Clinical decision support system

From Ganfyd

Jump to: navigation, search

Clinical decision support systems (CDSSs, clinical decision assistance tools, clinical assistance support systems etc) have evolved to differ in type and complexity[1]. The types are:

  1. Passive
    • Users explicitly make a request for support (as you have done if you have accessed this wiki directly)
  2. Semi-active
    • An algorithm triggers in the clinical context to give an option to access the tool but the user still has to request the assistance
  3. Active
    • The algorithm triggers in the context and either presents information without further input or makes a decision without further clinical input
    • It is known that such systems are more effective than the other options but are more intrusive[2]

The complexity varies from

  1. Simple
    • Usually check input, verifying that value allowable or reasonable and may identify contraindications
    • Output may be alert or reminder
    • Usual form implemented in clinical order systems or e-prescribing
    • Challenge is to not interrupt clinical flow unnecessarily
    • Usually easy to understand by trained clinicians
    • Even something as simple and obvious as allergy checking systems may have less evidence for effectiveness than many assume[3]
  2. Mid-level
    • Prognostic calculators, usually using a scoring system
    • Clinical pathway aids
    • Challenge is that usually do interrupt clinical flow as likely to have interface complexity and require input
    • Usually easy to understand by trained clinicians
  3. Complex
    • Use artificial intelligence, data mining or mathematical modelling to classify or predict a diagnosis or clinical state
    • Usually difficult to understand how decision has been reached by trained clinicians
    • Implicate ethical and other issues
    • Tend to difficult to integrate into routine practice

In many jurisdictions such support systems could be classified as medical devices. Problematically clinical impact of such systems can be difficult to demonstrate[4].

LogoWarningBox4.pngLinks from this page can not be checked for compliance against medical device law applicable to stand alone software in relevant jurisdictions:
60px-Flag of EU.png

In Europe a medical algorithm or clinical decision assistance tool implemented in software (eg an app) that serves a therapeutic purpose such as calculating a drug dose, if made generally available or marketed is regarded as a medical device. In house medical devices manufactured by a health organisation and not generally available or marketed are exempt.

Flag of USA.png

The FDA has consulted on a similar requirement but the outcome has been slightly different and appears to depend upon if a human (however ill qualified to do so) can interpret and is necessary to implement the software output. Closed systems are completely regulated. A draft implementation document was published in August 2016