Automation takes on fine details, bigger picture
July 2002 Eric Skjei
Automation is shifting gears.
It is now slowly succeeding in the lab as a modular, task-specific, highly targeted solution,
after passing through an earlier phase emphasizing total lab automation.
Evaluating the impact of an automation system of any scale means
measuring its ability to improve turnaround times, reduce errors,
and justify its investment costs or return on investment. In most
cases, this is a function of its impact on lab FTEs.
"The way we look at lab automation is as a facilitation process,"
says Tom Adkins, vice president, sales and marketing, centralized
diagnostics, Roche Diagnostics. "We don’t view ourselves as selling
lab automation; we try to understand the needs and concerns of the
laboratory and recommend a solution." Facilitating this process
of listening, understanding, and recommending a solution can lead
to different outcomes for different laboratories. Adkins explains,
"Recommended solutions vary widely, ranging from the relatively
modest, focusing on preanalytic sorting or perhaps back-end archiving,
to the much more comprehensive, possibly involving automating an
entire test line or even several lines."
LAS to LIS?
The LIS can make or break an automation project of any scope;
automating a lab, even in a specific, modular way, invariably requires
some degree of interface between the LIS and what is sometimes referred
to as the LAS (lab automation software).
For example, a lab may install a postanalytic automation system
that knows the exact location on the exact rack in the back-end
archiving refrigerator where any given tube is stored, as well as
how much specimen is left in every tube stored there. But the LAS
may not have been designed to share that information with the LIS;
the data may be sitting in the LAS, but not in a way that makes
it readily accessible from the outside. When the LIS receives an
add-on order, it needs to somehow obtain information about the tube,
the specimen, and the storage location to determine whether there
is adequate specimen to perform that test—ideally without
involving a human being.
"And that is the rub," says Hal Weiner, president of Weiner Consulting,
Florence, Ore. "Even today, when many LISs do have their own specimen
inventory management tracking systems, that does not mean that they
can effectively interface to the robotics system to tell it to check
the tube, pick the correct specimen out of the refrigerator, and
put it back on the appropriate instrument." Despite the very real
potential for this particular process to be completely automated,
in most cases human beings still need to be involved. And while
an interface can be created between the LIS and LAS for this specific
purpose, custom interface design is notoriously expensive and not
all that applicable to other installations, given the embryonic
state of the development of standards for exchanging information
between the LIS and LAS.1
In short, even successful targeted, modular automation installations
in the lab may raise a host of secondary implications, questions,
and issues about information exchange with the LIS. Solving these
secondary issues may involve additional, costly customization or
require continued human attention.
The lab’s ability to answer these questions efficiently is strongly
correlated to the age and design of its LIS. Because of their design,
older LISs are much more likely to fall short when asked to interact
with automation software. "Many of the more traditional LIS systems
cannot handle automation correctly," says Bill Blair, vice president,
sales and marketing, SIA Corp., which markets the Molis LIS. "Molis
was designed in the 1990s, and when we visit sites in Europe that
use Molis, the extent to which the LIS and automation system are
integrated is striking." In these sites, Blair reports, it is not
uncommon to see relatively few lab employees and to find that those
who are present typically are attending to exceptions rather than
handling rote chores. One simply doesn’t see a lot of people tied
up at keyboards, accepting normal results. "It’s pretty amazing
to watch an automated lab that is doing 6,000 or 7,000 orders in
a very short period of time and see maybe one or two technologists
at consoles who are only reviewing abnormals because the LIS is
successfully managing all the routine tasks required to keep the
line moving," he says.
Yes, agrees Gilbert Hakim, much of the success of any lab automation
project depends on the capabilities of the LIS. Hakim is CEO of
SCC Soft Computer. "Much of the gain, with respect to automation
in general, comes from LIS-driven autoverification, from the ability
to post results without human intervention. And that gain depends
on the ability of the LIS to tightly capture all the information
produced by the instruments in the LIS." However, LIS autoverification
alone is not adequate to accommodate the complex requirements of
a modern laboratory environment.
"An LIS should also provide a sophisticated rules capability to
properly manage the LAS-LIS-LAS interaction," Hakim continues, "particularly
in high-volume labs handling the work of a large multifacility health
care system. The database needed to handle these robust rules-based
decisions is simply too extensive to be handled by robotics or process-control
software. In a large installation, it may be necessary for rules
to pertain to more than one facility, since patients often move
among different locations, and in some instances, even things as
basic as their medical record number may change as they do.
"In the future, the LAS may reach a point where it can handle
perhaps 60 to 70 percent of this need," Hakim says, "but then it
will encounter other information management requirements, such as
the need to review the last five years of patient history with diagnosis
codes and test results, or accommodate QC across instruments from
different manufacturers at multiple locations. While it may be possible
in some robotics systems to establish a simple rule or two (i.e.
rerun a test), the breadth and depth of database needed to truly
manage patient information is just not there." This is where an
LIS plays a vital role.
