A Literature Review on the Safety of CPOE Systems and Design Recommendations

ARRA has allocated $20 billion to stimulate the adoption of health IT over the next several years. The largest share of the funding will be for incentive payments through CMS to encourage providers and hospitals to implement health IT systems that demonstrate meaningful use that includes the implementation of CPOE systems. Unfortunately, few tools exist to provide informatics specialists best practice configuration guidelines when designing or evaluating CPOE systems to reduce medication errors. This article includes a consolidated source of the current literature on the safety of CPOE systems and offers design recommendations based on the literature. A 46-item CPOE design checklist is provided as a tool that can be used during software selection, design or evaluation. Utilization of a checklist such as this, coupled with a strong evaluation program, can provide organizations with the best chance of reaching their error reduction goals. FoCus Quality outComes and Patient saFety 26 jhim n FAll 2010 n volume 24 / number 4 www.himss.org evidence (and no guarantee) that they reduce errors—a primary marketing slogan. As these systems have emerged and have become more sophisticated over the last decade, several studies have been conducted that indicate a reduction in medication errors. On the other hand, there recently have been studies providing evidence that these systems can actually increase errors, which are attributed to poorly designed CPOE implementation and system configuration. Both types of studies provide valuable information to organizations as they undertake CPOE implementation and its evaluation. A significant problem appears to be that there is no consolidated source of evidence that provides organizations with tools to design the safest CPOE system possible. The purpose of this article is twofold: To offer a compilation of the CPOE literature as it relates to the reduction of medication errors; and to offer a CPOE design checklist based on the literature that can be used by all levels of informatics specialists to promote error reduction. What the literature SayS For almost a decade, The Institute of Medicine’s To Err is Human has been a driving force for improvements in patient safety nationwide. The research quoted clearly indicates the seriousness of the situation in terms of lives lost and money wasted as a result of medication errors.4 Solutions to this problem have been offered by many authors, organizations and politicians, but the resounding theme is the use of technology and the application of CPOE systems. Searching peer-reviewed journals for studies that identify specific CPOE components that lead to the best chance of medication error reduction is challenging. Most articles using the search terms of “hospital information systems or CPOE and medication errors” were quasi-experimental studies that either supported CPOE with an outcome demonstrating a reduction in medication errors, or with an outcome demonstrating an increase in medication errors. Both types of studies are reviewed here with the belief that each offers a perspective on CPOE system configuration that has the potential to add value to a growing body of knowledge. evidence SuPPorting cPoe Multiple studies have concluded that CPOE systems reduce medication errors. In a systematic review by Kaushal, Shojania and Bates, a total of 12 studies were reviewed. Included were two nonrandomized controlled trials, one randomized controlled trial, three observational studies with controls and four observational studies without controls.5 Five of the 12 studies evaluated CPOE systems with clinical decision support (CDS), while the remaining seven were studies evaluating a CDS system without CPOE (primarily, antibiotic management systems). Overall findings reported that CPOE and CDS systems can reduce medication error rates, but identified that more research is needed to evaluate commercial systems given the majority of systems in this review were built in house. The configuration included in the CPOE group with CDS was considered basic, i.e. drug-allergy checking and drug-drug interaction checking. In the studies that involved CDS systems only, configuration included antibiotic drug advice and dosing programs to assist the physician with appropriate dose ordering. Two additional systematic reviews reported similar findings, yet make the appeal for higher quality in study methodology and reporting.6,7 A study by Franklin, O’Grady, Donyai, Jacklin and Barber found a significantly positive impact on error reduction post implementation of a closed-loop electronic prescribing and administration system on prescribing errors.8 This study included bar coded administration of medications, but did not provide details regarding the configuration of the CPOE system that supports the bar code technology. Shulman, Singer, Goldstone and Bellingan collected data on medication errors from a CPOE system and compared it to traditional hand-written medication orders.9 They found a significant reduction in errors with the use of CPOE, even without clinical decision support. Components of the CPOE configured in this study included a full audit trail, mandatory data fields such as dose, units, frequency and signatures. They also reported that dose errors were still prevalent with CPOE as a result of physicians choosing an incorrect drug template or selecting from multiple options in a drop down list. One prospective cohort study in a pediatric ICU found a significant reduction in prescribing errors after the implementation of a CPOE system.10 This homegrown system contained several types of clinical decision support including drug-allergy alerts, dose checking, drug interaction alerts, order sets and links to evidence-based literature sites such as the National Library of Medicine’s PubMed site. Continuing with studies that support the use of CPOE systems, Raebel et al., conducted a randomized trial including more than 11,000 women enrolled in Kaiser Permanente centers in the Denver area.11 They were attempting to determine if a computerized physician ordering system with alerts for drugs contraindicated during pregnancy could reduce the number of potentially harmful situations for mother and fetus. Their findings were statistically significant in favor of the new system but it did not completely stop inappropriate ordering. The recommendation was made to couple data from the information system with the knowledge and skill of physicians and pharmacists. Kim, et al. evaluated safety of chemotherapy administration to pediatric patients before and after implementation of a CPOE system, looking at data that spanned over a three year course of time.12 They incorporated the process of Failure Mode and Effects Analysis (FEMA) into the screen design for chemotherapy ordering. This included limiting choices by pre-defined menus instead of free-text entry, mandatory entry of height and weight data, provision of alerts for abnormal values, enforcement of protocol specifications and automated calculations. Results indicated that after CPOE deployment, daily chemotherapy orders were less likely to have improper dosing. In another study by Vaidya, et al., the safety and efficiency of a CPOE system was evaluated for continuous medication infusions in a pediatric ICU13. This was another homegrown system that was designed for ordering continuous drug infusions for pediatric ICU patients. The study was conducted in a “simulated test environment” with scenarios to test the safe ordering of IV infusions. Configuration of this system included default standardized concentrations based on the patient’s daily height, weight and Quality outComes and Patient saFety www.himss.org volume 24 / number 4 n FAll 2010 n jhim 27 fluid intake. All calculations were performed automatically by the system. They found a statistically significant reduction in errors and concluded that this type of system will enhance user acceptance and rapid adoption by the physicians. evidence citing iSSueS With cPoe While the studies cited above support the benefits and enhanced safety features that can be realized with a CPOE system, there have been recent articles published that indicate a potential negative side of CPOE. A popular, controversial study that was published in the Journal of the American Medical Association in 2005 by Koppel, et al., reported that CPOE systems actually facilitate medication errors.14 This article ran contrary to what most believed, but identified, using quantitative and qualitative methods, 22 types of medication error risks associated with the use of CPOE. These risks were divided into two categories: information errors: fragmentation and systems integration failure; and human-machine interface flaws: machine rules that do not correspond to work organization or usual behaviors. While many of the 22 types of error risks involve workflow process, there were some that were attributed to poor system design including the inability to see all the patient’s current medications listed on one screen, the ease in which one can select the wrong patient due to a small font size and alphabetic sorting of patient names. This study clearly points out that human factors need to be considered when seeking to create a CPOE system to reduce errors. In one qualitative study by Campbell, Sittig, Ash, Guappone and Dykstra, researchers attempted to identify the types of unintended consequences seen with the implementation of CPOE systems.15 This study involved an expert panel using an iterative process that took a list of adverse consequences of CPOE, and sorted them into categories. The category labeled “Generation of New Kinds of Errors” included organizational culture and business process or workflow problems, but also indicated that new kinds of errors appear when CPOE is implemented. Examples of items in this category include juxtaposition errors when users select an item next to the intended choice; a wrong patient being selected; desensitization to alerts (alert overload); confusing order option presentations; and system design issues with poor data organization and display. Users get frustrated trying to find the right spot to enter a particular data element and end up entering orders on generic screens, bypassing any rules and alerts configured; all potentially leading to medication errors. This study and a companion study by Ash, Sittig, Poon, Guappone, Campbell and Dykstra, that investigated the extent and importance of unintended adverse consequences of CPOE, do not provide statistical analysis to indicate that these items are the true cause of medication errors; rather they provide information based on qualitative methods about categories of potential issues.15,16 In a retrospective study by Walsh, et al., researchers attempted to determine the frequency and types of pediatric medication errors attributable to design features of CPOE configuration.17 The rate of identified computer related errors was 10 errors per 1,000 patient days, and the rate of “serious” computer related errors was 3.6 errors per 1,000 patient days. The main types of errors identified included duplicate medication orders, drop down