Measuring the impact of automated dispensing cabinets implementation on data and inventory management of human normal immunoglobulin in an acute teaching hospital: A pre- and post- intervention study

Abstract

Background

Human Normal Immunoglobulin (HNIg) is a complex plasma-derived blood product used to treat a variety of medical conditions. Global supply problems have increased focus on HNIg stewardship, including mandatory recording of HNIg usage on the National Immunoglobulin Database (NIgD). Local departmental audits identified significant inconsistencies in data uploaded to NIgD. Inventory management issues caused a number of stock losses in the last financial year. A paper-based HNIg batch number recording system was replaced with the use of Automated Dispensing Cabinets (ADCs), which generated an electronic batch number report, with matching fields required for electronic upload to NIgD.

Aim

To measure the impact of ADC implementation on the accuracy of data uploaded to NIgD, HNIg stock control and staff time associated with processes of HNIg data and inventory management.

Method

Three months of pre- and post-implementation HNIg dispensing data was compared to the data uploaded to the NIgD for discrepancies. Inventory reports were used to identify unexplained stock adjustments. Time and motion methods were used to quantify staff time associated with HNIg activities.

Results

Pre-implementation: 20.7 % (3762.5 g/18,217 g) of HNIg by volume (23.7 % of dispensing episodes; 66/279) were inaccurately uploaded or absent on the NIgD and three stock adjustments were made (loss of >£15 k). Post-implementation: 12 % (2325 g/19,347 g) of HNIg by volume (10.8 % of dispensing episodes; 31/286) were inaccurately uploaded or absent on the NIgD and zero stock adjustments were observed; Mean time for dispensing per HNIg prescription reduced from 17 min to 8 min; Time spent uploading data to NIgD per month reduced from 5 h to 1.75 h.

Discussion/Conclusion

The use of ADCs improved accuracy of NIgD data upload. Following implementation direct observations have cited unregistered or finished patient episodes, dispensing procedural compliance, and user familiarity with the system as common reasons for incomplete data uploads to NIgD. The new ADC process had consequences of forcing dispensing procedure compliance to improve uploads, whilst reducing dispensing times. It led to improved stock control and removed upload burden from dispensers. ADC stock discrepancy alerts allowed staff to proactively resolve discrepancies in real time. Time taken for additional processes to support management of stocks in ADCs, including monthly stock cycle counts, were important considerations for implementation. The key advantage of using ADC batch number reports is the ability to upload as a single monthly batch without having to manually access individual patient records on the NIgD. It facilitates early identification of failed upload attempts and supports resolution of stock discrepancies.

1. Background

Human Normal Immunoglobulin (HNIg) is a complex plasma-derived blood product used in the treatment of a wide range of medical conditions. Treatment can be short term or long term and can be for replacement (e.g. in hypogammaglobinaemia) or immunomodulation¹ (e.g. acute myasthenic crisis). There have been concerns in the United Kingdom over the continuity of supply to the National Health Service since 2006.² The COVID-19 pandemic interrupted plasma collection putting further pressure on supply. In addition, emerging evidence for benefit in a growing list of indications continues to fuel demand. With ongoing global supply problems,³ demand management initiatives continue to be vital in managing the distribution and use of HNIg nationally. This increased focus on HNIg stewardship has included UK pharmacy initiatives such as strict stock control and mandated recording of indication and HNIg batch numbers on the National Immunoglobulin Database⁴ (NIgD) for each individual patient administration in England. Complete and accurate data input to the NIgD is crucial as it enables national forecasting and determines HNIg stock allocation and financial remuneration for each acute National Health Service (NHS) hospital in England. Furthermore, as HNIg is a blood product, recording the batch number is vital to allow for traceability.

In the study setting, the uploading of the HNIg dataset to the NIgD is a pharmacy dispensary-based process, completed at the point of supply against a prescription. It is the responsibility of the pharmacy dispenser to label and select the correct number of vials for the prescribed dose and enter individual batch numbers for vials dispensed directly into the NIgD patient treatment record. If a new patient is not yet registered on the NIgD, details of dispensing are entered into a paper based HNIg record book for upload at a later date, as part of a rostered weekly task for the dispensary team. Local departmental audits in 2019 and 2021 had identified significant delays and inconsistencies in the data uploaded to NIgD. The lack of familiarity with the recording process, staffing shortages, missing paperwork and challenges accessing the NIgD from dispensary computers were cited as reasons that had contributed.

