Expert Systems For Clinical Decision-Making

The integration of patient or personal data with large amounts of stored information is widely used to perform computations in order to aid in decision-making, not only in medicine but also in several other fields.  Financial institutions use this strategy to check credit worthiness, or insurance companies use it to weigh their exposure to risk.  Such computer programs are often referred to as decision-support systems or expert systems.  The encapsulation of domain knowledge such as all known facts and treatment options on infectious diseases and antibiotics can be put together by a collection of implications.  Based on programming rules such as "if-then" scenarios, probabilities can be forecast.  In other words, if certain conditions exist, then a result can be defined.  This represents a more flexible type of expert system and is called the rule-based system.⁶  Today we know successful applications in medicine, such as the INTERNIST/QMR system in medical diagnosis⁷ or the Pathfinder project in pathology.⁸  They are based on causal or probabilistic reasoning and decision-theoretic schemes.  Later they were refined using so-called Bayesian methods.  These methods provide formalism for reasoning about partial beliefs under conditions of uncertainty.  This means that propositions are given numerical parameters that signify the degree of belief accorded them under some body of knowledge.  These parameters are then combined and manipulated according to the rules of probability.⁹  The primary limitations of a rule-based approach are that logical rule sets usually make rigid, binary [yes or no] decisions as to whether to classify given findings as important and there is no sense of a continuous degree of confidence.

However, only a few expert systems or other types of mathematical models have been developed for general dentistry or prosthodontics.  An example of such an expert system is the Oral Radiographic Differential Diagnosis (ORAD) software program that is designed to evaluate radiographic and clinical features of patients with intrabony lesions in order to assist in their identification using Bayes' theorem.¹⁰  Following the recognition of a dental problem and invoking the program, patient-specific information is entered to characterize the lesion in question.  ORAD's output provides a listing of each of the diseases in the program associated with the probability for that described condition.  It also computes a pattern match estimating how closely the set of entered characteristics match the typical presentation of each of the considered program's conditions.  This information is helpful to the clinician in the decision-making associated with arriving at a final diagnosis.  An online version of this program is available at http://www.orad.org.

Another example of an expert system for dentistry is the development of a simulation model of the caries process.  The model examines the influence of time relevant factors on health gain from restoring posterior approximal tooth surfaces.  This simulation model is described by Downer and Moles who demonstrated there is no automatic benefit from restorative dental treatment.¹¹  An advantage of using this decision model is that it highlights clinically relevant data that dentists need in order to make treatment decisions conducive to achieve a health gain for the patient.  They conclude substantially correct predicted outcomes, but within broad confidence limits and propose further investigation of cost-utility of restorative dental treatment.

These are only two examples of several tasks associated with diagnosis and treatment planning and represent only a fraction of dental skills needed to assess the patient's health status.  A dentist is trained in systematic workflow in order to gain insight into the patient's health status from clinical, radiographic, and health history findings.  Expert systems allow the dentist to offer maximum benefits to the patient in terms of time invested, dental specialty, and monetary outcome.  For the sake of completeness, additional skills in communication, education, training, and research are explained below.

Click Here To Navigate Within Article >> Abstract Introduction Decision-Making Models Expert Systems For Clinical Decision-Making The EOHR and Capturing Patient Care Data Diagnostic Technology Treatment Planning, Pretreatment, Treatment Dental Implications Recall and Outlook Conclusion References About the Author
	Page 4 of 16
Citation Number: Vol.  3, No.  1, Page 030