menu selection errors, incorrect abbreviations and order set errors (orders selected from the order set that were not clinically appropriate for the patient). No comparison was made between the data obtained and the previous state of errors. They concluded that serious pediatric computer-related errors were uncommon, but present and recommended the refinement of CPOE system design features. In another retrospective chart audit at a 658-bed academic hospital, 6,019 medication administration events were studied. This study utilized descriptive statistics and no significance was reported, yet the conclusion was that medication administrations do not consistently occur as ordered with an electronic CPOE system. Recommendations included the avoidance of duplicate medication orders, developing methods to alert nurses when medication orders have been entered by the physician and developing a safe means for nurses to shift medication dosing schedules.18 Additionally, Berger & Kichak, downplay the often quoted studies cited in the IOM report, To Err is Human.19 They report that the methodology of these studies has been challenged and cite the fact that they were conducted in the 1980s, making them not relevant in today’s more technologically advanced world. This article contends that there are relatively few controlled studies that provide evidence that these types of systems decrease serious adverse drug events and points out that medical facilities considering the implementation of CPOE should realize that the literature on the medical and economic effects are still evolving. evidence from exPertS or oPinion leaderS on cPoe Additional information surrounding CPOE and medication errors can be found in the literature in the form of expert advice and opinion. In an article by Kuperman, et al., the role of CDS combined with CPOE is recommended to reduce medication errors and improve patient safety.20 This article while informative and providing good background information does not provide evidence that any of the recommendations actually reduce medication errors. They suggest that clinical decision support should include the basic configuration of drug-allergy checking, basic dosing guidance, formulary decision support, duplicate therapy checking and drug-drug interaction checking, as well as advanced configuration of drug-pregnancy checking, drug-disease contraindication checking and guidance for medication–related laboratory testing. These concepts of advanced CDS configuration are echoed in an article by Bobb, et al., in addition to recommending pharmacist involvement in the medication ordering process.21 In 2001, 13 “experts” on CPOE from around the world gathered at a two-day conference to develop a consensus statement on successful CPOE implementation. This qualitative research approach was used to generate and validate a list of categories and considerations to guide CPOE implementations.22 In addition to workflow, training and technology considerations, a list of CDS configuration items was reported to improve patient care quality. These include: drug-drug interaction, drug-allergy interaction, duplicate medication checking, drug level monitoring, order sets and active surveillance in place. 28 jhim n FAll 2010 n volume 24 / number 4 www.himss.org Another CPOE opinion leader, The Leapfrog Group, represents a coalition of healthcare purchasers that has been a driving force in the improvement of healthcare quality and advocate use of CPOE systems. They have developed a CPOE “standard” including a requirement that organizations operating CPOE systems should demonstrate (via testing scenarios) to ensure that their inpatient CPOE system can alert physicians to at least 50 percent of common serious prescribing errors. In an article by Kilbridge, Welebob and Classen, the development of these standards are described including the description of a framework with 12 different categories of CPOE-based decision support that have the potential to prevent a prescribing error.23 Organizations are asked to conduct this test in a development or practice database that is similar to their production CPOE system. Configuration elements of the system that are tested include duplicate ordering, singleand cumulative-dose levels, allergy checking, drug-drug interaction, contraindications based on patient diagnosis, contraindications based on relevant laboratory values and dose levels based on radiology studies. This test is Web-based and self-administered, but can only be taken if the organization also participates in Leapfrog’s general hospital survey. evidence from the field of human comPuter interaction and cPoe Several articles have highlighted the concept of human computer interaction (HCI) or human factors design as a major consideration when looking at causes of errors.24 HCI is an applied science that is focused on minimizing human errors and optimizing performance where human beings interface with a device.25 Gainer, Pancheri and Zhang mention a “heuristic evaluation” as a technique to identify usability problems by identifying violations of well-established end user configuration that good design should follow.26 This symposium proceeding identified that an evaluation of a commercial CPOE system found numerous usability problems linked to user satisfaction and patient safety, although no study data were included. In a case study reported by Horsky, Kaufman and Patel, usability studies were performed using CPOE scenarios and found that there are commercial systems that introduce unnecessary cognitive complexity leading to errors.27 They purport that poor configuration