Inventory management is important to ensure that any HNIg stock received from the national allocation is appropriately tracked, monitored and made available for patient treatment locally in acute trusts when needed. As a tertiary centre and large users of HNIg, the associated volumes received monthly make this a challenging undertaking when completed manually. Considerable amounts of staff time are taken locating stocks, confirming counts and investigating discrepancies that arise within the department. This is further complicated by frequent changes to HNIg brand and vial sizes being received as part of monthly allocations based on national availability. The development of an automated process to upload batch numbers for HNIg using data extracted from a pharmacy dispensing system to NIgD has been shown to increase the accuracy and completeness of data as well as reduce administrative time and risk associated with financial losses.⁵ There is otherwise a paucity of published literature on the topic.

The use of Automated Dispensing Cabinets (ADCs) in hospitals and other healthcare settings is becoming increasingly common. ADCs offer an electronic medicines management system where medicines can be securely stored, stocks tracked, and individual patient-specific dispensing data logged.⁶ This solution aligns to address pharmacy problems identified locally with HNIg management so was the natural choice of technology to deploy in practice. This study aims to evaluate the impact of ADC implementation on the accuracy of data uploaded to NIgD, HNIg stock control and staff time associated with processes of HNIg data and inventory management. The ambition was to move from a semi-manual laborious paper-based HNIg recording system to utilisation of ADC technology to deliver efficiencies in data and inventory management.

2. Method

2.1. Study setting and design

The study was conducted at the main site of a large London teaching hospital with tertiary immunology, neurology and haematology services, all utilising HNIg for their complex caseloads. The hospital dispenses around 6000 g of HNIg per month.

A pre-and post-intervention evaluation design measured the accuracy of the data uploaded onto the NIgD, as well as the number, and financial value, of unexplained stock adjustments. The ADCs were implemented in January 2023 and data from two different periods for three months each were compared: May to July 2022 (pre-ADC) and February to April 2023 (post-ADC).

In addition, assessment of impact to pharmacy staff time was made using a time and motion study design to record time spent on activities associated with HNIg data and inventory management before and after implementing ADCs. Data from two different periods for a weeks' HNIg dispensing each were compared: November 2022 pre-ADC and April 2023 post-ADC.

This study was considered a service evaluation by the local pharmacy research and audit group. No patient identifiable data were recorded. The need for ethics approval was waived.

2.2. Process mapping

Before ADC introduction, an order to administer would be clinically screened by a pharmacist and received by the dispenser. The dispenser would then produce labels for the vials to be dispensed ensuring the correct brand and smallest number of vials to make up the dose was selected. They would then collect the physical vials from central pharmacy stocks located in the pharmacy store. Individual batch numbers for the vials dispensed would be entered either directly into the NIgD patient treatment record at this point if possible or entered into a paper-based HNIg record book together with a dummy label of the dispensing, for upload at a later date. The HNIg would then be final accuracy checked by a pharmacist or an accredited accuracy checking pharmacy technician (ACPT) and released for patient administration. NIgD data review and upload was a rostered weekly activity for dispensary staff to ensure any outstanding records entered in the paper-based HNIg record book were uploaded onto the NIgD. Any outstanding patient registrations or uploads beyond a one month timeframe were escalated to the trust immunoglobulin lead pharmacist for resolution.

Key stakeholders were identified and invited at an early stage of the initiative to support process mapping of all existing and proposed HNIg workflows. This included pharmacy staff from dispensary and stores, the trust immunoglobulin lead pharmacist, and external partners from the ADC supplier and NIgD developers. A process map identifying all steps required to complete the HNIg dispensing, NIgD data uploads and stock management as workflows both pre-ADC and post-ADC, shown in Fig. 1, Fig. 2 respectively.

Fig. 1.

Pre-ADC implementation workflow map for each process of HNIg dispensing, data and stock management of HNIg.

Fig. 2.

Post-ADC implementation workflow map for each process of HNIg dispensing, data and stock management of HNIg.

2.3. Intervention

A five cell ADC was repurposed and deployed within the main pharmacy dispensary footprint for HNIg in January 2023. All stocks of HNIg from stores were relocated and transferred into the ADC or fridges in the dispensary (which were networked to the ADC using a flex-lock) depending on their storage requirements. The ADC is configured to capture hospital number, patient name, HNIg brand and vial size, quantity dispensed and date of issue.

After ADC introduction, upon receiving a clinically screened HNIg prescription a dispenser would label the correct number of vials for the prescribed dose on the pharmacy dispensing system. At the cabinet interface, the dispenser would enter the patient details and select the correct brand and number of vials for the supply from stocks located in the ADC. When retrieving stocks from the ADC, the cabinet would enforce a stock count to be undertaken prior to the issue of the product. If the stock level entered does not match what the cabinet expects in stock, the dispenser is alerted and a recount is immediately undertaken. If the stock level still does not match a discrepancy notification is generated and staff escalate to a senior member of dispensary team for investigation. Individual batch numbers for vials selected are required to be entered into the ADC as part of the dispensing process. The HNIg would then be final accuracy checked by a pharmacist or an ACPT and released for patient administration.

An ADC report was developed in conjunction with the external supplier and NIgD developers to match fields required for monthly data uploads to the NIgD. Data fields included hospital number, HNIg brand and vial size, quantity dispensed and date of issue. This report would be downloaded monthly and transferred into the approved NIgD upload template, to permit a direct bulk data upload to the NIgD.

2.4. Data collection and analysis

The following primary outcomes were considered in data collection and analysis: percentage HNIg dispensed uploaded accurately to NIgD and the number, and financial value of HNIg stock adjustments. Secondary outcomes captured included reasons for unsuccessful, inaccurate or absent uploads to NIgD.

HNIg dispensing data was retrieved from the pharmacy dispensing system. This was compared with the data entered into the departmental paper HNIg record book and the data which had been uploaded onto the NIgD within one month of dispensing. Each individual dispensing was compared including the dose (in grams) and brand. Each month any inaccurate or absent NIgD uploads were investigated by a senior member of the dispensary team to confirm the reason for failure. Post-ADC implementation, the HNIg record book was continued for all dispensing episodes to enable validation of data from the electronic reports. A Pearson's Chi Square test of independence was performed to determine if there is a link between the use of ADCs and HNIg data recording accuracy.

An unexplained stock adjustment is where an amendment to the pharmacy inventory balance has been made on the dispensing system to match the physical quantity on hand with no clear documentation of the reason for increase or decrease in levels. Electronic inventory reports were reviewed pre and post ADC implementation to identify the number of unexplained stock adjustments, and their associated costs, during the defined data collection periods.

An observational time and motion method was utilised to record the total time per HNIg dispensing from start to finish, for all dispensing episodes over a one-week period before (November 2022) and a one-week period after implementation (April 2023), following a period of training. Total dispensing time spent was recorded as the dispensing workflows pre-and post ADC implementation were vastly different and individual steps were therefore not directly comparable. A mean time for dispensing processes to be completed per patient was calculated for comparison.

In addition, a self-reported time and motion study methodology was used to investigate impacts of the new workflows on time spent completing activities outside direct dispensing procedure. Data was collected for three consecutive months pre and post implementation and a mean time for each process was calculated. Pre-ADC dispensary staff self-reported the time taken to complete the monthly (manual) upload of the dispensing data recorded in the HNIg record book onto NIgD. Post-ADC dispensary senior staff self-reported the time taken to upload the monthly data using the new electronic report to NIgD. Pre- and post-ADC, the immunoglobulin lead pharmacist self-reported time spent reviewing and resolving any outstanding patient registrations or upload issues remaining once dispensary upload procedures were complete. Post-ADC introduction, the time taken for additional tasks including cycle counts and ADC restocks were also captured via observational time and motion methods.

3. Results

3.1. HNIg data accuracy

In the three months before ADC implementation (May to July 2022), a total of 18,271 g of HNIg was dispensed against 279 prescriptions. 20.5 % (3762.5 g/18,217 g) of dispensed HNIg (23.7 % of dispensing episodes; 66/279) was inaccurately uploaded or absent on the NIgD. The mean time from dispensing to uploading onto the national database was 82.5 days.

In the three months after ADC implementation, a total of 19,347 g of HNIg was dispensed against 286 prescriptions. 12 % (2325 g/19,347 g) of dispensed HNIg (10.8 % of dispensing episodes; 31/286) was inaccurately uploaded or absent on the NIgD – see Fig. 3. The mean time from dispensing to uploading onto the national database of 30 days.

Fig. 3.

A graph to show the mean percentage of HNIg dispensing's accurately uploaded to the database pre- and post-ADC implementation.

In relation to accuracy of uploads, the Pearson's Chi Square test confirmed the relation between these variables was significant, X² (2, N = 565) = 17.62, p = 0.0001493.

Reasons for inaccurate or absent data uploads to NIgD pre-and post-ADC implementation can be found in Table 1. The most common reason for failed upload was no treatment episode approved on NIgD both before and after ADC implementation.

Table 1.

A table to show the reasons for not uploading a HNIg dispensing episode to the NIgD.