Hakim points to several SCC installations that support his perspective.
For example, over the last four years, in a two-stage process directed
by Ralph Dadoun, St. Mary’s Hospital Center, in Montreal, Quebec,
has automated many of its preanalytical processes.2
Today, 80 percent of the 2,400 specimens processed daily at St.
Mary’s require little or no human attention. Turnaround times have
been sharply reduced and exhibit much lower variability than they
did before the automation project. Physician phone calls to the
lab have declined by about 80 percent. Standard productivity measures,
including those promulgated by the College of American Pathologists,
show improvements ranging between 26 and 30 percent. According to
Hakim, a key factor contributing to this success is the capability
of the St. Mary’s SoftLab LIS. "We control the process," he says.
"We actually physically drive the robotics in terms of telling it
how many aliquot tubes we need, what volume we need, what the priority
of the tube is, and so on."
Dadoun notes that much of the benefit of automation pertains to
transition time, a general term for the hand-off intervals that
take place between each step in a process. For example, a technologist
who brings a rack of tubes to the centrifuge but sees that it is
already processing a previous load of specimens and still has six
minutes to run is likely to set the rack down and move on to another
task rather than idly stand by while the current spin cycle finishes.
And having moved on to another task, she or he is not likely to
return to the centrifuge exactly six minutes later—which means
the centrifuge itself will sit idle until the technologist returns.
Automation, by contrast, can ensure this never happens, that the
period of time between the end of an automation device’s cycle and
the transfer of specimens to the next step in the process is never
more than some predetermined transition time—often a minute
or less. In short, in the absence of automation, time is unnecessarily
lost in transitions, which in turn leads to increased variability,
more outliers, and turnaround time problems, lowering all productivity
levels.
The net result at St. Mary’s is that staff time is now mainly
directed at tasks such as moving racks from preanalytical stages
to analyzers and performing occasional reruns or reflex testing;
being liberated from mundane tasks has led to higher job satisfaction
and lower turnover rates among lab employees, reports Dadoun, who
is vice president, corporate and support services for the hospital.
A larger perspective
Focusing on step-by-step, task-specific, modular automation begs
a bigger question: Can automation help solve larger health care
issues? At least one laboratorian is not only raising this question,
but creating real-world tests to try to answer it.
"The bottom line is that by 2007 the cost of medicine is going
to be $2 trillion in this country," says Robin Felder, PhD. "It’s
going to be a staggering figure—essentially unaffordable."
And much of that cost is attributable to the fact that human beings
still do what machines and software can do better. Cost control
may be the benefit that finally makes automation a priority in health
care. "If I was ever sure of anything, it is that medical automation
will become a major focus of the health care industry in the next
five years," Dr. Felder declares. "We have to move from where we
are now, which is essentially a feudal approach to medicine, to
a more integrated approach." Dr. Felder is professor of pathology
and director of the Medical Automation Research Center, University
of Virginia.3
Certainly this is true in the lab. "Many of the cost factors in
the lab, in excess of the cost of the tests themselves, arise in
the handling of specimens, from the time they are taken from the
patient through running them through analyzers and storing them,"
says Dr. Felder. Eliminate human specimen handling and you eliminate
about 70 percent of the cost—and most of the error as well,
he argues.
As the St. Mary’s story demonstrates, much of the productivity
gain from automation is linked to the transitions between steps
in a process. And those transitions, those points where a human
being must hand off something—a chart, a specimen, a patient,
a doctor’s schedule—to the next step in the process, are ubiquitous
in any health care system. Automating the individual components
of the process without looking at the whole process is akin to paving
a cowpath instead of building a new road. Automating isolated pieces
of a process or a facility leaves gaps, handoffs, and bottlenecks—places
where simple, repetitive work must still be done by human beings.
That’s hardly a winning strategy for the long haul.
While manufacturing clearly has much to teach health care about
making this transition, there are unequivocal differences between
the two. "Health care differs from manufacturing in many ways,"
says Dr. Felder. For one thing, medicine does not deal with standardized
components, particularly with respect to patients. Moreover, the
underlying objective of the automation process in medicine is, in
a sense, the diametrical opposite of that of a typical assembly
line. A factory seeks to assemble parts into a whole; medicine,
by contrast, often must first "deconstruct" the already completed
"part," the patient, through processes and techniques ranging from
drawing blood to biopsies.
And in this process, notes Dr. Felder, no patient should feel
that he or she is simply a cog in an impersonal assembly line. "It’s
important to understand that the patient not feel that he or she
is in a factory sense while they are being treated," he says. What’s
interesting is that Dr. Felder sees automation as a key means to
this end. "What they should experience, we believe, in the highly
automated hospital of the future, is a much more personal, higher-quality
interaction, because automation will make it possible for the individual
who is interacting with them, treating them, to have more time to
spend with them and understand their problem."
