Return
to Dr. Willie's Web Page
Decision-making as discussed here refers to the clinical process of
making diagnoses and selecting treatments based on that diagnosis. Psychiatric
decision-making is subject to the same constraints as medical decision-making
in general, and in addition must deal with 1) difficulties inherent in
the state of psychiatric knowledge, as reflected in the literature, 2)
the subjective, cultural, and individual nature of psychosocial problems,
3) ideological conflicts within the field, and 4) the context of psychiatric
treatment and evaluation, which may involve the exploration of personally
private and sensitive matters.
There are severe limitations in this process. First, the decision process may involve multiple variables. The limitations of the human mind in this regard are well documented. Studies have shown that in a variety of areas, clinical judgment, the evaluation of multiple variables by the unaided human mind, is inferior to actuarial decision making, where evidence is combined and evaluated according to known rules and established statistical methods (Dawes, et.al., 1989).
Secondly, the recall of data is subject to numerous errors as applied to clinical situations. Fig.1 is a model for information accuracy as applied to the clinical decision. It begins with the data obtained in a study. The data is filtered as it is put into publication, and some of what may have been learned as clinically useful may not be included in the publication for various reasons. As the information is applied to the clinical situation, its relevance may be in question: the population in the study may not reflect the characteristics of the patient. For example, patients selected for an antidepressant medication trial may exclude those with a concurrent personality disorder. In clinical practice, however, depression and personality disorder frequently present together. Thus what was learned from the drug trial may not be applicable to the individual patient.
Further knowledge loss occurs as the information is reproduced in publications or presentations. The summarization procedure will inevitably leave out some material from the original publication. In some cases, what is selected will be that which is relevant to the clinical case. In others, what is left out may be deemed to be irrelevant to the purposes of a review or presentation, but may turn out to be essential to a particular patient.
Fig 1
Overall losses in knowledge occur because access to the information varies: information in papers or books that the practitioner has not read is not available to the decision-making progress (Huth, 1989). In this case, the information simply does not exist. If the practitioner has read or heard the material, errors of comprehension may enter: the material or its significance may be misunderstood (Evidence-Based Medicine Working Group, 1991). Finally, in the actual clinical situation, recall of the material is entirely dependent on the memory of the clinician. This is inconsistent at best, and may be decreased due to overwork, stressed, or preoccupation.
Once information is imparted to a patient, as in a recommendation or a prescription, additional knowledge losses occur (Fig. 2). The patient may not be presented with all the information that is relevant to the situation, may not hear or understand all that is presented, may not remember what was understood, and may not be motivated to follow what is remembered.
When viewed in toto, even small knowledge losses at each step add up to a potentially frightening degree of misinformation at the time that a treatment is administered.
Fig 2
The Coupler begins with a series of questions that are answered by the
clinician, or when practicable, by the patient. A click of a button creates,
without further effort such as writing or dictation, a record of the findings
for that patient. The list of findings is used by the program to vote for
various diagnostic and management options. Both the findings and the options
are present with appropriate comments which may be printed out and presented
to the patient.
"Support providers" are used widely, especially in hospital and community psychiatry, and include psychologists, social workers, occupational and recreational therapists, psychiatric nurses, community mental health workers, and others. Although they may use a common medical record, and may participate in regular interdisciplinary treatment planning meetings, disparities in treatment philosophies, ideologies, and the arena of contact with the patients may lead to varying and conflicting views of the patient and his or her problems.
Problem Knowledge Couplers address these conflicts by providing a common framework for evaluating a patient, and providing a problem list shared by all providers. Couplers are able to integrate knowledge from multiple disciplines, so that all providers have relevant information from other disciplines as diagnostic and treatment decisions are made. A means of including input from all providers (an 800 phone number) allows the Coupler to become a mechanism for the cumulative and cooperative acquisition of knowledge across professional boundaries.
In addition to remembering a base of technical knowledge, applying it to a clinical problem involves the correlation of multiple variables of the clinical situation (history findings, symptoms, physical examination, mental status, and laboratory results) with the variables presented in the literature. A given problem may involve dozens of diagnostic possibilities each with several findings, many of which overlap. The number of combinations exceeds the capacity of the human mind to carry out a complete correlation. Mental strategies used in such situations inevitably involve simplifications, abstractions, and other methods that omit information, and carry the risk of introducing error into the decision making process.
Accurately correlating multiple variables is something that computers can do well. Although much research and discussion have gone into discussing the representations of knowledge (algorithms) that are best used, it has been shown that even the simplest models (linear models or simple but accurate lists) can aid accurate clinical decision making (Dawes and Corrigan, 1974).
