ity and frequency ranking and generalized patient data delineated only by age and weight. Such intelligent mechanisms would generate alerts specific to a patient’s unique characteristics and needs; physician prescribing, ordering, and error patterns; and evidence-based best-practice guidelines. For example, a patient might be allergic to one medication in a drug class but not others (Abookire et al., 2000). The software configuration should recognize the patient’s unique drug allergy without requiring that an alert be generated for every drug ordered in that class. As another example, a physician might have a preference for therapeutic duplication in transitioning a patient being prepared for discharge. The software configuration should accommodate the duration of the therapeutic duplication and the specific dosing transition of the two drugs without issuing repeated alerts requesting the same information. For this type of intelligent prompting to be possible, drug-related technologies must be linked not only to each other, but also to more comprehensive clinical information systems.

Incorporating rule-based physician monitoring features within prescribing systems is considered important for safety and learning. Anton and colleagues (2004) designed a more structured ranking of message severity according to seven categories, and system capabilities for monitoring based on storage functions and unique numbers for each prescription, provider, and patient. The system creates warnings using the incorporated rules and maintains a record of every occasion on which an alert is displayed. Each message can be linked to the user, the individual prescription key, and the outcome of the warning. Queries of the data were used to assess providers’ proficiency in preventing errors with the system and overall skill in using it (Anton et al., 2004).

In addition, evidence-based decision-support algorithms are necessary to ensure the adequacy of software configurations that incorporate specific protocols for real-time decision making and clinical action (Cole and Stewart, 1994; Sawa and Ohno-Machado, 2001; Fields and Peterman, 2005; Miller et al., 2005). Ideally, the algorithms should be developed according to three principles: (1) they should be system tested before full implementation; (2) they may have to be facility tailored based on process and workflow; and (3) they should be monitored and updated over time (Sawa and Ohno-Machado, 2001; Bates et al., 2003; Reichley et al., 2005). The algorithms, similar to any computer program, can never be finished or finalized as medicine is always changing; therefore, expiration labeling or update notices may be helpful to maintain currency.

One method for testing systems is to develop and test software configurations as well as train clinicians using simulation programs. The Anesthesia Patient Safety Foundation is the first medical community to adopt this technique successfully and apply it to anesthesia information management systems (AIMS) (Sawa and Ohno-Machado, 2001; Weinger and Slagle,



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement