EHR Development In Healthcare PDF

78 Joshua Lee tems generated by healthcare entities to more commercially available products that define the “state-of-the-art” functionality. To some, the computer has become the equivalent of the stethoscope in healthcare: the ubiquitous, necessary tool that enables providers to embark on the...

78 Joshua Lee tems generated by healthcare entities to more commercially available products that define the “state-of-the-art” functionality. To some, the computer has become the equivalent of the stethoscope in healthcare: the ubiquitous, necessary tool that enables providers to embark on their diagnostic and thera- peutic journeys. Although it took a while for the stethoscope to reach a peak of adoption and universality, so too will there be a period of adjustment as the electronic health record (EHR) reaches maturity in the marketplace, in the quality of the product, and in the sci- ence of correct implementation. As of the writing of this chapter, no consensus exists about the appropriate “lexicon” of functionality of the EHR or about the ideal manner of its adoption.1,2 EHR implementation is a struggle facing all who specialize in healthcare infor- matics; it also should spur additional inquiry in this field from all healthcare practitioners invested in improving the current state of medicine in the United States and elsewhere. One of the key transformations in the EHR is that, in place of static presentation of data for review by healthcare practitioners or even a venue in which to order pharmacologic or diagnostic interventions, the record has also become a place of discourse between health- care providers. Whether it is messages from a phone call to a physician’s office about an appointment or a dialogue between nurse and pharmacist about an optimal drug dosage regimen, the EHR now provides a place of record for these communications and also places them in the correct context of the record. An even more exciting frontier is the area of patient–provider communications. Many homegrown and vendor-provided EHR products now provide secure Web-based portals for use by patients. These portals allow for secure messaging between patients and clini- cians in issues of symptom evaluation, medication queries, and review of results. They also allow patients to keep track of individualized immunization schedules as well as their age- appropriate screening interventions. 6.2 HISTORICAL DEVELOPMENT OF THE ELECTRONIC HEALTH RECORD The first electronic records were aggregations of observations laid down on paper, either through full narrative entries or scanned copies of notes. The initial term for these records was “electronic medical records” (EMRs) in that they dealt with the record of disease and interventions to cure them. However, as the repository came to include elements of health- care maintenance and preventative care items, it became a full “health record” spanning the continuum of care, hence becoming an “electronic health record,” or EHR. Elements in the following key areas made online records more advantageous3: Accessibility. By making records available through applications distributed on desk- tops across an enterprise linked to a central server or by using an Internet-based application, patient data became available at any time and nearly any location, with- out the need for medical records file rooms and file clerks, thereby diminishing the specter of a lost record. Legibility. The high degree of variability in provider handwriting has often been iden- tified as among the root causes of medication errors.4,5 The EHR circumvents this by  Hospital Information Systems 79 using a typewritten interface and by limiting the kinds of data that are allowable (see Chapter 4). These goals are accomplished by restricting the range of a numeric value to prevent “keying” errors and by making providers choose from a list of allowable values rather than entering variable text. Use of discrete data. By forcing the use of discrete data elements in representation of the elements of history, medical problems, medications, and even social history documentation, the record allows for aggregation of data across populations and the creation of association with other findings such as laboratory values or radiological findings. A good example of this is the correlation of certain disease states with the appropriate prescribing of classes of medications. Patients with left ventricular heart failure should be prescribed an ACE-inhibitor or angiotensin receptor blocker. If the presence of a left ventricular heart failure is detected but one of these medications is not on the active medication list, the patient’s physician can be prompted to prescribe one by sending an electronic alert to him or her. The first use of EMRs occurred in single institutions that sought to develop systems to support the “business enterprises” of hospitals—namely, the capture of physician orders and their appropriate routing to departments such as laboratory, pharmacy, and radiology and then to processing the associated fees that should be charged for those services. The side benefit of these systems is that they often provided clinicians with access to results (e.g., laboratory and radiology reports) electronically, allowing them to move beyond paper printouts. The manner by which these results were aggregated formed the basis of the first tenet of comprehensive electronic records: an integrated view of patient data across time and specialty. Individual systems often used proprietary formats to display laboratory and radiology values, which were very useful and tailored for their own systems but were not under- standable to a larger interface. Drawing on the experience of early programmers who leveraged the application programming interface (API) and standards initially created by the American Society for Testing and Materials (ASTM), developers learned to cre- ate universal messaging protocols to allow for interchange of these data among different computer systems. Thus, individual hospitals would not be required to do interface programming them- selves when they sought to integrate individual departmental systems to their core EMR. This standard, now called health level 7 (HL-7), references the highest level of integration of information and is the industry leader. It is in use at over 1,500 healthcare institutions in the United States.6 Initially, this allowed providers to see relevant clinical information for a particular inpatient stay. The standard has emerged to allow review of clinical information across time and across modality (see Figure 6.1). To represent data elements, it also became important to standardize not only the ways in which data were transmitted, but also the way in which each individual data item was rendered to achieve the vision of “discrete” data as identified before. Terminologies were created to control the representation of data, each for its own individual area: diagnoses, 80 Joshua Lee FIGURE 6.1 Review of results from many modalities, labs, and radiology across time. (Courtesy of Epic Systems, Madison, Wisconsin.) procedures, psychiatric diagnoses, clinical observations (e.g., laboratory values, vital signs), and medications. The review of all of these is beyond the scope of this chapter (see Chapter 4 for details), but it is helpful to outline the most common standards used to represent these elements in electronic health records. 6.2.1 Diagnoses Diagnoses are usually managed by the International Classification of Diseases and its Clinical Modifications (ICD-9-CM), which serves as the “lingua franca” of diagnostic terms in U.S. hospitals. It is used for clinical decision support and for billing support pur- poses. It is now out of sync with the rest of the world, which has moved to ICD version 10, which is slated to be implemented in the United States by 2013. To aggregate a group of related diagnoses, the concept of diagnosis-related groups (DRGs) was created; this allows for a smaller number of diagnostic groups to define a given hospitalization, facilitating reimbursement for similar care across hospitals. For example, there are many ICD-9 terms for bacterial pneumonia (e.g., 482.83: pneumonia secondary to Gram-negative organisms, and 482.31: pneumonia secondary to streptococcus). However, many of them are rolled up into larger groups to create a more rational basis for compen- sating hospitals (e.g., DRG 89: pneumonia with complications, and DRG 90: pneumonia without complications). 6.2.2 Utilization and Procedures The American Medical Association keeps a master dictionary of procedures (Current Procedural Terminology) that encompasses the universe of diagnostic and therapeutic proce- dures done by providers to patients.7 Although its use is almost exclusively in the reimburse- ment realm, it has uses among health services researchers to understand patterns of care.  Hospital Information Systems 81 Furthermore, it is often the way in which requests for procedures (i.e., a laboratory or radiol- ogy test) are “ordered” by the core electronic medical record to the recipient ancillary system. 6.2.3 Laboratory Findings and Observations Although it is not broadly deployed, researchers at the Regenstrief Institute in Indianapolis, Indiana, developed a system of structured data for laboratory findings and later for other observations (e.g., vital signs, electrocardiographic findings). This came to be known as the Logical Observation Identifier Names and Codes (LOINC) terminology.8 6.2.4 Nursing Terminologies To provide structure to nursing documentation at the bedside, a schema of problem (unique to direct bedside care) catalogs of the expected outcomes and the interventions to achieve those outcomes were developed. The most broadly deployed are the North American Nursing Diagnosis Association (NANDA), Nursing Outcomes Classification (NOC), and Nursing Interventions Classification (NIC).9 However, it is key to note that there is no correlation between the more classical “medical” diagnoses utilized in ICD-9-CM and these nursing diagnoses. Consequently, most inpatient records have two problem lists at any time: those identified by the physician providers and those laid out by the nursing professionals. As more integrated EHRs are implemented, these disparate problem lists will likely need to be harmonized for the sake of interdisciplinary care. 6.2.5 Drug Codes The challenge in creating structured terminology coding for medications is that the cur- rent standard from the U.S. Food and Drug Administration (FDA) is a dictionary enti- tled the National Drug Codes (NDCs) that is driven by the manufacturer and not unique to a specific drug, dose, and route. Rather, it is very much influenced by manufacturer and packaging. Although the National Library of Medicine (NLM) has sought to cre- ate a universal standard for transmitting medication information (RxNorm), to date it has not been widely adopted commercially. Most hospital systems rely on commercially prepared, proprietary drug databases with attached clinical decision support informa- tion (e.g., checking drug–allergy interaction and drug–drug interaction). The two most common providers in the United States are First Data Bank (San Bruno, California) and Medi-Span (Indianapolis, Indiana). 6.2.6 Implementation of EHRs Structured terminologies and the manner to link these elements between ancillary systems and the core EMRs could enable institutions and, eventually, vendors to tackle the issue of using EHRs not only to display data, but also to capture observations, notes, and charges and to provide real-time decision support to clinicians. Initial records that emerged from early hospital information systems were pioneered at several academic medical centers in the 1980s and 1990s. Among the most notable were the HELP system developed at the LDS Hospital (Salt Lake City, Utah) and the Regenstrief System developed at Wishard Memorial Hospital (Indianapolis, Indiana). These systems were the first to use the rendering of  Pharmacy Information Systems 97 Patient File Prescriber File Drug File Outpatient Pharmacy Information System Insurance and Dispensing Billing Automation e-Prescribing FIGURE 7.2 Outpatient pharmacy information system architecture. Outpatient pharmacy information systems can be integrated into hospital information systems or other points of care to allow transmission of prescriptions from the prescriber directly to the pharmacy. The majority of computer systems in community pharmacies are stand-alone systems designed to support the function of an individual pharmacy or chain of pharmacies. Many outpatient pharmacy information systems are commercially avail- able and selection of the most appropriate system can depend on whether the pharmacy is a chain or independent pharmacy. Over 50% of pharmacy chains in the United States use pharmacy software developed by their corporate office or by the PDX Pharmacy System.2 Chain pharmacy systems have devel- oped the ability to transfer and share prescription information electronically across the chain rather than relying on telephone transfers. Enabling patients to fill their prescriptions at any location enhances convenience for patients. Electronic transfer of prescription information markedly streamlines this process for pharmacy chains. Smaller, independent pharmacies typically purchase a stand-alone system (e.g., QS/1 or McKesson Pharmacy Systems). The main components of an outpatient pharmacy information system are similar to those of the hospital systems, but functions specific to the outpatient pharmacy setting, such as insurance billing for prescription claims, are also supported (Figure 7.2). Generally, these computer systems consist of the following components: patient file; prescriber file; drug file; insurance information file and billing interface; e-prescribing interface; and dispensing automation interface. 98 Daniel T. Boggie, Jennifer J. Howard, and Armen I. Simonian The patient file consists of patient-specific demographic and clinical information, such as patient name, date of birth, sex, address and telephone number, medication allergy and drug reaction information, chronic conditions, insurance information, and a pre- scription profile. The prescriber file includes demographic data identifying the prescriber, including the provider name, address, and phone numbers and the Drug Enforcement Administration (DEA) and/or National Provider Identifier (NPI) number of the provider. The drug file contains all of the drugs that the pharmacy dispenses, including product names, dosage forms, routes of administration, strength, unit of measure, and procure- ment and inventory information. Drug database products can be purchased from companies such as First DataBank and Medispan and incorporated into the pharmacy information system to provide descriptive drug information, unique identifiers, and pricing information, as well as clinical decision support. The insurance file contains a list of all supported insurance companies as well as a billing interface for real-time adjudication of prescription claims. Finally, automated dis- pensing equipment interfaces may send completed prescription data to a robotic dispenser for filling. Outpatient pharmacy information systems may be similar in structure to hospital sys- tems, but the process used to verify and fill prescriptions is quite different from start to finish. In the hospital setting, medication orders are entered by the provider into the com- puter system via CPOE. In contrast, prescriptions in the outpatient setting are most often handwritten by a provider and carried by the patient to the pharmacy. New prescriptions can also be faxed to the pharmacy or called in by the provider’s office. Handwritten, faxed, or telephoned prescriptions must be transcribed into the pharmacy information system by the pharmacy staff once the prescription is received. The pharmacy staff is required to verify all pertinent patient information, provider information, and insurance and benefit information and select the appropriate drug. E-prescribing is the process by which a provider can electronically send an accurate, error-free prescription directly to the pharmacy information system from the point of care. E-prescribing greatly enhances patient safety and in the future will be the standard for transmission of new prescriptions from the prescriber to the pharmacy. These prescrip- tions are automatically transcribed into the pharmacy information system and displayed for pharmacist processing. The prescription is ready for processing by the pharmacist once it is entered into the pharmacy information system. It is during this step that allergy, drug–drug interac- tion, and drug–disease interaction checking automatically occurs. With CDS and order- checking processes, the pharmacy information system can determine whether the dose of the drug is appropriate for the patient’s weight, age, or renal function. The pharmacy information system can also screen for potential problems with prescribed therapy, including duplicate therapies, drug–drug, drug–food, and drug–disease interactions, and notify the pharmacist. Unlike hospital pharmacy information systems, real-time online insurance claim pro- cessing and adjudication occur during outpatient prescription processing (see Figure 7.3). If the patient has prescription drug coverage on file, the pharmacy information system  Pharmacy Information Systems 99 Rx Processed Prescription Claim Transmission Billing Interface Payer Pharmacy Computer Adjudicated System Response Adjudicated Response Received FIGURE 7.3 Online insurance claim processing. will send the prescription claim information to the insurance company’s central computer system through a billing interface. The prescription is then checked against stored insur- ance coverage data to determine whether it can be approved for payment. The pharmacist is immediately notified if the prescription is not covered by the patient’s insurance, if it is too early to refill the prescription, if a prior authorization (PA) is needed, or if a drug utili- zation review (DUR) is required before the fill can be processed. Once the prescription has been processed by the pharmacist, a label can be printed and placed on the prescription. Alternatively, the prescription information can be sent to an automated drug-dispensing machine, which then fills and labels the vial. Automated dis- pensing machines are discussed later in this chapter. All prescriptions must be checked by a pharmacist before they are dispensed to the patient. This can be a manual process, or it can be facilitated by the pharmacy information system. Many pharmacy information systems print a bar code on the prescription label that is scanned at the point of pharmacist checking. The bar code scan brings up a digital image of the drug for the pharmacist to aid in verification of the correct medication. Pharmacies are required to keep documentation of patient counseling according to applicable state and federal laws. Most outpatient pharmacy information systems now have electronic signature pads to capture patient signatures for documentation of patient coun- seling. An electronic signature can be stored within the pharmacy information system for as long as required. In the outpatient pharmacy setting, the prescription is dispensed directly to the patient or his representative. This is most often accomplished by the pharmacy staff handing the prescription to the patient at the pharmacy window. Many pharmacy information systems have point of sale (POS) software incorporated into their systems that help manage inven- tory. Some systems have the ability to track the status of a prescription (e.g., whether the prescription is currently being filled or is sitting on the shelf). There is also automation for dispensing to patients when the pharmacy is closed. These pharmacy dispensing “kiosks” are discussed in the next section of this chapter. Once a prescription has been filled by a pharmacy, refill prescriptions can be entered into the pharmacy information system manually, received by an automated telephone sys- tem, or processed through a Web-based program that interfaces with the pharmacy infor- mation system. 100 Daniel T. Boggie, Jennifer J. Howard, and Armen I. Simonian 7.4 PHARMACY AUTOMATION Technology companies have developed a large number of products to automate portions of the medication use process. Software logic has been added to machinery to create automa- tion tools that help alleviate the large amount of manual labor associated with the prepa- ration, distribution, and administration of medications in both hospital and outpatient settings and, at the same time, promote safer medication practices. Inpatient technologies have centered on the unit-dose cart fill, IV batch fill, distribution to nursing care areas, and safe administration of drug products to the hospitalized patient. Outpatient devices have helped to automate the traditional tasks of counting, pouring, and labeling of prescriptions and the distribution of the finished product to the patient. 7.4.1 Inpatient Pharmacy Automation Before the concept of unit-dose dispensing became a standard, bulk packages of medica- tions were kept in locked medication rooms in nursing care areas. Often referred to as ward stock, individual doses were poured by nurses from the bulk bottle into cups from which these doses were administered to individual patients. With the advent of the 24-hour cart exchange process, a patient’s daily doses were added to a cassette and delivered to the care area at a given time each day. Medications were dispensed directly by the pharmacy and the pharmacy consequently had a need for individually packaged medications. Early automation included packaging machines with labeling and database software to create and document production lots of unit-dose tablets and capsules from the manufac- turer’s bulk bottles. The next step in automation was to interface the packaging machine to the pharmacy information system, enabling the machine to package and group all of the patient’s 24-hour supply of medications. These were then placed in the unit-dose cassette, ready for delivery after verification by a pharmacist. 7.4.1.1 Automated Dispensing Cabinets An alternate concept of drug distribution eventually emerged, and these new machines returned to the old idea of ward stock. In this form of automation, the ward stock was in unit-dose packaging and placed in discreet locations within the drawers of a locked cabi- net. The cabinet drawer locks were controlled by a computer and electromechanical system that was able to open drawers, indicate the location of the medication within the drawer, sense the return of the drawer to a closed position, and record the entire dispensing trans- action including a date and time stamp and identification of the practitioner (usually a nurse), patient, and medication dispensed. This technology is now known as the automated dispensing cabinet (ADC). Competing vendors have taken different approaches to the dispensing methodology of ADCs. One refinement has been to restrict access to a drawer full of medications by expos- ing only one drug at a time to the practitioner. This feature has been accomplished in vari- ous ways, such as individually covered, locked locations (or pockets) that open only for the selected drug when the drawer opens. Another method is the use of a rotating mechanism or sequential drawer that opens to only one medication or dose at a time. Finally, some  Pharmacy Information Systems 101 ADCs have mechanisms similar to those on vending machines, which allow the device to gather an individual, selected medication and drop it into a receptacle from which the practitioner can retrieve the dose. This method eliminates the need to pick the dose from a location in the drawer and avoids errors associated with the selection process. Medication dispensing safety has been enhanced with ADCs by the introduction of gatekeeper logic, which maintains a database of active orders for each patient and will only allow dispensing of medications for which valid orders exist. Though the ADC may con- tain the top 300 medications used in a particular area, the practitioner is only able to select medications for removal from a list of active orders for a given patient. With this gatekeeper logic activated, the ADC may provide a bypass mechanism to allow for dispensing in emergent situations or other special circumstances. The software may allow an exception or override list of medications that are always available for dispensing to any patient. In these cases, the practitioner may select a medication for which a valid order has not yet been recorded for the patient within the ADC database. Standard reports are generated to help the pharmacist follow up on all of these special dispensing events and verify that a valid order for the medication was recorded in the patient’s medical record. Additional medication safety enhancements include rules and alerts. Some ADC soft- ware can alert the practitioner in cases where the patient has a recorded allergy to the drug selected for dispensing. Rules and alerts can also be developed to warn the practitioner of possible adverse outcomes and ask for a response to document the justification for continu- ing with the dispensing event after a warning has been displayed. Specialized ADCs have been created for areas such as anesthesiology, where the ADC replaces a medication tray or cart in the operating room. The anesthesiologist is afforded quick access to medications and simple methods for documenting administration during the procedure. Software for these cabinets is geared toward multiple dispensing events on a single patient over an extended period of time in contrast to the typical “one patient–one drug” removal of medications by nurses using traditional hospital ADCs. 7.4.1.2 Smart Pumps When IV infusion pumps were first introduced, their main function was to provide infu- sion of piggyback, syringe, or large-volume medication at a specified rate. Eventually, soft- ware and interfaces were added to these devices to create what are now known as intelligent or smart pumps. These newer devices are able to store a database of standard IV prepara- tions, allowing the nurse to select from a list of predefined items when ready to administer an IV solution to a patient. Preparations listed in the database can be assigned default rates with CDS parameters that can warn of rates too low or too high. With both ADCs and smart pumps, detailed information is saved with each transac- tion. In addition to the recording of date, time, patient, practitioner, medication, and other basic data, the triggering of alerts and warnings, responses to override or bypass questions, and breaches of dosage limits may be recorded. These data can be used to evaluate medi- cation usage, monitor for drug diversion, investigate potential and actual adverse events, and help identify targets for quality improvement initiatives. One of the major criteria for 112 Ashley J. Dalton 8.4 BCMA MEDICATION MANAGEMENT FOCUS 8.4.1 Prescribing and Pharmacy System Order Entry The success or failure of a bar code scan at the bedside often relies on the way in which an order was input into the hospital or pharmacy information system. Depending on how an order is entered, the BCMA system will anticipate what types of data the bar code for that order should include. This could include an NDC or designated drug code that has been created for a customized drug record in the pharmacy information system. For example, a customized drug record may be needed for noncommercial drug items that are com- pounded by the pharmacy. These items do not have an NDC number and must be assigned a drug code to complete the drug record in the pharmacy information system. Another source of bar code data that the BCMA system may anticipate is one that has been forced as a recognizable entry in the pharmacy information system or computerized BCMA library. A forced entry can occur when the bar code data for a manufacturer bar code is unrecognizable to the system. Many BCMA systems are equipped with mapping tables that allow the user to map various bar code data streams to a specified drug record. Another source of bar code data may come from patient-specific bar codes generated by a pharmacy information system. Patient-specific bar codes are primarily used for com- bination products, such as compounded intravenous solutions containing more than one ingredient, and may contain both patient identification and medication order data. BCMA implementation requires that pharmacists have knowledge of how specific medications and preparations are correctly entered to ensure successful verification at the bedside. In addition, invalid medication order entries into the BCMA system that must be reentered correctly can cause cluttering of the eMAR. Order entry errors or duplicate medication orders can cause unintended entries on the eMAR that require an additional action by the user. For example, an order for a one-time dose of acetaminophen 650 mg orally is sent to the pharmacy. The pharmacist enters the order into the pharmacy information system for two 325 mg acetaminophen tablets to be given now. However, after processing the order, the pharmacist realizes that the patient has a nasogastric tube and the liquid formulation would be a more appropriate form of medica- tion for the patient. The pharmacist then reenters the order using the liquid formulation, resulting in two entries for acetaminophen on the eMAR. In this scenario, the pharma- cist and the nurse must communicate to ensure that two doses are not administered, and the nurse must also perform the separate action of documenting the invalid entry as not administered to the patient. Invalid entries on the eMAR can cause BCMA user bar code overrides, phone calls to the pharmacy, and a loss of faith in the BCMA system. All elements of order entry in the phar- macy information system and how they affect the eMAR must be carefully reviewed prior to BCMA implementation. It is imperative that pharmacists have an understanding of the phar- macy information system used by their organization and how it interfaces with BCMA. Another critical factor in BCMA that can have an impact on both nursing and pharmacy is the administration time of medications. One of the five rights of drug administration includes giving the medication at the right time. All scheduled medications in the BCMA  Bedside Bar Coding Technology and Implementation 113 system will have a specific time for administration. In the environment of the hospital, the time that the medication is specified to be administered may not be the time at which it can be administered. For example, one tablet of aspirin (325 mg) is due to be administered to a patient at 0900, but the nurse is unable to administer the medication because the patient is getting an x-ray and is not in the patient care unit. Although most healthcare institutions have standard times of medication adminis- tration, it is up to the institution to determine how to handle the retiming of medica- tions. These decisions may force the organization to determine definitions for what may be considered “late” or “early” administration. Most BCMA systems allow some latitude on the administration time of orders, prompting users with a warning if the administra- tion is outside the defined time parameters. Some systems employ indicators, such as color changes on the patient eMAR, signifying that a medication order is late and should be administered in a timely fashion. Depending on the pharmacy information system and its interaction with the BCMA system, the pharmacy may be responsible for the retiming of medication orders. Prior to implementation, a structured methodology regarding the appropriateness of medication order retiming should be implemented by the institution. This could include instructions on proper notification of the time change to the pharmacy or a way for nursing to “catch up” to the correct administration time without permanently changing the entire time schedule. 8.4.2 Dispensing Analysis of the dispensing of medications needs to encompass every activity involved with a medication from a “door-to-patient” perspective. This analysis can include drug procure- ment, repackaging, and relabeling. One of the biggest challenges to the pharmacy is decid- ing how to achieve the goal of placing a bar code on every medication that is intended for direct patient administration. A few methods can be used to achieve this goal. Medications can be procured from many different vendors. The pharmacy procurement focus changes after an organization begins to implement BCMA. It no longer includes only cost and what is inside the package, but now incorporates the packaging itself. Most wholesalers have been astutely monitor- ing the development of BCMA and, in anticipation, have developed categories within their purchasing catalog databases that include an indication of products available in unit-dose form. Although this does not give the user an indication whether the product has a read- able bar code or even a bar code at all, it is a helpful resource. Many manufactured products are available in unit-dose form. When looking at manu- facturer unit-dosed products, the pharmacy must take into consideration the cost differ- ence between bulk and unit-dosed medications, the increase in costs that the department is willing to accept, and the need for pharmacists and technicians to perform activities other than the preparation of unit-dosed bar coded (UDBC) products. In addition, the concept of packaging quality assurance must come into consideration. Pharmaceutical manufacturers and third-party repackaging entities must abide by standards set forth by regulatory agen- cies for good manufacturing practices. It is very difficult for a hospital pharmacy to have the same quality assurance practices as a manufacturer when medications are packaged. 114 Ashley J. Dalton In preparation for BCMA implementation, most pharmacies examine their shelves to estimate the amount of labor for manual preparation of UDBC products from products packaged in bulk. Many pharmacies are pleasantly surprised to find that 60–80% of the products on their shelves are available in unit-dose form. However, in a survey conducted by the Institute of Safe Medication Practices, 6% of respondents reported having diffi- culty obtaining unit-dose medications from manufacturers that had previously provided unit-dose products.9 These manufacturers had changed their packaging back to bulk form, forcing hospitals to unit-dose bar code the product themselves. In the same survey, 75% of respondents reported having difficulty with the packaging of received manufacturer UDBC products. The 2004 FDA bar code ruling does not require manufacturers to unit- dose products. In addition, if a manufacturer does choose to provide UDBC products, the ruling does not place strict requirements on bar code symbology or bar code content. Manufacturers use many different types of symbology and many of them place data other than the NDC in the content of the bar code. Additional data often include product lot and expiration date, but they can also be internal data relevant only to the manufacturer. The receiving of UDBC products from outside sources forces many pharmacies to intro- duce new quality assurance procedures into their product receiving processes. Typically, once a product order is received, the products can be placed on the pharmacy storage shelves or directly into automated dispensing cabinets. In an environment with BCMA, each product must be checked for a readable bar code at the unit-of-use level. This means that at least one of every UDBC product received must be scanned and compared to the hospital BCMA software to ensure that the bar code is correctly recognized. This quality assurance step can cause major pharmacy work-flow changes. The diagram in Figure 8.4 is one representation of a typical pharmacy receiving process. The boxes shaded in light gray represent pharmacy procedures without BCMA. Each box shaded in dark gray represents a new step in the product receiving process. Instead of or in addition to purchasing manufacturer UDBC products, institutions can choose to repackage and unit-dose bar code every medication internally. According to a survey by the American Society of Health-Systems Pharmacists, 93% of hospitals repack- age some oral medications into a unit-dose form.5 Methods for unit-dose bar coding oral medications range from almost completely manual to highly automated. Manual processes usually involve a pharmacy technician placing tablets, capsules, or specified quantities of pharmaceutical liquids into unit-dose cups or bags. Labels or stickers are then generated using UDBC software. At a minimum, the labels usually contain the medication name, strength, dosage form, lot number, beyond-use date, and a usable bar code. Highly automated methods can include the use of high-speed packagers that can pro- duce UDBC medications in large quantities in a very short period of time. Some packaging machines can accommodate large canisters for storage of 100–500 different oral solids and have packaging speeds of 20–60 doses per minute. Many high-speed packagers have the capability to store medications in canisters and use the machine’s bar coding technology; this allows the user to scan the bulk product and a bar code on the canister to verify correct placement of the oral solid. In either style of repackaging, the data encoded in the bar code  Bedside Bar Coding Technology and Implementation 115 Order Received & Divided Manufacturer UDBC Available UDBC Unavailable- Items Repackaging Bar code Quality Assurance Solids Liquids Pharmacy Scans Item for Correct Drug, Dose, & NDC Item not Unreadable Scan Accurate Recognized by or BCMA System Inaccurate Bar code Pharmacist to Force Recognition of Bar code Double Check of Bar code Recognition by Scanning in BCMA System Tested Product Accepted into Inventory or Automated Dispensing Machine FIGURE 8.4 Pharmacy product receiving for BCMA. can be configured based upon UDBC software’s capabilities and the institution’s choice of BCMA software. Alternatively, all UDBC responsibilities could be outsourced to a repackaging company. Yet another option for achieving 100% bar coding for all medications is for an institution to purchase as much manufacturer UDBC product as possible, repackaging or relabeling products only when absolutely necessary. Each of the methods described previously has advantages and disadvantages that must be matched to the institutional needs. Most phar- macies choose a combination of the three methods. An additional challenge presented with BCMA occurs when the pharmacy compounds a product with multiple ingredients. This is most common with intravenous (IV) admix- tures and formulations compounded for infants and children. The challenge is to decide what data should be included on the bar code of the compounded product. Many admix- tures contain several components in varying amounts (e.g., parenteral nutrition) that could not all be represented in a single bar code. 116 Ashley J. Dalton To mitigate this challenge, many pharmacy information systems will assign a number to compounded products. This is similar to a prescription or order number. The system will generate a patient-specific bar code to be attached to the finished product, which can then be scanned against the patient record for verification. Alternatively, the pharmacy infor- mation system may encode both a patient identifier and an assigned order number into the labeling for patient-specific compounded items. This method also allows the compounded product to be scanned against the patient’s record for correct product verification. Regardless of how the pharmacy chooses to place a bar code on medications, it is impor- tant for it to strive to achieve 100% bar coding on all medications. The number of medica- tions bar coded may have an effect on nursing compliance in scanning medications at the time of administration. A Dutch study found that when not all medications for a patient were bar coded, the nurse was more likely to enter all of the medications into the BCMA system manually.10 This study highlighted the importance of having a readable bar code placed on all medications. 8.4.3 Drug Administration Nurses are often recognized as the last line of defense in medication error prevention. However, although 38% of medication errors occur at the point of administration, one study found that only 2% of these errors were caught.2 BCMA implementation has demonstrated success in increasing prevention of medication errors. In a post-BCMA implementation study in a large healthcare system (>500 beds), over one-third of nurse users acknowledged that they had avoided an error using their BCMA system.11 Although the error reduction potential has been acknowledged, the procedures and processes associated with BCMA impose tremendous changes on traditional nursing practice. Traditional medication administration involves many manual processes. Manual pro- cess can include nurse preparation of medications from bulk containers (e.g., preparing doses from bulk suspension bottles), hand-writing an entire MAR, having paper docu- ments at hand prior to medication administration, and manual documentation of medica- tion administration on the MAR. Prior to BCMA implementation, current nursing work flow should be documented in detail and compared with the anticipated BCMA integrated work flow. Nursing leaders can then focus on the largest areas of change and work toward solutions and staff education. Performing a careful work-flow analysis is essential in addressing nursing perceptions of the changes supported by BCMA and anticipating BCMA procedural “work-arounds.” One of the most often cited oppositions to BCMA is that it will increase the amount of time that nurses spend on medication administration activities. However, a time and motion study performed at a 735-bed tertiary care hospital found that, prior to BCMA implemen- tation, nurses spent approximately 26.9% of their time performing duties associated with medication administration. This compared with 24.9% of their time spent on medication administration after BCMA implementation—not a statistically significant change.12 Actual or perceived inefficiencies can lead to nurses’ deviations from the established BCMA procedures. These often include overriding bar codes on both mediations and patient ID bands, scanning medication bar codes away from the bedside, printing multiple patient  Bedside Bar Coding Technology and Implementation 117 ID bands to bypass scanning one physically attached to the patient, saving bar codes from the packaging of frequently used medications, and manually keying in data located in a bar code rather than physically scanning the bar code. A bar code “override” refers to the act of the user intentionally not scanning the bar code on the medication or patient ID and selecting the option of bypassing that safety check. Most BCMA systems allow overriding of bar codes, but require the user to input a reason for the occurrence (e.g., hardware failure, ripped bar code). Implementation teams strive for procedural excellence; however, care must be taken to prevent bar code overrides and work-arounds and to develop ways to combat them. Despite the potential for drastically reducing medication administration errors, improper use of a BCMA system negates any potential benefits. It is imperative that nurses support and are fully satisfied with the BCMA system selected by an organization. 