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

Abstract

Having found that an unintended consequence of computerized provider order entry (CPOE) implementation is “changes in the power structure” of the organization, we sought a deeper understanding of what was happening and why. If such consequences can be anticipated, they can be better managed. Using qualitative methods to study five successful CPOE sites, a multidisciplinary team found that CPOE enables shifts in power related to work redistribution and safety initiatives and causes a perceived loss of control and autonomy by clinicians. With recognition of the extent of these shifts, clinicians can anticipate them and will no longer be surprised by them. Greater provider involvement in planning, quality initiatives, and the work of clinical information coalitions/committees can benefit the organization and provide a different kind of power and satisfaction to clinicians.

INTRODUCTION

Computerized provider order entry (CPOE), defined as a system that allows a decision maker to directly enter medical orders via computer, has been shown to decrease medical errors [1–2], but it has often been resisted by users [3]. One reason for clinician skepticism has been the discovery of many consequences of CPOE implementation that are negative and were not intended, such as the additional time it takes to enter orders. To identify the full range of unintended consequences (UCs) related to implementing CPOE, we first held a conference of experts, and then we conducted fieldwork at five hospitals that had successfully implemented CPOE. Rigorous analysis of the qualitative data found nine types of UCs of CPOE. These categories have been described in prior work [4] (workflow issues, new kinds of errors, changes in communication patterns and practices, more/new work for clinicians, never ending system demands, changes in the power structure, overdependence on the technology, emotions, and paper persistence). The category “changes in the power structure” is perhaps the least obvious, and is rarely discussed openly. Although some shifts in power are usually expected, the extent is often greater than anticipated.

Power is “the ability to influence someone else,” while influence can be defined as “the process of affecting the thoughts, behavior, and feelings of another person” [5]. Power is closely related to leadership, authority, hierarchy, and control. There are many theories about sources of power in organizations, but most are based on that described by French and Raven [6]: reward power (control of rewards one sees as valuable), legitimate power (based on position and mutual agreement), referent power (personal power based on liking and respecting someone), and expert power (personal power based on expertise). Of interest to those in informatics, Robbins has added information power (having access to and control over information) to the list [7]. Power can certainly be abused, but in actuality, it is the reason most work gets accomplished, so it also may rightly be viewed positively [8]. To answer the question “how does CPOE change the power structure in organizations?”, we conducted a detailed analysis of all instances of UCs related to power in our data.

METHODS

Site and Subject Selection

We selected sites based on reputation for excellence in the use of CPOE, geography, type of organization (academic or community), kind of system (commercial or locally developed), and length of use. Sites included a county hospital in Indianapolis using the locally developed Regenstrief system (Wishard Memorial), three Partners hospitals in the Boston area (Brigham and Women’s and Massachusetts General Hospital using local systems, and The Faulkner Hospital using MediTech), and a community hospital in Burlington, North Carolina (Alamance Regional Medical Center, which has Eclipsys). Observations were conducted in a wide variety of inpatient and outpatient settings; subjects selected for observation were clinicians at work in those settings. We conducted hour-long oral history interviews with clinicians, medical records technicians, pharmacists, lab workers, information technology administrators, and others suggested by local principal investigators. The study received human subjects approval from Oregon Health & Science University and each individual study site.

Data Gathering and Analysis

A multidisciplinary team of 5 to 6 researchers visited each site for three-day periods. We completed 390 hours of observation of approximately 95 clinicians and 32 oral history interviews, using methods described elsewhere [9]. Qualitative data analysis software (QSR N6) assisted in analysis of 1,849 pages of transcripts and fieldnotes resulting from the expert conference and fieldwork. Once the instances of UCs were identified, a card sort method [10] was used for categorizing them. This was done iteratively in a series of 36 meetings until we reached agreement on the nine major types. For this paper, we then analyzed in detail the 44 instances of “changes in the power structure” using a process of axial coding, which is “intensive coding around one category” [11].

