Society for Medical Decision Making

45th Annual North American Meeting
In-Person Short Courses

Sunday, 22 October, 2023
AM Courses: 8:30am - 12:00pm

Modeling Approaches for Analyzing Health Care Problems – an Introductory Overview and Comparison
Course Faculty: Dr. Beate Jahn, PhD; Jagpreet Chhatwal, PhD; Prof. Uwe Siebert, MD, MPH, MSc, ScD

Course Level: Intermediate

The course starts with a short introduction to decision-analytic modeling. Alternative modeling approaches will then be introduced followed by an interactive discussion.

This session covers decision trees (DT) and state-transition Markov models (STMM), two widely used methods. STMMs are based on a set of health states and have been applied in decision analyses addressing questions about prevention, diagnosis and treatment of chronic diseases.

The application of microsimulation/individual level simulation allows investigators to model individuals and evaluate heterogeneous populations using various approaches including individual-level state transition models or discrete event (DES). This session gives a general introduction based on their applications in the social sciences, health care and politics.

DES is a microsimulation method in which entities (e.g., patients) interact and compete for resources (e.g., hospital beds or organ transplants). We will cover the primary components of DES such as entities, attributes, resources, and queues.

ABM is a relatively new approach to modeling autonomous, interacting agents. The fundamental feature of an agent is the capability to make independent decisions, and complex network structures and contact behaviors can be implemented. ABMs have been used to examine  emerging behavior, economic and health-care research questions (e.g. COVID-19 vaccination). We will cover the role of agents as active model components.

SD is a powerful modeling method that involves both qualitative and quantitative approaches. It takes a "whole system" view, demonstrating how a small change in one part of a system can have major unanticipated effects elsewhere, an aspect that is particularly suitable for healthcare applications.

By the end of this course, participants will

• understand the role of decision-analytic modeling in health care

• know the key concepts of six different modeling approaches

• be able to describe advantages and disadvantages of different modeling approaches

• be able to critically discuss model selection

Introduction to Data Analysis and Visualization in R
Course Faculty: Dr. KD Valentine, PhD

Course Level: Beginner

The material in this course will introduce participants to the R programming language. We will cover the structure of the software and the use of a free integrated development environment (RStudio) to improve utilization of the software. The goal of this course is to provide an understanding and foundation of the functions and structure of the R software, including downloading and installing R, the main windows and menus in RStudio, packages and functions in R, importing and exporting data in R, manipulating datasets in R, basic data analysis in R, identification and correction of common errors in R, use of R markdown documents for data transparency and reproducibility, creating of reproducible tables in R markdown, and creation and alteration of simple data visualization (histograms, scatterplots, boxplots, line graphs, bar graphs) including grouped and repeated measures data visualization. The instructional approach will include some software demos and will illustrate applications of data use and visualization through sample datasets. Participant are welcome to bring their own data to practice techniques with during the course.

Objectives

• Understand the basic functions of R and RStudio

• Run basic commands in R including importing, exporting, and manipulating data

• Create data visualization including histograms, scatterplots, boxplots, line graphs, and bar graphs.

• Utilize R markdown documents for transparent and reproducible analyses

Infectious Disease Modeling Using Dynamic Compartmental Models in R
Course Faculty: Kyueun Lee, PhD; Theresa Ryckman

Course Level: Beginner

Mathematical models are a useful tool to understand disease trends and evaluate health policies. Unlike non-communicable diseases, the risk of acquiring an infectious disease depends on the population size of people infected with that disease and their mixing patterns with the susceptible population. These transmission-dynamic components of infectious diseases necessitate a special set of methods for modeling infectious diseases.
In this course, we focus on a commonly used form of infectious disease models: dynamic compartmental models governed by differential equations. Our course has two parts: 1) providing a brief overview of basic concepts in infectious disease epidemiology – such as the basic reproductive number, contact matrixes, and herd immunity, and 2) demonstrating how to translate these concepts into model assumptions and parameters in R. In the model implementation section, we will start with the classic “SIR” (susceptible-infectious-recovered) model and describe embellishments to this model that capture additional characteristics of infectious diseases, such as latency and waning immunity. We will simulate disease control interventions such as quarantine and vaccination, highlighting how they alter the disease dynamics. Through this course, participants will learn both the theoretical underpinnings of dynamic compartmental models as well as how to implement them in R for real-world applications. The course will include hands-on examples covering a variety of infectious diseases and interventions. We will also briefly discuss the limitations of dynamic compartmental modeling and other model structures that may be more appropriate for certain diseases or situations.

The objectives of this course include:

• To understand the fundamentals of infectious disease epidemiology.