8.4.4 Monitoring One of the most important steps in the medication use process is monitoring the effects of the medication. In other words, what was the outcome of administering the medica- tion? BCMA systems often have configurable monitoring components to facilitate or even prompt monitoring. These system settings can include forced or passive documentation. Examples of forced documentation include a requirement to obtain a patient’s blood sugar measurement prior to insulin administration or checking blood pressure prior to administering an antihypertensive. An example of passive documentation is capturing the volume of parenteral fluid given to a patient through the administration of multiple intravenous medications for inclusion in total patient fluid intake. Often, this recorded information will be translated for other parts of the electronic health record at the time of documentation for integrated or interfaced systems. This is beneficial for nurses because it can save time by eliminating redundant documentation. Monitoring components of a BCMA system can play an important role in a patient’s treatment plan. For example, in treating a patient having difficulties with pain management, most institutions require the nurse to ask the patient to rate the pain before giving a pain medication. In some BCMA systems, the nurse is able to document a patient’s pain score and, in some cases, respiratory rate at the same time as administering the analgesic. As part of monitoring the medication’s effects on the patient, a second pain score and respiratory rate may be taken after a predetermined amount of time has passed since the drug was given. This effectively enables providers to assess the safety and efficacy of a particular medication. BCMA and an eMAR are particularly useful to providers when medications that require pharmacokinetic monitoring are administered. At the moment a medication is adminis- tered, the exact time of administration can be captured by the BCMA system and recorded on the eMAR. Instead of manual documentation or searching through multiple paper MARs, the information needed to relate drug administration times with the subsequent serum drug concentrations can be accessed from a computer. 8.4.5 Education In many healthcare institutions, nurses are the first line of information for patients regard- ing their medications. In addition, many regulating bodies require that patients be told 138 Joseph E. Scherger and Grace M. Kuo in this setting have been conducted. The frequency of errors associated with medication administration is estimated to range between 6 and 20 errors per 100 doses.2 10.2.4 Medication Errors in the Home At home, medication errors occur when patients take the wrong drug or wrong dose or take the right drug and dose at the wrong time.25 In a study of 6,718 elderly home- care patients, 30% had potential medication errors when either the Beers criteria or the home health criteria were applied.26 The Beers criteria are widely used to “identify patterns of medication use that unnecessarily place older persons at risk of adverse drug reactions.”26 The home health criteria were developed to “identify home healthcare patients whose patterns of medication use and signs and symptoms provided sufficient evidence of risk of a clinically important adverse drug effect to warrant reassessment of the patient.”26 Errors also occur when patients do not take their medications. Medication noncom- pliance is estimated to cost $100 billion in the United States each year and is becoming a public health concern.27,28 The community pharmacist is in an ideal position to moni- tor medication use in the home and provide education to patients and families. A robust information system that helps pharmacists review patients’ medications and provides an interactive communication tool with patients’ physicians (e.g., http://MedActionPlan.com) can greatly facilitate this process.3,29 10.3 CAUSES OF MEDICATION ERRORS Medication errors are caused by many factors. To Err Is Human was so named because it is normal for human beings to make mistakes. The culture of healthcare has tradition- ally held the belief that physicians, nurses, and pharmacists are so well educated that they would not make mistakes if they were careful. This is an inappropriate and dangerous belief. Human beings, no matter how well educated, make mistakes 2–5% of the time, especially when doing repetitive tasks. The error rate increases when the professionals are tired or hurried.30 10.3.1 Root Cause Analysis Root cause analysis (RCA) is a formal method used to investigate an error.30–32 RCA is used by industry but has only recently been applied in healthcare settings. Performing an RCA involves an individual or a team analyzing all the steps leading to an error event. Three types of factors are analyzed: human, organizational, and technical. Human factors involving one or more persons causing the error can be uncovered through nonthreatening interviews. Everyone involved in the error is interviewed. Organizational factors include structural problems, such as an insufficient staff, inadequate supervision, or inadequate training for a task. Technical factors include having to work with malfunctioning equip- ment. Once all of these factors are analyzed in detail, a report is written about the RCA with recommendations for improvement. The final critical step with any RCA is to ensure that the recommendations for improvement are put in place and maintained.  Avoiding Medication Errors 139 10.4 TECHNOLOGIES THAT ENHANCE MEDICATION SAFETY Three emerging information technologies are being studied for their potential to reduce medication errors: computerized provider order entry, bar coding, and electronic prescrib- ing. These specific technologies may exist separately or as part of a common informatics platform, the electronic health record (EHR). Pharmacists play a key role in the selection, adoption, and successful application of these technologies in any setting (see Chapters 7 and 9). These technologies, decision support programs, and automated dispensing systems can reduce rates of medication errors.11,33 However, technologies can also cause errors, so diligence in their application is critical, and pharmacists play a major role in monitoring their use. Each technology is discussed separately next. 10.4.1 Computerized Order Entry Often called computerized provider order entry (CPOE), this technology requires the person placing an order for a patient to do so using an information system platform rather than hand-writing or verbally requesting an order. CPOE is one of the most studied new infor- mation technologies and its success in reducing medication errors has been mixed. On one hand, some CPOE systems have resulted in dramatic reductions in medication errors in both pediatric and adult settings.34–42 On the other hand, CPOE systems have resulted in medica- tion errors (e.g., by allowing providers to select the wrong medication or medication direc- tions from pull-down lists), showing that to err is not always human.39,43–46 Nevertheless, the influential Leapfrog Group considers CPOE an important quality indicator. Their 2008 sur- vey of hospitals found that only 8% of the institutions surveyed had implemented CPOE.47 Information technologies such as CPOE should not be put in place without healthcare professional oversight because the application must make sense in the clinical context and clinicians are vital in identifying misuses of the technology. Pharmacists are in an ideal position to help implement the recommendations and monitor the outcomes of CPOE and are able to identify and avoid potential points of errors. With their expertise in drug infor- mation, pharmacists are vital for achieving maximum medication safety.48 Think of this situation as similar to aviation safety, where the pilot and automatic pilot work together. 10.4.2 Electronic Prescribing To Err Is Human called for the elimination of handwritten prescriptions and other orders. This has not happened yet, mainly due to the high cost of infrastructure and equipment for implementing systemwide electronic prescribing; however, its implementation is growing rapidly. Electronic prescribing (also called eRx) requires the prescriber to use a computer system that is compliant with standardized eRx features that interface with pharmacy computerized systems. Through eRx, the patient’s name, the medication name, and dosage are far more likely to be correct. Using computer systems in the prescribing of medications offers great potential for implementing safe medication practices. When these systems are well designed and tar- geted to certain patient populations and clinical conditions, the safety results are impres- sive.49–54 Here, again, professional oversight, especially by a pharmacist, ensures that these 140 Joseph E. Scherger and Grace M. Kuo systems perform properly and are maintained and improved based on feedback and mea- sured clinical experience. 