RESULTS

In general, we found that all of the sources of power outlined by French and Raven, plus information power, were related to CPOE implementation and use. Legitimate reward (and punishment) power was clearly used by administration when CPOE use was made mandatory. Referent power was exerted by clinical opinion leaders and champions when, because these clinicians were liked and respected, others followed them in using the system. Clinicians themselves hold expert power to such a degree that they can sometimes refuse to use the system or they can convince other health care workers to enter orders on their behalves. Finally, information power is held by information technology and administrative staff and is evidenced by their having information about clinician ordering patterns or restricting access to data for research.

We found three patterns, all of which were shifts: 1) shifts in the power structure through forced work redistribution and mandated safety pursuits; 2) shifts in control with a perceived loss of clinician control; and 3) shifts in autonomy and a move towards coalitions.

Shifts in the Power Structure

The power held by hospital administrators is formal and carries the ability to influence and thereby bring about change. CPOE enables those who hold power to redistribute work and make changes as needed to assure safety.

Forced work redistribution

Usually, policy changes, and even changes to an organization’s bylaws, are needed when CPOE is implemented, and they force changes in ways of doing work. Power is often related to the ability to impact the workflow of others. Since administrators generally make the decision to implement CPOE, they are responsible for any shifts in work caused by CPOE. Discussing planned shifts in work responsibilities, one administrator mused “who is going to be the arbiter of that negotiation and who is the power, but also who has the sense of the organization as to what’s fair? There is a lot of shifting of work that takes place and that is a real problem.” One example of an inadvertent shift is that, because clinicians usually enter orders for several patients into the system instead of jotting handwritten orders after each, they may sit for long periods isolated at a computer. “You used to have the respiratory therapists round with the doctors and nurses. . .we have a big change now. The doctors have to write the orders into the computer, and the RT isn’t usually there, so they are either asking the nurses what the settings are, or they have to go find the RT or go check the ventilator themselves.”

Mandated changes for safety pursuits

Many UCs related to power shifts are caused by clinical decision support implementation. CPOE can enforce clinical practice guidelines in many ways, including requiring that particular fields be used for entering data, having default selections in lists that are the least expensive option, and tracking the ordering patterns of individual clinicians to find out if they are practicing according to recommended guidelines. “There’s all these rules that come from pharmacy. . . this has become a political power thing, is how I read it, and they get all emotional, oh, this is for patient safety!” Clinicians are sometimes wary of these: “he used the data for the purpose which he publicly proclaimed he was going to use it for but . . . in the mind of some people these data are being gathered for purposes that people aren’t being public about.” Though information technology staff members merely implement major administrative decisions, they are viewed as powerful. A clinician, speaking of safety concerns, said “We’ve said in the past that CPOE should be a clinical project and not an IT project, but it’s still amazing how much I think it comes from the IT department.”

Perceived Loss of Control and Autonomy

Control is based in power, but is somewhat different in that it also infers monitoring and decision making to alter course. Much of the monitoring involves controlling behavior in both overt and subtle ways.

Loss of control

Alert fatigue is an unintended consequence of decisions that users think are made by administrators (though usually coalitions decide). One informant stated: “an administrator could think that it’s good, it’s okay to send out twenty alerts if there’s one that’s gonna be right on target but if you ask the physicians to vote on that and if they’re not employed by the institution they’re not gonna vote on a twenty to one ratio.” Another way that administration might control users with CPOE is by providing cost information about tests and medications to discourage use of expensive ones. One user noted, however, “I know this lab costs more so I just ignore the other costs of the labs.” In another cost-related discussion, a clinician expressed resentment about selection of the system: “They went off and got a system with little or no physician input, got something that few docs wanted and ignored the rest of us. This was totally administration-driven to drive down costs.”

Control even extends to choice of terminology, which influences reimbursement fees: “I didn’t realize how important nomenclature was in ordering. . .who drives naming them? Is it driven by your interfaces with your other systems? Is it driven by your coders and how you need to charge? Who exactly determines the nomenclature is a big deal.”