As psychiatric knowledge expands and is put on a more scientific basis, the application of that knowledge remains in a centuries-old mode. The number of effective interventions that have been demonstrated continues to increase. For example, the number of antidepressants presently available on the market is close to twenty, with a half dozen classes of augmenting or alternative agents, and many medications that may be used for comorbid conditions. Although all are effective against depression, the literature lists a number of demonstrated indications and contraindications for each agent and class of agent. Published guidelines may attempt to rationalize the selection of medication, but these leave out details that may be relevant to an individual patient, and again rely on the memory and processing capability of the clinician to implement them.
Knowledge Couplers written for psychiatric problems are able to create the relevant links between the ever-expanding psychiatric literature and the unique presentation of the individual patient. Beginning by asking those questions which should be asked for every problem, a Coupler is able to produce the links to all relevant knowledge from the literature, including the rare illness, the disorder that is the province of another medical specialty, and the multiple causes that may be present in a complex case.
Psychiatric implementation of this involves an educational component that helps to relieve stigma, utilize resources, such as self-help groups, that lie outside the province of formal psychiatric care. Respect is maintained for the patient's ability to determine the combination of symptom severity, disability, side effects, risks, benefits, and costs that he or she wishes to live with.
Knowledge Couplers provide a key link in transforming the center of
health care decision making. Their use in the earliest phases of evaluation,
and the provision of information that has been selected to be specifically
applicable to the problem at hand insures that the importance of the consumer
is recognized. As the patient reviews the Coupler output with the provider,
a collaborative, rather than passive and dependent relationship is established.
Computer systems have the ability to track large amounts of information,
and to provide for regular updating of that information. Implementation
of the PKC system includes regular updating of information as it is reviewed
by in the literature by Coupler authors and as suggested by users of the
PKC system.
Consistent input is applied to every case, regardless of the practitioner's specialty, discipline, training, orientation, or practice setting. While diagnostic classifications, clinical guidelines, and computer- based knowledge systems cannot reduce the variation associated with individual patients, consistent input can be used to ensure that all providers are working from the same database of information about the patient, and from the same base of knowledge derived from the literature.
Limiting costs in this context involves limiting services in some way, either directly through caps on services, or indirectly by limiting fees or taxing benefits. Little discussion is given to the role of preventive care, other than in a few areas such as prenatal care, where a clear cost benefit advantage has been shown. Although there is discussion of the role of guidelines to limit the inappropriate use of expensive technology, there is no discussion about how guidelines or any other relevant medical information is actually used in the patient-physician interaction (Weed and Weed, 1994). Indeed, the topic of the patient-physician interaction is seen as a taboo subject, and the role of information technologies in this interaction is a matter left to a few specialists.
Computer software can be used to select treatment that is appropriate by matching the characteristics of the individual patient to those that have been shown to identify safe and effective treatments. Superficially, this may appear to be the same process as that utilized by managed care. The difference is that the coupling of information takes place as part of the process of clinical diagnosis and treatment, rather than an additional procedure that is imposed upon it. For example, cognitive therapy may be a recommended treatment for some depressions. If some other form of therapy were being practiced by the therapist, the recommendation for cognitive therapy by a managed care representative may be seen as an invasion into the therapeutic process. If this recommendation appeared at the time that the treatment was planned, it could be reviewed as an option by provider and patient together, taking into account the provider's skills and patient preference.
Utilizing continuously updated software provides a way of rationally integrating newer treatments into clinical practice. Techniques that the clinician may not have learned during professional training can be brought to the clinician's attention at the time when it is most appropriate to consider them for application, rather than at a distant time that may be convenient to the schedule of the clinician or the presenters of the information. It also allows newer treatments to be evaluated appropriately in the context of information about established treatments, with both risks and benefits available. The over-prescription of newer medications because of pharmaceutical company marketing efforts, and their under-utilization, due to reluctance to employ unfamiliar treatments and lack of adequate information, can be overcome by software that can provide the precise amount of information that the situation requires.
R. Vance Fitzgerald, M.D., of the Medical College of Ohio at Toledo, reports that men seem to be more accepting of the use of computers than women. He writes:
"For some patients I deliberately elect NOT to use the Coupler in their presence. I am not entirely clear about how I arrive at this decision. I think it has something to do with the patient's sensitivity and compelling need for an interpersonal connection and, perhaps, to my sense that a patient has a bias against things mechanical. I may be particularly sensitive to this aspect of possible patient reaction because of knowing my wife's feelings and those of other women I talk with about computers. I don't mean this in a sexist way. Most of the patients I chose not to use a Coupler with are women. Many men and many women are different in regard to feelings about this procedure.
"Men, generally, seem more impressed. To them it's more logical
and indicates that the expert is one up compared to other experts who do
not use modern technology (Fitzgerald, 1992)."