• To learn how to construct and parameterize dynamic compartmental models of infectious disease transmission and simulate disease control interventions using R.

• To choose an appropriate model structure and set of assumptions for a given use case.

An Introduction to Research Prioritization and Study Design Using Value of Information Analysis
Course Faculty: Natalia Kunst, PhD; Anna Heath, PhD; Jeremy Goldhaber-Fiebert, PhD

Course Level: Beginner

Value of information (VOI) is a key concept in decision analysis that can be used to determine research priorities, inform resource allocation for potential further research and design proposed research studies. This course will introduce the general concepts behind VOI, present several key VOI measures and highlight where they can be most useful in directing future research. It will also demonstrate key graphical presentations of these measures and critically evaluate VOI analyses and their underlying assumptions.

The purpose of this course is to introduce VOI measures and their use in decision modelling. The course will introduce these measures, discuss their presentation and assumptions. By the end of the course, participants will be able to:

• Interpret the Expected Value of Perfect Information (EVPI)

• Interpret the Expected Value of Perfect Partial Information (EVPPI)

• Interpret the Expected Value of Sample Information (EVSI)

• Interpret the Expected Net Benefit of Sampling (ENBS)

• Discuss key assumptions that impact a VOI analysis

Napkin Math: Developing model intuition for the digital age
Course Faculty: Alyssa Bilinski, PhD; Josh Salomon, PhD; Jeff Imai-Eaton, PhD

Course Level: Intermediate

With recent advances in statistical techniques and computing power, decision science coursework has increasingly focused on complex and computationally intensive modeling. This course focuses on an often-neglected skill that remains critical for policy models: developing intuition for simple mathematical models (i.e., napkin math) that can answer policy questions as well as inform development and critical evaluation of more complex simulation models. This competency can help researchers to identify important potential areas of research, find and understand errors in code or thinking, and make conclusions clear and robust. It can also provide consumers of research with a starting point for critical evaluation of modeling papers. Course objectives include:

• Identifying how simple mathematical models can add value at different steps of the research process (e.g. idea generation, gut-checking results, understanding model robustness)

• Reviewing useful techniques for napkin math (e.g. understanding when simple arithmetic and algebra are sufficient to provide basic intuition, applying Bayes rule, recognizing geometric series)

• Working through practical examples inspired by recent peer-reviewed papers

• Discussing tradeoffs and interplay between simple and more complex models in terms of capturing essential drivers of system dynamics, understanding and communicating mechanisms, and incorporating uncertainty

Full Day Course: 9am - 5:30pm (12:00-2:00pm break)

An Introduction to Structural Equation Modeling for Medical Decision Making Research
Course Faculty: Adam Carle, MA, PhD

Course Level: Beginner

This short course will provide participants with a basic understanding of SEM concepts, with an emphasis on application. In the first half of the course, we will introduce several widely-used SEM approaches relevant to medical decision making research. These include factor analysis, path analysis, SEM-based mediation and moderation, and longitudinal latent growth curve modeling. For all topics, we will provide real world examples relevant for medical decision making researchers. In the second half of the course, participants will conduct SEM analyses in R with support from the instructors. We will discuss code, output, and interpretation.

In this course, students will:

• Learn fundamental SEM concepts and techniques.

• Understand the advantages of SEM over traditional statistical modeling.

• Be able to interpret the results of SEM studies.

• Be able to conduct and interpret SEM analyses in R.

PM Courses: 2:00 - 5:30pm

Transparent Programming: Important Habits for Reproducibility and Research Integrity
Course Faculty: Jacob Jameson, M.S.

Course Level: Beginner

This course, "Transparent Programming: Important Habits for Reproducibility and Research Integrity," is a hands-on exploration of robust and ethical programming practices. It focuses on cultivating proficient use of R and Git/Github for collaborative, open-source environments. Participants will delve into the ethos of reproducible research, comprehend the criticality of transparency in coding, and understand how these practices interweave with scientific integrity. A cornerstone of the course is the creation of a reproducible simulation, offering an applied perspective on version control, debugging, and transparent coding.

Course Objectives:

• Gain insights into the ethical implications of code transparency and the necessity of reproducible research in health decision science.

• Understand the essential elements of creating a straightforward, efficient coding narrative that allows others to follow the logic and progression of your research.

• Learn how to efficiently utilize Git/Github's features like 'Forking,' 'Pull Requests,' and 'Code Review' for seamless collaboration and continuous integration.

• Understand how to write well-commented, modular code in R to promote ease of understanding, maintenance, and scalability.

