Abstract
Purpose: To compare and evaluate 2 methods of inventory management in automated dispensing cabinets (ADCs). Methods: Ten profiled ADCs had 2 inventory management models implemented over 2 months. Implementation of the models on each ADC involved adjustment of par levels (desired accessible quantities of medication) and removal of medications not used in the past 90 days or more. The par levels of 5 ADCs were adjusted using a formula developed based on the economic order quantity model. The par levels of the other 5 ADCs were adjusted using a formula based on historical average daily usage. The study endpoints include stock out rate, vend:fill ratio, quantity of expired medications, and inventory carrying cost. Results: The total of number of medications stocked in the 10 ADCs was reduced from 3035 in a 2-month pre-implementation period to 2932 in a 2-month post-implementation period yielding a reduction of inventory carrying cost by $11 011. The mean stock out rate in both study groups increased and vend:fill ratio decreased after implementation. The quantity of expired medications increased in the modified economic order quantity formula inventory management model and decreased in the average daily usage inventory management model. Conclusion: The implementation of 2 inventory management models on ADCs had a negative impact on stock out rate and vend:fill ratio, a mixed impact on quantity of expired medications, and a positive impact on inventory carrying cost reduction.
Keywords: automated dispensing cabinet (ADC), vend:fill ratio, inventory management, par level
Introduction
The use of ADCs amongst medical institutions has seen a widespread increase. Survey data from 2014 indicates that 97% of hospitals use ADCs in their medication distribution systems, a dramatic increase from 49% in 1999. 1 The benefits of widespread ADC usage are well documented including: reduced drug waste, medication error rates, and drug diversion.2 -5 Despite these benefits, optimal inventory models have not been fully derived. This study aimed to optimize ADC inventories by decreasing stock out rate, inventory carrying cost, expired medications, and increasing vend:fill ratio through the implementation and comparison of 2 inventory management models.
Background
The Johns Hopkins Hospital (JHH) is a non-profit, 1154 bed academic medical center located in Baltimore, Maryland. JHH uses a mixed model of drug distribution that is heavily reliant on ADCs. JHH has 321 ADCs, located in various locations such as patient care units, operating rooms, and hospital-based clinics providing nurses access to medications. Despite the increase in use of ADCs at JHH, little research has been done to optimize their medication inventories and par levels. In this study, ADC optimization was defined as the manipulation of ADC inventory to increase operational efficiency. This optimization included removal of infrequently used medications and the adjustment of inventory levels to desired accessible quantities of medications. These inventory level adjustments were made in order to decrease supply exhaustion between refills as well as inventory removal to reduce the likelihood that a medication would expire.
For nearly 3 decades, ADCs have been providing safe and effective medication use to all patients at JHH. Nevertheless, recent analyses indicate that there is room for improvement regarding ADC inventory management. Carrying costs are increased as additional medications are stored within ADCs. Furthermore, if not used to provide patient care, these medications are left to expire, incurring further costs. The process of refilling ADCs can be burdensome and inefficient. Medication stock out rates are suboptimal and have the potential to negatively impact patient care as medication administration times and critical medications are delayed. Currently, there is no standard of determining inventory levels within ADCs at JHH. Inventory levels are set by pharmacists’ experience and discretion, thus may vary widely. The implementation of an inventory management model would serve as a means of standardization.
Despite increased utilization of ADCs in recent decades, there is a scarcity of literature investigating how ADCs can be optimized. Inventory level is critical to the efficient performance of an ADC. Inventory levels are usually determined by setting minimum and maximum periodic automatic replenishment (par) levels. The minimum par level is the level at which an ADC will generate a request to be refilled as unit doses are exhausted. The maximum par level is a quantity based on predicted need to ensure sufficient quantity exists to last between refills. Par levels are the focal point of ADC related activities and efficiency. Par levels that are too low can cause patients to potentially miss doses, delay doses of critical medications, and increase work volume for pharmacy technicians who must frequently restock the ADC. In contrast, higher par levels can cause the medication to remain in the ADC until it expires, utilizing the limited space within the ADC inappropriately or unnecessarily, and causing a reduced variety of medications within the ADC.
