Digital-health

As it happens for all new phenomena, a bit confusion and doubts can arise on their real usefulness and how the doctors will accept them.

A Clinical Decision Support System (CDSS) is meant as a software designed to assist, at the point of care, doctors and other health professional in their decisional process.
Basically, there are two types of CDSS, the knowledge-based ones and the others. The first type is generally composed of three elements: a knowledge base, an inferential engine and a communication tool. As additional element a user interface can visualize the results (which can also be shown directly on the screen of the clinical system integrated with the CDSS). By means of the communication component, the CDSS sends the patient clinical dataset to the inferential engine, which uses rules and algorithms to extract the information from the knowledge base and return them to the clinical system.

The CDSS which are not knowledge-based generally use the machine learning, a sort of artificial intelligence (AI) which let a system learn from the past experience and/or find specific patterns in the clinical data. In this way the need to write rules is surpassed, even if these systems do not make explicit the process beyond the suggested results, so operating like black boxes.

Following the increasing interest in CDSS, a lot of software tend to attribute themselves the definition of decision support system. Generally, every application which generates information helping the doctor in his/her decisional process can be presented as a CDSS. It is the case of some bibliography data banks with medical contents; electronic medical record and prescription software which manage clinical information and have some control algorithms.

Real CDSS, in addition to the different architecture, imply two main points: the type of information they send back; the interface mode and the integration with the clinical systems. For what concerns information, being CDSS designed to be used at the point of care, the information they must return must be conceived for this purpose. A doctor assisting a patient cannot spend time to read articles, he needs clear and synthetic information which can alert or suggest him what to do. More than access to a plethora of publications, he needs prompt and specific suggestions.

Another crucial aspect is the CDSS integration and activation mode. One of the most diffused standards is HL7 Info-button, a protocol which allows a clinical system to get information on a knowledge base by executing a query.

By simply pressing a button, the doctor can access contents more or less related to the patient he is treating (in relation to the research string set by the clinical system and the ability of the knowledge base to filter and extract a group of pages).

The main limit of this approach is its own “pull” logic, that is passive. The doctor has to push a button when he needs a support and is aware not to know. We should consider that over 800,000 articles are published on more that 5600 medical journals every year! It is clear that a doctor cannot be updated on and acknowledged of all published scientific evidences.

For this reason, an efficient CDSS must work with a “push” logic, actively signalling to the doctor that there are information, suggestions, advises and alarms he should be aware of. Among the available standard operating in this way there are HL7 vMR and, under development, CDS Hooks, based on FL7 FHIR.

The most relevant question on CDSS is if they are really efficient and useful for the doctors and if the doctors are truly intentioned to use them. For my experience, I can say that generally doctors are aware they need to be supported by DSS systems based on the Evidence Based Medicine (EBM). The sectors where the interest is greater are: the medicine prescriptions (interactions, side effects, contraindications, dosage), included the therapy reconciliation; the differential diagnosis; the therapy. Except few cases when the doctor is in competition with the CDSS or refuses a-priori an EBM-based approach, doctors are generally interested in these innovative tools.

Concerning the CDSS efficacy, a consideration should be made. Differently from the system in use, Electronical Medical Records included, different studies have been made on CDSS using scientific criteria (i.e. randomized clinical trial), aimed to evaluate their impact and utility in the clinical practice. Moreover, a lot of systematic revisions compare more studies to take conclusions and give indications. Of course, the trial result strongly depends from the context, the used CDSS, the valuated indicators. Generally, the CDSS usefulness has been scientifically demonstrated in many studies. These instruments add a great value to the clinical solutions, increasing usefulness and efficacy of investments on electronic medical record and CPOE (Computerized Physician Order Entry).

In a further article I will present you some scientific studies on CDSS.