HMI 8571: Decision Support Systems for Healthcare (SP11)Instructors:
HMI 7410 (design of health and human service systems), HMI 7430 (health informatics), HMI 7544 (epidemiology), a statistics course, or instructor's consent.
This course will apply, integrate, and supplement the following prerequisite skills:
- Basic understanding of epidemiology and statistics.
- Knowledge of the health care system and organizational structures.
- Understanding the relationship between health care consumers, clinicians, administrators.
- Familiarity with health care information systems.
This course is designed to provide an overview of decision support systems in health care, with a particular emphasis on design, evaluation and application of clinical decision support systems (CDSS). The course explores the background and state-of-the-art of CDSS. Students will understand the mathematical foundations of knowledge-based systems, learn to identify areas which might benefit from a decision support system and evaluate the challenges surrounding development and implementation. The course also includes a detailed discussion of issues in clinical vocabularies and other important issues in the development and use of CDSS, and provides guidance on the use of decision support tools for patients.
- Develop basic skills (information evaluation, management, and knowledge representation) to translate decision support techniques to the benefit of patient care.
- Understand the impact of scientific evidence, and its acceptance and use in clinical medical practice.
- Design effective guideline implementation for well substantiated clinical practice recommendations.
- Learn to identify areas which might benefit from a decision support system and evaluate the challenges surrounding developing and implementing such a system.
- Be able to discuss the ethical and legal issues surrounding clinical decision support systems.
- Develop a sufficient knowledge of decision support systems to make an intelligent purchase suitable to your company's needs from a decision support vendor.
REQUIRED TEXT AND READINGS:
- Berner, ES. (ed.). Clinical Decision Support Systems: Theory and Practice. Springer-Verlag, New York. 2007(ISBN 0-387-33914-0)
- A series of articles and reports relevant to each topic will be listed for each lesson.
APA format is required for all your submissions in this course. Refer to APA manual for details but here are some points for your convenience. Your paper/report must be word-processed, Times New Roman/Courier, 12 pt, Double-spaced, 1" margins on all sides. Please include your name and clearly mark your page numbers. Your paper/report should include bibliographic citation of any material that you use.
Submission of assignments will be accepted through the Moodle courseware. You are expected to post your assignments in the “assignment” section, and if the assignments are required to be shared with the class, you also need to post them in the corresponding section in the forum of Moodle. Due to the limitation of university email account capacity, no email submission is accepted.
Late submission of assignments will be accepted with a 10% reduction of score per day of delay. No submissions are accepted later than one week after the original due date of the assignment. Extensions may be granted but ONLY due to extreme causes (such as illness and traffic accidents) if supporting documents (doctor appointment slips, police records of car accidents, etc.) are provided to the professor.
Incomplete grades may be granted in cases of illness or other difficult circumstances. Incompletes must be requested in writing by the student to the professor before the final assignment/exam/paper due date.
METHODS OF EVALUATION:
The course grade will be based on the following:
1. Decision Support Reasoning Exercise (15 points) (15% of grade)
The purpose of this exercise is to apply three different reasoning approaches to a given problem. The idea is to get a little practice approaching a problem with a reasoning method and to evaluate the relative advantages and disadvantages of the different methods.
Scenario : Everyone should be well-aware of the significance of errors in the delivery of heath care services. Whether measured in dollars, lives, lost reputations, law suits, or dry statistics, the problem is important and a great deal of attention is now paid to devise solutions.
The Board of Directors of the Perpetual Order of Saints Hospital, where money is no object, has hired your consulting group to develop an information system to reduce errors and improve the quality of health care. The Board of Directors wants the system you design to produce real, measurable results. In particular, they want to see clear improvements in deciding:
- What to do when something goes wrong.
- How to decrease the number of things that go wrong.
