Ok. Thank you.

Good afternoon, ladies and gentlemen (everyone). I’m very grateful to you all for taking the time to
be here today. Ok. My name is Cresentia Carina. I’m from Universitas Gadjah Mada Yogyakarta. In this
opportunity, I will present my research with Mr. Irwan Endaryanto and Mr Fajar Adi Kusumo which
entitled Hybrid Continuous and Periodic Review Policy Model for Drug Inventory Management.

INTRODUCTION

The motivation of this research is the phenomena stockout and overstock of drugs that often occurs.
For example,

1. In General Hospital Sekayu during January to June 2015 experienced stockout at the same
time for 45 of 800 types of drugs. And It caused the hospital must make emergency order,
which is, the cost is higher than regular order. In addition, the patient satisfaction decreased.

2. Then, During October 2016 – March 2017, Islam Hospital Surabaya was in stockout
condition; about 50,855 drugs were ran out. Also, at the same time They was in stagnant
condition (or overstock condition) with amount 76,254 drugs. It gave impact that the hospital
suffered a great loss.
*The question is, why did all this happen?
General Hospital Sekayu used ineffective management of drug inventory in pharmaceutical
warehouse. They only used one system policy: consumption method. Also, they were lack of human
resources in monitoring inventory.
Islamic Hospital Surabaya were lack of attention to the management of drug inventory. They did’nt
evaluate the current system policy, however the system policy was not appropriate and gave the not
optimal impact.
The inventory system policy affect hospital budget, it might lead great loss for the hospital,
particulary the medication supply., like in Tunisian Hospital during 2020 used decentralised
management system. it resulted the increasing of the stockout of drugs and the inventory
management costs for the whole pharmaceutical supply chain. The efficient inventory system policy
is required to manage the system in a sustainable way, to optimize the stockout of drugs and the
inventory.
Stockouts of drug supply leads to : decrease of patient satisfaction and hospital need to pay extra for
emergency order. On the other hand, the overstock causes : hospital incurr higher costs in terms of
holding costs. More over, leads to the risk of expired drugs. Therefore, It is necessary to have an
appropriate drug inventory sytem policy to make a balancing Supply & Demand ; Service and Cost.
So, this study offer the drug inventory model with hybrid policy to optimize inventory management
cost

Objectives : This study is conducted to

• investigate the characteristics of the hybrid policy drug inventory system with deterministic
demand through mathematical modelling;

• solve the drug inventory model with deterministic demand using Branch & Bound and Kunh-
Tucker method.

There are several previous studies related to hybrid policy model:

• Akin (2010) combined (s,Q) policy and (S,T) policy.

• Roni et al (2014), combined (R,Q) policy with two types of replenishment (regular and
emergency) .
 • Rosales et al (2014) combined a periodic (s,S) policy and a continuous (R,Q) policy also
integrates them with the technology Automated Dispensing Machine (ADM) to minimize
long-run average cost per unit time. Rosales et al worked in two types of demand
(deterministic and stochastic).
Rosales solve the model with deterministic demand by scalar search method

This research refers to the hybrid policy deterministic model resulted by Rosales, but use a different
method. we will investigate the characteristics of the objective function of the model with deterministic
demand. Why is the model with deterministic demand? Because it is quite interesting, we can write the
objective function in integer non-linear problems. So, we will investigate the characteristics of the
objective function and then choose the suitable methods to solve the model. Based on its characteristic,
we can solve this model analytically.

Model Construction :

This is the description of the hybrid policy model discussed in this research. Each hospital ward is
equipped with ADM, which is a storage cabinet capable of recording incoming and outgoing drug
transactions and other information. Drug supplies in ward are filled from the central warehouse with
two types of refill.

In a periodic review, at the time of changing nurse shift or the beginning nurses shift, use the (s,S)
policy. It means if inventory position reach or below the reorder point s at periodic review time, then
a refill is purchased to raise the inventory position up to S.
While, out of the periodic review period, which will be referred to out-of-the cycle, using the (r, Q)
policy. It means that whenever the inventory position is less than or equal the reorder point R at out-
of-cycle, then a refill for Q units is placed.

Model Assumption :

There are some assumption that used in this model.

When periodic order is replaced, a renewal cycle occurs. Each renewal cycle consists of one periodic
refill and 𝑗 times out-of-cycle refill for an integer 𝑗 greater than or equal 0. The length of the renewal
cycle is 𝑘𝜔 for an integer 𝑘 greater than or equal 1.

We define several variables: Tp represent the length of time to exhaust (S-s) unit.

To is the length of time to exhaust Q unit.

Each Tp and To will be divided into 2 parts based on the costs incurred (holding cost or penalty cost)
in that part, so that it becomes Tph, Tb, Toh and Tob.

The figure 1 shows that there is remaining parts that not represent by Tph, Tpb, Toh or Tob. This
model hoped that there will be no this remaining parts. Therefore to accommodates this objective. ,
we add the constraint : k.w = Tp + j.To

Based on the definition of the variable and the model construction, we can write Tph, Tpb, Toh, Tob
as shown in equation (1) and (2). Then we can write s,S,R and Q as shown in equation (3) and (4)

The objective of the model in this paper is minimize the long-run average cost per unit of time (hour).
Due to the inventory position recurs pattern every renewal cycle, so that the long-run average cost
can be calculated for one renewal cycle. The components of the total cost are periodic & out-of-cycle
refill costs, holding cost and shortage penalty cost.
Performance Criteria

The average inventory during the renewal cycle (or I bar) is shown in equation (5), and The average
backorders during the renewal cycle (or B bar) is shown in equation (6). So, we have the objective
function shown in equation (7) subject to the constraint. If we substitute the constraint into the
objective function, we have a new objective function shown in equation (8).

Optimality of Tp and To

The optimal value of function C is found by using the first-order partial derivative and the optimal
value is obtained when To=Tp=k omega/(j+1) so that the C function becomes a function in j and k.

Next, the optimal value of j and k will be searched. Problem (9) can be written in the form of a
nonlinear integer problem. We will solve this problem by using Branch&Bound method. The
relaxation problem of (10) is shown in problem (11). We will investigate whether the objective
function of the relaxed problem is quasi-convex or pseudo-convex function using the Bordered
Hessian matrix test.

B&B Kuhn Thucker

The following is the definition and some properties of quasi convex and pseudo-convex function. The
definition 5 give the definition of Bordered Hessian matrix of a function Then, the theorem 6 stated
we can know the quasi-convexity a function by calculated the all principal minor of Bordered Hessian
Matrix. We use this theorem to investigate the characteristic of the objective function.

Consider that C is twice differentiable continuous on E. The Hessian matrix of C is shown in equation
(14) and the bordered Hessian matrix of C is shown in equation (15).

For all (j,k) in E, the all determinant of the principal submatrix of B at the point (j,k) is less than equal
to 0. Furthermore, the gradient of the function C whose value is 0 is only satisfied when Kp=Ko
(which means contradicts the model assumptions) or for the value of j=-1 (which means (j,k) are not
in E)

Thus, the gradient C is not equal 0 for every (j,k) in the interior of E (interior of open set S).
Consequently, the objective function on the (11) is quasiconvex on E based Theorem 2. In this part,
we already get the characteristic of the objective function.

After we form relaxed problem, we get the nonlinear programming which the objective function is
quasi convex. We will solve the nonlinear program by using Kuhn Tucker. But, what is the guarantee
that the result from Kuhn Tucker is optimal solution? This is the definition of Kuhn Tucker condition
and we have the theorem 8 that stated If the objective function and constraint function are
quasiconvex; and there is lambda and x* such that satisfied Kuhn Tucker conditions also the first
partial derivative in x* is not equal 0, then x* is the optimal solution.

Consider the objective function and constraint function on Problem (11). We already know that C is
quasiconvex and the first partial derivative in (j,k) is not equal 0 for all (j,k) in the domain of C. The
constraint function is also quasi-convex. It can prove by using the definition. So, if we can find the
lambda and x* such that satisfied the Kuhn Tucker Conditions, then x* is the optimal solution of
Problem (11).

And then we use the Branch&Bound algorithm to find the solution of Problem (10), until the fathoming
condition is satistifed.

Example Cases
We provide calculation of some example cases. The example case use different parameters, there
are the demand rate (lamba)=30,50,60, (penalty cost) b=5,9, omega =4,8,12, periodic refill fixed
cost=24,48 and out-of-cycle refill fixed cost=48, 80 but we use a fixed holding cost per unit per hour h
= 1.
So, the total example cases are 36 scenarios.

Based on Table I,
• the optimal s and R have the same value as a result of T p = To.
• j represents the optimal number of out-of-cycle refill and kw represents the optimal length
renewal cycle.
• the number of out-of-cycle refill increases with the value of w; the optimal length renewal
cycle and the length time of periodic review have the same value.
• And then the optimal long-run average cost per hour increases with the value of w, b and
fixed cost refill.

Conclusion
Here we are in the final part. The conclusion of this research. First (baca) :

Closing

That’s the end of my presentation. I hope you enjoyed it. If there is any question please feel free to
ask. Thank you very much.