Optimal Health Care Inventory Management Using Analytics Neeraj Agrawal Prakarsh Paritosh Ashish Paralikar Dibyajyoti Pati General Electric JFWTC, 122 EPIP Bangalore, India 560066 neeraj.agrawal@ge.com ABSTRACT Inventory Management is an important issue for health-care systems because it influences clinical and financial decisions. Before selecting, adapting and implementing a process for Inventory management, it is important that the various factors affecting Inventory management are carefully considered. The objective of this paper is twofold. First, it proposes an analytic model for hospital inventory management commodities, which would be able to predict the future demands of various inventory commodities. The model takes into account previous demand, population and geographic Location and other factors to successfully predict the future demand. Second, the paper suggests an optimization model that would minimize the cost involved in supply chain & logistics management so that the required commodities can be made available to the hospitals at the minimum possible cost. A web based portal is proposed to facilitate the hospitals to optimally distribute future inventory requirements among the different vendors. It is observed that in case of hospital inventories, concepts of holding and ordering cost is not very suitable [4]. Awareness of logistics management is becoming more widespread and many initiatives and studies dealing with supply chain integration have been undertaken [1]. However, internal supply chain remains the weak link in process integration and optimization [1]. The lack of systemic approach to internal supply chain management is reflected in huge costs in materials management and low service quality delivered to patients [1]. In the current scenario of increasing health care costs, systems inventory must be optimized without sacrificing the level of service provided. KEYWORDS Health-care Information System, Analytic Modeling, Supply Chain Management,Inventory Optimization, Hospital Vendor Portal 1 INTRODUCTION For the optimal use of hospital resources and the subsequent need to restore budgets, managements are responsible for finding solutions to achieve more operational efficiency in hospital processes [1]. Some of the areas of hospital management are that of controlling the inventory of medical gases, drugs and medical supplies. According to the classical inventory models the optimum inventory levels are maintained with the objective of minimizing the sum of cost of excess and under stocking. Figure 1. Proposed Inventory Management Model. Good inventory management is essential to the successful operation of any health care organization. One of the most important is the proportion of the organizations budget that represents money spent for inventory. Although the amounts and dollar values of the inventories carried by different types of health care providers vary widely, in a typical hospital s budget 25 to 30 percent goes for medical supplies and their handling. Because the inventory of medical supplies may comprise a significant portion of a health care organization s total assets, reducing its inventories significantly ISBN: 978-1 -941968-20-8 2015 SDIWC1 37
Vendors Vendor1 Vendor2 Deamand 1 Deamand 2 Analyze Demand and Usage of Hospital Inventory Data Develop and Analytical Model to predict future demand. Optimally divide the demand among different vendors Web based portal to display, order inventory to the different vendors. Figure 2. Different stages of proposed inventory model. raises its return on investment, and hence its position in the financial markets [2]. Since the revenues of hospitals are impacted by the carrying costs and ordering costs of pharmacy, medical-surgical supplies, and medical gases, it is important to create a mathematical model to minimize the total cost and simultaneously ensuring that the commodities are available in time [4]. Proposed inventory management model ant its stages are shown in Figure 1 and Figure 2 respectively. The objective of this paper is to come up with a predictive analytical model to optimize the inventory of pharmacy, medical-surgical supplies, and medical gases from the database available in the hospital information system. For this purpose the usage pattern of various commodities were analyzed and a model was designed to predict the usage for the upcoming month and give economical solution to choose among available vendors keeping in mind their capacity and storage capabilities. Model has been designed using various parameters such as previous demand of surgical supplies, oxygen and other anesthetic gases, geographical location, population density and population distribution of the area, seasonal variation for different hospitals. A web based interface was developed for different hospitals to display the optimal future demand of inventory. This portal will also aid in placing the order to the different available vendors. 2 FACTORS TO BE CONSIDERED FOR DESIGNING ANALYTIC MODEL The Present system of Inventory management suffers with several shortcomings resulting in its inefficiency. Availability of consumable on time is of utmost importance in health-care industry, where delay of a few seconds can cost a life. The challenge is even greater as the number of expected patients are unpredictable; suppliers are unreliable and costs are rising [3]. Analytic model should be able to reduce this uncertainties and make consumable available on time Cost of medical supplies has been spiraling up, greater numbers of patients are demanding high quality and reasonably priced health-care services. State-of-theart inventory management analytic model should automate the inventory reconciliation and visibility process. This would reduce the amount of work performed by hospital staff to maintain correct levels of inventory contrary to an archaic inventory management system. ISBN: 978-1 -941968-20-8 2015 SDIWC2 38