Nurses have subtle but effective ways of controlling physicians. “Most nurses are good at encouraging doctors to enter their own orders.” “She will take phone orders from physicians when they are in their offices; if the physicians are on the unit, they are responsible for their orders.” A nurse at a non-teaching hospital said “The MDs are our guests, and can’t be ordered to do things.” A pharmacist noted “wrong patient errors happen daily. Nurses ask physicians to re-enter all of these orders. . . when errors are made by physicians, they are politely pointed out but infrequently corrected on their behalf—this is seen as a learning opportunity.”

Because power and control are sometimes informal and based on referent power (you like and respect someone), they can be held by individuals in positions that are not usually considered powerful. For example, ward clerks can informally exert a good deal of control: “Ms. X [the ward clerk] learned to do it [enter orders]. People are afraid of Ms. X, in a good way. Ms. X runs the unit. You know, she’s in control, and so you get somebody like her comfortable with the system and that can sell the merits.” Front-line information technology staff members have subtle strategies for influencing physicians as well. One information technology worker said “One of the mistakes we made, we decided to hide their order sheets. We thought, well, we’re just going to make this really difficult. If we’re not going to mandate it [CPOE], we’ll at least hide their order sheets. Well, that was a mistake, because what happened was, the ward clerk got orders on a napkin. You can’t have a napkin in the medical-legal chart.” Finally, at sites that had commercial systems, we were told how much control the vendor exerts over the hospital. “We’re dependent on the vendor to fix things that are outside the scope of our control.”

Loss of Clinician Autonomy

Autonomy implies independence and that one’s actions are one’s own choices. Physicians have traditionally been highly autonomous, because they have been so respected as the ultimate decision makers in clinical situations. Their autonomy is being challenged in many ways by the health care system as a whole, but CPOE is certainly a factor. As one interviewee stated explicitly, “the whole issue of physician autonomy, we don’t talk about it much but it needs to be recognized because it’s cherished by physicians. . .whether it’s the government, whether it’s payers or whether it’s CPOE, they resist lessening of their autonomy.”

In many respects, CPOE is a threat to the autonomy of the providers who are expected to use it, but it is likewise a threat to other health care professionals, such as nurses and pharmacists. Often, physicians do not directly rebuff administrative power but instead use their power over nurses. “The physicians are saying well, I didn’t go to medical school to be a secretary, and the nurses are saying well, I didn’t go to nursing school to be a secretary, and the unfortunate thing is the buck always stops with the nurse.”

Decision support can be a threat to physician autonomy. We were told a story about another hospital: “The mindset of the medical staff, that they were very much in the belief that they—they—were in control of the hospital, and the practice of medicine in that community. . . so that’s why the administration was subversive [about CPOE] because they really believed if they went at it on top of the blanket, it never would have gotten anywhere. The subversiveness was considered necessary, given the culture of that community.” Alert fatigue, a reaction to receiving too many alerts, may partially result from annoyance about being told what to do. One administrator noted: “they [physicians] view them [alerts] as a nuisance, they don’t want to be told how to practice, they don’t want a system to suggest practice.”

Coalitions gain power

While individual physician autonomy is weakening to an extent that surprises clinicians, clinical information systems committees of many kinds are gaining power. Coalitions, groups which band together for support so that they can influence other people, have been described as an effective power tactic [7]. The UC here is that with the proper composition, these coalitions can become extremely powerful. They are often the vehicles for implementing clinical decision support. All of the successful sites that we studied had interdisciplinary committees that provide oversight of clinical systems, including CPOE. Although they include clinicians as members, other clinicians sometimes remain skeptical, thinking that too much power is given to the committees: “There are committees to create order sets for each specialty based on best practices and they say ‘this is what you’ll use’ and there’s very little way for people to get around that and I don’t like that, I don’t trust them.” On the other hand, it seems that the clinicians who are most accepting of decisions to implement decision support are closely tied to these committees and highly involved in making decisions about what should be implemented.