	Reason	Number of Dispensing's
		Pre-implementation	Post Implementation
Absent upload	No active treatment episode available for upload	22	21
	New member(s) of staff unaware of HNIg dispensing process	7	0
	Inability to log onto NIgD database	5	0
	No documented reason for failed upload	19	0
Inaccurate Upload	Incorrect dose uploaded on NIgD database	6	5
	Wrong brand uploaded/booked out	2	1
	Duplicate dispensing's / Returns	2	1
	Incorrect dose booked from pharmacy dispensing system	3	2
	Part dose returned (e.g. infusion reaction)	0	1
Total		66 (23.7 %)	31 (10.8 %)

3.2. Number of unexplained stock adjustments

In the three months before ADC implementation (May to July 2022), there were three unexplained stock adjustments of HNIg. The total cost of these adjustments equated to greater than £15 k. In the three months after ADC implementation, there were zero stock adjustments of HNIg and no associated financial loss.

3.3. Time and motion

The observational time and motion study results (see Table 2) showed that the mean time to dispense HNIg per patient reduced with the new ADC workflow from 17 min per patient to 8 min. The time taken for data management processes including upload and resolution of issues by pharmacy staff reduced post-ADC implementation from 300 min per month down to 105 min per month.

Table 2.

A table to show the time taken (minutes) for each stage of the of HNIg dispensing, data and stock management of HNIg pre-ADC implementation.

HNIg associated processes	Mean time taken (minutes)
	Pre-ADC	Post-ADC
HNIg dispensing (per patient)	17	8
NIgD data upload by dispensary staff (per month)	180	45
NIgD data review and upload by lead pharmacist (per month)	120	60

The observed time to complete new ADC activities as part of the post implementation workflows in the dispensary is presented in Table 3: HNIg monthly stock cycle count 30 min; HNIg monthly restocks 30 min; and NIgD discrepancy investigation and resolution 90 min per month.

Table 3.

A table to show the mean time taken (minutes) for new activities post-ADC implementation.

New ADC activity	Mean time per process (minutes)
Monthly Stock Count	30
Discrepancy Review and Rectification	90
HNIg ADC Restock	30

4. Discussion

The use of ADCs to store and dispense HNIg combined with partnership development of a linked electronic ADC batch number report that can be downloaded monthly has resulted in improved accuracy and completeness of mandated data uploaded to NIgD. This outcome is supported by Pearson's Chi Square test which showed there was a statistically significant difference and therefore can assume that the implementation of ADC had a positive impact on HNIg data recording accuracy. The key advantage of using ADC batch number reporting is the ability to upload data in bulk once a month without having to log into individual patient records on the NIgD. The batch number reports could also allow for ease of tracing HNIg as a blood product if needed. This has negated the need for dispensers to manually enter information into the national database as part of the supply process, reducing associated time required to dispense HNIg significantly. By automating and standardising the process for upload, the errors related to the processes for upload were also reduced. The time required for the weekly NIgD upload from the paper-based HNIg record book has also been removed. These results align with outcomes shown in another study, where an alternative method was used to electronically log and upload the mandated information to the NIgD, with similar success.⁵ The downloaded ADC report required manual data manipulations in Excel to remove information that did not match the upload template, such as time of dispensing from the ‘date’ field, ADC brand names to match that approved naming in the NIgD template and the removal of returns ahead of information transfer to the NIgD data template for successful uploads was time consuming. Further work should consider the full alignment of naming conventions on the ADC batch number report to those on the NIgD upload template, which would further minimise the manipulations required to transfer data into the approved upload template, delivering additional efficiencies.

Although the new report allowed easy and accurate upload to the database, the lack of treatment episode approved on NIgD remained the main reason for unsuccessful data upload pre and post ADC implementation. However, the monthly bulk uploads using the NIgD template has facilitated easy and complete identification of these failed upload attempts by highlighting episodes which did not upload successfully. In turn, this enabled early communication with the trust immunoglobulin lead pharmacist, who was able to support resolution of issues associated with delays and potential non-compliance in the clinical approval process. The trust immunoglobulin lead pharmacist is addressing improvements required in the clinical approval process separately through the trust immunoglobulin panel. People factors including staff not being familiar with HNIg dispensing process, on-call pharmacist not complying with HNIg dispensing process and staff not registered with NIgD database were cited as reasons for absent uploads pre-ADC. For many absent uploads, it was difficult to determine a reason for lack of upload, due to lack of documentation by dispensers.

Inaccurate data uploads were identified pre- and post-ADC implementation (n = 13 and n = 10 respectively). The reasons for errors were similar before and after implementation, and all errors apart from part returns were reduced post implementation. This is thought to be because the new workflow did not require dispensers to upload information as part of the dispensing process. Overall, the numbers were small and further data would be needed to enable robust conclusions to be made.

Returns continued to be a major challenge for NIgD data uploads. Where returns were due to a dispensing error, supply would need to be returned via the ADC as well as on the pharmacy dispensing record. These returns needed to be removed manually from the monthly data set prior to NIgD data uploads. Returns from clinical areas for doses not administered to patients, which occurred after the bulk database upload, were individually processed on the database by entering each individual patient record, and this process could not be automated.

Whilst overall dispensing time was reduced, entering information into the ADC, such as hospital numbers or batch numbers, is time consuming and has an associated error risk. As HNIg batch numbers are between 10 and 16 characters long, entering the batch number manually for each individual vial is time consuming for the dispenser. Due to this, on-going training is vital in assuring data quality and completeness at the ADC interface.⁷ Further accuracy and efficiency could be achieved by introducing barcode scanners integrated with the ADC, which allows the batch number to be populated automatically following scanning. In addition, interfacing the ADC with the pharmacy dispensing system⁸ to reduce the need to enter patient details manually or to select the correct drug from the list of medicines in the ADC would also improve data quality and workflow efficiency. This could remove the need for manual entry of batch numbers, patient information, and support reduction of brand selection errors.

It must be acknowledged that there is a requirement for additional time to support new ADC management activities in the dispensary to maintain stock inventory e.g. monthly stock cycle counts and restocks into ADC, and NIgD data uploads. At the study site, ADCs were already in use for the management of controlled drugs, therefore dispensary staff were already familiar with its use. In practice, the time taken to undertake additional activities described may be less than for a trust implementing ADCs for the first time. A limitation of the study was that it was not possible to measure the time spent on stock management pre-implementation, as this was undertaken on an ad-hoc basis by the pharmacy stores team, therefore no formal comparison using time can be made. The move of all HNIg stocks into dispensary-based ADC allowed the dispensary team to control the entire ADC management and provides full transparency of inventory within the department which has also facilitated the prompt investigation and resolution of any discrepancies flagged. As part of the supply process, the ADC were configured to enforce a blind stock count prior to the issue of the product for each individual dispensing. If the ADC noticed a discrepancy between stock level entered and inventory expected, the ADC alerted staff in real time to proactively resolve discrepancies.

Monthly cycle counts are a three-way stock level check between the pharmacy dispensing system, physical stock level and the ADC expected inventory level. Although the cycle counts are a time-consuming process, they are an essential part of accurate inventory management. Prior to ADC implementation, stock counts were undertaken on an ad hoc basis when significant discrepancies were identified. Due to the time lapse of some investigations, it was not always possible to identify the reason for stock losses or gains. Access to ADC electronic reports has also enabled easy identification of dispensing errors and stock discrepancies. The new cycle counts process identified seven discrepancies, which have been easily rectifiable due to early detection, and ease of investigation by comparing the dispensing system log against the ADC electronic log. Prior to ADC implementation, discrepancies were more difficult to investigate. There has been improved stock control with no unexplained losses (stock or financial) in the post-ADC data collection period.

When considering the use of ADC, it is important to consider cabinet storage capacity for product lines. For HNIg management, the frequent changes to HNIg brand and vial sizes being received as part of monthly allocations based on national availability prompts a cabinet reconfiguration task to be undertaken as part of cabinet restocks when stocks are receipted in the department.

A noted limitation of the study is when staff in the study attempt to change or improve their behaviour simply because it is being evaluated – The Hawthorne Effect. All members of the dispensary team were aware the study was being undertaken and of the intervention being introduced, with potential for improvement in results due to focused efforts on the correct use of ADC. The likely impact of this is higher in the time and motion studies where staff were directly observed. To minimise this risk, the method of data collection was kept the same pre- and post-ADC implementation and therefore the effect would have minimal impact on our study findings. The study design also involved data gathered automatically through ADC use indirectly that would also have supported overcoming this.

5. Conclusion

The use of an ADC to manage HNIg dispensing, data and inventory management reduced the number of absent and inaccurate data uploads to the ADC. The greatest improvement was seen in inventory management and a subsequent reduction in financial losses. The implementation brought with it benefit in time efficiency, staff engagement and empowerment in managing the whole process. This work has also highlighted the need for all HNIg processes to be reviewed and aligned (e.g. the clinical referrals) to get maximum benefit from the implementation.

CRediT authorship contribution statement

Alice Pinfield: Writing – review & editing, Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation. Kit Lai: Writing – review & editing, Writing – original draft, Supervision, Resources, Methodology, Investigation, Data curation, Conceptualization. Shelley Jones: Writing – review & editing, Writing – original draft, Validation, Supervision, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.