Integrated decision support should prevent patients from receiving the wrong drug or the wrong dose when prescription errors are made. However, not all errors are caught because alerts are turned off, are not read, are misinterpreted, or are wrongly overridden. In , the process of order entry, interpretation, and handling of drug safety alerts and the emergence of medication errors are presented schematically.
Process of order entry and handling of safety alerts in computerized physician order entry.
may help explain why overriding does not always result in a medication error. There can be good reasons for overriding (justified overriding), for example, when the benefits of the drug (combination) outweigh the disadvantages and potential adverse effects can be monitored.19
Conversely, a cancellation or change of a drug order due to a drug safety alert can itself result in a medication error. Overriding a safety alert is often seen as a problem in itself, a system violation, but it should be emphasized that only unjustified
overriding (ignoring alerts, misinterpretation, wrong selection) poses a problem.
Justified overriding may be patient related or can occur when an alert is based on erroneous patient information. Patient-related reasons include, for example, clinically insignificant alerts, a limited treatment course, patient tolerance of the medication or dose in the past, discussion of potential adverse events with the patient or monitoring thereof, absence of a good alternative, and the benefits of the drug outweighing the disadvantages.19
Examples of erroneous patient information, justifying overriding, include inaccurate allergy information or medication lists that are out of date.19,30
Appropriate alerts can be defined as true positive alerts, alerts that are correct and current for the patient at hand. It does not imply that appropriate alerts are always perceived as useful.
The authors suggest that problems of safety alert overriding can be explained with the help of Reason's model of accident causation. This model is applicable to complex sociotechnical systems that require coordination of a large number of human and technologic elements and focuses on person, team, task, workplace, and organization.10,31
Alerting systems in CPOE are an example of such a complex sociotechnical system.
Reason distinguishes between active failures, error-producing conditions, and latent conditions.10,31,32,33
Active failures are errors (slips, lapses, and mistakes) and violations of an individual having an immediate adverse effect. Error-producing conditions are factors that affect performance of individuals, thus provoking active failures. These factors can originate in the environment, team of care providers, individual, or task at hand. Latent conditions are defensive gaps, weaknesses, and absences that are unwittingly created as the result of earlier decisions made by system designers, builders, regulators, and managers. Latent conditions can originate from organizational processes or management decisions. Reason's model shows that accidents result from a concatenation of several contributing factors at different levels: active failures, error-producing conditions, and latent conditions; individual and organizational factors. Simultaneous alignment of gaps or absences within diverse and redundant defenses results in accidents.10,31,32,34
shows how active failures leading to medication errors are the result of error-producing conditions in alerting system and physician and latent conditions in the organization.35
It shows how suboptimal decision support can reduce physicians' motivation, thus provoking active failures in alert handling.
Reason's model applied to drug safety alerts in computerized physician order entry.
Studies that linked overriding to adverse events showed that overriding did not result in adverse drug events in more than 97% of cases.15,19,20
shows that overrides might result in adequate therapy as well as medication errors. A medication error is any error in the process of prescribing, dispensing, or administering a drug, whether or not there are adverse consequences.36
Unjustified overriding of a drug safety alert (medication error) does not necessarily have adverse consequences. Overriding may, for example, result in suboptimal treatment or may be annulled by a dispensing or administration error. Conversely, adverse drug events can also result from justified overriding,20
or intrinsic drug toxicity.37
The relationship between overridden alerts and adverse drug events, presented in , shows that appropriate alerts can be overridden and that overriding does not necessarily result in adverse drug events. It shows also that justified overriding (adequate therapy) cannot always prevent adverse events.
Relationship between appropriate alerts, overridden alerts, and adverse drug events.
Two studies describe how expert panels review patient charts and score adverse events.15,19
For justification of overriding, it is easier to score override decisions, which gives more information about reasons.20,38
Requiring entry of reasons for overriding is more feasible for daily practice, triggers physicians to rethink the potential unsafe situation, gives pharmacists and other caregivers insight in the considerations, and can help adjust the knowledge database.
Alerting systems may contain several error-producing conditions increasing the risk of a medication error.