537 Home

Objectives

Syllabus

Assignments

Grading

BHI Home Page

BHI Research Methods

MEBI 537, Fall, '09

Assignment Calendar:

I will draw material from three textbooks for this course:

Shadish, WR, Cook, TD, & Campbell, DT. (2002). Experimental and
Quasi-Experimental Designs for Generalized Causal Inference.

Cohen, PR. (1995). Empirical Methods for Artificial Intelligence.

Friedman, CP & Wyatt, JC (2006). Evaluation Methods in Biomedical Informatics.

The Shadish textbook or the Friedman text is required; Cohen is a less relevant text, and I will provide selections as PDF files.

All other reading assignments from the primary literature can be retrieved from the course Eres pages. (Please note: there are more readings on this site than those that are actually assigned here!)

General expectations:

All written deliverables must be handed in electronically, via the course Catalyst drop box.

There are three categories of assignments in this class: (a) reading assignments, (b) written essays, and (c) the final project. For the readings, I expect you to come to class prepared -- with questions, ideas or discussion points. Typically, I will ask for your opinions before lecturing or providing my own views on the material--and your responses will contribute to the "class participation" portion of your final grade. See below for more on the written essays--my teaching goal for these is to improve your written communication. To help with that goal, you can expect that I will grade homeworks promptly (within about a week) and be available should you want to discuss them. See the final project page for information on this assignment.

Finally, I expect you all to be adult learners. I am assuming you want to be in class, and that you want to learn the material I'm advertising to teach. I will therefore also expect to learn from you -- teaching and learning work best when it is a two-way street.

Reading assignments:

In addition to (obviously) reading these prior to class, I also expect you to be able to discuss them in class. In some cases, this may require multiple readings; it certainly means that you must think about what you read, and perhaps taking some notes to help you offer discussion points and ideas during class. (As specified in the grading page, class participation is 10% of your final grade.) Reading assignments are either from textbooks, where an aspect of research methods is presented in a didactic manner, or readings from the primary literature. For the latter sort of reading, the idea is that the papers are cases or examples of different sorts of research methodologies. Thus, you should focus on these methodological aspects of the readings.

Written (essay) assignments:

First, please read my web page on a "reaction essay", as designed for the KR course (MEBI 550) that I taught in Winter '05. Although the essays for BHI Research Methods are not all "reaction essays", this description should give you a view of my perspective on writing skills and the value of a short, well-constructed essay. All of the essays for this course have the same page limit: 2 pages, 12 pt font, single-spaced.

All of the essays are also somewhat open-ended. My approximate grading rubric for these essays is as follows:

In addition to helping you think and learn about research methods, an important secondary goal of these assignments is to be able to write clearly and convincingly. So that you can learn about my expectations and improve over the course of the quarter, you can expect that I will grade homeworks promptly (within a week) and be available should you want to discuss them.

Essay #1:

Using the three assigned papers, either defend or reject the idea that all three describe research that is appropriately part of the single scientific field of "biomedical & health informatics". Given your 2-page limit, you should not summarize each of the articles, but instead focus on particular aspects of the papers that support your argument.

Essay #2:

The article by Freidman et al. is a good example of a systems evaluation paper. It is a formal, quantitative study, yet has no driving hypothesis to be proven (no causal connection to be established), nor does it evaluate system use in the "real world" (it is a laboratory-only evaluation).

Use this work as a springboard for some imaginative work: Write a piece of fiction that assumes some medical institution (hospital) is actually using the MedLee system to code narrative text in that institution's EMR system. What issues / problems does the institution encounter, and / or what opportunities for experiments (possibly hypothesis-testing ones) might there be, given the installation of this technology? What benefits have accrued? (To do best, this assignment might require some background reading in EMRs and / or the "coding vs. free text" issue. You can get this from the Shortliffe text, parts of chpts 2 or 12, in 530.)

A required component of this assignment is that your fictional work must make appropriate reference to the assigned MedLEE paper. Use as much as you can of the paper as "background" to your installation paper, discussing details of its implementation, the precision and recall results reported, etc.

Broadly, I am asking you to demostrate how systems evaluation research can be built upon and used in more mature, hypothesis-driven or real-world evaluation research.

Essay #3:

How to we decide if qualitative work is good or useful? In 1998, Poses & Isis published a rather inflammatory "perspective paper" about the use of qualitative research and its relevance & significance to practitioners of health care. [The '98 issue of J Gen Int Med also includes an explicit rebuttal by various authors. Although fun to read, I didn't find it as useful. It's also on the eres pages; consider it an optional reading assignment] Imho, one of the best rebuttals can be found in JAMA, in a pair of articles by Giacomini & Cook (2000). I've assigned the first of these articles that focuses on the quality of qualitative work.

Given these two articles, take a stand on the issue: How do these authors' arguments apply to BHI research? My strong preference is that you not simply say "BHI needs to have both qualitative and quantitative research", but instead argue in favor of one approach or the other. One teaching goal for this assignment is to help you realize that although one can do good research in either tradition, in some ways the two world views are indeed in conflict with one another. You certainly need not be as dogmatic as Poses & Isis, but I want you to emphasize and argue for the strengths of one of these approaches over the other.

Essay #4:

For this essay, I'll simply ask for a regular, open-ended "reaction essay" on the Raebel paper. By now, you should have read enough and thought about methods enough to have lots of good ways to organize your essay. For example, you could compare this work to the validity framework of Shadish. You could also react to the notion of research that uses a hypothesis driven experimentation methodology (since this is the first of these we've read so far), in contrast to qualitative or evaluative methodological frameworks.

Last Updated:
Oct, '09

Contact the instructor at: gennari@u.washington.edu