DISCUSSION

Our goal was to gain a deeper understanding of the changes in the power structure that take place when CPOE has been implemented. There are definite shifts in power and control and a loss of individual physician autonomy taking place. The UC is that the shifts are surprises, especially to the clinicians most impacted by them. As the redistribution of power occurs, its ownership becomes less clear, causing discomfort among clinicians and confusion among health care team members. Such blurring of role boundaries as a result of information technology has been noted by others. Reddy et al found that introducing a wireless alert pager system in an ICU changed hierarchical boundaries. It caused shifts in control and blurring of responsibility when an attending physician and a resident received the same alert for the same patient, for example [12]. Saleem et al warn that role confusion between nurses and providers can become a problem when implementation strategies like clinical reminders are used [13].

Legitimate hierarchical power is exerted by decision makers at the highest level when they make the decision to implement CPOE. CPOE changes work practices in different ways depending on the system, and those in the organization make many decisions over time that affect workflow. Some decisions are made not because of CPOE, but they are enabled by CPOE. Users often perceive that IT and Quality Assurance have gained power, IT because it carries out the enabling process, and QA because CPOE enables safety measures not possible in a paper system. CPOE is the enabler of power redistributions, not necessarily the cause, so blaming CPOE, and especially the technical aspects of the system, is misplacing the culpability. Rather, a better understanding of what is really going on might be more productive. One strategy for avoiding surprises is to clarify roles and describe changes in power that may occur when CPOE is implemented as part of the planning process.

Whether something is intended or unintended depends on one’s perspective. The UCs related to the use of power, through CPOE, to make the hospital safer, are unintended from the clinical point of view, but perhaps fully intended by those in power. It is striking that the consequences we found closely mirror all five of the recommendations outlined by Amalberti et al for lowering barriers to achieving ultrasafe health care: limiting clinician discretion and autonomy; moving toward an equivalent actor mindset (no star surgeons, for example); optimizing safety strategies, and simplification of rules [14]. If these are organizational goals, they need to be understood by all stakeholders. A useful strategy would be to involve clinicians in the process, because only then will they be able to identify what they will gain and lose.

The control mechanisms we saw were softer, gentler, power shifts. The doctor-nurse game, much discussed in the 1960’s [15], seems to have changed with CPOE. Instead of a strict hierarchy between doctor and nurse, the boundaries are breaking down. In many cases, the nurses become the CPOE experts, but they encourage the physicians to enter their own orders. Nursing expert power is increasing. Mechanic et al noted control over workplace issues is especially important to physicians, but that control is eroding. Loss of autonomy is one of the main reasons for physician discontent [16], as CPOE forces guideline and rule adherence. Again, Amalberti et al have said that other industries have experienced a reduction in autonomy along with improved safety [14], so they would consider this a positive consequence of CPOE. Coalitions of an interdisciplinary nature in the form of clinical information systems committees are powerful and have great influence over CPOE. DiPalma, in an insightful study of power in hospitals, describes these complex coalitions as webs that crisscross hierarchical boundaries and she calls the result “webbed power” [17]. A final strategy would be to encourage such webs.

CONCLUSIONS

Concomitant with CPOE implementation and use is a shift in the power structure of the organization. Much of the redistribution of power is made possible by the ability of the system to provide clinical decision support, which can be used to monitor as well as guide clinician behavior. Because information technology and quality assurance staff tend to gain power as CPOE succeeds, clinicians perceive that they are losing power and autonomy. Coalitions are being formed which have considerable decision making power over clinical systems and decision support. Clinician involvement in planning for the shifts is needed, and heavy involvement in the work of the coalitions will, in fact, give power to clinicians once again but in a different form. Once the shifts in the power structure become acknowledged and anticipated, they can be managed.