Computer decision support systems in practice

Nevertheless, they are difficult to implement, require significant resources for implementation and maintenance and uptake by physicians is often suboptimal.

CDS are probably most useful if they are integrated in the electronic health record and computerised physician order entry (CPOE). Ideally the CDS algorithms take into account individual patient information (e.g. microbiologic data, body weight, renal function, concomitant medication, comorbidities etc.) and epidemiologic data (e.g. institutional or unit-specific antibiograms) to guide the prescriber to choose optimal antibiotic therapy with regard to antibiotic choice, dosing and duration and prescribe appropriate microbiologic and imaging tests.

In practice the integration of individual patient information is difficult to achieve since databases may not be easy to link or data may not be coded to a standard terminology (e.g. it becomes difficult to use microbiologic data if there is no clear code identifying the terms “S. aureus”, “Staphylcococus aureus”, “Staph. aureus”, “MSSA” etc. as the same concept).

User friendliness, functionality and seamless integration into the clinical workflow are probably also key for successful implementation of CDS. Any “alerts” of the CDS system should have a high positive predictive value (i.e. few false alarms) in order to avoid “alert fatigue” and subsequent ignoring of alerts by prescribers.

In summary CDS systems currently still face significant obstacles. With the widespread adoption of electronic health records and the ever increasing availability of electronic devices at the point of care it remains to be hoped that CDS systems will become a key pillar of antibiotic stewardship in the future.