One study regarding ADC efficiency and optimization strategy exists. Researchers at the University of North Carolina Medical Center (UNCMC) studied a total of 8 ADCs located within perioperative and labor and delivery settings. Their optimization efforts included development of a formula to optimize par levels. 6 To test their formula, they split the 8 ADCs in their study into 2 groups of 4. In one group, par levels were calculated as a range of 3-day minimum and 7-day maximum values based on the daily mean number of vends over the previous 2 months. 6 A vend is a transaction or dispensing of a medication from an ADC. In the other group, they took into account the average daily vend count, the highest daily vend count, and the maximum time between deliveries to calculate the quantity of safety stock and reorder quantity. 6 From these formulas they derived subsequent formulas determining the minimum and maximum par levels. The previously described formulas by O’Neil and colleagues 6 considered the mean average quantity of medication dispensed between deliveries in 24 hours and the standard deviation of these values, the order lead time, and an assigned z-score to capture 99% of inventory needs. Their study was applied in perioperative and labor and delivery ADCs. 6
Methods
The study was split into 2 three-month time periods: pre- and post-implementation of 2 inventory management models. The study was exempt from institutional review board review as a quality improvement project. In the pre-implementation time period dispense data were collected from September 1, 2017 to November 30, 2017. Post-implementation data were collected from April 2, 2018 to June 2, 2018. All data were collected using CareFusion’s integrated analytic web-based system (Knowledge Portal for Pyxis, CareFusion Corporation, San Diego, CA). From the pre-implementation data standard stock medications (Box 1) and medications not accessed in 90 or more days were identified. Par levels were calculated using the Modified Economic Order Quantity (EOQ) Formula developed by researchers at UNCMC and a Day Supply Formula in which dispense data collected over a 3-month period were used to determine a daily average. The daily average was then used to determine a 3-day minimum (daily average multiplied by 3) and a 7-day maximum (daily average multiplied by 7). The calculated par levels were manually set via an ADC computer software interface console (Pyxis MedStation 4000 Console, CareFusion Corporation, San Diego, CA), and optimized to suitable locations within the ADC based on quantity and size. If the formulas calculated a minimum par level of less than 3, the par level was set at a minimum of 3 in order to allow sufficient time for refilling to avoid compromising patient care. The maximum level in these instances was set to 7. When the quantity of medication within the ADC was at or below the minimum par level, a signal was transmitted to the central inpatient pharmacy to refill the medication on the next delivery, using traditional central pharmacy refill automation. Deliveries were performed once daily for all nursing units.
Box 1.
Standard Medications Stocked in Automated Dispensing Cabinets at Study Site.
Medication |
---|
Albuterol 0.083% 2.5 mg/3 mL Nebulizer Solution |
Aspirin 81 mg Chewable Tablet |
Insulin Aspart 100 units/mL Vial |
Furosemide 20 mg/2 mL Vial |
Furosemide 40 mg/4 mL Vial |
Furosemide 100 mg/10 mL Vial |
Hydralazine 20 mg/1 mL Vial |
Labetalol 100 mg/20 mL Vial |
Metoprolol Tartrate 5 mg/5 mL Vial |
Midazolam 2 mg/2 mL Vial |
Nitroglycerin 0.4 mg Tablet |
Racepinephrine 2.25% 0.5 mL Nebulizer Solution |
Ten profiled ADCs (Pyxis MedStation, CareFusion Corporation, San Diego, CA) located within internal medicine and medical/surgical floors were chosen by the research committee for inclusion in the study and implementation of the inventory management models. The 10 ADCs were split into 2 groups of 5, in which an inventory management model (Modified EOQ Formula or Day Supply) was implemented. The chosen ADCs encompassed a total of 4 patient care areas (A, B, C, and D). Similar patient care areas (one medical/surgical and one internal medicine) were chosen for each model (Table 1). The inventories of each ADC in a patient care area (A, B, C, or D) were mirrored, meaning that they contained the same line item inventory. However, each ADC’s individual utilization was considered for inventory calculations to determine par levels.
Table 1.
ADCs by Inventory Management Model.
Day Supply Model | Modified EOQ Formula Model | |
---|---|---|
ADC Type | Medical/Surgical 1A | Medical/Surgical 1B |
Medical/Surgical 2A | Medical/Surgical 2B | |
Medical/Surgical 3A | Medical/Surgical 3B | |
Internal Medicine 1C | Internal Medicine 1D | |
Internal Medicine 2C | Internal Medicine 2D |
Note. ADC = automated dispensing cabinet; EOQ = economic order quantity.