can lead to placing heavy cognitive demands on users and they urge system designers to utilize guiding principles and design solutions when they are developing complex interactive systems in an effort to reduce errors. While the authors admit that what constitutes optimal configuration may be an elusive concept, they offer a conceptual methodology to evaluate configuration of the user interface that emphasizes clarity. critical revieW of data SourceS and methodS In these studies reviewed, the measurement of medication errors as they relate to the use of CPOE systems has been less than optimal; in fact, no two studies have used the same definition of what constitutes a medication error. Some studies collected data on actual adverse drug events, while others collected data on medication errors that may or may not have led to an adverse drug event. In a study by Gandhi, et al., medication errors were defined as, “Any error that occurred in the medication use process including ordering/prescribing, dispensing, adherence, and monitoring.”28 Shulman, et al. defined a medication error as having occurred when a prescribing decision or prescription writing process resulted in either an “unintentional significant reduction in the probability of treatment being timely and effective or an unintentional significant increase in the risk of harm when compared with generally accepted practice.”9 The definition provided by Potts, et al., stated that a medication error occurs when an order is found to be incomplete, incorrect or inappropriate at the time of physician ordering.10 The National Coordinating Council for Medication Error Reporting and Prevention, an industry leader in error prevention, provides the following definition, “A medication error is any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the healthcare professional, patient or consumer.”29 Each study and each professional organization reviewed had a different definition of what comprised a medication error. Instead of providing a global definition of a medication error, some studies provided definitions of types of medication errors. Koppel, et al., categorized errors as being one of two types as previously mentioned.14 Shulman, et al. categorized medication errors into 12 types including incorrect drug, drug needed but not given, dose error, incorrect route and frequency omitted.9 These inconsistencies and lack of standardization in measurement hold true for the concepts of medication error severity ranking and measurement of outcomes of adverse events related to medication errors. Many studies reviewed customized their definitions of medication error, type, severity, cause and outcome to reflect their particular organization’s current practices and study methodology. All used varying scales, scores and tables of measurement. No two studies appear to have used a similar methodology for data collection or measurement. It is now clear why there have been no meta-analysis studies conducted related to medication errors and CPOE. A few of the studies cite as a limitation to their work, the fact that there are no standards for measurements or a common taxonomy when conducting and reporting on medication errors and recommend further study. Most studies clearly indicate that their results are not generalizable. Many of the studies simply reported that they compared a “CPOE system” to a handwritten system, omitting the details regarding the specifics of CPOE configuration. Many included the fact that they used various components of clinical decision support, but then failed to mention how they were configured. The fact that a system is reported to have drug-allergy checking configured does not indicate whether or not the alert will hard stop a user when attempting to order a medication that the patient has a reported anaphylactic reaction. The fact that a system is reported to have duplicate order checking configured does not indicate whether the alert will fire if the same medication is ordered within a two-hour window, a 12-hour window or a 24-hour window. A typical description of an organization’s CDS components reported in studies reviewed is found in an article by Cordero, Kuehn, Kumar and Mekhjian that states, “Numerous clinical www.himss.org volume 24 / number 4 n FAll 2010 n jhim 29 Category Description/Comments Source CLINICAL DECISION SUPPORT Allergies and cross allergies. Display alert when an allergy has been documented or an allergy to another drug in same category is documented. Provide alert of potential allergy at time of order entry, not order submission. [14, 23, 29, 31, 32, 33] Duplicate order checking. Display alert when the same medication is ordered and when separate doses of the same medication are to be given within a “closely spaced time”. [17, 18, 21, 23, 30, 33, 35, 36, 37] Single and cumulative dose limits. Display alert when ordered dose exceeds recommended dose ranges for both single and cumulative doses. [14, 23, 30, 36] Contraindicated route of administration. Display alert when order specifying a route of administration that is not appropriate for the ordered medication. (e.g. Antifungal topical cream ordered with route of IV) [23, 30. 