Figure 3. The overall ranking of top 25 drugs from the HDU across all years with the inclusion of Methotrexate (MTX) and Tacrolimus (TAC). Inventory standards can be measured by the ability to deliver utmost quality service to the patients. It can be measured in terms of services offered, price charged and perceived value [3]. A hospital will have to make decisions regarding the storage of supplies or direct delivery at the point of use. It deals with the implementation of CPFR (Collaborative Planning, Forecasting and Replenishment) among hospitals and suppliers and manufacturers of medical supplies. Good Inventory Management also depends on the warehouse s location, production level of goods, etc. It also deals with finalizing the distribution network between the suppliers and hospitals to minimize transportation and inventory costs. Hence, Optimization of distribution network between suppliers and hospitals is needed for an efficient system of Inventory Management. 3 INVENTORY MANAGEMENT MODEL To accommodate the factors explained in Section 2, analytics has been used to predict the future requirements well in advance. Predicted inventory requirement is distributed among available vendors using an optimization model. 3.1 Medical Inventory Analytic Model For designing the analytical model, following features from hospital database were taken into consideration: Previous demands of various pharmacy such as surgical supplies, Oxygen and other anesthetic gases Geographical location of hospital Population density and population distribution of the area Month of year No of surgery and patient admitted in an year ISBN: 978-1 -941968-20-8 2015 SDIWC3 39
Figure 4. Trends in the percentage of persons using prescription drugs: United States, 1999-2008 For model designing, the results of various surveys involving the demand of inventory commodities against various factors were studied and analyzed. For instance, The graph in Figure3 shows the ranking of top 25 drugs based on their uses [5]. It can be clearly seen from graph, usage shows linear relationship with year. In Figure4, graph shows trends in the percentage of persons using drugs, this graph also shows linear increase in usage with time [6]. The trend of various other hospital drugs, commodities and anesthetic gases were plotted against time for different graphical location. It was seen that the usage pattern of majority of hospital commodities varies linearly with time. The pattern of various hospital drugs used varied linearly with logarithm of population growth. Hence, a regression model can be developed on hospital database after clustering. Clustering can be done to group different population density region and to cluster data depending on graphical location. A random sample data with 500 observation was created to predict the demand of different gases. This set was used to train the model. Sample data set is created for a region with less population density. Attributes used in this data set are year and the corresponding gas demand in last month. The Results of regression model developed for predicting the oxygen demand is shown in Fig 5. Figure 5. Summary of regression model developed to predict the demand of oxygen gases. Different regression model were build for different commodities and different gases. Generalized equation for commodities prediction is given in (1) P redictedv alues = A LastMon.Demand + B Y ear + C Month + D log(p opulation). (1) 3.2 Medical Inventory Optimization Solution The primary function of Medical Inventory Optimization solution is to allow hospitals to effectively fulfill demand and identify how to gain additional profits. Improved efficiency through effective resource management and optimization lead to an increase in service level, improved performance against patients requests. Manufacturer Distributer 1 Distributer 2 Distributer 3 Hospitals Figure 6. Stages of optimization model for inventory distribution among vendor for a hospitals Hence, an important aspect of efficient inventory management requires that the forecast de- ISBN: 978-1 -941968-20-8 2015 SDIWC4 40
