Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Optimizing Multi-Item Inventory Management Decisions in Healthcare Facilities

Esmaili, Nazanin (2017) Optimizing Multi-Item Inventory Management Decisions in Healthcare Facilities. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

Download (2MB) | Preview


Healthcare costs in the United States continue to grow at a significant rate. In many healthcare settings material supply and inventory management represent significant areas of opportunity for managing healthcare costs more effectively. In this dissertation, we explore three topics related to these areas.

In the first chapter, we propose methodologies to help clinicians store medications and medical supplies optimally in space-constrained, decentralized Automated Dispensing Cabinets (ADCs) located on hospital patient floors. This is significant for many reasons: first, locating and storing medical supplies and pharmaceutical products within automated dispensing devices on patient floors is often not done efficiently and these devices are not utilized optimally. The primary purpose of an ADC is to ensure ready access of pharmaceuticals and medical supplies at floor locations within a hospital. However, the allocation of the limited space within an ADC to these items is typically not planned systematically and this often results in wasted staff effort as clinical personnel must expend effort in locating and retrieving them from a hospital's central pharmacy/storage location. A second major issue in using these devices is human error associated with the selection of pharmaceuticals from floor storage.
These problems are addressed via two different mixed integer programming (MIP) models. In the first model, we only focus on the tradeoff between storing many of a few items and storing smaller quantities of many items and in the second model we also consider how to reduce medication dispensing errors by designing appropriate storage layouts. We also propose valid inequalities and continuous relaxations to facilitate solving instances of a scale that represents real-world applications. Based on computational tests using actual data, these refinements can reduce the run time to well under 10\% of the time of the base model and thereby allow for large, real-world instances to be readily solved.
Our results indicate that using simplistic space allocation and inventory management policies, rather than our modeling approach, could result in about twice as much work for medical staff while still leaving unused space in the ADC. The second (position-based) model decreases risks associated with medication errors by at least 38\% over simpler methods.

In the next chapter, we investigate a class of inventory control systems which are used in inventory management systems at points of use (POUs) in hospitals. This class of inventory control systems is characterized by stochastic demand, periodic reviews with fractional (or very small) lead time, expedited delivery when stockouts occur, limited storage capacity, and service level requirements. We develop discrete time Markov chain models of different inventory control systems that deal with all of these characteristics while minimizing the total expected replenishment effort at POUs. We have derived closed form solutions and propose an exact algorithm to calculate the limiting probability distribution by locally decomposing the state space. We investigate the structural results and based on our approach we propose an algorithm that is much easier to use in practical applications compared to solving the steady state equations in Markov models, and the computational effort required for finding the replenishment policy parameters is reduced.

In the final chapter, we address the management of inventory for multiple non-perishable medical supplies in floor storage by selecting the optimal inventory policy for each item along with its corresponding operating parameters. In practice, hospitals tend to assign the same overall inventory control policy to all or the majority of the items. This simplistic approach often leads to wasted staff effort and ineffective policies. The objective of our research is to minimize the average labor effort required to count and replenish all of the items, while providing an acceptably high level of service (avoiding stock outs) and taking into account constraints on available space. We consider four policies: PAR, $(R,s,S)$, $(R,s,Q)$, and a two-bin Kanban system. We illustrate the model with actual data from a healthcare setting and propose some practical insights and guidelines on how to choose a hybrid inventory system based on demand and system characteristics.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Esmaili, Nazaninnae22@pitt.edunae220000-0002-0605-4996
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee CoChairNorman,
Committee CoChairRajgopal,
Committee MemberMay,
Committee MemberProkopyev,
Date: 1 February 2017
Date Type: Publication
Defense Date: 9 November 2016
Approval Date: 1 February 2017
Submission Date: 23 November 2016
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 155
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Industrial Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Mixed integer programming, computational optimization, two-staged two dimensional knapsack problem, valid inequalities, healthcare operations, automated dispensing cabinets (ADCs), discrete time Markov chains, local decomposition, closed form solutions hybrid inventory control system, periodic inventory policies, expedited deliveries, point of use locations
Date Deposited: 01 Feb 2017 20:22
Last Modified: 01 Feb 2019 06:15

Available Versions of this Item


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item