W H I T E    P A P E R   O N
A U T O M A T I O N    I N    P H A R M A C Y 

ACCURACY AND SAFETY

It may seem reasonable to evaluate automated components of the drug distribution system only in terms of the accuracy or quality of the output delivered by the automated component. However, such a compartmentalized approach to the evaluation of a complex human-machine system ignores the possibility of secondary effects, such as when the presence of automation makes the humans complacent in the performance of their own tasks.

In the review that follows, the reader should note whether the study evaluated the accuracy or error rate of the automated component of the drug distribution system, or medication administration errors as an outcome of the entire drug distribution process. The studies described here are limited, with some exceptions, to those using an observation technique to detect medication administration errors.

In general, a medication error is defined as a dose administered or prescription dispensed that deviates from the physician's order. Medication errors have long been counted and used as an outcome measure of the quality of the entire drug distribution system in hospitals, nursing facilities, and more recently, in ambulatory pharmacy settings.^6-8 Each error in medication administration or dispensing is viewed as an instance of a system failure, and it is appropriate to consider the impact of automated components of the drug distribution system on the medication error rate.

The first measure of the impact of automation on the quality of a drug distribution system was reported in 1969. A hospital re-engineering study resulted in the design of a patient profile-linked dispensing envelope system. Each unit dose arrived at the patient's bedside at the time due for administration in an individual envelope labeled with the physician's order. Data regarding medication errors were collected by direct observation during 192 eight-hour work shifts for the old system and 64 work shifts for the new (automated) system. The error rate declined from 13% to 1.9%.9 In 1975, Means and colleagues reported that introduction of a similar dispensing envelope system reduced the error rate from 7.35% to 1.61% (omission errors were not studied).¹⁰ Interest in such envelope systems has been rekindled today because they can be barcoded inexpensively.

In 1984, a controlled hospital study examined the use of a bedside automated dispensing system that sounded an alert to notify the nurse when a medication was due for administration; the device allowed access to only those medications due for administration at that time. The devices were randomly assigned to patients on one medical-surgical nursing unit, with a crossover during the two-week study period. Medication errors declined from 15.9% to 10.6%, a result of a reduction in wrong time and omission errors.¹¹

Kratz and Thygesen, in a 1992 study from patient medication profiles of a computer-driven automated pharmacy system, reported a 99.98% filling accuracy. The errors found included an extra tablet, half tablet, or missing tablet.¹²

In 1993, Maliekal published an evaluation of an automated dispensing system in an ambulatory pharmacy. Two hundred seventy prescriptions filled by the machine during normal operation were pulled off the line, and the count was compared against the computer-labeled amount. Thirty-eight discrepancies were discovered (14%); 22 were overfills, 14 were underfills, and two could not be classified.¹³

Klein, Santora, and Pascale compared medication cart filling using an automated dispensing system with traditional cart filling by technicians. Filling errors were found in fewer than 1% of the doses; 0.84% for the traditional system and 0.65% for the automated system.¹⁴

Cooper and colleagues compared medication administration error rates in a long-term care facility using three different distribution systems: single unit-dose packaging, patient med pak packaging, and blister card. All medications were checked by a pharmacist before delivery to the facility. Medication administration error rates, measured by observational audits, decreased from 8% using blister cards, to 2.5% with single unit-dose packaging, and 0% with the patient med pak system. These rates included all identified errors, many of which are independent of the drug distribution system (e.g., improper administration technique, documentation errors). The authors reported that errors that may be related to a drug distribution system (e.g., wrong drug, wrong time, drug omission, and wrong dose) decreased from six (2.1% of medications observed) in the blister card system to zero in both the single unit-dose and patient med pak systems.¹⁵

In 1995, Barker and Allan reported a study of medication administration errors involving a hospital nursing unit-based automated device used for narcotics and selected first-dose medications. The system configuration allowed nurses to obtain any medication stored in the device for any patient; the system was not integrated with the patients' computerized medication profiles. The error rate for all doses retrieved from the automated device was 16.3% (20 errors per 123 opportunities); 11 errors were administration of twice the dose ordered, while seven doses had no chart order. The error rate for doses retrieved from the patient's medication drawer or from nonautomated floor stock was 5.4% (43 errors per 796 opportunities). The observers noted that the nurses typically administered drugs from the automated device without checking them, whereas those taken from the patient's medication drawer were typically checked.¹⁶

Ray and associates compared technician filling error rates before and after installation of an automated medication station on a 26-bed adult medicine unit. Six weeks before and six weeks after conversion to the automated medication station, the medication cassettes filled by technicians were checked by pharmacists against the pharmacy-generated medication administration record. There was a significantly lower error rate in filling the automated medication station (0.61%) than in filling the traditional unit-dose cassettes (0.89%), a relative difference of 28.7%.¹⁷