36] Drug-drug interaction. Display alert when there is a potentially dangerous interaction when the medication is administered with another medication. [23, 30, 32, 38] Drug-food interaction. Display alert when there is a potentially dangerous interaction when the medication is administered with a particular food group. [23, 32, 34, 38] Contraindication/drug based on patient diagnosis. Display alert when drug is contraindicated based on the patient’s diagnosis. [21, 23, 32, 33, 34] Contraindication/dose limits based on patient diagnosis. Display alert when dose ordered is a contraindication based on the patient’s diagnosis or the diagnosis affects recommended dosing. [21, 23, 32, 33, 34] Contraindications/dose limits -based on patient weight. Display alert for over dosing or under dosing based on patient weight. [10, 17, 23, 30, 33] Inaccurate weight or height data entry (used to calculate medication dosages). Display alert that warns clinician that patient weight or height just entered is 10% < or > the previous value entered. Clinicians can accidently add or delete a digit when entering a new patient weight (or enter lbs instead of kilos) and then is subsequently used for drug calculations. [39] Contraindication /dose limits -based on laboratory studies. Display alert when a drug is ordered for which laboratory studies must be considered prior to dosing. [12, 23, 32, 33, 34, 37, 40] Contraindication/dose limits -based on radiology studies. Display alert when drug is ordered which may have a potential interaction with contrast medium (in ordered radiology study). Ex, Metphormen [23, 33] Potassium order appropriateness checking. Display alert informing users ordering potassium when there has not been a serum potassium value recorded in the past 12 hours or if the most recent potassium value is greater than 4.0.This would reduce the likelihood of ordering potassium when the patient is hyperkalemic. [35] Provide automated mechanism to calculate dosages. Do not rely on manual calculation of dosages. [33] Create an alert informing users ordering potassium when there has not been a serum potassium value recorded in the past 12 hours or if the most recent potassium value is greater than 4.0. This would reduce the likelihood of ordering potassium when the patient is hyperkalemic. [35] Alerts should clearly indicate the problem and offer a solution. Error messages should be expressed in plain language (no codes), precisely indicate the problem, and constructively suggest a solution. Poor conceptual graphical representation of alerts makes it difficult for CPOE users to find certain information leading to inefficient searches for information provided by the system. [41, 42] Alerts from Clinical Decision Support should fire at the time of potential error and allow the user to recover without the loss of data entered. Loss of orders entered prior to alert require users to start over leading to frustration and potential loss and inaccuracy during repeat order entry. [42] ORDER FORM CONFIGURATION Mandatory fields for drug, dose, frequency and route. Cannot submit order without completing these fields. [12, 32, 37, 43] All ordering screens should be designed in a similar fashion. Fields for drug, dose, route, frequency, etc should be in the same place on all screens. There should be consistency of screen control, behavior, clarity of meaning and data labels for fields. Order entry screens for IV bolus injections should be configured similar to those for IV fluids with medication additives. [35, 41] Do not use field labels that require a negative answer for a positive response (e.g. Is IV contrast contraindicated?). This is confusing. Ensure clear understanding of all data field labels. [39] Evaluate field labels for potential user-designer mismatch. Ensure consistent meaning of data labels and make changes if not clear (e.g. Total Volume – of what?, Date – for what?, etc). [35] All data elements needed for an order should have an entry field that is easy to find and screens should be consistent and easy to navigate. If a busy clinician does not immediately see where to enter ordering data, they may enter information into the comments or miscellaneous section. This creates problems and potential errors in categorizing, cross checking and applying clinical decision support. [15] Include display of routine or default times for orders written to be administered q4h, q6h, q8h, etc. If not clearly displayed, providers may not be aware when first dose of a medication will be administered and may inappropriately order a STAT or Now dose. [18, 39, 43] If a drug dosage is calculated by the system, display the calculation used to derive the dose. Although automated calculations can assist users in deciding upon a drug dose, without showing the calculation used clinicians are sometimes prone to recalculate manually to check the computers work. [42] Table 1: CPOE Design Checklist. CONTINUED ON PAGE 32 30 jhim n FAll 2010 n volume 24 / number 4 www.himss.org decision support tools are integrated into the ordering pathways. These include, but are not limited to, drug allergy, drug-drug interactions, order duplication, corollary orders, weight-based dosage, maximum dosage, and drug route restriction.”30 Another study described their configuration as the incorporation of CPOE with a moderate level of CDS.30 Three systematic reviews that in total analyzed 50 CPOE studies all reflect on the heterogeneity of the methodologies and systems under study.5-7 With the “best” way to configure a CPOE system yet to be defined, it becomes clear that we are at a phase in the evolution of CPOE that still contains much trial and error work at the organizational level. It becomes clear that tools are needed based on what has been learned and documented in the literature; tools that can be used to evaluate CPOE systems and provide guidance to informatics specialists as the iterative process of improvement continues to cycle. The importance of this becomes paramount as governmental incentives require organizations to adopt CPOE systems. the