 Computer Simulation Modeling Of Laboratory Processes: A
Powerful Tool For Process Reengineering
|
Michael S. Gannon - Director, BCI-Clinicon Consulting Group,
Beckman Coulter, Inc.
Labor is, by far, the largest expenditure in
a modern laboratory's budget. It is also the most significant
source available for cutting costs. To economically justify a
reengineering project, it must show that it can generate
sufficient savings in labor costs to justify the laboratory's
investment in time, effort and money. The most accurate way to
measure the potential labor cost savings associated with a
project is to construct a model of the new operation and then
determine the level of staffing required to operate the model
under a given set of constraints. In a diagnostic laboratory,
these constraints are related to quality of service.
Laboratory operational models that are able
to predict manpower requirements with any degree of accuracy are
difficult to construct using static analytical methods. Workflow
is complex, involving specimens that cross process boundaries.
Processing is conditional on specimen type, quality and the test
orders attached to them. In an integrated operational model,
resources are shared by different processes. Labor effort and
cost drivers change at different stages of the processing cycle.
For example, the number of orders and the patient status (known
or unknown) on a test order form are primary labor drivers at
order entry. During specimen processing, the drivers are the
type of specimen, its condition and its processing requirements;
during analysis, it's the number and types of tests ordered.
Measuring the impact of changes in an
operational model on manpower utilization requires an analytical
methodology that permits the examination of behavior under
varying degrees of workload stress. Such a methodology is
available in computer simulation, which allows for the dynamic
analysis of the behavior of a system over time. Combined with
appropriate analytical methods, computer simulation can allow an
analyst to seek the optimal level of resources required to
deliver service given a certain set of constraints.
BCI-Clinicon uses advanced modeling and
optimization to study complex systems requiring sensitivity
analysis and goal seeking. Sensitivity analysis is required when
the effect of a variable, such as resource levels, on the
quality of the system's output must be determined. Goal seeking
allows an analyst to determine the minimum amount of a resource
required to guarantee a certain level of system performance.
The modeling process begins by importing a
dimensioned floor plan of the proposed laboratory into the
layout module of the simulator. Individual instruments and
workstations are defined as locations where processing takes
place. The processing steps involved and their labor
requirements must then be defined, including any conditional
processing logic that relates to the types of order forms,
specimens or tests that are processed at the location. Resources
that will perform the processing are then assigned to the
processes.
A resource is defined by its availability and
cost of usage. Resource availability is specified by assigning
individual resources to shifts and processes. For example, a
daytime technologist on weekdays assigned to automated testing
would be unavailable during his or her off-shift hours, lunch
break and rest break periods. The person would also be
unavailable for processes not related to automated testing.
Cross-coverage rules are used when resources
may be shared by more than one process. Such rules define in
what order resources are "captured" by specimens. Such
a rule might take the following form: "Capture the next
available technologist assigned to automated hematology first.
If unavailable, capture the next available floater." The
primary resource cost used is the fully-loaded cost (base hourly
salary plus benefits) of a compensated man-hour for the
resource.
Finally, entities and their arrival patterns
are defined. "Entities" are the things that get
processed and their attributes. In the simulation "world
view", it is the entities and their attributes in the model
that drive cost in the system. This is because they must
"capture" costly resources in order to be processed.
In the laboratory the principle entities are order forms,
specimens and tests. The principle attributes vary from entity
to entity. For a specimen these include specimen type, priority
and condition.
During a simulation entities enter a system
according to a defined arrival schedule that matches the known
or estimated workload of the laboratory. According to their
type, priority and condition, they capture the necessary
resources required for their processing cycle. When an entity
enters a location for processing and a resource is unavailable,
the entity enters a wait state until the next enabled resource
for the process becomes available. Waiting time is calculated
separately from processing time. In addition, the reason for the
wait state is captured. Wait states can be caused by the
unavailability of a resource, the unavailability of a location
which is at capacity or the occurrence of an event that blocks
the entity's passage to an available location. These statistics
are helpful for capacity planning or to pinpoint the cause of an
unacceptable turnaround time.
As the entity is processed, it collects
processing time and processing cost according to its usage of
resources. This allows the analyst to capture the cost of
processing for a given entity based on the processing activities
and resources it consumes. At the end of its processing cycle,
the entity's transit (turnaround) time is calculated as well as
its total processing cost.
As entities capture and consume the available
time of a resource, resource utilization is calculated. Utilized
time is compared to available time in order to calculate idle
time.
One of the key factors in preventing a model
from becoming intractably complex is to determine what level of
detail is required in specifying the activities of a resource.
In a laboratory it is impossible to capture all the activities a
resource may be engaged in. Time spent conferring with
colleagues , answering calls or assimilating skills are required
activities but would be difficult and unnecessary to model in
detail . It is important, however to make allowances for these
activities. These activities are accounted for without making
them explicit in the model by adjusting the availability of the
resource. For example: A dayshift technologist is available for
7.5 hours. He spends 15% of his available time engaged in
required activities that are not explicitly modeled. The
simulation reports the resource's utilization at 50%. In order
to account for the activity that is not explicitly modeled, the
analyst upwardly adjusts the reported utilized time to 65% and
reduces the idle time by an equivalent amount.
The cost of the resources consumed by the
processes is assigned to the individual processing activities.
This allows the cost of each individual activity to be
determined according to its utilization of resources. This is
important information because it results in a cost model that
allows laboratory management to determine the economic impact of
making changes to the process with a high degree of accuracy.
In order to determine an optimal staffing
plan for a laboratory, the analyst examines the utilized and
idle times of individual resources and adjusts them upward or
downward in a series of simulations to determine the minimum
number of resources required to deliver service within a given
set of constraints such as maximum acceptable turnaround time or
maximum acceptable error rate. BCI-Clinicon utilizes a powerful
mathematical optimization program which effectively automates
this exhaustive process through the application of an
evolutionary genetic search algorithm. The algorithm generates a
number of possible solutions and then examines them. The best
solutions are retained and are used to generate better solutions
in a process very similar to natural selection in the biological
world. The number of solutions is gradually narrowed down until
a best (optimal) solution is found.
The economic impact of change is determined
by translating the simulation output into an economic model and
comparing it with the economic model of the current operating
model.
|