Borel and Rascati compared medication administration error rates just before and two months after implementation of a medication profile-linked version of an automated pharmacy system. The automated pharmacy system was interfaced with the hospital's medication records, and nurses had access to only those medications approved for a specific patient. However, it was possible for the nurse to obtain certain medications (a provision for emergency situations) without prior approval by a pharmacist. The authors reported that such overrides occurred during the study, but did not state if any were associated with medication administration errors. The medication administration error rate for the traditional unit-dose system was 16.9% (148 errors in 873 opportunities); the error rate for the automated pharmacy system was 10.4% (97 errors in 929 opportunities)-a significant reduction in errors. The automated pharmacy system decreased the relative frequency of omission errors, but unauthorized-drug errors and wrong-dose errors were not affected; the rate of wrong-time errors increased. The authors did not observe any instances where the automated pharmacy system was filled incorrectly.¹⁸

REDUCING ERRORS WITH AUTOMATED PHARMACY SYSTEMS

Automated pharmacy systems can help to reduce medication errors significantly, as anticipated. However, automation may indirectly contribute to human errors when it engenders complacency. There is some evidence that nurses are less likely to check medications retrieved from an automated pharmacy system than those removed from a medication cart.¹⁶ It must be recognized that automation is only one component of a human-machine system, and the education and training required for the human part must not be neglected.

Successful implementation of automation requires reengineering of the process to be automated. The following recommendations for the redesign of health care delivery systems to reduce the likelihood of error have application to automated pharmacy systems.¹⁹

• Reduce reliance on memory through the use of checklists, protocols, and computerized decision aids.

• Improve information access by displaying information where and when it is needed, in a form that permits easy access.

• "Error proof" tasks where possible, such as design of computer systems that do not allow entry of lethal doses or medications to which a patient is known to be allergic.

• Standardize processes wherever possible, such as drug doses and times of administration.

• Provide training in procedures and problem-solving with an emphasis on possible errors and how to prevent them.

• Build "buffers" into each system so errors can be identified in time to be intercepted before they can cause harm to patients, such as requiring pharmacist review of medication orders before administration.

• Redesign work schedules, division of responsibilities, task descriptions, and work environment to reduce unnecessary stresses.

DESIRABLE FEATURES FOR REDUCING ERRORS IN AUTOMATED PHARMACY SYSTEMS

Features desirable for reducing errors in automated pharmacy systems, as supported by relevant research and theory to date, are outlined below.²⁰ Some of these features apply only to institutional settings.

• Controls are comprehensive-Controls extend from point of order-entry to point of dispensing or administration (including confirmation of charting); systems are integrated with the pharmacy or facility information system.

• "Error-prone" components are controlled-The system imposes maximum controls over "error prone" dosage forms, such as injectables, oral liquids, and I.V. admixtures.

• Unit doses are dispensed-Almost all doses are delivered in ready-to-administer form that requires no (or minimal) preparation or measurement by the person administering the medication. Each dose must be individually packaged and labeled so it retains its integrity to the bedside or other point of care.

• Dosing time is signaled-The caregiver or patient is alerted when the dose is due.

• Labeling-Labels are machine-printed on the unit-dose package.

• Bar code identification is used-All components, including the medication, patient, and person administering the medication can be identified using bar code technology.

• Access to medication is limited and controlled-Medications are accessible only at the time needed, and only by authorized personnel.

• Dose administration is captured-To reduce reliance on memory, medication administration is documented at the point of care (bedside), and at the time of administration. Charting done at the "bedside" (point of administration) reduces reliance on memory and improves accuracy.

• Drug use information is provided-Important drug use information needed for correct administration is provided at the time of administration. Access to the needed information must be immediate and at the point of care. Lack of information was identified in a recent study as a root cause of errors.²¹

• Controls are not easily compromised-System overrides are signaled visibly and/or audibly at the time of the event, and electronically documented. This "buffer" assists in the detection and correction of potential errors.

RECOMMENDED FOCUS FOR QUALITY ASSURANCE

The development and implementation of appropriate policies and procedures, and oversight through quality assurance and continuous quality improvement programs, are required for the safe and effective use of automated pharmacy systems. The following areas should be the focus of both policies and procedures and quality assurance activities.

• Order entry-Order-entry errors are particularly troublesome because of the risk that an input error will be repeatedly carried out without detection. Automated pharmacy systems integrated with patient medication information will only be as accurate as the data entered.

• Downtime and system failures-A contingency plan should be developed and rehearsed periodically.

• Filling and refilling of medication containers-Bar code technology should be used to verify accuracy of these tasks.

• Expired and recalled medications-Policies and procedures must be in place to monitor and track expired and recalled medications.

• Storage-Storage conditions (temperature, light, humidity) must be monitored to protect drug integrity.

• Education and training-All persons working with automated pharmacy systems must receive appropriate training, including the purpose and capabilities of the system, how system failures can occur, and the competencies required of the operators. A formal orientation and training program should be developed and standardized, and its effectiveness routinely evaluated.

• Machine errors-Automated pharmacy systems are capable of producing unanticipated errors, which may be due to faulty software or other defects. Accuracy and reliability data should be sought from other users, particularly to characterize the failures peculiar to the system in question. All systems should be thoroughly tested before implementation.

• Labeling-Labeling should be machine-printed, contain all required information, and be easily readable. Packaging should provide sufficient space for readable labeling.

• Servicing equipment-Policies and procedures must be in place to allow for routine servicing of automated pharmacy systems without disruption of work flow.

• Medication error detection-Medication dispensing and administration errors should be measured, monitored, and evaluated as part of the quality assurance program. Medication administration should be evaluated using the observation method; the validity of this method is well established.^{6, 22}

BARRIERS TO USE OF AUTOMATED PHARMACY SYSTEMS

Regulatory

A major barrier to the widespread implementation of pharmacy automation is the variety and unpredictability of state regulations. Many state pharmacy practice acts contain definitions or provisions that preclude the appropriate and efficient use of automated pharmacy systems. For example, there is a need for a new definition of "dispensing" in many states to allow the use of automated technology. Language concerning automation that is enabling rather than restrictive, and that focuses on outcomes rather than process, is needed.

Currently many state boards of pharmacy are struggling with the development of regulations for automated pharmacy systems. Most pharmacy practice acts either do not address automation, or contain provisions that preclude the effective and efficient use of automation.

The Automation in Pharmacy Initiative, in cooperation with the National Association of Boards of Pharmacy, developed language on automation in pharmacy for inclusion in NABP's Model State Pharmacy Act and Model Rules. With a group of NABP members, the Automation in Pharmacy Initiative drafted the following "enabling" language, which was submitted to NABP for approval at its May 1996 annual meeting: "The use of automated pharmacy systems shall require that adequate and appropriate policies and procedures and quality assurance programs are in place prior to implementation to ensure safety, accuracy, security, and patient confidentiality."

At its 1996 annual meeting, NABP approved the appointment of a Task Force on Automation to draft model regulations for automation in pharmacy. In cooperation with NABP, the Automation in Pharmacy Initiative drafted model regulations that are outcome-focused rather than process-focused. This task was accomplished through an initial review of current or proposed regulations concerning automation in pharmacy, identification of areas of particular concern to boards of pharmacy, and drafting of model regulations to address these concerns. Working from the materials submitted by the Automation in Pharmacy Initiative, the NABP Task Force on Automation drafted model regulations, which were submitted and approved at NABP's 1997 annual meeting (Appendix 1).

Economics

Another barrier to the use of pharmacy automation is the cost of the systems. In 1995, the University Hospital Consortium surveyed its members regarding barriers to automation implementation. High cost, computer system interface, and space requirements were considered the most significant barriers.²³ Many pharmacies do not have the financial resources to purchase automated pharmacy systems; however, some vendors now offer both purchase and lease options to relieve this problem.

THE NEED FOR NATIONAL STANDARDS AND PROFESSIONAL GUIDELINES

There are no national standards for automated pharmacy systems. Some state boards of pharmacy have, or are in the process of writing, regulations for the use of automated pharmacy systems, which may result in de facto standard setting. However, this state-by-state approach has obvious disadvantages: individual boards of pharmacy lack the resources and expertise to adequately evaluate automated pharmacy systems, and regulatory differences among jurisdictions may discourage investment in new or improved automated pharmacy systems.

It is appropriate that standards for automated pharmacy systems be developed by organized pharmacy, in cooperation with the automation industry and NABP. It is also incumbent on the national pharmacy associations to develop guidelines for selection, evaluation, and appropriate use of automated pharmacy systems in the different pharmacy practice settings. ASHP is planning to develop voluntary practice guidelines on the use of automated technology in health-system pharmacies.

Until such standards are available, the pharmacist has a professional responsibility to ensure that appropriate policies and procedures and quality assurance programs are in place prior to implementation of automated technology, to ensure safety, accuracy, security, and patient confidentiality.

The only comprehensive evaluation of automated pharmacy systems to date was conducted by ECRI (formerly known as the Emergency Care Research Institute), a non-profit health services research agency. ECRI's monthly publication, Health Devices, provides independent, objective judgment for selecting, purchasing, managing, and using medical devices, equipment, and systems. Each article undergoes extensive review by ECRI's interdisciplinary professional staff, outside clinical experts, and the participating manufacturers. ECRI does not accept funds, directly or indirectly, from the medical device industry. No members of the staff are permitted to engage in outside consulting work for, or own stock in, medical device or pharmaceutical firms. The results of ECRI studies are prohibited from use for any promotional purposes.

The ECRI evaluation of automated pharmacy systems, which was limited to hospital-based, automated, decentralized pharmacy dispensing systems (ADPDS), was published in December 1996.²⁴ The study evaluated three ADPDS's from two suppliers, and focused on the ability of the systems to store and dispense medications in a safe, secure, and effective manner. Criteria were evaluated and tests performed in the following areas:

• Inventory and Stocking Functions-Inventory, stocking, return and waste of unadministered drugs, refrigeration.

• Dispensing Functions-Pharmacy profile and electronic medication administration record (MAR) interfaces.

• Security and Quality Control

• Reporting/Documentation

• Accessing and Interacting with the System-System accessibility, dispensing cabinet displays and controls, pharmacy workstations.

• System Reliability and Manufacturer Service and Support

The criteria developed by ECRI for the evaluation of ADPDS's may provide the basis for the development of standards for these types of automated pharmacy systems.

MORE