Printer Friendly

The need for collaborative engagement in creating clinical decision-support alerts.

Clinical decision support (CDS) encompasses a broad array of technology and approaches, all of which involve the provision and use of clinical information in medical processes. (1) Medication-focused CDS is frequently used in the context of inpatient computerized prescriber order entry (CPOE) and outpatient electronic prescribing. CDS for medications includes active computer-generated alerts, interactive decision trees, and passive limits on order sets and selection options, all intended to guide users to the most appropriate medication therapies. (2)

While active, interactive, and passive CDS for medications are all important, this article focuses exclusively on active CDS for "common alerts," meaning drug-allergy, drug interaction, and dose-check alerts that are based on predefined drug knowledge bases such as those provided by commercial vendors including First Databank (South San Francisco, CA), Cerner Multum (Denver, CO), and the MediSpan (Indianapolis, IN) subsidiary of Wolters Kluwer Health. Institution-specific active alerts based on custom rules and logic are excluded from the scope of this discussion.

Hospital and health care organization concerns

From a user perspective, the configuration of common alerts for medications is conceptually simple: The computer should be programmed to present clinicians with relevant alerts at the proper time; irrelevant alerts are unhelpful and should be eliminated. When the rate of unhelpful alerts far exceeds the rate of relevant alerts, clinicians begin to get distracted and attribute their inattentiveness to alert fatigue. (3,7) Alert fatigue is more than just an annoyance because it increases the risks of harm to patients and liability to providers. Often, hospitals try to avoid excessive alerting by disabling low- and moderate-severity alerts as classified by the commercial vendors. While this approach reduces the number of alerts, it may not change the ratio of true-positive alerts to false-positive alerts, and it may also result in the failure to display some clinically relevant alerts, creating the potential for harm.

Although commercially available for over 30 years, CDS systems enabled by drug knowledge bases have not been adequately refined with regard to "simple rules" used for common alerts. Simple rules are rules using simple conditional logic (e.g., IF the patient takes drug A, AND drug B is ordered, THEN an alert is triggered). The technology of simple rules is reliable, but such simple rules often lack enough specificity to be clinically relevant. With the proliferation of information available in the medical literature and electronic medical records and the need for this information to be synthesized with patient-specific data, common alerts could be improved by incorporating more complex logic that goes beyond the logic of the most basic, general comparisons. For the purposes of this discussion, "logic" is defined as an algorithm containing inclusion and exclusion criteria that are used to detect a patient's susceptibility to an adverse event. The inclusion and exclusion criteria may use patient-specific data, such as diagnosis, conditions, and laboratory results, or order-specific information such as route of administration.

We believe that it is critical for drug knowledge base and electronic health record (EHR) vendors to work collaboratively and aggressively to increase the clinical relevance of common alerts for medications. Improvements in both knowledge representation and knowledge management are required, along with software tools that provide health care providers with additional control over common-alert configuration.

Knowledge representation is an area of artificial intelligence focused on how best to use a set of symbols to represent a set of facts. CDS alerts relating to dose limits, for example, might be improved if dose information were represented not only in relation to the patient's age but also with regard to weight, organ function, tolerance level, and other patient factors.

Knowledge management involves a group of organizational processes for creating, curating, updating, distributing, and applying collected knowledge. In the context of medication-focused CDS common alerts, knowledge management challenges include the use of outdated drug knowledge base files due to difficulties loading updates.

The changes advocated in this article will lead to an increase in meaningful alerts and a consequent reduction in less relevant alerts. To be fully effective, common alerts for medications must save clinicians time and improve their workflow. We believe these outcomes and benefits are not being achieved in most health care organizations today.

As CPOE is being implemented on a widespread basis, prescribers are encountering alerts that did not exist in paper-based systems. Prescribers are now experiencing the sort of medication alert fatigue that pharmacists have experienced for decades. The difference is that prescribers have proved to be less willing to tolerate numerous unhelpful alerts and, as a result, have sometimes resisted CPOE altogether or demanded that the most frequently triggered common alerts be disabled. Health care organizations are struggling to use common alerts appropriately due to both concerns over liability if alerts are turned off and the lack of common-alert standards. The improvements recommended in this article will, if broadly implemented, support compliance with the initial requirements for "meaningful use" of EHR technology mandated under the American Recovery and Reinvestment Act of 2009 by reducing the amount of unproductive alerts for drug-allergy and drug interaction checking, thus improving physician acceptance of these systems. (8)

The need for greater collaboration

Just as it takes all the legs of a three-legged stool to support the seat, collaboration among three parties is needed to support safe and effective CDS common alerts. Hospitals (and health care organizations) that perform the local configuration of EHR systems form one leg of the stool; drug knowledge base vendors and EHR vendors form the second and third legs, respectively. In order for common alerts to work well, these three parties need to align, as must the legs of the stool. If standards for alert content, triggers, and use were developed, those standards would serve as a necessary stabilizer to reinforce collaboration. Standardization is a well-known, effective approach for creating and sustaining predictable outcomes, but its impact is always limited by the scope of adoption. The current environment does not permit the widespread adoption of common-alert standards via the sharing of common-alert configurations among organizations in order to create a consistent nationwide safety net.

