An Unintended Consequence of CPOE Implementation: Shifts in Power, Control, and Autonomy.

From Clinfowiki
Jump to: navigation, search

Ash JS, Sittig DF, Campbell E, Guappone K, Dykstra RH. An Unintended Consequence of CPOE Implementation: Shifts in Power, Control, and Autonomy. AMIA Annu Symp Proc. 2006;:11-5.

Background: A recent paper by Campbell et al. looked at the unintended consequences that occur with the implementation of a computerized physician order entry system (CPOE) From this study nine unintended consequences of CPOE implementation were found. This paper focuses on the dynamics behind one of those unintended consequences- Shifts in power, control and autonomy.

Objective: To answer the question “How does CPOE change the power structure in organizations?”

Methods: The authors selected several sites based on their reputation with CPOE, geography, type of hospital setting and whether they had an in-house CPOE system or a commercial one. They visited each of the sites for 3 days to perform a qualitative study via direct observation in the inpatient and outpatient arena as well as structured interviews with users of the system. The combined 390 hours of observation, 32 user interviews, 1894 pages of field notes and transcripts were analyzed with Quality Analysis software (QSR N6) to identify any unintended consequences of CPOE implementation. Out of this analysis there were nine categories of unintended consequences found. For this paper, the 44 instances categorized as an unintended consequence of a change in power structure were analyzed in detail.

Results: The first change in power structure was forced work redistribution. Policy changes are needed to be made by a health system’s administration when CPOE is implemented. The result is forced changes in the workflow and workload of a clinician. The second change in the power structure was an increase in the power of the Quality Assurance department through mandated safety goals. The Quality Assurance department can use the CPOE system to enforce clinical practice guidelines by requiring certain data fields be filled out or having default selections in lists with the least expensive item.

The other dynamic noted in the study was the perceived loss of control and autonomy by clinicians. Clinicians felt they had less control over the types of alerts they received, the default values chosen for labs and medications, or even the type of clinical terminology that drove billing fees. Parallel to this, clinicians felt that their role and autonomy as the ultimate clinical decision maker was being usurped by the CPOE system.

As a counterbalance to these shifts in power, the authors note that physicians who are involved with and sit on CPOE oversight committees regain some of the individual power they lost by becoming a part of a coalition. All the sites reviewed in this article had an oversight committee and many of the decisions regarding decision support came from them.

The authors conclude that with the implementation of a CPOE system there is an unintended consequence of a shift in the power structure of the organization. The CPOE system doesn’t so much cause these changes as it enables them. Much of the redistribution of power is made by the ability of the system to provide clinical decision support to monitor as well as guide clinician behavior.

Commentary: The implementation of a CPOE system is in some respects the “Holy Grail” of clinical information systems implementations. The organizational commitment and resources needed to succeed are immense. This qualitative study shows that there is a shift of power with the clinicians feeling a loss of control and autonomy while the administration, information technology department, and quality assurance department gain power. True to its nature of being an unintended consequence the people interviewed were surprised by these changes in power. These changes in power need to be anticipated and planned for.

D. Klingen

Overriding of Drug Safety Alerts in Computerized Physician Order Entry

Helen Van Der Sus et al. Overriding of Drug Safety Alerts in Computerized Physician Order Entry. J Am Med Inform Assoc. 2006; 13: 138-147.


Many computerized physician order entry (CPOE) systems have integrated clinical decision support (CDS) with the potential to reduce medication errors and improve patient safety. The article looks at how physicians handle safety alerts when medication orders are processed. The review focuses mainly on the conceptual analysis of the determinants of overriding drug safety alerts.

Research Questions

The authors posted the following questions: “How often and in what situations are safety alerts overridden? Why do physicians override them? What effects ensue? What understanding of alert overrides can lead to improved alerting systems?”


The authors searched for literature on publications involving CPOE on Medline and Embase databases from January 1980 to December 2004. They examined and identified publications addressing overriding of drug safety alerts during medication order processing. Seventeen publications on overriding safety alert in CPOE were selected and nine provided quantitative information on overriding.


Safety alerts are overridden 49% to 96% of cases with lower override rates for drug interactions and contraindications. Though not completely confirmed, some studies showed that low-level alerts appeared to be overridden more often than high-level alerts. Drug renewals, drug interactions with topical drugs and poorly defined drug allergies were among those with high override rates. Adverse events were observed in 2.3% to 6% of overridden alerts.


CDS when integrated with CPOE systems improve patient safety by reducing medication errors through safety alerts. Alerts, depending on whether they are justified or unjustified, low or high specificity and sensitivity, could cause adverse events when overridden. Too many unnecessary safety alerts will cause alert fatigue and increase overrides. If poorly designed, decision support in CPOE can compromise patient safety. Improving alerting will decrease overriding and improve safety. Alerts need to be monitored, should have high specificity and sensitivity, do not unnecessarily interfere with workflow, and should have clear information.