cPoe deSign checKliSt As hospitals and care provider groups move forward in their journey to adopt electronic health records with CPOE, they will need resources and tools to ensure that systems are created to meet meaningful use criteria. Simply installing a vendor’s version of a CPOE system or building one in-house requires knowledge of best practices and evidence from the literature that support the configuration of a quality system. “Meaningful CPOE configuration” needs to be added to the list of required criteria along with evidence of an ongoing evaluation plan. The checklist offered in this article, has pulled quality evidence together in one tool. It is intended to be an instrument for informatics specialists at all levels that can be used to conduct an assessment of a CPOE system’s configuration from an error reduction standpoint. It is easy to use and based on available evidence. The tool is called the CPOE design checklist and can be used by organizations involved in the initial selection or design of their system as well as by organizations who wish to evaluate their current system. The checklist is organized into functional sections that include: clinical decision support, order form configuration, human computer interaction factors and workflow configuration. Each item on the checklist is listed with a description and includes its source or reference from the literature (quantitative studies, qualitative/observational studies and documentation from opinion leaders and healthcare informatics governing bodies). A few of the items on the list are CPOE design configurations that have been applied at the author’s organization, the National Institutes of Health, Clinical Center in Bethesda, MD, as a result of lessons learned and corrections made. (See Table 1.) The CPOE design checklist contains 46 items. Sixteen items are categorized as clinical decision support, such as allergy checking, duplicate order checking and single/cumulative dose checking. Each configuration item is described in the checklist, yet it is realized that granular details on each one is not included, nor is it included in most of the CPOE literature. Each CDS item can be configured in any number of ways typically based on a need identified by the organization. None of the studies provided details on how a specific CDS component was configured making it difficult to do anything other than add it as a general CDS component on the checklist. Numerous evaluation opportunities exist to determine the effectiveness of specific components of CDS along with their unique configuration. There are 22 items on the checklist in the category of order form configuration including mandatory fields for drug, dose, frequency and route; inclusion of relevant lab values; consistency of form design; displaying of calculations; and use of tall man lettering. This category provides more specific information and detailed configuration features than the CDS category. These items were typically included in the literature as a result of errors previously made and an electronic solution offered. Again, the evaluation of these components once implemented will ultimately answer the question—did it help? There are four items each in the categories of human computer interaction and workflow process configuration. One could probably make the case that all items on the checklist are related to HCI in one way or another, but the focus for the items listed are believed to highlight the interaction between human and machine. Items that are listed to support workflow include providing a way to alert caregivers to new orders and providing access to medication reference information that is convenient and does not disrupt workflow. The importance of supporting workflow to reduce errors is also seen as an area of further study as CPOE systems evolve. Reviewing limitations of the checklist, the literature review was in all probability not exhaustive. While the terms and subject headings used in the search were felt to be inclusive of any articles that pertained to CPOE and medication errors, articles were subsequently found that did not fall into these headings, but were found using reference lists from other articles or sysIn light of the fact that significant ARRA funding is being allocated for the adoption of CPOE systems, it is imperative that informatics specialists at all levels, from novice to expert, have quality tools to implement and evaluate these systems. www.himss.org volume 24 / number 4 n FAll 2010 n jhim 31 Category Description/Comments Source Provide the ability to search from medication lists which use “Tall Man” letters. Tall Man letters help to differentiate similar looking medications (NIFEdipine vs. niCARdipine or predniSONE vs. prednisoLONE). [33] Only allow clinicians to order the maximum “split size” dose allowed by hospital policy, as multiple individual doses. Example – if an 80 mEq dose of potassium were ordered, pharmacy would need to split it into four 20 mEq dose IV bags. The system should only allow users to order one bag of 20 mEqs per order. [18] For multiple potassium chloride infusions, provide a graphical display of the patient’s serum potassium and creatinine values of the last week. Could also be shown a list of those medications that the patient is taking that might affect potassium excretion (e.g. spironolactone or ACE inhibitors). Consider the system requiring ordering of stat serum potassium levels between each infusion. [18] Make “stop after” field mandatory. If forgotten, patient may receive more doses than required. [39] Provide flag or notification to provider when