Input : You have three primary sources of data. First, Perpetual Order of Saints Hospital has a sophisticated electronic medical record system. Second, Perpetual Order of Saints Hospital has a system inspired by the MU Patient Safety Net that lets anyone (patients, families, and staff - including administrators, doctors, nurses, therapists, etc.) report something that either did or very nearly did go wrong. Third, your system can interact with a decision maker and collect information from her.
Output : The system you design should deliver information to a person who makes decision about 1 and 2 above.
Strategy : Your consulting group has expertise in a variety of decision support methods and will explore the possible utility of 1) probabilistic, 2) rule-based, and 3) case-based methods on meeting the Board of Director's goals. Remember, in a generic sense, your objective is to take stuff you already know (in the form of statistics and likelihoods, heuristic or logical rules, or experience of similar cases) and use it to improve how you handle a new situation.
- What are some situations (in the above context) for which the methodology promises to work well? Think of some examples that might illustrate the utility of the method. For each example, consider what kind of input your method will require and what kind of output it will produce.
- Consider adverse drug reactions. The Perpetual Order of Saints Hospital Board of Directors will want your system to improve decisions at the time adverse drug reactions occur and to decrease the incidence of their occurrence. How can you method contribute to this? Consider issues like:
a. Who needs to know an adverse drug reaction has occurred, when, and how will they find out?
b. How will the system use you method to help prevent adverse drug reactions?
- What are the advantages and disadvantages of your method - in particular, with respect to:
- Understandability - arcane skills required to trace how it works and to modify it? Is it bite-sized, modular?
- Maintainability - is it complex and interconnected, vulnerable to bugs, easily updated or improved?
- Compatibility - does it "fit" with the processes, people, formats, etc. where you propose to use it?
Example responses to the decision support reasoning exercise will be shared with the class.
2. Vocabulary/Coding System AND/OR Clinical Decision Support System Review(10 points) (10% of grade) (Class Presentation, Class & Online Discussion and Executive Summary) : During the first session each student will select a piece of paper from a cup that will determine which A) vocabulary/coding system or B) existing clinical decision support system on which to present in class according to the schedule and an executive summary (not to exceed 250 words). All presentations files in PPT format and executive summaries in DOC format will be shared with the class on the discussion forum.
A. Vocabulary/Coding System - The coding systems to choose from include: LOINC, CPT, ICD-9/10, DRGs, SNOMED, HL7, Arden syntax, RxNorm, and UMLS. Each topic will cover the following areas. (1) Overview (What organization produces or created it? Any info on organization? What is the most current revision? What new features have been added? Other introductory material as you deem appropriate; (2) (DRG only) Be sure to relate DRGs to the Case-Mix Index and payment for a given disease; (3) Level of acceptance and use (Who has adopted it and how widely is the classification used?); (4) Classification Structure; (5) Limitations; and (6) Reference List.
B. Clinical Decision Support System - The list of possible decision support programs to research include: MYCIN, OPAL/ONCOCIN, Leeds Abdominal Pain System, DxPlain, Internist-1/Quick Medical Reference (QMR), Iliad, EON/Protege, Isabel, Casnet, Germwatcher, Help, and ATTENDING. The topics should include the following sections: (1) Overview (who or what company designed the system, when was the system used and is it still in use, how much does the system cost if it is sold); (2) What form of data was input into the system (eg. signs and symptoms, clinical guideline rules, etc.); (3) what type of reasoning model does it use (how does it work - does it use rules, fuzzy logic, neural networks, etc. to go from the raw data to its conclusions); (4) What made this program novel compared to other programs; (5) What was the level of acceptance and use and/or are their any studies showing the systems effectiveness in improving outcomes or diagnoses or at least comparing the system to a gold standard; (6) what were the limitations of the decision support program; and (7) reference list.
3. Journal Club Critique of a Decision Support System (5 points) (5% of grade) (Class Presentation, Class & Online Discussion and Executive Summary): Each student will choose a clinical journal study of a decision support system to evaluate. The paper to be evaluated should be an original study and not a literature review or meta-analysis. The student will need to present in class pointing out the strengths and weaknesses of the paper and study. The student will also write an executive summary (not to exceed 250 words) of the study and lead discussions in the forum/classroom. All executive summaries in DOC format and critiques in PPT format will be shared with the class on the discussion forum.
