US Pharm. 2012(37):47-49.

Computerized clinical decision support (CDS) offers the potential to reduce prescribing errors at the point of care. As noted in the September 2011 edition of this column, clinicians can employ a computer, usually an electronic health record (EHR), to prepare prescriptions.1 In some instances, the prescription will be forwarded to a printer in the clinician’s office and a paper prescription will be generated, which will then be manually signed by the clinician and handed to the patient to bring to the pharmacy. In other instances, the prescription will be transmitted to the pharmacy electronically, usually via an intermediary using a national data standard. The latter process is electronic prescribing. Both processes utilize the same computer software to prepare the prescription, thereby allowing the computer to apply a series of edits to check for potential errors that could harm the patient. Any potential error can then be addressed and corrected by the clinician before the prescription is printed or transmitted to the pharmacy. In the event that an edit is not clinically relevant to the patient, the clinician can override the edit.

The most commonly employed edits in commercially available EHR software are drug-allergy edits, drug-drug interaction (DDI) edits, and drug-dosage edits. Most EHR software systems allow clinicians to enter patient-specific data about medication allergies and adverse drug reactions (ADRs). The events are classified as either an allergy or an ADR, and the nature of the reaction, if known, is entered into the patient’s record. The severity of the reaction can also be described. For example, if a patient is allergic to penicillin, the record entry might indicate: penicillin allergy (severe); patient developed rash and anaphylaxis. Once a patient’s profile is created in this manner, all future prescriptions for that patient will be referenced against this computerized database. Any agent that might trigger a reaction will then trigger an alert. In this example, a prescription for penicillin will trigger an alert and the clinician will be informed that the patient previously had a severe reaction with rash and anaphylaxis. Agents that could potentially cross-react will also trigger an alert for the clinician’s review. For example, a cephalosporin prescription will trigger an alert. In this case, the clinician might know that the patient can tolerate the cephalosporin, so he or she will override the alert. For medical-legal reasons, it is often a good practice for the clinician to document why an alert is being overridden, particularly when the potential adverse outcome is serious. The utility of the computer edits for allergies and ADRs is highly dependent on having a complete and accurate database for each patient. If an adverse reaction is known (for example, in the patient’s old paper chart) but is not entered into the computer database, the software edits will fail to alert the clinician.

The second commonly employed decision-support edit is the DDI edit. The successful application of this edit requires two types of databases: 1) an accurate and complete medication list for the patient; and 2) a database of known DDIs. We will first review the patient’s electronic medication list.

Patient-Specific Medication History

EHR systems allow the clinician to populate the patient’s record with medication history information.    There are a number of potential sources of information that can be used to build this database for each patient. The patient interview is critical because the patient can supply current information about which agents he or she is taking. This can include information about herbal preparations, OTC agents, as well as prescribed medications. In addition, EHR systems will automatically keep a record of all prescriptions that have been entered into the prescription software. This includes prescriptions that were printed as well as prescriptions that were electronically prescribed. In an enterprise EHR system that is used across an entire clinic or system of clinics, the data from all prescribers in the system will be aggregated into the patient’s database. 

The process of medication reconciliation is critical because an electronic database might contain information about agents that were prescribed but either not taken or discontinued by the patient. In addition, the electronic database might be missing data about OTC and herbal preparations being taken by the patient. The medication reconciliation process is best done in a face-to-face encounter with the patient by a health care professional, such as a pharmacist or nurse. Each item can be reviewed and validated with the patient. This is also a good opportunity to reinforce the need for adherence to chronic medications and to answer any patient questions about medications.  

An EHR system can automatically capture prescribing information from all prescribers in a given office, clinic, health care delivery network, or integrated delivery system; however, the EHR will often not have information from prescribers outside of the system. The role of health information exchange (HIE) in this context is critical and can correct this deficiency. There are a number of vehicles available in the commercial marketplace to mediate HIE. For example, e-prescribing intermediaries such as Surescripts offer a medication history service that integrates information from both pharmacy benefit manager (PBM) paid-claims databases and retail pharmacy transaction databases. This provides a rich source of information regarding prescriptions from a large universe of prescribers. Medication history may also be available through regional or statewide HIEs. Even these various electronic data sources are not guaranteed to be 100% complete, however, because they might lack information on self-pay prescription transactions from certain pharmacies that do not report such data to retail databases. Nevertheless, a download of medication history from a source such as Surescripts can be a very valuable tool and can help ensure that the electronic database in an EHR is as complete and up-to-date as possible.

Drug Interaction Databases

The second required component for successful computer editing of potential DDIs is a drug-interaction database. Most of the available databases are produced by commercial entities. Some well-known examples are the Drug Therapy Monitoring System from Wolters Kluwer Health and the National Drug Data File Plus from First Data Bank; these products are supplied in the form of a module that developers can incorporate into EHR software. The First Data Bank product offers three severity levels and clinical effects subcategories. The product also incorporates drug-interaction monographs that summarize the relevant information regarding the interaction. Developers can utilize these features to design filters and presentation formats for clinicians.

The commercial databases are generally very comprehensive and include a wide variety of drug interactions, ranging from serious, life-threatening interactions to minor interactions of questionable clinical significance. Some interactions are specific to route of administration, and others relate to timing of administration of two agents. For example, the coadministration of calcium with levothyroxine can interfere with the absorption of levothyroxine. These commercial databases also frequently provide routine dosage recommendations.

The impetus behind the development of CDS with drug-allergy, DDI edits, and dosage recommendations is improved patient safety. One major study evaluated the application of a robust e-prescribing (e-Rx) system with decision support in the community ambulatory setting and compared it with traditional paper prescriptions.2 There was a sevenfold reduction in medication errors of all types using the e-Rx system. Although many errors probably would not have impacted patient outcomes, a small number of errors were more serious, including incorrect dosages. This study is an important proof-of-concept and shows the potential of the point-of-care computerized CDS software to improve outcomes for patients.

The Danger of Alert Fatigue

Given the wide variety of potential interactions in the commercial databases, many clinicians become overwhelmed with the frequency and volume of alerts that are of questionable clinical significance. This is the so-called phenomenon of alert fatigue. In one large, multistate study of more than 2,000 clinicians, it was found that clinicians override approximately 90% of alerts.3 However, clinicians did accept about 10% of the alerts, which implies that they found some value in certain circumstances. Nevertheless, alert fatigue may lead the clinician or an organization to turn off the entire DDI software package, which of course defeats a major patient safety initiative.4 These data suggest that the current approach to DDI checking is more of a hindrance than a help to clinicians. The challenge is to develop an approach to DDI checking that clinicians will find valuable more often than not. This will allow clinicians and patients alike to benefit from the potential error reduction as demonstrated in the Kaushal et al study.2

Recognizing the challenge to develop a more clinically relevant approach to the problem of identifying important DDIs, the federal Office of the National Coordinator for Health Information Technology (ONCHIT) has awarded a grant to the RAND Corporation and Harvard/Partners Healthcare to study the problem and develop a solution. The grant program, called “Advancing Clinical Decision Support,” has four goals (see SIDEBAR).5 The RAND Corporation and Harvard/Partners Healthcare are planning to take on the challenge of developing a “clinically important drug-drug interaction list, as well as a legal brief about the liability implications of using the clinically important DDI list.”

In addition to the grant program, ONCHIT has sponsored a national discussion workshop in which the problems relating to computerized CDS have been discussed by national experts.5 There was a consensus among these experts that there needs to be critical analysis of “how, when, and why each CDS intervention will occur.” In addition, clinical processes and workflows need to be redesigned to take full advantage of computerized CDS. Furthermore, participants concluded, “raising the specificity of alerts reduces the ‘noise’ level of nuisance alerts. This in turn would be expected to shift user perception of CDS from being annoying to being very useful. Discussions noted that identifying the scenarios and triggering data points for when CDS interventions will display requires significant effort by clinicians who will be using the system, and that clinician involvement is critical to the long-term success of CDS implementations. Additionally, discussions surfaced about the idea that the specificity of CDS interventions should be maximized for different localities with differing disease patterns and health priorities.”


There is great promise in the application of computerized clinical decision support in ambulatory clinical practice. Drug-allergy, DDIs, and drug-dosage edits can greatly reduce medication errors, as demonstrated in recent studies. However, there is a great need to fine-tune the clinical applications so that only clinically relevant alerts are presented to clinicians. This is necessary to avoid “alert fatigue,” which leads to situations where clinicians simply turn off all alerts. There is now a major effort in the federal government, in academic circles, and in the commercial pharmaceutical database industry to address this issue and refine the EHR applications so they will add value to clinicians and their patients.


1. Figge HL. What really is electronic prescribing? US Pharm. 2011;36(9):HS-35-HS-38.
2. Kaushal R, Kern LM, Barron Y, et al. Electronic prescribing improves medication safety in community-based office practices. J Gen Int Med. Accessed November 24, 2011.
3. Isaac T, Weissman JS, Davis RB, et al. Overrides of medication alerts in ambulatory care. Arch Intern Med. 2009;169(3):305-311.
4. O’Reilly KB. Doctors Override most e-Rx safety alerts. American Medical News. March 9, 2009. Accessed November 24, 2011.
5. US Department of Health and Human Services. Office of the National Coordinator for Health Information Technology. Clinical Decision Support. Accessed November 24, 2011.

To comment on this article, contact