research of a clinical decision

Clinical Decision Support System: QMR
Student’s Name
Institution Affiliation
Clinical Decision Support System: QMR
Lack of sufficient information for medical practitioners affects the practice negatively. For example, it can lead to the wrong diagnosis. Having a lot of information available for the practitioner can also cause adverse effects in the profession. When there is too much information at the practitioner’s disposal, extracting the relevant information becomes a challenge. Clinical decision-making process thus becomes slow or inaccurate.
The problem of having too much information was the problem came about in the 19th century. There was a rapid explosion in the volume of medical information physicians had to peruse before making a decision. This made it hard to address emerging issues and questions in the field sufficiently. Most of the problems were either poorly answered or not answered at all. There were worries that those factors would lead to bad decisions being reached (Miller et al., 1986). Making a wrong decision in the medical field can have adverse results. Those factors necessitated the development of clinical decision support systems (CDSSs).
CDSSs are computerized systems that assist physicians in making clinical decisions. Most of them take in a patient’s information, compare it with a knowledge base or algorithm generated report, then give recommendations to the physician. One example of CDSS is the Internist-1 and later the Quick Medical Reference (QMR).

The Internist-1 experiment started as an academic project at the University of Pittsburgh. The idea of the project was conceived in 1972 and 1973 at the university (Lemaire et al. 1999). The principal intention of the project was to come up with an automated system that would make decisions like a clinician. The system was to accurately predict a diagnosis when fed with patients’ data such as symptoms and historical information. To do this, the contributors first developed a knowledge base comprised of medical facts which the system would use to make decisions.
An audit of the Internist-1 project was conducted in 1981. After the review, it was revealed that the diagnostic performance of the system was impressive. In qualitative terms, the performance was similar to that of physicians in several teaching hospitals. Despite this, the audit recommended that the system was not ready for deployment in hospitals. The recommendations were based on the shortcomings of the project. The primary deficiency of the project was its accuracy. The ability of the system to produce the right diagnosis was unreliable. The results it produced differed significantly with those from an expert physician.
The shortcomings of the Internist-1 project lead to the development of the QMR system. The QMR is a computerized CDSS used in internal medicine. It acts a quick reference for medical practitioners. QMR is an extension of the Internist-1 project. The QMR project aimed at adopting the success of the Internist-1 project while addressing its shortcomings (Miller & Masarie, 1989). The contributors to the project had gained valuable experience when building Internist-1.
Internist-1 engaged data collectors in labor-intensive activities. Users were subjected to abstracted access to the system’s knowledge base. It is, therefore, right, to say that it was designed to drive a high-powered system to solve complex systems in internal medicine. The QMR system aimed at developing a system that would meet the practical needs of medical practitioners.
The original Internist-1/QMR systems are hardly in use today. It has been a while since the Internist-1/QMR information base was updated. Nevertheless, other systems cloned from the projects are still in use today. They have been modified to address the shortcomings of the Internist-1/QMR and include emerging trends in the health sector.
What Data Was Input?
As earlier stated, the QMR system adopted the knowledge base of Internist-1. Most of the contributors to the project were volunteers. Each of the contributors would select a medical condition that was not previously covered. The contributor would then go into a period of extensive research on the topic. The aim was to go through as much literature review on the condition as possible. The contributors were allowed to present verified scientific facts only.
After the collection of the data, the contributor would then come up with a list of findings from their research. The conclusions included sign and symptoms, clinical abnormalities, and demographic data. These results were then combined to form a disease profile. Besides the disease profile, the contributor would then come up with a list of linked diagnosis. This would help in finding the related infections.
The disease profile was then reviewed by a team of experts who would resolve any inconsistencies or omissions in it to design a flawless system. The profile was not adopted until it underwent a further round of testing. This time, the profile was tested against a real case of a patient with the disease under examination. The profile was then adopted depending on the results (Lemaire et al. 1999). Each profile was to be tested against classic example every two years.
The Reasoning Models and Functionality
The QMR system had various uses. The primary function was returning results related to the typed input. Its functionality was almost similar to modern search engines. The user interface had a field for one to input the keywords. After entering the keywords, the system would then do a binary search on the knowledge base and display a list of relevant results (Miller et al., 1986). The results included disease profiles and links to related infections.
Apart from displaying raw results, the QMR system could also show differential diagnosis for the keywords entered. The list was a suggestion of the most probable diagnosis in descending order. When doing a predictive analysis, the individual conditions and the related data were first examined. The system would then combine the results. For example, if the keywords were fever and abdominal pains, the system would first investigate the profiles with fever then find all profiles with abdominal pains. The profiles that had both of this system would appear top in the suggested lists.
The Internist-1/QMR systems were also used in performing simple and complex case analysis. In a simple case analysis, the practitioner sought confirmation for a diagnosis that existed in the database. They would input all the positive results from the examined patient. The system then searches the knowledge base and returns the profiles that exactly match the positive results. In the complex case analysis, there is no precise diagnosis (Lemaire et al. 1999). The physician, therefore, inputs different negative and positive verdicts and the system returns the most probable diagnosis.
The Basis of Results Ranking
The system employed two methods of ranking the relevant search results. These two techniques were returning of possible case reviews and displaying a single diagnosis by applying the Internist-1 approach. The Internist-1 algorithm used a quasi-probable scoring scheme to examine the knowledge base. The results were then ranked depending on the positive findings. The system also past attention to past similar cases when generating the results. To enable this, the user had an option of confirming the correct diagnosis from the suggested list. The system would then match the entered findings to the diagnosis and save the relationship in the memory. The global approach worked on the assumption that several diseases can exist in one patient. The system, therefore, generated a general hypothesis which linked the infections.
What Was Noble about Internist-1/QMR?
At the time when the Internist-1​/QMR systems were being developed, the prevalent method in use was the MYCIN. The system combined various rules to advise physicians on the most probable diagnosis. The MYCIN was never deployed for use in medical facilities. Nevertheless, it paved the way for the development of more CDSSs (Miller & Masarie, 1989). The noble thing about the Internist-1 project was its knowledge base. The contributors managed to build an extensive information bank whose diagnostic performance almost matched that of an expert clinical officer.
QMR was developed from Internist-1. The elegant thing about QMR was its user-friendliness. QMR was initially based on microcomputers. It had a user interface that could be used by novice users. Using the QMR was just a matter of keying in some keywords and reading the output from the screen.
The Adoption and Use of The Internist-1/QMR
As mentioned before, the Internist-1 was never deployed into real use. An audit report on its accuracy recommended against using it in medical institutions. However, its successor, the QMR has been used in health organizations. Though it was one of the dominant CDSSs in the 1980s and 1990s, it did not enjoy the level of success that was expected.
The performance of the QMR system has been a subject of multiple studies. One of them is a recent study by the National Center for Biotechnology Information(NCBI). The study involved 70 medical experts. Each of the medical experts was given a previous case with a known diagnosis. The experts were then asked to prepare a diagnosis without using the QMR (Miller et al., 1986). After saving the results of the first experiment, they were asked to make a diagnosis using QMR tools and save the results. The results were then compared.
The QMR displayed the correct results in only 7% of the cases where the experts had arrived at a wrong diagnosis in the unaided experiment. In the cases where the experts got the correct diagnosis in the unassisted test, the QMR displayed the accurate diagnosis in its top 10 suggestions in 86% of them. It was therefore right to conclude that the predictive ability of the QMR was almost consistent with that of the medical experts.
A similar study was conducted by the Conjoint Research Ethics Board of the University of Calgary in 1996. The research found that the QMR produced the correct diagnosis in its top 5 suggestions in 41%-50% of the cases. The QMR suggested the accurate diagnosis in the top 10 predictions in 65 % of the cases. The study concluded that the QMR had limited effectiveness thus the medical practitioner’s input was necessary when coming up with a diagnosis (Lemaire et al., 1999).
Core Benefits of Systems
One of the immediate impacts of the Internist-1/QMR projects was the significant reduction in the decision-making time. One profile on the Internist-1/QMR would take weeks to compose. The facts that a physician could have the information displayed in a matter of minutes reduced the decision-making time. With time, the projects brought down the cost of treatment. This is because the medical practitioners were saved the hassle and costs of collecting data.
Furthermore, profiles were carefully examined and cleared of all inconsistencies before being added to the knowledge base. The records thus had a high degree of accuracy. Thought the efficiency and effectiveness of the system’s prediction systems were limited, the database was perfect. That meant that when a clinical officer added their expertise in interpreting the results produced by QMRs, they would build more accurate diagnosis.
Compared to earlier CDSSs such as Internist-1 and the MYCIN, the QMR had some unique benefits. First, it had a user-friendly interface. The QMR could thus be used even by novices (Miller & Masarie, 1989). In addition to that, it was based on a later generation of computers which had a higher processing speed. QMR systems, therefore, took a shorter time to display the results.
Limitations
The most significant limitation of the Internist-1/QMR systems is that if a disease was not in the database, it could not be returned as a diagnosis. This presented a new challenge. There was need to keep updating the profiles to add any emerging medical information. This was tedious. Furthermore, the profiles included related diagnosis. A single piece of information could impact multiple profiles. Updating the information base became hard.
The other limitation of the Internist-1/QMR was the limited effectiveness in producing the correct diagnosis. When the QMR system was being developed, the aim was to create a system that would provide accurate results, unlike the Internist-1 (Miller & Masarie, 1989). Thought the accuracy increased, the output of the system could not be trusted without input from a medical expert.
Finally, the Internist-1/QMR could not support natural languages. Natural languages enable systems to draw relationships between different languages. Lack of native language support meant that the methods could not return relevant results for synonyms of condition that existed in the knowledge base.
References
Lemaire, J. B., Schaefer, J. P., Martin, L. A., Faris, P., Ainslie, M. D., & Hull, R. D. (1999). The effectiveness of the Quick Medical Reference as a diagnostic tool. Canadian Medical Association Journal, 161(6), 725-728.
Miller, R. A., & Masarie Jr, F. E. (1989, November). Quick Medical Reference (QMR): An evolving, microcomputer-based diagnostic decision-support program for general internal medicine. In Proceedings of the Annual Symposium on Computer Application in Medical Care (p. 947). American Medical Informatics Association.
Miller, R. A., McNeil, M. A., Challinor, S. M., Masarie Jr, F. E., & Myers, J. D. (1986). The INTERNIST-1/Quick Medical Reference Project Status Report. Western Journal of Medicine, 145(6), 816.