| Decision Science in Rehabilitation
SESSION SUMMARIES
August 29, 2007
Session 1
It was decided that the seminar would begin at 1:00 PM for those who can be there. However, since not everyone is able to be there until 1:30 PM and others have obligations at 2:30 PM, it was decided that the hour between 1:30 and 2:30 would be ‘Prime Time’ for discussions/presentations.
Dr. Deogun suggested that the next steps should involve the formalization of key concepts so that all participants have the same understanding of these concepts. For example, it is important that all understand the roles and relationships between the patient, the clinician and the treatment. It is anticipated that “prime time” for September 5 will focus on the point of view of the clinicians.
The Project activities were recapped keeping in mind the purpose of the grant is to develop a collaborative relationship across the various disciplines.
I. Build an intelligent system to support clinical decision making in rehabilitation. The following 4 points can be thought of as a heuristic formulation
1. complete domain/methods ontologies
2. develop concept and design for DSS
3. evaluate empirical analysis for pathways and decision nodes
4. evaluate current DSS for human interface
II. Begin documenting and publishing
_____________________________________________________________________________________________
July 11, 2007
Planning Session Summary
Below is a list of the Core Group of Seminar Participants and their email address. In addition to the Core Group, we can look forward to the participation and contributions of an ad hoc group of collaborators and consultants, two of whom are identified below. We anticipate that the membership of this informal group will vary over time as the project cycles between complex modeling and implications for design and development of clinical data management and decision support systems.
Core Group of Seminar Participants
Principal Investigator
Will Spaulding, Ph.D., UNL Dept. Psychology (wspaulding@neb.rr.com)
Co-investigators
Calvin Garbin, Ph.D., UNL Dept. Psychology (cgarbin@unlserve.unl.edu)
Bill Shuart, Ph.D., Inst. Rehab. Research & Engineering, Madonna Hospital (bshuart@madonna.org)
J. Rock Johnson, J.D., Nebraska Mental Health Assn. (jrock10@sprynet.com)
Modeling/display specialists
Jitender Deogun, Ph.D., Modeling/display specialist (deogun@cse.unl.edu)
John Flowers, Ph.D., Modeling/display specialist (jflowers1@unl.edu)
Hamid Vakilzadian, Ph.D., Modeling/display specialist (hvakilzadian1@unl.edu)
Research Staff
Jeff Nolting, M.A., Project Manager, UNL Clin. Psych. Training Program (jnoltin2@bigred.unl.edu)
Felice Reddy, B.S. Research Assistant, UNL Clin. Psych. Training Program (lfr@bigred.unl.edu)
Melissa Tarasenko, B.A. Research Assistant UNL Clin. Psych. Training Program. (mtarase1@bigred.unl.edu)
Ad hoc collaborators and consultants
Vijay Dewan, M.D., Psychiatrist, Lincoln Regional Center (vijay.dewan@hhss.ne.gov)
Joe Swoboda, PhD., Psychologist, Community Mental Health Center (jswoboda@ci.lincoln.ne.us)
July 11, 2007 Planning Session – Summary
The specific aims of the project and the role of the distant collaborators were reviewed. (To review the specific aims go to http://www.unl.edu/dsc, click the Research button, then the DSR link at the top of the page.) The “distant” clinical rehabilitation researchers will provide expert input to various components of the DSR Project and compare their findings of outcome-related patterns of longitudinal behavior change in rehabilitation databases to those developed by the PI and the co-investigators. (A more complete description of the clinical rehabilitation research collaborators can be found in the grant narrative, pp. 61-62.)
Main Objectives – It is important to keep in mind that the objectives can only be realized through an iterative process
1. Get more $$$. It is anticipated that the activities in this project set the stage for an R01 grant – a prototype of a complete decision support system for NIMH.
During the final year of the project the seminar collaborators will address three specific topics and generate project proposals to pursue those given the highest priority, as determined by the results of this project’s analyses and their implications for clinical informatics design:
A. design and development of a next generation clinical decision support system;
B. testing the ability of human clinicians and consumers to correctly interpret and understand measurement, or to identify change patterns associated with specific desired outcomes;
C. testing the effectiveness of clinical decision support systems in optimizing clinical decisions, enhancing consumer involvement in the decision process, and improving outcome.
2. Formulate domain ontologies for the informatics system.
The ontology describes ‘what may exist’—what are the possible components of a protocol (e.g., eligibility criteria that determine whether an intervention or clinical trial actually applies to a given patient, an algorithm that specifies the sequence of treatments to offer, tests to perform that may predicate decisions in the clinical algorithm) and what are the relationships among those components (e.g., that steps in the clinical algorithm are related temporally). The ontology does not describe any particular protocol (i.e., a specific plan of care for a specific problem), but instead lays out the classes of entities that exist in protocols in general.
Examples might include:
A. phenomenology of the “expert”
B. empirical findings
C. legal, political, economic, and healthcare systems
D. rehabilitation treatment team
3. Formalize the articulation of “expertise.”
There is substantial evidence that, as humans become experienced in an application area and repeatedly apply their know-how to specific tasks, their knowledge becomes compiled and thus inaccessible to their consciousness. Experts lose awareness of what they know. The knowledge that experts acquired as novices may be retrievable in a declarative form, yet the skills that these professionals actually practice are procedural in nature ( Anderson, 1987). The inability of experts to verbalize these compiled associations is well accepted (Nisbett and Wilson, 1977; Lyons, 1986). The consequence is that the special knowledge that we would most like to incorporate into our intelligent systems often is that knowledge about which experts are least able to talk. Johnson (1983) has identified this phenomenon as “the paradox of expertise.”
The problem for builders of intelligent systems is that, when professionals are asked to report on their compiled expertise, they often volunteer plausible answers that may well be incorrect. In experimental situations, subjects have been shown to be frequently (1) unware of the existence of a stimulus or cue influencing a response, (2) unaware that a response has been affected by a stimulus, and (3) unaware that a cognitive response has even occurred. Instead, subjects give verbal reports of their cognition based on prior causal theories from their nontacit memory (Nisbett and Wilson, 1977). Furthermore, because Western culture mistakenly teaches us that accurate introspection somehow should be possible ( Lyons, 1986), people freely explain and rationalize their compiled behaviors without recognizing that these explanations frequently are incorrect.
Developers of knowledge-based systems thus are not mining preexisting nuggets of knowledge. They are creating de novo theories of professional problem-solving behavior and representing those theories in terms of electronic knowledge bases. The developers serve the important function of detecting gaps in the articulated knowledge of their informants and of helping them to fill in those gaps by defining plausible sequences of actions that can achieve the necessary goals. The intelligent systems that result from this work may not achieve the same level of nuanced performance associated with the procedures actually used by domain experts. The underlying knowledge bases, nevertheless, can be observed, extended, and easily disseminated to other people in need of advice. It is simply incorrect to view an electronic knowledge base as an embodiment of actual professional knowledge; knowledge bases instead represent only models of surface-level behaviors—models that attempt to approximate, but that do not reproduce, the actual problem solving steps used by humans (Clancey, 1989). (Musen, 1998.)
|