4. Decision Support System Project (55 points) (55% of grade)
- Proposal (5 points)
- Progress report (5 points)
- Individual paper (45 points)
Presentations from each project are required. The presentation files in PPT format and executive summaries in DOC format will be shared with the class on the discussion forum.
Students may team up (up to 3 members in a team) with your classmates. However, each student needs to write an individual final paper based on the project.
a. Choose an area in health care where an application of decision support technology is of interest to you. Your selection must be a good problem in an area that you can reasonably expect to solve this term with a decision support process. Review the literature on the area and your chosen problem with particular attention to how decision in this area are made, what efforts at decision support have already been tried, and what the need is for decision support in this area. (What you learn should justify your choice of problem.)
b. Choose a type of decision support methodology (e.g., rule-based systems, artificial neural networks, Bayesian belief networks, case-based, etc.) that you think will work on your problem in this area. Review the literature on the use of that methodology with particular attention to two key areas: the historical development of the methodology and its current state of the art (e.g., latest or typical uses). You must be able to make a case that this methodology is reasonable to apply in solving the sort of problem you have chosen.
c. Create a decision support model to apply this methodology on your problem.
d. Write a paper that integrates your reviews of the literature, justifies your choice of a decision support methodology, and your model for a solution. You will also want to address the following questions in your paper: What are the organizational implications of building and implementing this DSS? Is it a guideline, a pathway, evidence based? Will it affect quality of care, and if so, in which direction and how? Where is the knowledge base going to come from and how will it be represented? What type of reasoning will be required in the programming of the system? Are there available vocabularies to support it?
e. Paper requirements: length 10-12 pages (not including title page, bibliographic references, tables and figures) (double spaced, font size 12, margins 1 inch on all sides, See APA format for details); bibliographic references should appear after the body of the paper; all tables and figures should be clearly referenced in the text (e.g., see table 2) but appear in order after the bibliographic references.
f. This project can be a team project (up to 3 team members). However, each team member needs to write an individual paper at the end. Each project needs to have: a project proposal and progress report (PPT presentations and executive summaries up to 250 words). The due dates are shown in Moodle . You will define the decision-making problem you hope to solve, as well as the decision support methodology you will apply.
g. Example topics:
- Decision support for testing and treatment of latent tuberculosis using a rule-based system
- Diagnosing acute rheumatic fever (Jones criteria) using an artificial neural network
- identifying women at risk for domestic violence using a Bayesian belief network
- Identifying potential dialysis patients in the emergency room when the admitting diagnosis is not kidney-related
- Determining whether a nursing supervisor for a ward should call more nurses to duty with a fuzzy logic system that accounts for the type of patient and their nursing needs
- A case-based system to advise patients on the best way for them to lose weight (or to solve some other health problem)
- See topics from previous students…
5. Discussion Performance (15 points) (15% of grade)
- Classroom Performance (10 points)
- Forum Performance (5 points + 5 bonus points)
- Discussion in each forum
- Contribution of facts and questions based on the textbook
Your activities on forum discussion and classroom discussion will be evaluated based on the questions you raise and answer. In addition, you are expected to pull out interesting facts and develop questions from the textbook. You may get an extra 5 bonus points by contributing good questions. Post your questions on the forum and be ready to provide the keys in the textbook. For example,
- Type T for true or F for false in the parenthesis
- ( ) In the early 1970s, the common perception was that continuous heart rate monitoring can protect the fetus from prolonged intruterine oxygen deprivation. Subsequently, several controlled clinical trials succeeded to demonstrate the clinical benefit of this technology.
- Which of the following method is NOT a knowledge-based approach for developing Clinical Decision Support Systems.
- a) Probabilistic reasoning
- b) Rule-based reasoning
- c) Case-based reasoning
- d) Artificial neural networks