Part II:Process Improvement - Saving $500K utilizing iGrafx and Design of Experiments

In Part I of this article we showed how utilizing iGrafx we could quickly map the business process flow, and begin to simualte the differences between the current system and the proposed new system.

In this week's article Part II we will look at the Factors, and create a design of experiments to analyze the process alternatives with regard to cost, cycle time, and Rolled Throughput Yield.

To recap:

• There is an opportunity to install a vault dispenser for medications on each floor, and remove the satellite pharmacies. The medications would be stored on each floor of the hospital in the specialized dispenser.


• The pharmacy would be contacted for a prescription, but rather than dispensing a medication (med for short), they would provide the nurse an authorization code to access the proper bin in the vault and extract the med.


• The pharmacy would keep status of each dispenser real-time and update each dispenser (4) times daily

Should the hospital invest time and money in improving the current process (e.g., by improving the chance the satellite pharmacy has the needed med), or by implementing the proposed Vault System.

Model Design

The model has two distinct processes. This is done to allow for direct comparison in the experiments. The left path is the current satellite prescription (Rx), or "As-Is", process. The right hand path is the proposed process utilizing the vault system.

The new process only utilizes components of the current process when the medication is not available in the vault dispenser. In that case, the main Rx is used to fill the order.

Responses

The responses of interest are:

• Time: How can we get the med to the patient most quickly; i.e., what is the shortest cycle time? Which process, current or proposed, has the shortest cycle time?


• Cost: Which process, current or proposed, is the lowest cost?


• Defects: What is the Rolled Throughput Yield (RTY) of each process? Each defect endangers patients, potentially in a life-threatening way; so it is critical to minimize defects.

Factors

The factors of interest are:

• Which process? Does the Satellite Rx, or the Vault, process provide the best responses?


• Immediate Med Availability?What's the impact of immediate med availability on the current and proposed process? The Satellite Rx method currently has an 85% chance of having the med. What effect would increasing the availability of the med in the Satellite Rx to 95% have? How does that compare to an 85% or 95% chance that the Vault has the med available?

Utilizing iGrafx we design the process follows as can be seen by clicking here.

Now we need to simulate the process for the Design of Experiments. Initially we used interview estimates of how long do certain processes take. For example it was estimated that activity #8 took beetween 1 and 4 minutes. But how can we more closely approxiamte this?

Doing a quick motion study we record 30 occurences of both the To-From Satellite process times and the Satellite pickup times. We then utilize iGrafx in conjunction with Minitab to determine the best approximate distribution.

As can be seen in the photo below we approixmate the distribution for the To-From Satellite times from the data.



The A-D Measure is the Anderson - Darling measure. The lower the A-D the better fit of the distribution to the data. From the tool we see that the best approximation for the observed data is the normal curve with the following parameters: Mean of 2.578 and Standard Deviation of 0.7940462.

Likeise doing the same for the satellite pickup time gives us the following.



Which shows that the best approximation is the Weibull distribution with the following parameters Shape of 8.988834 and Scale of 2.652587.

We now take those values and replace the previous interview estimates of task durations witht these distribution curves.

Utilizing this data we want to create a design of experiments that provides us answers to the following questions:

1. Which process do we utilize the current process or the new vault; and


2. What availability level should we have the medicine 85% or 95%.

We use a 2-Level full factorial design. This means that we take four measurements of the cost, average cycle time, and rolled throughput yield (RTY) of each setup of the factors at different levels: Vault at 85%, Vault at 95%, Satellite at 85%, Satellite at 95%.

Typically this would be very expensive and time consuming to do in real life, but since we are utilizing iGrafx to simulate the process and Minitab to track the outputs it does not have a cost nor time impact.

After running the simulations the results look as follows ordered by least cost:



As you can see from the results the best least cost solution is running the Vault Dispenser at 95% for a cost of $570K with an RTY of 99.975% and an average cycle time of 303 as opposed to running the Sateille at 95% at $1,087,504 with an RTY of 99.498% and an average cycle time of 495.

As you can see by creating a design of experiments we can quickly simulate the annual impact and identify the proper settings for the factors of the business process.

This translated into $500K less costs per operating year and a reduction of processing time by 40%.

Hawaii Business Consulting is the Value Added Reseller of in Hawaii. If you need to be able to quickly analyze your business processes and identify new solutions, iGrafx is the software package you need. It will enable you to quickly model, simulate, and analyze your data similar to what we have shown in the article. The software starts at $500 and that includes an initial training session with Hawaii Business Consulting.

Give us a call at (808) 224-0470 and we will create a free high-level model of your business process and identify ways in which your business could be saving money through process improvement.