• Foster an attitude of continuous learning and improvement in coding and stay abreast of the best practices in open-source software development.

• Recognize common pitfalls in programming that may compromise reproducibility and learn strategies to avoid them.

• Develop an understanding of test-driven development to ensure code reliability and validity.

Causal Inference and Causal Diagrams in Using Big Real World Observational Data and Pragmatic Trials
Course Faculty: Prof. Uwe Siebert, MD, MPH, MSc, ScD

Course Level: Intermediate

This course introduces to the principles of causation, causal diagrams (directed acyclic graphs, DAGs), and methods for causal inference. We will use the framework of "target trial emulation" with counterfactual "clones" to apply causal methods such as “cloning-censoring-weighting” to big real-world observational data and pragmatic trials and link this framework with causal decision-analytic modeling.

The objectives of this course are to:

• define causal interventions and actions, draw and interpret causal diagrams, and distinguish causal from non-causal statistical associations.

• decide which biostatistical/epidemiological methods must be used in different situations to derive causal effect parameters.

• understand the conceptual approaches of causal decision modeling

• know how these methods are applied in big real-world data and pragmatic trials

Published examples from different fields will be used to demonstrate how to:

• Adjust for time-independent confounders (i.e., confounder affects both risk factor and disease), where standard stratification, regression analysis, propensity score methods and matching/balancing approaches yield valid causal effects if all confounders are measured;

• Adjust for time-dependent confounding (i.e., the confounder simultaneously acts as an intermediate step in the causal chain between risk factor and disease), where standard regression analysis fails and causal g-methods such as g-formula, inverse probability weighting of marginal structural models or g-estimation of structural nested models must be used;

• Adjust for non-adherence or treatment switching in randomized clinical trials, where both the intention-to-treat and the naïve per protocol analyses can fail to yield the true causal intervention effect;

• Assess the “fallibility of estimating direct effects” when adjusting for intermediate steps;

• Apply the concept of “target trial emulation” and counterfactuals using “clones” to perform a valid per-protocol analysis of sustained treatment regimens with big real-world data and pragmatic trials with post-randomization confounding.

• Discuss different biases such as time-independent confounding, time dependent confounding, selection bias, and immortal time bias etc.

Advanced Computation of Value of Information Measure to Determine Optimal Research Design
Course Faculty: Natalia Kunst, PhD

Course Level: Advanced

This course presents computation methods for the Expected Value of Sample Information (EVSI), a decision-theoretic measure of the monetary value of collecting additional information through potential future research. Participants will discuss EVSI and how it can be used to design research studies. The course will then give a demonstration of how to efficiently compute EVSI in practice with accompanying code provided in R.

The purpose of this course is to introduce EVSI as a tool for research prioritization and study design. The course will introduce several recent methods for the calculation of EVSI alongside R code to calculate and present these measures. By the end of the course, participants will be able to:

• Distinguish four recently developed calculation methods for EVSI

• Decide which EVSI calculation method is suitable for a given health economic decision model

• Calculate EVSI in R for two different health economic models

• Present EVSI analyses using standardized, publication-quality graphics

• Discuss key assumptions for calculating the Expected Net Benefit of Sampling (ENBS)

• Design efficient future research studies by determining their optimal sample sizes

How Might We...? An introduction to design thinking in medical decision-making research
Course Faculty: Dr Alison M Buttenheim, PhD MBA

Course Level: Beginner

This short course provides a rapid but substantive introduction to the principles and practices of Design Thinking and Human Centered Design. The course is intended for medical decision-making researchers who want to incorporate user experience and innovation methods into their research – whether as part of intervention development, evaluation, or implementation. In a structured sequence of hands-on, participatory activities, learners in the course will explore design thinking as a set of mindsets and methods and will leave with practical skills that can be immediately applied to a program of research in medical decision-making.

While design thinking has been used for decades in the innovation and technology worlds, it has been slower to diffuse through the academy. Research on health and health care can benefit from design thinking in myriad ways. Not a substitute or a replacement for our current ways of conducting research, design thinking can supplement and complement our research processes with new modes of seeing, asking, thinking, deciding, and doing.

Objectives of the course are:

• To gain familiarity with design thinking, user experience design, the double diamond, and their application to research processes.

• To understand, through case studies and reflection, how design thinking has been used successfully in academic and applied research on decision-making, including in grant proposals.

• To learn, through active engagement, how and why to scope and carry out activities in each phase of design thinking: Discover, Define, Develop, and Deliver.

• To experience and practice design exercises including briefs, user journey maps, opportunity statements, ideation, prototyping, and pitches.

• To make a plan for incorporating design thinking into a program of research.