Much has been written about medical decision-support systems and there is a strong convergence of thinking about what makes for useful decision-support systems. Payne writes that ‘clinicians are more likely to use electronic decision-support systems if they give patient-specific recommendations, save time, and are incorporated into the workflow of clinic, office or hospital’. Some of the most important features of a medical decision-support system to be useful in solving medical diagnostic problems, identified in the following:

1. Good Performance: The algorithm has to be able to extract significant information from the available data. The diagnostic accuracy on new cases has to be as high as possible. In the majority of learning problems, various approaches typically achieve similar performance in terms of the classification accuracy, although in some cases some algorithms may perform significantly better than the others. Therefore, a priori almost none of the algorithms can be excluded with respect to the performance criterion. Rather, several learning approaches should be tested on the available data and the one or few with best estimated performance should be considered for the development of the application.
2. Dealing with Missing Data: In medical diagnosis, very often the description of patients in patient records lacks certain data. Medical decision-support algorithms have to be able to appropriately deal with such incomplete descriptions of patients.
3. Dealing with Noisy Data: Medical data typically suffer from uncertainty and errors. Therefore, machine learning algorithms appropriate for medical applications have to have effective means for handling noisy data.
4. Transparency of Diagnostic Knowledge: The generated knowledge and the explanation of decisions should be transparent to the physician. He should be able to analyze and understand the generated knowledge. Ideally, the automatically generated knowledge will provide to the physician, a novel point of view on the given problem, and may reveal new interrelations and regularities that physicians did not see before in an explicit form.
5. Explanation Ability: The system must be able to explain decisions when diagnosing new patients. When faced with an unexpected solution to a new problem, the physician shall require further explanation; otherwise he will not seriously consider the system’s suggestions. The only possibility for physicians to accept a ‘black box’ classifier is in the situation where such a classifier outperforms by a very large margin all other classifiers, including the physicians themselves in terms of the classification accuracy. However, such situation is typically highly improbable.
6. Reduction of the Number of Tests: In medical practice, the collection of patient data is often expensive, time consuming, and harmful for the patients. Therefore, it is desirable to have a classifier to be able to reliably diagnose with a small amount of data about the patients. This can be verified by providing all candidate algorithms with a limited amount of data. However, the process of determining the right subset of data may be time consuming, as it is essentially a combinatorial problem. Some medical decision-support systems are themselves able to select an appropriate subset of attributes, i.e. the selection is done during the learning process and may be more appropriate than others that lack this facility.
7. Integration into Other Medical Information Systems: A good clinical decision-support system should be able to access and to use of relevant clinical information on the patient derived from other systems as well as information directly entered as a part of the medical decision-support process.