Medications not used in 90 or more days in all mirrored ADCs were removed after vetting and approval by pharmacy management, unless they were a standard stock medication (Box 1). The par levels of standard stock and controlled medications were also adjusted according to the formulas. Each ADC was inventoried in order to activate the par levels and remove excess stock in order to avoid confounding results. All 10 ADCs had the inventory management models implemented via this method in a 2-month period by student pharmacists, pharmacy technicians, and pharmacists who were given print-out reports of the newly calculated par levels. Pharmacy technicians in the inpatient central pharmacy were familiarized with the project and instructed to refill only exact quantities to reach the maximum par levels in the selected ADCs.
Study Endpoints
The endpoints of the study include stock out rate, refilling efforts by pharmacy technicians as captured by vend:fill ratio, quantity of expired medications, and inventory carrying cost. The stock out rate was calculated as a percentage by dividing the total number of stock outs by the total number of vends for one ADC. Vend:fill ratio for a whole ADC is simply all vends from an ADC divided by all refills. A low vend:fill ratio signifies more refilling efforts for a number of vends, whereas a high vend:fill ratio signifies less refilling efforts for a number of vends. For instance, a vend:fill ratio of 9 means the ADC vends 9 units of medication before a refill is needed. Usually, higher vend:fill ratios correlate with increased efficiency. However, a vend:fill ratio that is too high has an increased risk that the medication may expire within the ADC. Inventory carrying cost was calculated using the wholesale acquisition cost of each medication within the ADC. Expired medications were quantified using the analytic web-based system.
Statistical Analyses
Descriptive statistics were reported to describe outcomes before and after inventory management model implementation in the 2 ADC groups. An independent student’s T test was used to identify any significant differences between the 2 inventory management models. All analyses were performed using Stata Version 13 (StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP).
Results
Ten ADCs stocked with a total of 3035 medications were subjected to inventory management model implementation from February through March 2018. Data collection for all machines ended in June 2018, with a total of 180 188 recorded transactions including vends and refills; 90 582 transactions occurred in the pre-implementation period, and 89 606 transactions occurred in the post-implementation period. Before implementation, each ADC had a mean of 304 medications. Post-implementation, each ADC had a mean of 293 medications. The total number of stocked medications was reduced from 3035 to 2932 medications in all 10 machines, a 103 item (3.4%) decrease.
The stock out rates increased after implementation. The Day Supply inventory management model had a 0.6% stock out rate before implementation and a 1.6% stock out rate post-implementation (Table 2). The Modified EOQ Formula inventory management model had a 0.5% stock out rate pre-implementation and a 0.9% stock out rate post-implementation. For stock out rate in the post-implementation period, the Modified EOQ Formula inventory management model was found to be significantly different from Day Supply inventory management model (P = .0032), indicating that the Modified EOQ Formula inventory management model fared better than the Day Supply inventory management model.
Table 2.
Characteristics Pre- and Post-Implementation.
Pre-Implementation | Post-Implementation | |||
---|---|---|---|---|
Characteristic | Day Supply (n = 5) | Modified EOQ Formula (n = 5) | Day Supply(n = 5) | Modified EOQ Formula (n = 5) |
Mean no. stocked medication | 314 | 293 | 297 | 290 |
Mean no. vends | 8488 | 7959 | 8130 | 7680 |
Mean vend:fill ratio | 9.9 | 10.7 | 7.6 | 8.7 |
Mean stockout % | 0.6% | 0.5% | 1.61% | 0.9% |
Mean no. expired medications | 214 | 111 | 110 | 155 |
Note. no. = number; EOQ = economic order quantity.
The mean vend:fill ratio for all 10 ADCs during the pre-implementation period was 10.3; whereas the post-implementation vend:fill ratio was 8.1. The vend:fill ratio decreased from 9.9 to 7.6 in the Day Supply inventory management model, indicating an increase of approximately 30 pharmacy refills for every 1000 vends. The vend:fill ratio also decreased from 10.7 to 8.7 in the Modified EOQ Formula inventory management model, indicating an increase of approximately 21 pharmacy refills for every 1000 vends. The post-implementation interactions between the Day Supply and Modified EOQ Formula inventory management models were found to be insignificant (P = .3483) in terms of vend:fill ratio.