10.4.3 Clinical Decision Support with Safety Features As eRx, CPOE, and EHRs become more common, efforts are underway to improve the clinical decision support systems (CDSSs) built into these tools. The first CDSS application that has been present for a long time is drug alerting: pop-up messages that are automated into the prescribing process. Computerized decision support tools embedded within CPOE are effective in reducing hospital-based and outpatient medication errors.55,56 Informatics tools can detect and prevent medication errors and adverse drug events.57 High-quality clinical decision support software is crucial in reducing errors (see Chapter 15).58 Early CPOE systems failed to protect against many medication errors and sometimes caused harm (e.g., due to lack of safeguard features or allowing for free text typing that resulted in spelling errors). Newer systems have a better track record. Also, narrow appli- cations of CPOE, such as in pediatric or neonatal critical care, seem more effective than in more complex situations like general adult medicine, where the patient mix and therapies are more variable. Prescribing alerts can be highly effective when targeted to specific medications and clin- ical conditions. For example, calculation of dosages of critical medications such as digoxin in elderly patients has enhanced safety with electronic support. 59,60 Computer signals related to medication safety alerts in Medicare enrollees at multispecialty group practices identified 53% of potential incidents.61 In a prospective analysis after one hospital imple- mented CPOE, the medication error rate (excluding missed doses) fell by 81% and non- intercepted serious medication errors fell by 86%.34 Though CPOE is generally beneficial, system “noise” from nonstandardized alerts and signals may be annoying and not helpful in reducing errors; therefore, standardized alerts and electronic prescribing standards are still needed to improve medication safety. However, if the alert threshold is set too low or implementation is poor, many drug alerts will be inappropriate or unhelpful.19,62,63 Physicians’ and pharmacists’ almost universal experience with pop-up drug alerts has been that very few of the alerts are clinically relevant to the patient being treated. The alerts are often triggered at such a low threshold that almost meaningless information comes up. This results in “alert fatigue” in which the physician or pharmacist ignores or bypasses the alert without even reading it or giving it serious thought. Alert fatigue is dan- gerous because the occasional alert that is important is likely to be missed. Effective clini- cal decision support tools must present alerts that are clinically important most of the time and not burden the professional with meaningless messages.62,64 Clinical decision support tools are in their early stages and have the potential for many more applications that will improve both the safety and overall quality of medical practice. (See Chapter 15 for a more complete discussion.) 10.4.4 Bar Coding Bar coding is used in many industries to prevent human errors of recording and calcula- tion and to ensure proper identification. Increasingly, patients are being assigned a bar  Avoiding Medication Errors 141 coded wrist band when they are admitted to a hospital or other healthcare facility. All medication and other orders such as requisitions for laboratory tests and x-rays are also bar coded to ensure that the right patient receives the correct intervention. Bar coding reduces 60–80% of administration errors65–67; however, as with any technology, careful oversight is important to avoid the potential for bar coding to cause errors.68 (See Chapter 8 for a more complete discussion.) 10.4.5 Other Tools Other electronic tools, such as medication error reporting programs using computerized forms, allow healthcare professionals to document errors that have occurred in their prac- tice setting. The reported errors, including prescribing, dispensing, administering, and monitoring errors and the type of medications involved, can inform the healthcare orga- nization where errors occur and what areas need improvement.69 In addition, automated diagnostic and pharmacy data systems can be used for surveillance purposes to track ill- nesses and assess medication use (see Chapter 16).70 An innovative method that links an ambulatory care EHR and a Web-based patient portal has been implemented for patients with diabetes mellitus. This has allowed physi- cians to make medication dosage adjustments in a timely manner and encouraged patients to be more engaged with their care plans.71 Other disease management tools and patient registries have also been successfully implemented in the EHR to enhance chronic disease management and improve the quality of care.72,73 10.5 PHARMACISTS’ ROLES IN MEDICATION SAFETY The contributions of pharmacists in preventing medication-related problems for adult and pediatric patients in both inpatient and outpatient settings is well documented.3,6,29,74–79 For example, for patients seen in adult outpatient clinics, pharmacist identification of drug- related problems through medical record review has helped prevent adverse consequences.8 For patients being discharged from the hospital, pharmacist counseling has helped prevent adverse drug events after they leave the hospital.80 However, the current pharmacist short- age is having a negative impact on medication errors.81 The role of the pharmacist is expanding with the advancement of informatics technology and the increasing demand for electronic health record systems to continue efforts of medi- cation error prevention. Increasingly, pharmacists are needed to contribute their expertise in drug management through designing and implementing healthcare informatics tools. In addition to their skills in dispensing and drug monitoring, pharmacists can help develop informatics decision-making tools with medication safety features designed to prevent errors.82 Computerized tools, in turn, enable pharmacists to expand their consultation poten- tial and improve the quality of healthcare provided to patients.83 In one study, 78% of poten- tially harmful prescribing errors were intercepted by pediatric pharmacists using CPOE.84 10.6 SUMMARY This chapter has explored many of the informatics applications that can enhance medi- cation safety. Optimal healthcare systems are committed to having the best information 218 Pieter J. Helmons 15.1 INTRODUCTION This chapter discusses the impact of clinical decision support systems on medication errors. Therefore, it is important to understand the definitions of “adverse drug events” and “medication errors” before discussing clinical decision support systems. Adverse drug events (ADEs) are defined as any injury secondary to medication use.1 These events can be divided into nonpreventable, preventable, and potential ADEs: Nonpreventable ADEs (also known as adverse drug reactions [ADRs]) are inherently asso- ciated with medication therapy. An example of a nonpreventable drug event is an allergic reaction following administration of a drug to a patient with no known drug allergies. Preventable ADEs are those that cause injury to the patient that could have been prevented. Using the previous example, if an allergy to the drug was known, but was ignored and the administration of the drug resulted in an allergic reaction in the patient, this would be a preventable ADE. A potential ADE is an ADE that could have occurred as a result of an error, but (for- tunately) did not. In the preceding example, if the patient were allergic to the drug and received it, but no allergic reaction occurred, this would be a potential ADE. Medication errors are defined as any mistakes in ordering, transcribing, dispensing, administering, or monitoring of medication.1 This is a very broad definition and although potential and preventable ADEs are medication errors, not all medication errors are ADEs. Both medication errors and ADEs are common, costly, and cause clinically important problems.2,3 Each year, an estimated 770,000 people are injured or die in hospitals from ADEs. Approximately 28% of adverse drug events are the result of medication errors and are therefore preventable. More than half of these medication errors occur at the drug ordering stage and are the result of insufficient patient-specific information at the time of prescribing1,4 (see Chapter 10). This chapter starts with a case that illustrates how medication errors can result from the lack of patient-specific information. Next, the same case is presented, but this time the healthcare provider is supported by a clinical decision support system, resulting in an entirely different scenario and patient outcome. Although this alternative scenario lacks a specific pharmacist intervention, the crucial role pharmacists play in designing and main- taining these systems will be discussed later in this chapter. 15.1.1 Case before Clinical Decision Support5 Patient X is a 62-year-old woman with diabetes, hypertension, and borderline kidney fail- ure. She has been seeing her primary care physician, Dr. Smith, for the past 3 years and has generally been pleased with her care. She arrives at the office for a visit, checks in at the front desk, and then is ushered into an examination room. A few minutes later, Dr. Smith enters the room to see her. He is carrying her paper chart, and he flips through it as they discuss her current issues. After some discussion and a brief physical examination, Dr.  