided.
Agent-Based and Hybrid Modeling Bootcamp and Incubator for Health Researchers 2019
Aug 19-24, 2019
An intensive, hands-on tutorial offering health science researchers with a practical & accessible introduction to agent-based & hybrid system science modeling for health, together w/an Incubator to give most participants the opportunity to leave with a well-crafted exploratory ABM/hybrid model for their areas of custom interest. This intensive course provides in-class exercises, step-by-step written exercises and challenge problems, guest lectures on application & methodological topics, and distributed many example models not otherwise included with AnyLogic. Most videos, screencasts, and audio are available on my Youtube Channel. Please consult this course for my most refined materials on Agent-based modeling in AnyLogic
I teach a graduate course that provides an introduction to dynamic (simulation) modeling for public health (CMPT 858). This course provides a basic hands-on introduction to the theory and practice of Agent-Based and System Dynamics modeling in the context of health issues. Two versions of this course are available, one focusing on Agent-Based modeling and the other on System Dynamics modeling. Screencasts and presentations from the lectures & tutorials of this class are available below. I hope that this material will be of value for those seeking to learn more about simulation modeling.
I have also taught a variety of smaller, tutorial-style courses related to modeling for Public Health. These sessions are taught using Vensim and AnyLogic software.
I further teach undergraduate courses in Software Project Management (CMPT 371) and co-located undergraduate & graduate courses on Advanced Software Engineering (CMPT 470 & CMPT 816).
I am glad to share the materials for these courses with interested parties.
Videos of some of my talks and selected tutorials can be found here.
I currently supervise a broad set of students trained in Agent-Based Modeling, System Dynamics Modeling, and in the use of ubiquitous portable, wireless sensor systems for health insight and decision-making. These students operate at the undergraduate, M.Sc. and Ph.D. levels, and offer a wide range of skill levels in the health sciences, Computer Science, Information Technology and mathematics. Many such students value internship opportunities and knowing about post-graduation career opportunities. Some students are also interested in consulting options. Please be encouraged to write if you have opportunities available, as my students may be interested in learning about them.
Please see my page on student supervision.
In order to ensure that I continue to offer quality supervision to existing trainees, I am unable to take on additional students until May 2020.
For those interested in knowing more about my life outside of work, I also maintain a bare-bones personal webpage.
|Last updated February 1, 2019. Please report problems to email@example.com|