Derek G Swanson
Deputy Director of Pharmacy
Royal Liverpool & Broadgreen NHS Trust
The Royal Liverpool University Hospital (RLH) is a large acute teaching trust with around 900 beds delivering a wide range of clinical specialties including a number at tertiary level. In 2001 a pharmacy service development programme was started, which included plans to implement enhanced clinical pharmacy activity, on-ward use of patients’ own medicines, a dispensary refurbishment and a One-Stop Dispensing (OSD) process for inpatients (also known as Dispensing for Discharge). The aim was to improve service efficiency and cost-effectiveness.
The milestone document for UK healthcare, A Spoonful of Sugar,1 which was published around the same time, recommended that hospital pharmacies implement dispensary automation to obtain the following benefits:
- More accurate picking (reducing dispensing error)
- More efficient storage (dispensaries are often too small)
- Improved stock management (reduced dispensary running costs, including removal of expired stock)
- Improved speed of pick (and prescription turnover).
In 2001, one or two early adopters of automated dispensing robots in UK hospital pharmacy demonstrated the reality of these benefits and their facilitation of other developments.2 We revised our dispensary refurbishment plan to include an automated dispenser. With pharmaceutical manufacturers migrating their packaging from traditional large bottles and tubs to ‘patient packs’ of a month’s course of dose units (eg. 28, 56, 84), we could see that Original Pack Dispensing would better support our new OSD process and so pursued the purchase of an original pack automated dispenser.
Like the other early adopters, we hoped to achieve both direct and indirect benefits of automated dispensing. The robot’s ‘deskilling’ of the medicines selection aspect of dispensing could realise the replacement of qualified and highly trained Pharmacy Technicians with unqualified but specifically trained staff. The salary of one full-time-equivalent (FTE) Pharmacy Technician could be converted to 1.33 FTE Dispensers, yielding more staff time for the same cost. More importantly, replacing one FTE Pharmacy Technician with one FTE Dispenser would release a sum that would enable the conversion of one FTE Pharmacy Technician post into one FTE Senior Pharmacy Technician post. These more experienced staff could be deployed to work at ward level supporting better management of medicines on the ward and most notably enabling the assessment and use of patients’ own medicines.
Having surveyed the market and compiled an output-based specification, an official journal advert of the European Union tender was placed in 2002 resulting in two bidders:
- The Rowa Speedcase, marketed in the UK by ARX Ltd
- The Swisslog Pack-Picker marketed in the UK by Swisslog Healthcare UK.
Both products were manufactured in Germany where dispensary automation is well established in community pharmacies. Both manufacturers were international companies specialising in warehouse and transport logistics, including lifts, conveyors, pneumatic tube systems and automated vehicles. The tender process required site visits to view the machines and understand costs, benefits, performance and opportunities. Both systems offered similar features and capabilities along with some differences. We viewed a Rowa Speedcase system in a UK hospital, which indicated how it would function in our department. There were no Swisslog Pack-Pickers in the UK at the time but visits to two independent community pharmacies in Germany demonstrated that this product could also work well in a hospital pharmacy. It was clear that, whichever robot was chosen, there would be a considerable amount of preparatory work to be done before the machine could be operational.
Making the choice
Based on our output-based specification, our observations, manufacturers’ information and the customers we visited, a product comparison chart was produced (Figure 1) to facilitate and evidence our purchase decision. A simple scoring system was chosen: 0=desired feature not deliverable; 1=desired feature deliverable; 2=desired feature exceeded that of the competitor.
Both products scored many ‘1s’ (desired feature deliverable), and each scored one ‘0’ (desired feature undeliverable). The Swisslog Pack-Picker scored more ‘2s’ (desired feature exceeded that of the competitor) than the Rowa Speedcase, the key areas being:
- Swisslog Pack-Picker had five storage/picking modules which offered better down-time resilience and faster picking than the two modules of the Rowa Speedcase for the same pack capacity. It also offered a single point loading system that fed packs into all five modules
- The Swisslog dispensary redesign delivered five dispensing commission/delivery stations, one more than the ARX design, supporting higher dispensing throughput
- The Swisslog Pack-Picker’s shape and size meant that it would fit into the allocated space within the redesigned dispensary without requiring structural changes, whereas, for the same pack-capacity, the Rowa Speedcase installation required the removal of a structural wall.