Paranoid patients may object to or falsely answer some questions
in the Coupler "Psychotic-like Behavior, Thinking, and Speech," the one
that is most likely to be used in their cases. This can be overcome be
running a Coupler that is more neutral for them, such as Anxiety or Depression,
first. The patient may then lose interest in specific questions before
the Coupler for psychosis is run.
To serve its function, the PKC system should be used on a regular
basis. I routinely do the appropriate Couplers on all evaluations. Occasionally,
on what appears to be a clear cut case, I will omit doing the Coupler.
Within a few weeks, I usually regret the decision, and do the Coupler at
that point.
One disadvantage of this method is that the user will inevitably find a few questions that were not asked during the interview. Since the use of a Coupler requires that the user answer all questions, the accuracy of the data generated will be compromised. One of the reasons that Couplers were developed was to be certain that all pertinent questions were asked, and this method is likely to defeat that purpose.
Another difficulty is that the results of the Coupler are not available at the time decisions are made. For a diagnosis, this may just mean adding a chart entry, but if it affects treatment, the patient may be unavailable or confused by an alteration of treatment plan at this point.
In short, running a Coupler after a clinical evaluation should
be done only as a training exercise, or in those rare occasions in which
the Coupler session for some reason cannot be completed in the patient's
presence.
This allows the clinician to evaluate the patient with little disruption of his or her usual style, and still allows access to the results of Coupling as treatment decisions are made. This also allows the clinician who may not entirely trust or understand the coupling process to form her own clinical opinion, against which the Coupler-generated options may be compared.
Patients may sometimes feel that the clinician is involved in
some secretive process from which he or she is being excluded. This may
be handled by inviting the patient to observe the computer screen as the
data is entered. This procedure also allows the patient to confirm the
accuracy of responses as they are selected.
Running the Coupler early in the interview may mean that the evaluation
process may be limited to the evaluation of the presenting problem. In
an emergency, this may be acceptable, and may allow for prompt evaluation
and initiation of treatment. There is, however, a risk that the evaluation
may end here, and that other essential tasks such as obtaining a complete
problem list or a thorough past history may be ignored.
As in the situation above, where the psychiatrist runs the Coupler, the clinician must decide when to run the Coupler. The major difference is that the discussion of Coupler results may not take place immediately after the Coupler is run because some of the content of the Coupler may be considered beyond the knowledge level of the interviewer. In a mental health setting, medical issues are more likely to be beyond the scope of a mental health worker, and in medical settings, psychotherapeutic issues may be unfamiliar to medical personnel.
Although this may be a more efficient use of the psychiatrist's
time, this procedure risks minimizing the importance of the coupling process,
and the Coupler results could even conceivably be ignored, just as critical
laboratory reports may be overlooked.
If the patient presents with multiple complaints, more than one Coupler may need to be run. I have found that two Couplers can be run without interfering with the evaluation process. If an evaluation requires more than two Couplers, it is best to schedule them over more than one session. Again, there is probably no additional increase in time involved, since these are cases that require a longer evaluation anyway.
The Personality Disorder Coupler is the only psychiatric Coupler
that may take more than 10 minutes to run, because it contains questions
for each of the individual criteria of all the DSM-III-R Personality Disorders.
Because of this, a self- assessing questionnaire, the PsyComNet Self-Assessing
Personality Disorders Inventory by Ivan Goldberg, M.D., has been adapted
for use with the PKC System. The patient completes the questionnaire, and
a clerical person can enter the data directly into the Personality Disorder
Coupler using the "Add to the List of Findings" command from the "Other
Functions" screen. This process takes about 10 minutes of clerical time,
not counting time to print out the report. The results of the questionnaire,
as with all self-assessing instruments in personality disorders, must be
interpreted with caution.
Dawes, R. M., Fust, D., and Meehl, P.E. (1989) Clinical versus actuarial judgment. Science, 243, 1668-1673.
Ferguson, T. (1987) Health in the information age: sharing the uncertainty. Whole Earth Review, 57, 130-133.
Fitzgerald, R.V. (1992) Personal communication.
Huth, E.J. (1989) The underused medical literature. Annals of Internal Medicine, 110, 99-100.
Hyler, S. E., et.al., A Comparison of Clinical and Self- Report Diagnoses of DSM-III Personality Disorders in 552 Patients, Comprehensive Psychiatry, 30 (2), Mar-Apr 1989, pp. 170-178.
Oxman A.D., Sackett, D.L., Guyatt, G.H. (1993) Users' guides to the medical literature: I. How to get started. Journal of the American Medical Association, 270, 2093-2095
Weed, L. L., and Weed, L. (1994) Reengineering medicine. Federation Bulletin: J. Medical Licensure and Discipline, 81, 149-183.
Weed, L. L. (1991) Knowledge Coupling: New Premises and New Tools for
Medical Care and Education, New York: Springer-Verlag.