medication orders have been auto discontinued by the system. Providers may not be aware that antibiotics or narcotics were discontinued by the system at a predetermined time based on policy. [14, 36, 37] Only enter default values for drug, dose, frequency and route if will always be correct. If the default is correct 80% of the time, then chances are the other 20% will use the defaults incorrectly. House staff relies on the system to indicate usual doses. If dose listed is based on pharmacy warehousing or purchasing decisions, not clinical guidelines could cause problems. Ex – if usual dosages are 20 – 30 mg and the pharmacy only stocks 10 mg doses, the 10 mg dose should not be the default. [14, 16, 17, 37] Do not require clinicians to determine the base solution on order forms for IV infusions of medications. House staff is often unaware of inappropriate combinations and can inadvertently choose the wrong one. [14] Use of White Space—not too busy. Keep design simplistic. Too much information on one screen can be distracting and cause users to miss important information. [41] Do not make prescribers enter orders based on formulary. Allow them to put in actual dose. Medication should be put in as the total dose, for example: If pharmacy stocks 10mg tablets of Prednisone, order should be Prednisone 80mg p.o. and not Prednisone 10mg, give eight 10mg tabs. This is confusing for many people and can lead to wrong dosages. [39, 42] For IV infusions—allow ability to enter the order in dose per time instead of drops per time. If drops per time is required, then you have to know what the quantity of IV solution the medication is going to be mixed in then you have to manually calculate the drip rate. Confusing. [16] Pull in relevant information onto order forms and include the date and time of specimen collection if it’s a lab value. Should be the most recent lab value available displayed. Examples: Current ventilator settings, lab values, height/weight, Potassium levels when furosemide is ordered.The user should not have to remember information from one part of the system while working in another. [16, 17, 32, 35, 41, 42] Minimize free-text entry. Care providers should be able to choose most information needed in an order from a list of limited choices. Free text entry of any of the five medication order rights can lead to errors. Also, free text entry is typically not part of clinical decision support configuration and important alerts may not function properly. [12] Do not put similar terms on top of one another in drop down lists. If list items are close together and similar in name, there is an increased risk of clicking on the wrong item. [17] Avoid abbreviations in drop down lists. Having similar choices right on top of one another in a drop down list such as IV and IP for route can create errors. [17] Length of items in drop down list should be manageable. If greater than 12 items, then group the most commonly used items at the top.. Lists that are too long make it difficult to view all the choices at once and hard to use and require scrolling down to see other items. [42] HUMAN FACTORS CONFIGURATION Use alternate line colors between patients to help visual separation of names. Helps keep rows separated visually – to reduce wrong patient choice [14, 42] Do not put patient lists in alphabetical order. This places similar patient names next to each other making it easier to choose the wrong patient [14, 17, 42] Provide method to verify correct patient prior to order entry (e.g. display the selected name in large letters on the screen). Prior to submitting orders provide validation that the patient is the right one [15, 16] Ensure that patient name appears on all order entry screens. To reduce wrong patient choice [14] wORKFLOw PROCESS CONFIGURATION Provide way to alert caregivers to new orders. Without face to face interaction, nurses may not be aware that a new medication order has been entered. Provide flag or notification that new order is pending. [16, 43] Provide access to medication information that is convenient and does not disrupt workflow. Medication knowledge deficiency was noted in one study to contribute to over half of the prescribing errors noted that could have potentially been preventable. [21] Provide method to access view of all patient’s medications (including dose and frequency) on one screen. Screens that list active medication orders also should list IV drip orders [14, 35] Do not use test patients with “cute” names in the production system. They may be mistaken for real patients and orders entered will not be carried out on the real patient. [15] Develop and use unique and consistent naming convention for test patients. CCNIHTEST, TESTPATIENT is more unique than Tester, Joe [39] CONTINUED FROM PAGE 30 32 jhim n FAll 2010 n volume 24 / number 4 www.himss.org tematic reviews. An additional limitation to the checklist is the fact that it does not represent a comprehensive list of configuration items that can lead to the reduction of medication errors. There are certainly others that are not published. It is hoped that this is the first iteration of a tool that will evolve and improve over time as evaluative studies add to the list of items. The checklist is also not detailed enough to provide informatics specialists the ability to replicate each configuration item. This is a more systemic issue as the literature lacks the detail to include in the checklist. It is however a good starting point for informatics specialists to assess their CPOE systems and identify areas where improvements can result in the reduction of medication errors.