Statistical Analysis of Administrative Health and Linked Longitudinal Encounter Data.

A graduate course (PSYC 513 - Statistical Methods) taught at the University of Victoria in the summer of 2017.

Course description

This course will focus on types and sources of administrative health data and on the permissions, acquisition, management, and statistical analysis of administrative health and linked longitudinal encounter data. Analyses will cover a range of methods, including descriptive statistics, longitudinal trajectories, visualization of encounter timelines, multilevel modeling of within-person change, growth mixture models, and predictive models to quantify risk of future health events. This course will lead to an ongoing collaborative interdisciplinary research group, including data, policy, and statistical analysis experts. Simulated data from Island Health will be used for purposes of this course (signed non-disclosure agreement required for access to information and analysis tools).

Resources

• Course Syllabus
• Lecture slides
• Book: Secondary Analysis of Electronic Health Records (open access)
• This repository: https://github.com/IHACRU/ihacru-2017-psy513
• A course on statistical modeling using R: uvic-2015-psy532
• Papers related to the use of electronic health records in clinical and academic research. (Please contribute!)
• R programming cheat sheets
• Guided Explorations
• Research Review Template
• Template for Research Protocol to initiate a project with Island Health
• Example of Research Protocol: Polypharmacy risk reduction.
• Research Approval Process
• DEPICTs - data requirements and data management description

Reports

Reports are available from the private repository. (Must be logged in with github account)

Schedule

Instructors

• Scott Hofer
• Kenneth Moselle
• Andrey Koval

Prerequisites

Basic familiarity with R is recommended. While we do not exclude other statistical packages, the primary data handling, modeling, and graphing will be written out in R scripts. Student engagement with the data and models will take place via RStudio. Students will be expected to learn how to read, modify, and apply existing R scripts to perform data manipulation, model estimation, and publish results as dynamic documents. If you are new to R, please explore resources for a different course, uvic-2015-psy532, which gives a good introduction to the programming and statistical tools we will be using. Don't be discouraged if you are new to R: all skill levels will be given a room for success and growth.

Grading

Class Participation (20%):

Students are encouraged to take an active part in the direction and content of the course by preparing for and contributing to class discussions of assigned readings, sharing with others their (direct or indirect) research experience, challenges and successes. The level and relevance of contribution to the course will be a prominent factor in final grading.

Research Project (80%):

• 1. Analysis of Encounter Data: Research Reports 1 and 2 (30%):: Statistical analysis will focus on simulated longitudinal encounter data from Island Health. Data and initial scripts for statistical analysis and plotting will be provided. Learners will be required to submit two reports which they will produce using software and scripts demonstrated in class. Course materials related to these reports are located at https://github.com/IHACRU/ihacru-2017-psy513.
• 2. Research Proposal (30%): A proposal (7-10 pages) will be developed that will be sufficient for a data request and ethics application. Details will be given in class.
• 3. A 10 minute presentation (20%): Students are asked to present their research proposal to the class at the end of the term.