Automation could, for example, provide medical variations on the
theme of just-in-time resource management. "Right now there is no
single entity that understands where the resources are in a hospital
at any point in time and how they’re being deployed," says Dr. Felder.
"We need to develop the technologies that will track resources,
which could be anything from the doctor, the nurse, and the receptionist
to the blood pump, the x-ray machine, the glucose meter." Such technologies
will also understand and track these resources, including their
capabilities and the information they provide. "Provided with real-time
data on what your resources are, what their capabilities are, where
they are, and what state they are in-active or inactive—the
challenge becomes a solvable process control software problem, that
of trying to deploy resources to obtain data and deliver services
in the most efficient, optimized manner," Dr. Felder says.
Here’s how a solution to that challenge might look. A nurse walks
into a patient’s room and picks up a glucose meter. That event automatically
triggers the glucose meter to turn on and self-calibrate. When the
nurse then carries the meter to the patient and holds it near him
or her, the patient and nurse are automatically identified to the
meter and to one another, without anyone having to enter a single
keystroke, and a glucose reading is taken from the meter. The data
obtained is automatically transmitted to the hospital information
system, appended to the patient’s medical record, and made available
to anyone—physician, pathologist, billing department—who
may need access to it. Because the patient has already been identified,
there is no way to test the wrong patient. If the nurse is not trained
to use the meter, it simply won’t operate. "What you’ve done in
this scenario," says Dr. Felder, "is perform a medical procedure
with a minimum of operator interaction, and you’ve obtained a high-quality
result while at the same time preventing a medical error."
The hospital of the future could, in short, help create an environment
that is more human and humane, not less so, one in which skilled
staff and clinicians are able to work at the pace that best suits
them. "People vary," says Dr. Felder. "Some are very efficient and
can get a lot done in a short time, while others may take longer
but deliver a slightly higher-quality product." It may also reduce
the amount of time patients spend in waiting rooms, rather than
being treated. A patient who appears in, say, the x-ray suite should
expect to find the staff and equipment ready and waiting for him
or her, so precise may be the system controlling the convergence
of patient, resources, and staff.
In 1995, to further explore the application of manufacturing principles
in medicine, Dr. Felder founded the Medical Automation Research
Center. Now employing about 25 people, MARC has transcended Dr.
Felder’s roots in laboratory automation and is now focused on the
larger process of bringing medical automation technology into the
health care system. "We’ve taken the broad view that the laboratory
ought to be a team player in the automated hospital," says Dr. Felder.
"While we spent the last decade or so looking specifically at laboratory
automation, we’re now taking a couple of steps back to ask ourselves
how the automated laboratory will fit into a continuum of automation
throughout the hospital."
Even the very idea of what constitutes a hospital is likely to
change, says Dr. Felder. It is likely to become a much more ubiquitous
part of the community and the environment. MARC is helping create
"smart home" technology for a housing project that will continually
monitor a number of quality-of-life measurements in its elderly
residents. "What we’re trying to do is actually forecast disease,
in the home, by doing a daily assessment," Dr. Felder says. "It’s
an ongoing evaluation of quality of life, which includes not only
health issues, but mobility and social activities." The monitoring
technology has been installed in a home near the University of Virginia
campus and is connected to Dr. Felder’s facility. Underway for about
nine months now, the project is moving into a second phase involving
a 31-unit low-income complex. Evaluating health and quality-of-life
factors in residents may eventually help delay the transition to
assisted care. Ultimately, a key goal is to help shape a viable
health care system by exploring how automation can control costs.
"For every month we can delay the day that our aging population
moves into assisted care facilities or nursing homes in this country,
we save approximately $2 billion," he says.
References
1. Only within the last two years or so did version
2.4 of the HL7 standard first incorporate a chapter (13) on lab
automation. It was intended to facilitate timely, accurate data
exchange between equipment and information systems, and covered
topics such as specimen status, equipment response, and equipment
test code settings. More recently, a joint effort of HL7 and NCCLS
has delivered expanded automation guidelines for HL7 version 3.x
and is continuing this work. See www.nccls.org
and www.hl7.org.
2. Dadoun, R. Case Study: Automation’s Impact
on Productivity and Turnaround Time. Medical Laboratory Observer.
2002;34:36-38.
3. Dr. Felder and MARC are sponsoring a nonprofit conference
in Washington, DC, on March 27-28, 2003 on the issue of medical automation and
cost containment. For more information, see http://medicalautomation.org.
Eric Skjei is a writer in Stinson Beach, Calif.
|
|
|