The quantity of expired medications decreased in the Day Supply inventory management model and from 214 to 110. The quantity of expired medications increased in the Modified EOQ Formula inventory management model from 111 to 155 (Table 2). There were no significant differences between the 2 models (P = .1038).
The reduction in inventory carrying cost associated with inventory management model of the 10 machines totaled $11 011. The Day Supply inventory management model had a greater inventory cost reduction, with total inventory cost reduced by $7032 (Table 3). Whereas, the Modified EOQ Formula inventory management model had a lower impact inventory cost reduction, with total inventory cost reduced by $3979 (Table 4). Upon further investigation, it was found that the Modified EOQ Formula inventory management model called for an increased quantity of several high-cost medications.
Table 3.
Day Supply Inventory Management Model Cost Reduction.
ADC Station | Pre-Implementation | Post-Implementation | Difference |
---|---|---|---|
Med./Surg. 4 | $4073 | $2517 | $1556 |
Med./Surg. 5 | $4179 | $2845 | $1334 |
Med./Surg. 6 | $4384 | $3074 | $1310 |
Internal Med. 3 | $4287 | $2945 | $1342 |
Internal Med. 4 | $4359 | $2869 | $1490 |
Total Cost Reduction = | $7032 |
Table 4.
Modified EOQ Formula Inventory Management Model Cost Reduction.
ADC Station | Pre-Implementation | Post-Implementation | Difference |
---|---|---|---|
Med./Surg.1 | $6401 | $6943 | -$542 |
Med./Surg. 2 | $6490 | $5325 | $1165 |
Med./Surg. 3 | $6522 | $5511 | $1011 |
Internal Med. 1 | $4218 | $3051 | $1167 |
Internal Med. 2 | $4304 | $3126 | $1178 |
Total Cost Reduction = | $3979 |
With JHH’s total of 321 ADCs, the organization could reduce approximately $451 454 in inventory carrying cost through the Day Supply inventory management model, and $255 451 through the Modified EOQ Formula inventory management model, assuming a constant cost-reduction across all ADCs.
Discussion
Contrary to expectations, almost all study endpoints were less optimal relative to inventory practices prior to implementation. This may be due to several reasons. In the UNCMC study, the ADCs optimized were in perioperative and labor and delivery patient care areas. The patients in these areas typically require routine, predictable medications, and thus the medications within those ADCs are more “protocolized.” Furthermore, the ADCs in that study had approximately half the inventory of the ADCs studied at The Johns Hopkins Hospital. The greater inventory amounts coupled with more variable medication needs of patients in medical/surgical and internal medicine patient care areas may be one explanation for the results observed. In addition, the ADCs at The Johns Hopkins Hospital were profiled, whereas the ones in the UNCMC study were not. With profiled ADCs, the medication needs may be more unique to the patient, creating more complexity when attempting to manage inventory. The par levels were adjusted for both controlled substances and standard stock medications in this study, and included in the analysis whereas in the UNCMC study, controlled substances were excluded from the analysis. It could be that the vending behavior of these 2 unique classes of medications, skewed the inventory management formulas. This study removed medications not vended in 90 days, whereas the UNCMC study removed medications not vended in 180 days. The UNMC study optimized based on 2 months of usage data whereas this study optimized based on 3 months of usage data.
However, UNCMC undertook a staggered approach, optimizing 2 ADCs each month and collecting data 2 months before and after each optimization. This staggered approach results in 7 months of temporal data, with different ADCs optimized at different time points. This study implemented the inventory management models on all ADCs at one time and captured 5 months of temporal data. There is a potential that the non-staggered approach in this study does not capture the seasonality of medication vending as well, ultimately affecting results. There is likely a temporal relationship between the season and type of medications used. Furthermore, there is a lapse between usage data collection (September 1, 2017 to November 30, 2017) to use in the inventory formulas, and when the results collected and analyzed (April 2, 2019 to June 2, 2019). Par levels were not adjusted for large medications that would not fit within the ADC, and is a limitation to this study.
Another possibility is that the inventory management models work for certain tiers of medications, but not others. For instance, inventory management of a high use medication such as acetaminophen may not benefit from the inventory management models, but a moderate or low-use medication may, or vice versa. Further analysis is required to elucidate any tiers or categories of medications that benefit from these inventory management models.
Most medications’ par levels were reduced by the inventory management models as evidenced by the cost data in Tables 3 and 4. Additionally, the number of medications in each ADC decreased post-implementation. Overall, there seems to be an obvious tradeoff between inventory quantity and cost reduction. Whether this tradeoff is worthwhile is dependent on an institution’s ADC replenishment workflow. If ADCs are replenished on a daily basis, a slight decrease in vend:fill ratio, may be worthwhile for reduced inventory carrying costs.
A limitation of this study is that it is short in duration. A study with a longer duration may have achieved more substantial results and reduced the impact of any medication usage fluctuations. Given that the data collection period post-implementation is only 2 months, it is difficult to ascertain whether this is an accurate portrayal of the models. A longer data collection period may be needed to truly assess expired medications. Not all par levels were adjusted due to the size of certain medications. Though efforts were made to standardize inventorying of ADCs to activate the par levels, consistency may have not been maintained due to the variability of personnel completing inventorying process. There were 2 instances in which the par levels were increased for medications based on patient needs. This is reflective of typical practice, as ADC inventory is subject to change based on fluctuations. Pharmacy technicians responsible for filling the ADCs were informed of the initiative and instructed to replenish exact quantities, but continual adherence to these instructions cannot be guaranteed. This study did not assess the impact on pharmacy technician or nursing workflows. Inventory model implementation of one ADC took an average of 6 hours. Widespread inventory model implementation of ADCs may not be feasible given staffing constraints. Further investigation as to whether several groups or tiers of medications are more positively impacted by the inventory management models is warranted. A limiting factor in this study was the amount of data. Looking toward the future, with the advent of big data and artificial intelligence, perhaps predictive analytics would be able to fine-tune such inventory management models to better assist the end users’ needs and improve the metrics discussed in this study.
Conclusion
Implementation of 2 different inventory management models in ADCs had a negative impact on stock out rate and vend:fill ratio, a mixed impact on quantity of expired medications, and a positive impact on inventory carrying cost reduction.
Footnotes
Authors’ Note: Christopher Boreen is now affiliated to Department of Pharmaceutical Care, University of Iowa Health Care, Iowa City, IA, USA. Edward Lau is now affiliated to Department of Pharmacy, UNC REX Healthcare, Raleigh, NC, USA. At the time of the study, all authors were affiliated with the Department of Pharmacy, The Johns Hopkins Hospital, Baltimore, MD, USA.
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article.
ORCID iD: Thomas Walczyk
https://orcid.org/0009-0008-6024-2892
References
- 1. Pedersen CA, Schneider PJ, Scheckelhoff DJ. ASHP national survey of pharmacy practice in hospital settings: dispensing and administration–2014. Am J Health Syst Pharm. 2015;72(13):1119-1137. doi: 10.2146/ajhp150032 [DOI] [PubMed] [Google Scholar]
- 2. Murray M, Shojania K. Unit-dose drug distribution systems. U.S. Department of Health and Human Services. Agency for Healthcare Research and Quality. Making health care safer: a critical analysis of patient safety practices. Evidence report/technology assessment, No. 43. 2001. Accessed May 1, 2019. https://archive.ahrq.gov/clinic/ptsafety/chap10.htm
- 3. Hynniman CE, Conrad WF, Urch WA, Rudnick BR, Parker PF. A comparison of medication errors under the University of Kentucky unit dose system and traditional drug distribution systems in four hospitals. Am J Hosp Pharm. 1970;27(10):802-814. [PubMed] [Google Scholar]
- 4. Barker KN, Heller WM. The development of a centralized unit-dose dispensing system for U.A.M.C. Part VI: the Pilot Study—medication errors and drug tosses. Am J Hosp Pharm. 1964;21(12):609-625. [Google Scholar]
- 5. Anyanwu C, Egwim O. The prevalence and determinants of controlled substance discrepancies in a level I trauma hospital. Am Health Drug Benefits. 2016;9(3):128-133. [PMC free article] [PubMed] [Google Scholar]
- 6. O’Neil DP, Miller A, Cronin D, Hatfield CJ. A comparison of automated dispensing cabinet optimization methods. Am J Health Syst Pharm. 2016;73(13):975-980. doi: 10.2146/ajhp150423 [DOI] [PubMed] [Google Scholar]