Clinical Decision Support Systems 219 Smith determines that patient X has a sinus infection. He glances at the medicines she is taking and his last written note about drug allergies, and then he hand-writes a prescrip- tion for an antibiotic. Patient X leaves the office with the written prescription and takes it to her pharmacy. The pharmacist enters the prescription into his computer system and then informs patient X that the antibiotic is not covered on her benefit plan. The pharmacist places a call to Dr. Smith’s office, resulting in the prescription of an alternative antibiotic. Patient X receives the antibiotic and instructions from the pharmacist about how to take the drug and then returns home. That evening she takes the first dose of the drug; an hour later, she develops severe vomiting. Patient X calls her doctor’s office to report the new problem. When the message reaches Dr. Smith, he considers that perhaps the drug was given in too high a dose given her age and kidney function. He lowers the dose of the anti- biotic and prescribes an antinausea medicine. The antinausea medicine eventually controls her vomiting but makes her very sleepy—so much so that when she gets up that evening to go to the bathroom, she stumbles and falls, breaking her hip. She is taken to the hospital by ambulance and undergoes surgery the next morning to have her hip stabilized with pins. 15.1.2 Case after Clinical Decision Support Patient X arrives for her office visit. The nurse brings her to the examination room and puts a preliminary diagnosis of “sinus infection” into the computer. Dr. Smith arrives to see her a few minutes later. After examining her and confirming the preliminary diagnosis, Dr. Smith clicks a button to reveal an evidence-based recommendation on the best antibiotic options for this condition. The computer returns a list of three antibiotic choices; next to each choice is an icon indicating whether that medication is covered on patient X’s plan. The first antibiotic is nonformulary, so Dr. Smith selects the second antibiotic. The computer checks the patient’s other active medications, and an alert window pops up indicating that the drug may interact with one of her diabetes drugs, resulting in vomiting. (In fact, it was this interaction, not the patient’s age or kidney function, that was responsible for her vomit- ing in the first scenario; in that scenario, the physician did not make this connection.) Dr. Smith contemplates giving patient X a reduced dosage of the drug and treating despite the risk of vomiting. To be sure, though, he clicks a button revealing her drug his- tory over the past 3 years. He notes that one of his partners gave a similar drug to her last year and the result was, indeed, severe nausea and vomiting. Armed with this highly rele- vant history, Dr. Smith cancels the drug order and selects the third antibiotic. No warnings appear this time, but the computer does recommend a reduced dosage based on her age and last measured kidney function, which Dr. Smith accepts. He confirms the prescription with a click, which directs the prescription to be electronically transmitted to the patient’s local pharmacy, and also prints a concise patient’s guide to the drug and its potential side effects. He reviews the prescription, dosage, and potential side effects with patient X and prepares to discharge her from the office. Before sending her home, however, he notes that the computer, which includes a full electronic health record as well as an electronic prescribing function, is recommending that the patient be placed on a cholesterol-lowering drug, based on her most recent choles- terol and LDL results and her diagnosis of diabetes; the system again shows which of the 220 Pieter J. Helmons applicable drugs is on the formulary of the patient’s plan. With two clicks, Dr. Smith pre- scribes this medication as well—again following the computer’s recommended adjustment for age and kidney function. The computer also recommends a follow-up blood test (cre- atine kinase) after 4 weeks of therapy because of the potential risk of muscle inflammation with this family of drugs. With one click, Dr. Smith orders this blood test and instructs the patient to return in 4 weeks to get the test done. The rest of patient X’s course remains uneventful and she recovers rapidly from her sinus infection without further incident. 15.2 INTRODUCTION TO DATA, INFORMATION, KNOWLEDGE, AND DECISION SUPPORT 15.2.1 Definitions In pharmacy informatics, the words data(base) and knowledge (base) are often used. To better understand the definition and function of decision support systems, it is essential to understand the difference between these terms (Figure 15.1). A datum (the word “data” is plural) is defined as a single observation that characterizes a relationship; in other words, it is the value of a specific parameter for a specific object (e.g., a patient). 6 Knowledge is derived from the formal or informal analysis of data. As an example, if the result of a single measurement of a patient’s blood pressure is 180/110 mm Hg, this is considered a datum. An analysis of a large number of blood pressure measurements in a population leads to the reference values of normal, high, and low blood pressures. This analysis has now resulted in knowledge on patient blood pressure. A database is a collection of individual observations without any summarizing analy- sis. A computerized medication record is primarily a database; only data on the patient’s medication are stored. However, if (medical) knowledge is added to these systems (e.g., reference values of kidney function or knowledge of interactions between medications), the computer may apply this knowledge to aid in case-based problem solving. The system is then a knowledge-based system or decision support system. This brings us to the definition of a clinical decision support system (CDSS)7: “software that is designed to be a direct aid to clinical decision-making, in which the characteris- tics of an individual patient are matched to a computerized clinical knowledge base and patient-specific assessments or recommendations are then presented to the clinician or the patient for a decision.” These systems convert patient data essential for the clinician to make the right decisions into usable information at the time of decision making. Typically, a CDSS is based on the following elements (see Figure 15.1): The knowledge base translates scientific knowledge (e.g., guidelines, treatment protocols) into computer-interpretable decision algorithms (e.g., clinical rules or algorithms) The rules engine retrieves patient-specific data, often stored in multiple databases, and checks whether the criteria set in the knowledge base are met. Software allows the user to create clinical decision algorithms and generates recommendations.  Clinical Decision Support Systems 221 BCMA Laboratory Software DATA Guidelines CPOE Rules Engine Guidelines Knowledge Base EKG Infusion Data Monitoring Output to User FIGURE 15.1 Elements of a CDSS. Clinical guidelines (knowledge base) are translated to computer interpretable decision algorithms (clinical rules). The rules engine is then used to match patient- specific information to the parameters specified in the clinical rule (e.g., the current dose of a medi- cation is matched to the renal function of the patient). If dosage adjustment is warranted according to the criteria in the knowledge base, the user is notified. CPOE: computerized provider order entry; EKG: electrocardiogram; BCMA: bar-code-enabled medication administration. 15.2.2 Why Are Decision Support Systems Needed? The Institute of Medicine report, “Crossing the Quality Chasm,” has documented the gap between what healthcare providers know and what they do.8 The report identified three types of quality problems: overuse, underuse, and misuse. Misuse (errors) has been the predominant focus of attention, but it is likely that underuse or overuse of practices and resources results in a larger portion of current quality problems.9 Surveys of clinicians indicate that a major barrier to using current research evidence is the time, effort, and skills needed to access the right information among the massive volumes of research.10 Each year, the National Library of Medicine indexes over 560,000 new scientific articles in the MEDLINE database. In addition, 20,000 new randomized trials are added to the Cochrane Library.9 This corresponds to 1,500 articles and 55 new trials per day! Even if the clinician is aware of the evidence, he or she needs to agree, adopt, and adhere to this evidence. For example, in one study, 90% of the clinicians were