If CDS common alerts that have proved to be effective could be more easily identified and shared, each health care organization could adopt a set of standard common alerts with some assurance that its configuration is complete, effective, and safe. Given that there are numerous combinations of EHR and drug knowledge base products currently in use, it is important that standard common alerts be developed in an open, collaborative manner. One way to accomplish that is to use a shared methodology for developing and maintaining common alerts for medications across EHR systems.

Opportunities. We have several recommendations on how to give health care organizations more control over common alerts in order to improve the value and usability of these alerts. Drug knowledge base vendors can also benefit from the establishment of reliable, professional consensus guidance, which could significantly improve the EHR user experience. The following are key areas where collaboration is needed to improve common alerts.

Alert content. Vendors and health care organizations should discuss how the clinical relevance of common alerts can be openly prioritized and endorsed by a group of clinicians. The American Society of Health-System Pharmacists (ASHP) has a membership of knowledgeable and experienced informaticists who, in conjunction with other clinical informaticists (physicians and nurses) might serve very well in this capacity.

User interface. The common-alerts user interface should have the following configurable elements: a notification method (e.g., display window, e-mail, text message, pager message), a predefined alert-escalation pathway (e.g., time- or rule-based escalation or cancellation), alert display-window style options (e.g., size, color) and alert display-window functionality options allowing the user to override or accept alerts or take other action as needed (e.g., buttons to remove the original triggering order, set a new triggering order, or override the alert), and a method for presenting relevant patient data (e.g., contraindicated medication therapies, laboratory values, patient conditions, medication orders) at the time of the alert. Common alerts should always provide quick and easy access to evidence of clinical relevance, so that clinician users can further evaluate an alert.

Data on previously triggered common alerts and responses should be readily available for users to review in the EHR.

Triggers. Alerts should be designed to allow triggering by a group or groups of inclusion and exclusion criteria, as defined by health care organizations, including:

* Medication characteristics (e.g., ingredients, generic name, therapeutic class, AHFS Drug Information class, controlled-substance schedule),

* Patient demographics (e.g., age, gender, actual body weight, ideal body weight, diagnosis, location of care delivery, clinical service),

* Clinical characteristics (diagnosis, indication for medication use, medical history, allergies, laboratory values),

* Trends in individual patient data (e.g., temporal analyses of laboratory values, calculated measures of organ function),

* EHR events (e.g., posting of laboratory values, entry of patient-specific information, charting of medication administration),

* Medication order details (e.g., dosage forms, administration routes, administration rates and frequencies),

* Place in workflow (e.g., order selection, entry, modification, discontinuance, administration),

* System user characteristics (e.g., service, years of experience, specialty), and

* Location and venue characteristics (e.g., emergency department, outpatient setting).

Actions. CDS systems should enable users to respond effectively to a common alert by providing direct action-taking capability (e.g., the ability to discontinue, cancel, suspend or modify current or interacting medication orders or to defer or forward an alert from within the alert). The alerting method should provide the capability for users to provide feedback, to provide a rationale for responses to alerts, and to readily access more advanced CDS systems.

Performance. Users should never have to wait for common alerts to appear or become functional, and alerts should not impede EHR system performance.

Documentation and outcome assessment

The alert system must log all alerts and user responses, including

alert triggers, overrides or acceptances, actions taken, reasons for actions, and other communications. This information should be shown at subsequent instances of the alert (e.g., physician responses at order entry are shown during subsequent pharmacist verification). All alert-related data must be available for further processing and analysis as required for quality audits and CDS governance. Patterns of alert response and subsequent actions taken in the EHR should be used to create standardized alert outcome reports.

Interoperability. To facilitate the sharing of common alerts, EHR systems must be capable of rendering individual alerts as text files with a standard format and style (similar to the Food and Drug Administration's XML-based structured product labeling standard and Health Level Seven's Arden Syntax Standard [Health Level Seven International, Ann Arbor, MI]).

Achieving greater collaboration. We propose that ASHP facilitate the development of standards for the clinically relevant content used to enable and govern common alerts related to medications. We believe that ASHP should call on its membership, working in conjunction with other clinical informaticists, to produce, validate, maintain, and provide the marketplace with evidence-based information on clinically relevant drug allergies, clinically relevant drug-drug interactions, and clinically relevant dose limits. We are in agreement with the recommendations on CDS that the ASHP Board of Directors reviewed in June 2012. (9)


The potential for CDS common alerts for medications to have a major impact on the quality, safety, and cost of care has been demonstrated by numerous studies. In order to realize the promise of common alerts, health care organizations and drug knowledge base and EHR vendors must come together to substantially enhance alert capabilities, so that alert systems provide flexible, patient-specific alerting that reduces both false-positive and false-negative alerts in favor of useful information to clinicians, and participate in the collection and maintenance of the tools to maintain these alerts.

The members of the American Society of Health-System Pharmacists 2011-12 Section Advisory Group on Clinical Information Systems and the section's executive committee are acknowledged for their help and support in the preparation of this document.

The authors have declared no potential conflicts of interest.

Copyright [c] 2013, American Society of Health-System Pharmacists, Inc. All rights reserved. 1079-2082/13/0102-0150$06.00.

DOI 10.2146/ajhp120435


(1.) Richardson JE, Ash JS, Sittig DF et al. Multiple perspectives on the meaning of clinical decision support. Proc AMIA. 2010:1427-31.

(2.) Chaffee BW, Zimmerman CR. Developing and implementing clinical decision support for use in a computerized prescriber-order-entry system. Am J Health-Syst Pharm. 2010; 67:391-400.

(3.) Glassman PA, Simon B, Belperio P et al. Improving recognition of drug interactions: benefits and barriers to using automated drug alerts. Med Care. 2002;40:1161-71.

(4.) Ash JS, Berg M, Coeira E. Some unintended consequences of information technology in health care: the nature of patient care information system-related errors. J Am Med Inform Assoc. 2004; 11:104-12.

(5.) Petersen JF, Bates DW. Preventable medication errors: identifying and eliminating serious drug interactions. J Am Pharm Assoc. 2001; 41:159-60.

(6.) Cash JJ. Alert fatigue. Am J Health-Syst Pharm. 2009; 66:2098-101.

(7.) Chaffee BW. Future of clinical decision support in computerized prescriber order entry. Am J Health-Syst Pharm. 2010; 67:932-5

(8.) American Recovery and Reinvestment Act of 2009, Pub. L. No. 111-5-4102 (n)(2)(A-2) (E) (2009).

(9.) American Society of Health-System Pharmacists. Official language of professional policies approved by the 2012 ASHP House of Delegates, June 27, 2012. Policy 1212: Clinical decision support systems. Policy/HOD/OfficialLang2012Policies. aspx (accessed 2012 Oct 31).

By David Troiano, Michael A. Jones, Andrew H. Smith, Raymond C. Chan, Andrew P. Laegeler, Trinh Le, Allen Flynn and Bruce W. Chaffee

David Troiano, BS Pharm, MSIA, is Director of Consulting, Dearborn Advisors, Chicago, IL.

Michael A. Jones, PharmD, is Informatics Pharmacist for Clinical Decision Support, University of Colorado (UC) Hospital, Aurora, and Adjunct Associate Clinical Professor, Skaggs School of Pharmacy and Pharmaceutical Sciences, UC Anschutz Medical Campus, Aurora.

Andrew H. Smith, BS Pharm, MHZ, is Pharmacy Clinical Applications Analyst, Novant Health, Inc., Winston-Salem, NC.

Raymond C. Chan, PharmD, is Informatics Residency Coordinator--Information Technology, Sentara Healthcare, Norfolk, VA.

Andrew P. Laegeler, MS, PharmD, is Pharmacy Informatics Operations Manager, Harris County Hospital District, Houston, TX.

Trinh Le, MS, BS Pharm, FASHP, is Clinical Pharmacy Manager--Informatics, Department of Pharmacy, University of North Carolina Health Care, Chapel Hill.

Allen Flynn, PharmD, is Solution Designer, Health Practice Innovators, Ann Arbor, MI.

Bruce W. Chaffee, Pharm D, is Coordinator for Strategic Projects and Adjunct Clinical Associate Professor of Pharmacy, Department of Pharmacy Services, College of Pharmacy, University of Michigan, Ann Arbor.
COPYRIGHT 2014 American College of Physician Executives
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Reprint of American Journal of Health-System Pharmacy article
Author:Troiano, David; Jones, Michael A.; Smith, Andrew H.; Chan, Raymond C.; Laegeler, Andrew P.; Le, Trin
Publication:Physician Executive
Geographic Code:1USA
Date:May 1, 2014
Previous Article:Safety science as second nature: training residents to use best practices instinctively to keep patients safe.
Next Article:Be prepared ... 2014 Summer Institute: July 9-13, 2014 Sheraton Seattle Hotel.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters