gms | German Medical Science

Background: Medication errors and resulting Adverse Drug Events (ADEs) are an important issue of global healthcare. Several researchers and institutions recommend implementing electronic prescribing ("CPOE", computerized physician order entry) to reduce the number of medication errors [1]. CPOE systems can offer advanced decision support, offering drug-drug interaction alerts, drug-allergy checks, checks on dosing errors, drug-lab checks etc. Systematic reviews show that these CPOE systems can indeed lead to a reduction of medication errors [2] and thus positively affect patient safety.

However, one problem recently observed is over-alerting ("alert overload") [3]: Clinicians tend to ignore alerts, if they seem not relevant for the particular patient and context. Over-alerting can be reduced by prioritizing and filtering alerts based on the given clinical context, so that only alerts important and relevant for a given situation are displayed. Reducing over-alerting could increase alert compliance and consequently improve patient safety. However, it is still focus of debate how "context" can be defined, and which types of "context" are relevant when deciding whether or not to display an alert to a user.

The objective of this paper is to define "context", and to present "types" of context.

This reseach is part of a larger EU project "PSIP - Patient Safety through Intelligent Procedures in Medication".

Material/Methods: We searched major databases (Medline, Embase), the top-10 health informatics journals (JAMIA, Int J Med Inform etc.) as well as proceedings of major health informatics conferences (MIE, Medinfo). We searched for CPOE papers and especially for those focusing on alerting. As keywords, we combined MeSH-Headings (such as "order entry systems") with free keyword search.

Results: Until now, we have identified 23 scientific papers that deal with context in CPOE systems. First structuring based on content analysis revealed the following major categories of contextual factors that should influence alert prioritization:

1.

The user: Area of speciality, knowedge and experience, earlier reaction to a given alert

2.

The patient: sex, age, recent diagnosis, earlier diagnosis, severity of illness, ealier reaction to drugs, multi-morbidity, number of drugs prescribed, overall health status, risk group, severity of potential ADE

3.

The drug: Type of drug, type of interaction to other drugs

4.

The hospital: department speciality, earlier ADE rates, earlier medication errors, type of patient treated.

Based on the literature review, we will now verify the level of evidence available for each of those categories. "Level of evidence" means: Is there any evidence, that CPOE systems that take into account a specific contextual factor will improve alert compliance and by this improve patient safety? We will report these results at the conference.

Conclusions: CPOE systems can have large impact on quality of care. However, over-alerting and alert fatigue are more and more seen as barriers for patient safety. Future CPOE system will have to take into account the clinical context of each prescription in order to prioritize each alert according to the given context. Our research defines "context" and identifies major "context categories". Based on an analysis of the available evidence, we will derive recommendations how to handle these contextual factors in future CPOE systems.