Sites and CPOE applications

CPOE evaluations have been concentrated at large academic medical centers, particularly at Brigham & Women’s Hospital in Boston and the Ohio State University medical center in Columbus. The unique, “homegrown” systems at Brigham, Vanderbilt University in Nashville, and the Regenstrief Institute at the Indiana University School of Medicine (Indianapolis) populated earlier literature—although, more recently, vendor systems have been studied. Studies of vendor systems include the Siemens, Eclipsys, General Electric (GE), and Cerner CPOE applications, in descending order of frequency.

Medication errors and adverse drug events

The relationship of CPOE to medication errors, adverse drug events (ADEs), and subsets of those categories is reported in 12 studies (see Table 1). Of these, three systematic reviews addressed the effect of CPOE on medication errors and/or ADEs. The systematic reviews concluded that CPOE (and isolated clinical decision-support systems) can reduce medication errors.^19–21 However, since these reviews were published, researchers have published conflicting conclusions about the topic.^{22, 23}

Studies on inpatient units in five different settings reported significant decreases in medication errors after CPOE implementation.¹¹ More specifically, all medication errors and non-missed-dose errors decreased in two sites,^{10, 18} and potential and nonintercepted ADEs significantly decreased with the homegrown application at Brigham and Women’s Hospital.¹⁵ Shulman and colleagues¹¹ reported a lower proportion of medication errors with CPOE, and King and colleagues¹⁸ reported a 40 percent reduction in errors. Transcription errors were reduced or entirely eliminated.^{17, 20, 24}

In contrast, two researchers found no differences in rates for ADEs with CPOE.^{16, 18} In the outpatient arena, no differences were found in total errors, ADEs, or rules violations,¹⁶ although the system in this one setting lacked any order checking for completeness or accuracy and lacked basic decision support. The differences in ADE detection may be due to underpowered studies¹⁹ and also to the differences in functionality discussed earlier and the differing definitions for ADEs. Moreover, studies used varying scales to rate errors and ADEs ranging from self-developed categorical scales of minor/major/serious to the American System of Health-Systems Pharmacists classification.

Increases in medication error rates after implementation of CPOE have also been reported. In 2005, Spencer and colleagues¹² reported an increase in one type of error (i.e., pharmacy processing) on two inpatient medicine units, while other medication errors were unchanged from pre-CPOE implementation. Unfortunately, this site had no system interface to pharmacy, and the researchers acknowledge that the increased error rate may have been related to the need to transcribe orders in the pharmacy. Without the interface or a method to ensure orders were complete and accurate, the only seeming advantage of CPOE at this site was the speed of communication to the pharmacy for transcription. Researchers in Portugal¹⁷ concluded that CPOE eliminated their transcription errors, but other errors continued—such as right class/wrong drug, and other errors likely solvable by the basic decision support they lacked (e.g., unclear orders, missing frequency, incorrect dosages, drug interactions, and duplicative therapies).

Life-threatening errors and serious ADEs were higher in the early years at Brigham and Women’s, when no decision support was installed. For example, a screen for potassium orders allowed new, potentially very serious errors to occur.¹⁵ These potential errors were intercepted by either nursing or pharmacy before the drugs were administered. Likewise, CPOE at one institution in London created three major errors that could have resulted in harm or death of a patient had they not been intercepted.¹¹ Again, this site had no decision support in place to prevent a reported error of 7 mg/kg of morphine being ordered instead of 7 mg, a potential overdose of 70 times the normal range. This site also saw an increase in minor, nonintercepted errors with CPOE, from 43 for handwritten orders to 93 with CPOE. That said, with all errors combined, the overall rate of errors was lower with CPOE. However, the details behind that overall rate show increases in potential major errors if decision-support capabilities are not available. This statement is in contrast to the conclusion by Chaudhry and colleagues,⁴ perhaps because his literature review concentrated on the years before these newer studies were published.

In all studies, the researchers did not include external forces, contextual variables, or organizational forces that may have contributed to changes in medication errors. For example, several studies extended over multiple years, through changes in chief information officers, national changes about patient safety, and increased emphasis on medication errors. Especially with the more recent studies, changes in error rates could have been due to these factors as well as information technology implementations.

The variability in conclusions may also be explained by the differences in available functions, particularly the lack of pharmacy interfaces and basic decision support. Therefore, the previous researchers’ conclusions about CPOE decreasing medication errors must be modified: CPOE can reduce medication errors if appropriate functions are available to prevent new errors. Transcription errors can be eliminated with electronic communication and interfaces together with structured order entry. CPOE can substantially reduce overall (and many serious) medication errors if (1) electronic communication and automatic order interfaces are in place, (2) basic order checks for completeness are present, and (3) decision support at its most basic level is available—checking for drug–drug and drug–allergy interactions and for dosing ranges.

Clinical efficiency measures—time and length of stay

Eleven studies examined the effects of CPOE on efficiency measures of time and/or hospital length of stay (LOS) (see Table 2). CPOE offers clear benefits in processing efficiency for orders management and availability of electronic laboratory and radiology results. CPOE reduced the time from order entry to results availability for laboratory and radiology orders in four sites.^{13, 25–27} Another clear benefit is that CPOE decreased the time from pharmacy ordering to medication administration time.^{13, 25, 26, 28}

Likely because of timely availability of results and faster order processing, patients’ hospital LOS was shorter.^{26, 29} One systematic review concluded that one of the benefits of CPOE is reduced LOS.²⁰ Other clinical efficiencies are related to use of CPOE. For example, Ohio State University hospital saw a significant improvement in the number of patients whose abnormal potassium levels were normalized within 24 hours.²⁴

On the other hand, order entry itself takes longer using CPOE than with paper. A systematic review of the impact of computers on time efficiency concluded that the use of central desktops for CPOE was not efficient, consuming 98.1 percent to 328.6 percent more time per working shift.³⁰ CPOE took 2.2 minutes longer per patient, but after duplicative tasks were removed, the extra time per patient was shortened to an average of 0.43 min.³¹ At Brigham and Women’s, CPOE took 44–73 minutes longer per day, especially for entering one-time orders.³² However, this study was done before order sets were widely used. At Regenstrief, an early CPOE application took interns 33 minutes longer during a 10-hour period.²⁹ Interns entered orders on microcomputers and then printed them, using them as traditional paper documents afterwards. Likewise, a more recent study at Massachusetts General Hospital³³ demonstrated an increase in medical interns’ ordering time, among other time-related variables. Prior to CPOE, interns spent 2.1 percent of their time ordering; after CPOE, they spent 9 percent of their time ordering. Two of these studies were published about early CPOE applications in the 1990s, and all four measured homegrown systems. None of the studies examined time for order sets or vendor-based solutions. Of note, CPOE may take providers somewhat longer to enter orders, but efficiencies are obtained later in the orders management cycle—in nursing, ancillary departments’ order processing, and in reduced time for results availability and administrative tasks.³³

Quality care variables

Three studies examined variables not reported by others, namely the quality of documentation and one particular patient outcome (see Table 3). In one study, a significant increase occurred in the number of documented consents for do-not-resuscitate orders.³⁴ In another study, researchers conducted a randomized controlled trial to examine the effect of medical students’ rotations in a CPOE site versus traditional sites on the quality of orders written on a fictitious patient⁵⁰ They found that the quality scores for orders during an academic examination were significantly higher for students using CPOE.

A third study produced alarming results that conflict with other, more promising benefits of CPOE. A recent article reported an increased mortality coincident to a CPOE implementation at a pediatric hospital.¹⁴ The researchers reported a direct association between CPOE and increased mortality among pediatric patients admitted through interfacility transport. However, this facility experienced substantial workflow changes in conjunction with CPOE installation. For example, no preregistration was available, delaying order entry until full registration was completed after the patient physically arrived. This process change delayed therapies and diagnostic testing.

Important human–computer interaction issues impacted treatment times. The new ordering system required substantially more order entry time during a critical period of patient care, and the wireless bandwidth capacity was often exceeded during peak periods. Crucial aspects of work organization changed with all medications being centralized, meaning that nurses were unable to access medications locally and the pharmacy could not process medication orders until they had been activated. Sadly, when the pharmacy accessed CPOE to process an order, other clinicians were locked out of the application. The researchers also reported a decrease in face-to-face communications that provided relevant information for patient care management post-CPOE.

In this study, CPOE was likely a proxy variable for significant, but untoward, process changes in that particular institution. The lesson from this article is that work processes must be thoroughly examined before CPOE “goes live,” and projected, substantial treatment delays are an excellent reason to delay going live until work design is safe for patient care. For critically ill patients in emergency departments (EDs), intensive care units (ICUs), and pediatric units, new processes cannot delay treatment. Workflow and usability analyses can preclude the kinds of impacts seen in this article.

Study descriptions and designs

Identifying and specifying the capabilities of CPOE is imperative. Future studies should indicate the exact functions in the article and abstract as well, and in the title, if at all possible. Chaudhry and colleagues.⁴ noted this same issue with general HIT studies. Careful conclusions are necessary when CPOE does not provide basic functionality such as pharmacy, laboratory, or eMAR interfaces. CPOE capabilities are not equivalent, so they should be treated like the different independent variables they are. Adequate descriptions of these study characteristics are needed, at minimum.

Broadening the definition of CPOE would allow researchers to conceptualize future studies differently. The term “order entry” misrepresents the concept by implying that only the entry portion of the whole process is important. Ordering is a process starting with entry, to communication, to processing by various recipients, and then to documenting actions against specific orders. By conceptualizing ordering in this way, future studies can be designed to measure impacts across the health team.

Potential external influences need to be taken into account in study design as potential confounding variables. Studying CPOE in a natural environment is challenging research. However, rather than ignore these variables, as has been done in the past, future researchers should want to identify and control, or at least measure, these variables. This notion is stressed by Snyder and colleagues.³⁸ External forces outside the institution should also be considered in study conclusions—for example, the influence of national trends for increasing patient safety with concomitant information technology installations.

Future research themes

Three major themes for future research emerged: (1) nursing impacts from computerized orders management, (2) human-computer interaction issues, and (3) implementation science. The concept of CPOE needs to be expanded to encompass an orders management cycle. To date, the concept has been studied primarily as order entry in quantitative studies. The ordering process is a complex, interdependent, and interactive process composed of at least these multiple, intersecting elements: systems design, interpersonal and intersystems communication, implementation processes, and organizational structures. Thus, orders management needs to be examined in the future as the interdependent, interdisciplinary, and interactive process that it is. A few authors of qualitative studies have started that process.

Because orders management is a complex process, identifying only simple outcomes variables does not do the phenomenon justice. Multiphased studies with multiple process and outcome variables are needed to begin to understand orders management and its impact.

Decrease in medication errors

The most commonly measured variable was the change in medication error rate. Four studies described the measurement of this variable: three showed a decrease, and one study recorded an increase in medication error rate post-BCMA implementation. Coyle and colleagues⁶⁶ described the implementation of a BCMA system at one VHA facility—some of the refinements and upgrades that the system had undergone based on nursing recommendations. The changes were made to address specific workflow issues to increase acceptance of BCMA in routine practice. A survey administered to the nursing staff evaluated the acceptance of BCMA technology after 3 years of implementation. Results showed that 97 percent of the nursing staff agreed that BCMA had decreased the risk of medication errors. Additionally, in the first year, the medication errors decreased by 23 percent, and, by the fifth year, by 66 percent. Consistent with these results, another VHA study by Johnson and colleagues⁶⁵ compared the overall medication error rates in 1993 and 2001 and found an 86 percent decrease over this period. Anderson and Wittwer⁶⁷ conducted a similar study in a nongovernmental setting and found that the medication error rate decreased to less than 50 percent of its baseline value within 6 months of implementation of the BCMA system.

Another study measuring medication error rates was conducted in the medical-surgical units of a Midwest government hospital.⁶⁸ This study examined whether there was a difference in the medication error rate 1 year before and after implementation of BCMA. The error rates at the administering and dispensing stages were specifically examined. The study showed an 18 percent increase in the medication error rate. However, this increase was explained by the ability of the BCMA system to record any discrepancies in the medication administration, such as late or missed doses, which were previously underreported or went undocumented.

Discrepancies in documentation

The documentation functionality of BCMA systems enables the creation of an accurate and complete MAR, which can then become part of the patient’s EHR. However, large discrepancies exist in medication documentation, even in an electronic environment. Only one study compared the discrepancies in documentation that arise upon implementation of a BCMA system. An examination of the discrepancies between MAR and patient billing records for large-volume intravenous solutions identified three types of discrepancies.⁶⁹ Failure to document administration to a patient occurred 38 percent of the time, the rate of failure to credit the patient for returned solution was 37 percent, and the rate of administration of solution to a patient other than for whom it was dispensed was 25 percent. This study also examined the potential for BCMA technology to decrease these discrepancies in documentation. A 19-percent improvement in the consistency of documentation was observed after introducing BCMA technology.

Impact on nursing workflow

A supplementary finding of the study by Barry and colleagues⁶⁹ was that the lowest scanning rates for items were achieved by the nursing personnel, largely because BCMA has such a huge impact on the workflow of nurses. In this study, the scanning capability was available only at the nursing station, requiring nurses to return to the station each time they wanted to scan an item. Lack of consideration of nursing workflow processes can result in low rates of adoption of BCMA technology. Another study of the medication administration process from the perspective of nurses found ways to make the technology less disruptive to nurses’ workflow.⁷⁰ The researchers described strategies to improve acceptance of the technology among nurses and hypothesized that a tangible measurement of this acceptance would be seen in the increase in scanning rates. Patient armband scans went up by 7 percent, and medication label scanning showed a 15-percent increase over a 5-month period. The increase in scanning rates was small, but the study lacked a clear description of what the exact intervention was, and hence it is difficult to make conclusions about why it failed to have a larger impact on the scanning rates.

Using human factors theories to guide an ethnographic evaluation, before and after implementation of a BCMA system, the analysis and process-tracing protocols derived five negative, unintended effects of introducing this technology.⁷¹ The investigators found that BCMA technology can lead to the creation of work-arounds that might result in new paths to occurrence of ADEs. This study is the first of its kind to conduct an in-depth analysis of the design modifications, organizational policies, and elements of training that need to be in place for BCMA technology to fit seamlessly into the nurses’ workflow.

Thus, the range of variables assessing the impact of BCMA technology on the workflow of nurses is broad and complex. An example of a simplistic variable is the measurement of the nurses’ acceptance of BCMA technology by number of medication and patient identification scans.⁷⁰ Complex variables surrounding nurses’ workflow have been measured using conceptual frameworks derived from the human factors engineering domain, such as recognition-primed decisionmaking (RPD), human-automation interaction, workload, authority-responsibility double-binds, and mutual awareness among members of the clinical team.⁷¹

Use of BCMA in the outpatient setting

Only one study described the use of BCMA in the outpatient pharmacy setting.⁷² This was a small feasibility study evaluating whether BCMA could be used for automatic verification of medications during dispensing. The study suggested that manual checks could be replaced by barcode technology to improve pharmacist productivity and increase cost savings; however, empirical evidence supporting these conclusions was absent.

Beyond medication error rates

Unlike CPOE systems, there is a fair amount of uniformity when describing a BCMA system. Also, contrary to CPOE interventions—which were conducted largely at urban, academic institutions—BCMA studies are primarily conducted in VA settings. However, the slow penetration of this technology has resulted in very few evaluation studies. In general, there is a paucity of empirical evidence supporting the implementation of BCMA systems. The BCMA technology is advocated as an important safeguard for reducing ADEs, but sufficient evaluation of how this technology affects the dynamics of a complex hospital setting, in ways other than the reduction of medication error rate, is lacking.

Impact on nursing workflow

One of the efforts recently suggested was the replacement of the laptop computer with hand-held devices. A field evaluation of usability of this technique with nurses revealed that, while this might be useful for the administration of pain medication and hanging IV fluids, it was not ideal for use with medications in general.⁶⁶ Such determinations made from actual field studies serve two purposes. First, they enable nurses to be involved in the process of development and deployment, thus fostering ownership of the technology. Second, evaluation of the technology in a naturalistic setting can help us understand the far-reaching impact that the introduction of a new technology can have on the workflow patterns—and prevent any new threats to safety that might be introduced by the technology itself.

Changing nurses’ perceptions about technology

Adoption of BCMA technology calls for a behavioral change. According to the Technology Acceptance Model,⁷⁵ such a change can be brought about by improving perceptions regarding the system, specifically perception of the ease of system use and its usability in routine practice. Tailoring interventions that are geared toward understanding nurses’ perceptions will promote adoption. System training should focus on educating nurses about how BCMA serves as a safety net and aids them in preventing errors. Also, organizational policies that support a transparent environment for the reporting of errors—rather than a culture of blame—can improve nurses’ perception of BCMA technology. As the nursing shortage increases in hospitals across the nation, nurses need to view safety checks, such as BCMA technology, as aids rather than impediments in their practice.

Enhancing interdisciplinary communication

A seamless integration between the CPOE and BCMA systems can enhance workflow and also build interdisciplinary communication. Information systems have the capability to serve as the common thread linking an interdisciplinary clinical team with the therapeutic decisions driving patient care. Expansion of BCMA into institutions will strengthen nurses’ relationship with other members on the clinical team and help other clinical staff to better appreciate nurses’ contribution to patient safety.

Medication error rate

A deeper examination is needed for the most commonly evaluated variable characterizing the medication error rate. No study reports an analysis regarding the type of ADEs, such as preventable and potential (often called near misses), that occurred while a BCMA system is in use. Such an evaluation would give us a deeper understanding of the ADEs that are being missed by the system and allow us to create modifications to better capture them.

Need for evaluation of economic outcomes

The policies of the FDA mandating barcoding of medications and the regulatory efforts of other patient safety organizations will, hopefully, encourage adoption of barcode technology. Research in this domain is limited and needs to be expanded to include examination of some core outcomes related to BCMA implementation. Quantitative estimations of return on investment following BCMA implementation and economic outcomes resulting from prevention of medication errors will expedite the adoption of this technology in more hospitals.

Outcomes such as reduced length of stay, decreased number of nursing full-time equivalents needed to perform medication administration, and decreased litigation following administration of incorrect medications are important economic considerations that hospital administrators evaluate when deciding in which technology to invest their health care technology dollars. These outcomes need examination with respect to BCMA technology.

Need for nurse involvement in BCMA implementation and design

As BCMA technology gets deployed in the medication administration process, it will have serious implications for nursing practice. Several issues surrounding the nursing workflow environment need to be examined when implementing a BCMA system. Issues ranging from the usability of the hardware and software to pragmatic issues, such as the ease of use of the barcode reader and availability of the portable computer, will determine nurses’ acceptance of BCMA in their patient care routine. This offers nurse researchers unique opportunities to provide leadership in the development and design of BCMA systems. Active collaboration of nurses with information technology personnel—to provide input on the display of alerts for urgent orders, reports of missing medications, or on recording the missed medications during a shift—can be invaluable to the deployment of BCMA systems.

Sociotechnical evaluation

Evaluation of BCMA technology from a sociotechnical perspective would help gain a deeper understanding of nurses’ use of BCMA systems. There is a paucity of literature measuring the sociotechnical issues of compliance with alerts, cognitive load, efficiency, productivity, and emotional aspects of using the technology.

Documentation discrepancies

Besides serving as a safety check for nurses in the medication administration process, BCMA systems also play a key role in creating electronic MARs. Discrepancies in the documentation process have also been evaluated. Such discrepancies arise when medications that have been dispensed by the pharmacy fail to be administered by nurses. The returned medications are not credited back into the patient’s billing account. Thus, discrepancies arise between what is dispensed, what is administered, and ultimately what the patient is billed for. Even though documentation discrepancies have been examined, there is no evaluation of the economic impacts of these discrepancies. Such discrepancies could lead to undesirable fiscal outcomes for the hospital, which might affect adoption of these systems. A systems analysis of how these discrepancies arise and what organizational policies can be put in place to inhibit them needs to be conducted.

The implementation of barcode technology can prevent the potential or near-miss errors that would not have been detected otherwise. Nurses must take an active and visible role in the development and deployment of BCMA technology. Participation is the key solution to implementing BCMA technology.

Direct decision support for nursing

Out of the 31 studies identified as relevant, 13 focused directly on nursing. Three studies concerned consultant systems for the prevention of pressure ulcers. All three were essentially qualitative or descriptive, presenting very little patient outcome data.^80–82 Decision support for the management of urinary incontinence has been studied as well. Petrucci and colleagues⁸³ found large increases in knowledge and decreases in episodes of urinary incontinence in a patient care unit where a consultation system was implemented, as compared to a unit in the same hospital where it had not been implemented. Three studies examined the performance of staff conducting telephone triage with the help of algorithmic decision-support systems. All found improved performance, although all three used the weakest design, a pretest, post-test evaluation.^84–86 Two other studies directly addressed alerts and reminders to nursing staff for preventive care. Both of these showed strong results, including one study that compared standing orders with alerts directed at nursing staff to alerts directed toward physicians.^{79, 87}

In another study, the effects of different forms of physiologic data displays in a neonatal ICU were examined in a CPOE environment.⁷⁸ In this study, all infants were randomly assigned during an 18-month period to one of four groups: (1) no display of trend data, (2) continuous display of trend data, (3) alternating 24-hour display of trend data starting in the first 24 hours, or (4) the same as the third group starting after the first 24 hours. The number of orders for colloid, blood gases, and ultrasound were measured, as were longer-range variables, such as total time on ventilation, total time on supplemental oxygen, length of stay, and death. No differences in patient outcomes were noted, although surveys found increased knowledge regarding neonatal physiology on the part of the staff. Because patients rather than clinicians were assigned randomly, it is highly likely that there was dispersion of the effects of the independent variable.

Finally, several studies on a clinical decision-support system (CDSS) that guides nurses in identifying patient preferences consistently found improvement in the degree that nurses were able to act in accordance with patient preferences. These studies employed a high-quality design, and although the program was never embodied in an EHR, it is conceivable that one day it might implemented with resultant, improvement in the continuity of care.^88–90

Indirect decision support for nursing

The remaining 18 studies were selected because they used a computerized intervention that either was instituted in a CPOE environment or used a well-established EMR. In addition, these studies focused in areas where nursing would likely be highly involved. One study⁷⁹ contrasted standing orders (to the nursing staff) versus computerized reminders for preventive care for inpatients, finding nearly twice the improvement with standing orders. Three studies focused on the use of guidelines to prevent deep venous thrombi in post-surgical patients, involving both nursing and physician activities. In two of the studies, significant results were found for provider compliance using the guidelines, but no difference was found regarding patient outcomes.^{91, 92} In the third study,⁹³ both compliance with guidelines and patient outcomes were improved.

Four other studies were conducted in critical care settings, involving mostly complex guidelines (e.g., ventilator support). In two other studies, the effects of a computerized guideline for the treatment of adult respiratory distress syndrome were the focus of the investigation. Both were conducted by investigators from the LDS hospital in Salt Lake City, which had an extensive EHR at the time, but not full-scale provider order entry. East and colleagues⁹⁴ examined the impact of the computerized guideline in a prospective multi-center randomized trial for 200 patients. No significant differences were found in survival or ICU length of stay between treatment groups. There was a significant reduction in morbidity as measured by a standard scoring system, as well as a lower incidence of over-distension lung injury. In a similar study, a pilot of the study published above at Memorial Hermann Shock Trauma ICU, McKinley and colleagues⁹⁵ randomized 67 trauma patients to either being cared for by the protocol or not. No difference was found between patients in terms of survival, length of stay, or morbidity.

In another study conducted in the outpatient setting in the VA, with a complete CPOE system, computerized guidelines for mental health screening resulted in significantly higher compliance than paper guidelines. It was not clear in the description how nurses (not including the advance practice nurses) were involved in the implementation, but they might have been the individuals actually receiving the alerts.⁹⁶

Several studies using qualitative techniques found similar issues. Karfonta⁹⁷ used grounded theory to examine the experiences of 23 nurses and 10 physicians using DSS systems in the ICU. Overall, all interviewees mentioned the role of DSS in assisting in forecasting the outcomes of decisions. Difficulties in learning the system, trusting the output, and understanding the technology were additional themes.⁹⁷ Lyons and colleagues⁹⁸ examined VA employees’ perceptions of guideline implementation and utilization in the VA’s CPOE system. Information technology issues were perceived as major barriers to effective guideline implementation. Patterson and colleagues⁹⁹ conducted a human factors qualitative analysis of the VHA clinical reminder system and noted that increased workload, training, time, and role divisions were key barriers to success.

The remaining studies all focused on preventive care reminders or hypertension followup, with only two focused mainly on nursing.^{87, 100} In most cases, but not all, the studies found improved compliance. Because the studies took place in the outpatient setting, the role of the nursing staff likely varied greatly, but is not elucidated. Further analysis would have to be done to determine if an increased role of nursing in providing followup and initiating immunization was a determinant contributor of success.

Challenges with research evidence

Three main themes can be extracted from the results. First, for the most part, nursing activity is simply not addressed in these studies. Nurses make decisions every day about pain and wound management, whether a patient’s symptoms are severe enough to notify a physician. Nurses often have to decide if a patient’s symptom is drug related, or if a drug might interact with other drugs before they give them. In many cases, nurses have primary responsibility for patient education and family support. Few decision-support interventions have been developed for any of these high-level, decision-based actions. It is as if nursing decisionmaking is invisible and nurses are viewed as data collectors, rather than decisionmakers. In addition, one of the main roles that nurses fill in an inpatient setting is that of an intermediary between the patient and other providers. This communication role is a crucial function in ensuring quality of care. However, communication has been significantly neglected in EHR designs, as was noted earlier in this chapter and by many other authors.^{101, 102} Most of the work to create DSS has focused on structured documentation, order review, or systems designed to force or track nursing actions. For example, one study examined the effectiveness of putting a signal on a nurse and tracking where they were at all times to ensure efficiency.¹⁰³ Another study examined the impact of opening locked medicine cabinets (in the room) only when medications were due, to ensure that nurses would give them on time.¹⁰⁴

Second, the mechanics of providing DSS for nursing in a regular CPOE inpatient setting has not been well explicated. In many settings, the computers are located at the nursing stations, making them unavailable at the time of care. The model of having decision-support software located on computers situated at the nursing station fails to support a nurse who is constantly on the move. Development and exploratory work has been published, examining the use of portable laptops or hand-held computers, but no high-quality studies have reported on an actual implementation directed at nursing.

Third, none of the studies have examined the mechanism of action for DSS interventions. This is true of the DSS literature in general. Because DSS interventions can range from alerting a clinician about something they already know (e.g., a reminder), to alerting the staff to where a patient is in a process (tracking), to providing new information that educates and informs, it is important to measure the intended psychological effect as well as the outcome.^{105, 106} Although most studies show significant increases in provider compliance, the effect is small, and the upper limits are in the low 40- or 50-percent levels.

Inclusionary terms

Online order entry OR computer-based physician workstation OR practitioner order entry OR computer-based medical record OR electronic health record OR computerized physician documentation OR computer medical records OR decision support computer program OR health maintenance reminder OR CDSS OR computer-aided OR computerized decision making support OR clinical decision support system OR computerized feedback OR computer-assisted dosing OR computer feedback OR predictive instrument OR computer-aided quality assurance OR computer alert OR clinician order entry OR provider order entry OR computerized reminder OR computer reminder OR computer-based monitoring system OR expert system OR computer-based medical decision support OR decision support system OR computer based order entry OR event reporting system OR electronic healthcare record OR electronic monitoring OR electronic health record OR electronic medical record OR electronic incident reporting OR electronic record OR electronic patient record OR electronic record keeping OR medical information system OR computer-predicted OR computer-based monitoring OR computer-based prompt system OR CPOE OR POE OR electronic journal OR medical reminders OR electronic reminders OR medical record alert.

Inclusionary Mesh® terms (for PubMed® only)

“Decision Support Systems, Clinical”[MeSH] OR “Hospital Information Systems”[MeSH] OR “Medical Records Systems, Computerized”[MeSH]

Exclusionary terms

NOT (X-ray OR biochemistry OR DNA OR RNA OR genome OR tomography OR dentistry OR dental OR simulation OR molecular OR animal OR psychiatric OR biofeedback OR HIPAA OR in-home OR data mining OR algorithm)

Exclusionary Mesh® terms (for PubMed® only)

NOT (“Validation Studies”[Publication Type] OR “Editorial”[Publication Type] OR “Letter”[Publication Type] OR “News”[Publication Type] OR “Comment”[Publication Type] OR “legislation and jurisprudence”[Subheading] OR “Libraries, Medical”[MeSH])

Limits

Articles considered were those published in English during the time period January 1, 1976 to August 1, 2004.