mand of hospital commodities is optimally distributed among the different distribution centers so that the cost incurred by the hospitals is minimum. Hospital medical optimization model can be broadly divided into three stages. The stages are shown in the Figure 6 The members participating in the model are n manufacturers as M1, M2, M3,..., Mn and i distribution centers as D1, D2, D3,..., Di for j products as P1, P2, P3,..., Pj and k transportation costs as T1, T2, T3,..., Tk. The value of each product is V1, V2, V3,..., Vj. The database holds the information about the stock levels as Q1, Q2, Q3,..., Qj of each product and lead time of products at each level. The Lower and upper bounds for each product are calculated based on the regression analysis model explained in Section 3.1. The Optimization is utilized to predict the emerging excess/shortage of stock levels which are vital information to be maintained in the future to minimize the cost and maximize the availability. Multi-echelon model has been used as a building block in the formulation of the entire model. The multi-echelon model will consider uncertainty, non-stationary demand and lead times of the products. The evaluation function is determined for each randomly generated individual. The cost function is given by 2 Cost = Di (T i + V i) Qi (2) Constraints for optimization model are given by 3 Qi <= Qmax, Qi >= Qmin (3) Evolutionary Algorithms (EA) like Genetic Algorithm (GA) is used for on the above cost equation to minimize the cost distributing the cost across various Distributors. Given the non-stationary nature of the demand process, there may be more inventories left over at the end of a period than is desired in the next period. This is because it is assumed that inventories may be reduced only through demand. Minimizing takes the interdependence of decisions between time periods through the use of cumulative supply variables and demand random variables. Figure 7. Authentication page for hospital and vendors. 3.3 A Web Based Portal Design A web based portal is designed to provide easy access to hospitals to order their inventory as suggested by optimization model. Portal will need authentication from hospitals and vendors as shown in Figure 7. After the authentication, hospitals can see the future demand of various commodities on portal as predicted by analytical model and distributed among vendors optimally by optimization model running on hospital server. Predicted demand for various gases is shown in Figure 8. When vendor logs into their account using portal, same demand will be shown to vendor. Vendors will have option to either decline or accept the requirement from hospitals. If vendor declines the request, optimization model will automatically calculate the next best best way to distribute the demands. Vendor page to accept or decline the request is shown in Figure 9. 4 CONCLUSION The Analytic model that was developed, provided a solution how much of the inventory commodities was needed to be order next month. The output from the analytic model allowed eliminating placing orders for commodities whose demand was not expected in ISBN: 978-1 -941968-20-8 2015 SDIWC5 41
Figure 8. Next month demands of different gases predicted by model. Figure 9. Vendor s page to decline or accept the hospital demand. the near future. The objective function and the constraints used to define the parameters for the Optimization Model found the optimal solution by distributing the demand among different vendors by taking their distance from the respective hospitals, cost of supply and storage capacities such that the cost incurred by the hospital was minimum. Finally a web based portal has been designed and proved to be effective and fast tool to order optimum inventory. The benefit of this model in an implementable manner, along with accurate and real-time data can provide a health care system an opportunity to reduce their inventory surplus and allow them to track what the optimal stock levels for different inventory commodities should be. Reduction in storage room and optimization of ordered items provides more visibility in supply chain management and it allows a health care system to be more responsive towards their inventory ordering system. The lead time to locate items was reduced by reduction in storage rooms and the storage rooms that were freed up of inventory can be utilized for other hospital purposes. REFERENCES [1] Raffaele Iannone, Alfredo Lambiase, Salvatore Miranda, Stefano Riemma and Debora Sarno. Modelling Hospital Materials Management Processes. International Journal Engineering Business Management, 2013, 5:15. doi: 10.5772/56607 [2] Varghese, Vijith, Manuel Rossetti, Edward Pohl, Server Apras, and Douglas Marek. Applying Actual Usage Inventory Management Best Practice in a Health Care Supply Chain. International Journal of Supply Chain Management 1, no. 2 (2012). [3] Sarda, Anil N., and Yogesh J. Gharpure. Trends in Inventory Management in health care sector in IndiaIssues & Challenges. [4] Kyong Kim,Yusuf Ahmed,Dr. Alper Murat,Dr. Dean Pichette, Inventory Management and Storage Optimization for Health Care System. [5] Jayaraman B and Barrett JS, A SAS-based Query and Exploitation Interface to Hospital Drug Utilization Data. ACCP Annual Meeting, 2008. [6] Gu, Qiuping, Charles F. Dillon, and Vicki L. Burt. Prescription drug use continues to increase: US prescription drug data for 2007-2008. NCHS data brief 42 (2010): 1-8.. ISBN: 978-1 -941968-20-8 2015 SDIWC6 42