The zero scores also informed our decision. The Swisslog Pack-Picker scored zero for the absence of a refrigerated module, which was not critical for us, especially because the temperature control capabilities of the Speedcase fridge module had not been formally validated at that time in the hospital pharmacy environment. In contrast, the zero for the Rowa Speedcase was for its inability to meet our specification to work across the two floors of our department. Our goods receipt area was on the floor below the dispensary so we specified that the loading of packs into the robot should take place here, whereas for speed of dispensing delivery we specified that the picking should take place in the dispensary on the floor above.
The Swisslog solution used a small lift between the single-point loading unit on the lower floor and the main body of the machine in the dispensary. This solution still works as specified. The Swisslog bid was accepted and financial terms agreed in late 2002 with installation planned for a date in 2003. The cost was significant at £300,000; however, this sum included all enabling works and the dispensary refit. Seven years on, that investment continues to deliver the desired benefits and, although the dispensary layout has undergone a few tweaks since 2003, the original design still works.
Here come the robots
The preparatory work required before the installation was very significant, comprising several work streams with task sets that took several months to complete before the robot could be built. Key areas included:
- Planning the building work schedule. This included: temporary relocation of dispensary activities and delivering work outside normal working hours to avoid disruption of daily work
- Identifying and defining the packs suitable for Pack-Picker storage; taking account of shape, size and weight/volume. For liquid medicines we decided on 300mL as the maximum due to its weight and momentum on the delivery chutes
- Purchasing patterns were changed for most efficient dispensing, moving away from 100 tablet packs to ‘patient packs’. This later led to us commissioning pre-labelled packs for fixed analgesic and antibiotic courses which could be stored in the robot and used for dispensing or stock supply to the Accident & Emergency department.
We chose to retain traditional storage of refrigerated items, controlled drugs and unlicensed medicines. At the end of the process we had identified 1,500 robot-capable lines, which accounted for 97% of the total pack storage volume.
Our pharmacy system from JAC Pharmacy Services Ltd (JAC) required much reconfiguration, most notably the allocation of eight or 13 digit European article numbering (EAN) codes to the robot-capable products and renaming ‘shelf locations’ to enable automated picks from Pack-Picker via e-messaging from the JAC system. The completed data set of the robot-capable items was exported from JAC into the Pack-Picker database, the EAN bar codes within both JAC and Pack-Picker databases being the link that enabled accurate picking on demand from the JAC system.
The Pack-Picker was assembled in the summer of 2003 and a considerable amount of testing was then carried out to confirm that the system and new processes worked as intended. Swisslog software engineers had worked with their JAC counterparts to configure and test the interface that delivers the pick request to the Pack-Picker software followed by the picking mechanics and conveyor delivery systems. Packs were loaded into the machine and test picks generated, directly from the Pack-Picker console personal computer and indirectly from the JAC system. There was quite a ‘buzz’ among the staff when the first test requested picks via JAC were achieved.
Staff awareness and training needed careful management. The concept of a ‘robot’ was alien to most staff whose initial thoughts were of a wheeled machine that would speed around the dispensary picking packs off the existing shelving as they did. Many dispensary processes had to change, requiring revision of our standard operating procedures. Some changes went through several iterations after implementation as we gained experience working with the Pack-Picker. Although most pharmacy staff were computer literate, the new screens and processes for managing the Pack-Picker caused new fears and challenges for some pharmacists who needed considerable coaxing to engage with the new systems. Training for loading packs was a particular challenge because there was a specified process of scanning the barcode, entering the expiry date and placing the packs on the loading unit. Failure to place packs properly resulted (correctly) in rejection of the pack since the dimensions were different (a built-in safety feature). A small number of staff were trained as ‘super-users’ with greater knowledge in correcting common faults. Maintaining a cohort of ‘super-users’ remains an ongoing task.
Having completed all of the preparatory work, the go-live process proved equally challenging, with a range of tasks delivered over a weekend – a very short time frame. We planned to load the Pack-Picker with as many packs as possible on the Saturday and Sunday to deploy Pack-Picker ‘live’ on the following Monday morning. Lists of Pack-Picker–capable items were printed and the items removed from the existing shelves. The stock quantity for each item was validated against the JAC record so that JAC and Pack-Picker stocks were aligned, adjustments to the JAC record being made as appropriate. The Pharmacy Department remained open for business on the mornings of both days and at times, so staff loading the Pack-Picker were competing for the same stock as colleagues wanted for dispensing. Despite a very considerable number of staff hours being deployed over the weekend, the target number of packs loaded into the Pack-Picker was not achieved. We understood that our chosen loading strategy was flawed, so we changed to load different sets of packs (eg, all inhalers, all oral liquids), returning stock to the dispensary pending their turn. With hindsight, it became clear that a staged deployment by different pack types would have been more successful and certainly less stressful.
The Pack-Picker went live on Monday September 22nd 2003. Over the following week, the combination of the loading backlog, new stock arriving, and the Pack-Picker picking, proved quite a challenge. Once the Pack-Picker contained a critical mass of packs, the loading task became easier and stability was achieved after three weeks.
Some expected benefits were seen immediately, such as:
- Improved accuracy of pick reduced the likelihood of dispensing errors, although labelling and errors from non-robot picks (eg, fridge items) were not improved
- More efficient storage and use of space was achieved since the pack storage density of the Pack-Picker far exceeded that of standard open shelves. Some shelving was retained for items that the Pack-Picker could not support and for part-pack stocks that could not be returned to the Pack-Picker.
- Improved stock management was not a significant feature of our project because tight stock control and purchasing had been imposed over previous years; the annual stock turn rate for the dispensary being around 18. Expiry date tracking was improved since a pack date expiry in the format ’10 2006′ could be entered via a calendar in the Pack-Picker’s loading screen. An on-screen report of expiry dates could be checked, listing all packs expiring by a given date. This check became a monthly task, which was performed one month in advance. The Pack-Picker showed improvements on standard shelving; the one pack left in stock always remained visible and accessible, which will continue; checking the stock levels in the Pack-Picker against the JAC data was a mouse click away. The new tidiness of the dispensary encouraged better housekeeping of the remaining shelves with the reduction of part-packs promoting this
- Improved speed of pick and prescription turnover became apparent immediately after the initial loading problems were resolved. Prior to Pack-Picker implementation, it was usual for the ‘late team’ to stay up to two hours after normal closing time to complete the day’s work. With Pack-Picker’s impact on dispensing speed, the overtime time reduced by about an hour.
Other benefits emerged over time, as seen by the early adopters. The Pack-Picker’s automated pick deskilled the pack selection part of the dispensing process. We were able to reduce the number of Pharmacy Technicians dispensing, replacing them with trained ‘Dispensers’.
This released Pharmacy Technicians to concentrate on final dispensing checking duties and ward-based medicines management. These changes improved the dispensing team’s cost-effectiveness and the on-ward relabelling and recycling of medicines began to reduce dispensary activity. An intangible but significant benefit was the enhanced reputation that an automated dispensary brought to the Pharmacy department, which had a positive impact on recruitment.
Being the first hospital in the world to install a Pack-Picker, there were some initial challenges due to the demand we were placing on the machine and its software. A typical community pharmacy in Germany would process around 300 packs per weekday with a similar number being loaded. We were picking 1000 packs per weekday and 600 over the weekend as well as loading around 1,500 packs per weekday. Swisslog engineers worked hard to ensure that the Pack-Picker stood up to our demands but the pharmacy staff bore the brunt of the frustration and delays when things went wrong.
Installing the Swisslog Pack-Picker was a very significant experience for the Royal Liverpool Hospital as it not only changed the way we dispensed, but also paved the way for other developments such as OSD and ward-based Pharmacy Technicians.
Today, the combination of the Pack-Picker and those developments means that on average we dispense only 40% of all items on discharge prescriptions, enabling much better dispensing throughput with fewer and less skilled dispensary staff than in 2003. The Pack-Picker is involved in 90% of all dispensing transactions and picks 70% of all packs dispensed. Since ‘go-live’ in September 2003, the Pack-Picker has loaded and picked over 7,000,000 packs.
In the intervening years, Swisslog has updated the system in several ways, improving its efficiency and effectiveness, most noticeably by the addition of an automated loader, which has freed up further staff time.
Although some staff were initially very concerned about using the technology and took time to acclimatise to the new ways of working, these concerns resolved as the benefits were seen. Seven years on, no one in the Royal Liverpool Hospital pharmacy can conceive working without Pack-Picker, and visitors to the department are consistently amazed at what it does.
- Audit Commission. A Spoonful of Sugar – Medicines management in NHS Hospitals. London: Audit Commission 2001
- Slee, Ann, Farrar K, Hughes D, Implementing an automated dispensing system Pharm J 2002; 268: 437–8