In early 2010, Peninsula Regional Medical Center (PRMC)—a 317-bed acute care hospital located in the hub of Maryland’s Eastern Shore—embarked upon a project focused on evaluating and potentially modifying our medication distribution system. At the time, our system was best classified as a centralized distribution system that employed a central robot to pick the majority of doses in the pharmacy, medication carts to deliver medications to the patient care units, and automated dispensing cabinets (ADCs) to house controlled medications and select first doses.
This traditional centralized distribution model with a 24-hour cart exchange process had been in use since 1999, but due to the age of both the robot and ADC technology, we faced a pivotal decision in determining what direction to take with our future distribution model: Should we simply replace the technology and continue to employ the centralized model or was a point-of-care model more in line with our future pharmacy management plans? We believed the decision should not be made lightly and that proper due diligence into our options was prudent. Furthermore, we knew a project of this scale was likely to stray from the established action plan in unexpected ways. In retrospect, we were able to learn from our mistakes and overcome some considerable obstacles in realizing our vision.
Measure Twice, Cut Once
The first step in approaching this project was to convene a multidisciplinary group to decide our path comprising representatives of pharmacy, nursing and nursing administration, as well as performance improvement engineers. In order to aid our decision making, the team developed a critical-to-quality matrix that focused on three broad categories: patient safety, operational efficiency, and cost containment (see Online Figure at www.pppmag.com/ctq ). After numerous site visits and additional research into how the various distribution models work, we decided to move forward with a point-of-care distribution model in late 2010 (as described in our article for the September 2011 issue of PP&P). As part of the process, we established a specific set of goals that included mandates for the following:
To augment our research and give our staff first-hand interaction with the types of automation and software we were considering, we conducted a vendor fair in late 2010 that focused primarily on ADCs and automated anesthesia cabinets. Patient safety and efficiency were the main categories scored by practitioners attending the fair (see Table 1). The results of this vendor fair survey were reviewed and all input was finalized by the end of 2010, which allowed for contracting to be completed by March 2011.
Nuts & Bolts of Process Change
Subsequent to contract finalization but prior to project initiation, the pharmacy department worked with facilities management on two key initiatives that would ultimately be crucial to the successful implementation of the new automation and distribution model. The first initiative focused on a renovation of the central pharmacy, which included the removal of the pharmacy robot and installation of the two new medication carousel units. Emphasis on improving central pharmacy workflow was a key driver to this particular effort.
The second initiative focused on facility modifications to the patient care units in order to accommodate the additional ADCs necessary to achieve a true point-of-care distribution model. This part of the overall project proved to be the most challenging due to the space limitations common to many units. We realized a single plan would not address every unit’s specific limitations, so we had to work with each clinical nursing manager individually on unit-specific modifications. Timelines for both major facility initiatives were established and aligned with product implementation timelines that were coordinated with the vendor.
Prior to automation and distribution model implementation, our team of pharmacists, nurses, and performance improvement engineers met to focus specifically on the possible impact of these changes on nursing workflow. Current state and ideal future state reports for nursing medication acquisition were developed, and a standard work reference for nursing was created. In the current state report, first doses were supplied from the central pharmacy—resulting in longer turnaround times—and routine medications were housed in medication carts. To bolster this process and provide more security, the ideal future state report suggested that a majority of medications, including routine medications, should be housed in ADCs to further mitigate turnaround times. We felt that standardizing medication acquisition by nursing was a key component necessary to a successful major distribution model transformation (see Table 2).
As part of the ADC workload analysis, we established a target ratio of each ADC serving 20 or fewer patients. This ratio would generally allow for one ADC to be available for approximately three nurses on a typical medical floor at any given time. As a result of our team meeting with nursing and performance improvement representatives, this approach was deemed a requirement in order to minimize queuing at units, which was a significant concern of nursing administration. After our new ADCs were implemented, any nurse queuing concerns were communicated directly to the pharmacy and addressed as warranted. It should be noted that no major nurse queuing events have been identified subsequent to full initiation of the point-of-care distribution model.
In mid-2011, the new ADCs were implemented using a phased approach. Among the 62 ADCs implemented, 22 were updated or retrofitted models and 40 were completely new. During this process, we also began installing medical-grade refrigeration units as part of each ADC implementation to enhance active temperature monitoring with wireless temperature probes. The pharmacy department used legacy medication dispensing data from both the central robot and existing ADCs to estimate proper medication types and quantities to be housed in each ADC. By compiling utilization data through the ADC system software, we were able to see which medications were being sent to any given nursing unit as well as their approximate daily use. This enabled us to custom stock the ADCs per nursing unit with the proper products and quantities to allow for a three-day fill.
We also used this opportunity to provide our pharmacy students with experiential education by tasking them with using the utilization data to stage multiple ADCs in preparation for deployment. Because numerous resources are needed to prepare even a single ADC for implementation, staging the implementation process helped minimize the number of additional pharmacy staff hours and overtime that would be required for the actual, hospital-wide implementation phase. After implementation of each ADC, the previous distribution model’s medication carts were removed from the patient care units and patient-specific medications were placed within the ADCs. Pharmacy workflow was modified to target once-daily replenishment for the majority of ADCs and workflow was load-distributed throughout the day and evening shifts to eliminate any workload peaks.
Based on the data generated after placement of the ADCs, medications loaded in each area were optimized to increase product availability in the devices. Optimization included loading additional high-use medications, removal of low-use medications, and revising medication quantities (max and par). We also established a target of less than one percent medication stockouts, which we felt was an aggressive, but attainable goal given the vast number of medication shortages nationwide. We monitored the remaining quantity of cart fill items to ensure a decline in the number of medications not available in the ADCs.
Shortly after the start of the phased ADC implementation, we introduced a remote software program that allowed nurses to queue medications in advance while the nurse was not physically located at the ADC. The multidisciplinary team felt this software was needed to further minimize the potential for nurse queuing during peak medication administration times—for us, 9am and 9pm. After adopting this software on several units, it was determined that its usefulness was overstated and further implementation of the software was abandoned. If the software is updated in the future to include more advanced functionality, we may review it again for potential adoption.
Automating Anesthesia Storage and Utilization
In late 2011, we installed 20 anesthesia cabinets, one in each surgical room. The new devices replaced the existing, standard medication carts that had been in use for approximately seven years. For the most part, a standard template was utilized to build the medication profiles in these cabinets.
Among the key drivers of this project was the incorporation of controlled medications as part of the medication profile on each cabinet—a prior limitation that had been identified as an obstacle to the provision of exemplary care in the OR. With anesthesia cabinets, we are now able to house controlled medications in each surgical suite.
All anesthesia providers and pharmacy technicians were required to complete a focused training program prior to implementation; this helped ensure a successful transition from our previous system, as staff members became comfortable with the new system before it went into service. Workflow in the pharmacy was modified to allow for anesthesia cabinet replenishment during the evening/night shift, a process that had predominately occurred during day shifts under the old distribution system. After implementation, we performed a stock level optimization analysis in coordination with anesthesia services, which helped streamline the storage and stocking process.
As part of the anesthesia cabinet implementation, we transitioned to prefilled syringes for several high-use medications to help improve safety. However, this process introduced new product sizes that we did not anticipate and required us to rearrange the storage compartments of the cabinets. Accordingly, it is important to consider the size and volume impact of prefilled syringes when determining the configuration of the drawers.
Lastly, each anesthesia cabinet was outfitted with a color laser printer and wireless network interface card to enable up-to-date and accurate reporting. Initially, the majority of these devices were hard-wired to our hospital network, but due to the benefits of device mobility in some cases, we are in the process of transitioning some of the anesthesia cabinets to a wireless network. This transition has allowed for the movement of cabinets to remote areas, such as MRI or endoscopy, when needed. Color printers were included to enhance product labeling by anesthesia providers. For identified medications, a color label with bar code (containing the medication NDC) will automatically print upon dispensing, and anesthesia providers have been specifically trained to apply the printed label immediately before moving on to another product. The colored labels adhere to the existing anesthesia color-coding standard from the ASTM, thereby helping to improve medication clarity and safety. In addition, the label can be applied to any syringes prepared by providers, which encourages adherence to established regulatory guidelines that all medications be properly labeled. With the pending implementation of an anesthesia information system to close the anesthesia administration and management loop, the bar-coded labels also will be used to scan products upon administration.
The Role of High-Speed Packaging
Concurrent to the implementation of the anesthesia ADCs, we placed two high-speed packagers in the central pharmacy to assist with unit-dose packaging of oral solids, specifically products that were not available in unit of use from the manufacturer. The two compact, high-speed packagers can work in tandem to mimic the level of output normally managed by one larger device, but offer flexibility in physical placement and in scale of packaging runs, as the packagers also can be run independently.
One of the essential elements of this project entailed the development of a list of medications that would be commonly loaded in each packager. To develop this list, we reviewed our entire formulary and included medications not available in manufacturer unit-of-use packaging or products with only one manufacturer of unit-of-use packaging (ie, sole source). Samples of these medications were then sent to the vendor to develop custom canisters to house each product. Furthermore, each packager has the capability to hold up to 128 different medications and as each packager can run independently, multiple fills can be processed at the same time.
Central Pharmacy Carousel Storage
One of the final technological steps of the medication distribution remodel was to implement two new carousel units in the central pharmacy. These carousel units are crucial to our new, point-of-care model because they make it possible to have a perpetual inventory system, help streamline ADC replenishment, and interface directly with our medication distributor, which enables quicker and more efficient medication ordering.
Due to the large size and technological requirements of the carousels, their implementation required a substantial reworking of job duties for both technicians and pharmacists. Scanning products into and out of the carousels has become an essential part of the new workflow, and we anticipate this process will increase ADC fill accuracy as compared to the previous distribution model.
By design, the majority of medications were selected for inclusion into carousel inventory. Exemptions included large medications (eg, PEG solution), chemotherapy, large glass containers, bulk liquids, and bulk bottles of oral solids. Maximum and reorder inventory levels were established using existing utilization reports, and once sufficient data becomes available within the system, the plan is to track fluctuating inventory levels dynamically based upon utilization and thus increase our inventory turns.
Distribution Changes Affect Staff and Workflow
Since implementing the point-of-care model, the accuracy of ADC replenishment has improved. With the frequency of replenishment increasing significantly, one might expect that opportunities for error also would increase; nonetheless, the number of reported replenishment errors actually has decreased.
As demonstrated by our analysis of the accuracy of medication distribution from the point of carousel removal in the pharmacy to administration at the bedside, the processes of scanning medication bar codes upon removal from the carousels (performed by technicians) and double checking the scans (performed by pharmacists) has directly mitigated errors. What was otherwise a relatively error-prone process under our previous manual system, has since transformed into a much safer one due in large part to the addition of automation. As of now, we are still in the process of implementing medication scanning during ADC restocking; an additional step we feel will further reduce errors.
Under the old centralized medication distribution model, turnaround times for medications were dependent upon four major activities: pharmacist order entry, pharmacy technician preparation, pharmacist check, and medication delivery. With that level of activity, turnaround time for the majority of medications averaged 45 minutes (see Figure 1). With our new point-of-care model, medication turnaround time now depends largely on a single component—pharmacist order entry; an activity that usually takes less than 10 minutes to accomplish. Given the importance of fast, efficient, and accurate medication distribution, our new system provides nurses with expedited access to medications, thereby enabling prompt medication therapy and contributing to an increase in both patient and staff satisfaction.
Likewise, the physical acquisition of medications by nursing has been simplified by transitioning to the point-of-care model. Nurses can now obtain the medications they need, when they need them from ADCs designed and optimized for efficiency; no longer is it necessary to shuffle through multiple medications in a patient cassette. It is worth noting that overall, the amount of time required for nurses to acquire medications from ADCs is relatively neutral as compared to the cart-fill system, but the new process is more standardized and less prone to error.
Introducing sweeping changes to the medication distribution model of a hospital pharmacy can and will impact many departments and individual members of the health care team. Although nurses, anesthesiologists, pharmacy technicians, and pharmacists are usually the most affected by this type of change, remember that physicians, risk managers, and facilities engineers all can and should play a role. Key to any project of this scale is patience and flexibility, as it takes time for all affected staff members to adjust to change and become comfortable with new automation and software systems.
The practice of hospital pharmacy requires a mindset dedicated to perfecting processes and systems, all the while understanding that a perfect system is unattainable. Nonetheless, the tenets of systemic improvement efforts, such as Lean and Six Sigma, seek to always enhance what is at hand. At PRMC, we recognized an opportunity to merge process improvement with realistic plans (both financially and logistically) for technology upgrades. While new automation and an improved system can help enhance medication safety, maximization of that safety lies in the minds and hands of technology’s human counterparts. Introducing standardization across the gamut of medication distribution functions—from the manner in which medications are stored to scheduling and maintaining training and education—is essential to successful and beneficial process change.
Dennis Killian, PharmD, PhD, received his graduate degrees from the University of Maryland-Baltimore School of Pharmacy in 1999 and 2001, respectively. He currently serves as director of pharmacy services at Peninsula Regional Medical Center in Salisbury, Maryland, and also as an associate professor of pharmaceutical sciences at the University of Maryland Eastern Shore School of Pharmacy.
William C. Cooper, PD, received his BS in Pharmacy in 1984 from the University of Maryland-Baltimore School of Pharmacy. He currently serves as the coordinator of pharmacy operations at Peninsula Regional Medical Center.
Adding New Items into the Carousel and Clinical System
A process was developed to identify and check any newly purchased items for bar code programming within the carousel and clinical system. Once a new medication is identified, the following process occurs: