Antibiotic alert—labs, pharmacies take on cost and misuse
From tiny Randolph County Medical Center in Pocahontas, Ark., to the Mayo
Clinic, health systems are wrestling with how to control antibiotic costs, which
can consume one-third of a typical hospital’s drug budget, and curb antimicrobial
In large part, they have attacked the related problems by asking the microbiology
laboratory to provide timelier data on bacterial cultures and response to antibiotic
therapy. The closer collaboration between microbiology, hospital pharmacy, and
clinicians is made possible by software programs, either vendor-supplied or
home-grown, that trigger computerized alerts when an antibiotic may be needed
or is being used inappropriately.
Typically, a pharmacist monitors the alerts and makes recommendations to the
clinician, although in some hospitals a clinician is the intermediary. The results
of such programs are not only lower antibiotic costs but also shorter lengths
of stay and, longer term, a lower likelihood of bacterial resistance to drug
Rural hospital use
Randolph County Medical Center, a rural community hospital with 50 licensed
beds, invested nine years ago in the bioMérieux TheraTrac 2 system, which sends
microbiology results to the pharmacy, along with "alarms" that warn when a patient
has an infection, when a bacterial culture is not responsive to the initial
antibiotic, or when dosing may be too high or too low. The pharmacist then relays
the information to the prescribing clinician, with a recommendation for therapy.
"TheraTrac 2 brings together all the pharmacy and patient information and
combines that with the lab side," says Mike Broyles, PharmD, director of pharmacy
and laboratory. As lab information such as type of bacterial culture or patient’s
renal function becomes available, "I can be notified by on-screen alerts, printer,
fax, pager, or e-mail."
With all the data in one place, he says, he can "formulate a plan [for therapy],
communicate it to the physicians, and follow up by tracking associated outcome
results, such as drug costs and length of stay." Before TheraTrac, he might
have performed one intervention per hour. "Now," he says, "I do six or seven
in 45 minutes." Since user-defined custom alarms are so easy to create, Broyles
can monitor all aspects of patient therapy.
The system works this way: A patient who comes in with a diagnosis of pneumonia
(DRG 79) might be prescribed two grams of Rocephin (ceftriaxone) a day. As soon
as the patient comes in, the physician orders a culture report from the lab.
It turns out the organism is resistant to Rocephin. "I’d get an alert on my
screen saying the therapy needs to be changed," Broyles says. "We would make
a recommendation to put the patient on another antibiotic."
Through an interface with the laboratory information system, TheraTrac can
deliver, for example, the culture results, including minimum inhibitory concentrations,
and the serum creatinine results, which can be an indicator of inappropriate
dosing. If the MIC of an organism is too high, the antibiotic needs to be changed.
If it’s low, the dose can be adjusted downward to reduce cost and the risk of
All this can save the hospital anywhere from a few dollars a day on antibiotic
costs to several hundred dollars a day or more for patients who can be discharged
earlier. Before purchasing TheraTrac 2, Randolph County Medical Center’s average
length of stay for DRG 79 pneumonia was 9.3 days and the cost of drugs per adjusted
pharmacy day was $93. Today, those figures are 3.9 days and $42. "Studies have
validated the significance of inappropriate antibiotic therapy," says Broyles.
In picking an effective agent, they look at each patient individually "to optimize
therapy and dose to improve outcomes, reduce selection of resistance, and lower
costs," he says. The aim is to select the most specific, narrow-spectrum antibiotic
Broyles says the physicians are never left out of the process. "My goal is
to be a resource for them," he says. Physicians respect laboratory data, he
adds. "When you provide them with good information, they will respond."
Mayo Clinic’s CBAM
On the other end of the size spectrum, the Mayo Clinic has developed an in-house
program called CBAM (computer-based antimicrobial monitoring) that’s used for
inpatients at two Mayo Clinic hospitals, which together have 1,500 beds, in
The major objectives of the CBAM program are twofold: to improve patient care
by detecting suboptimal antimicrobial therapy early and to reduce inpatient
antimicrobial costs by identifying the most therapeutically effective and cost-effective
antimicrobial. The system is not designed to be comprehensive. "It focuses selectively
on detecting clinical scenarios that occur frequently and have potential for
excess drug expense, drug toxicity, or poor patient outcome—for example,
untreated bacteremia," says John Wilson, MD, an infectious disease specialist
at Mayo Clinic.
The system’s chief components are a set of algorithms that define suboptimal
antimicrobial use based on consensus expert opinion, as well as pharmacy, laboratory,
and microbiology databases that are updated continually. "About 35 algorithms
are processed each day for every hospitalized patient at Mayo," Dr. Wilson says.
Each algorithm identifies clinical conditions that, when met by a patient, warrant
review by an infectious disease specialist or pharmacist specializing in antimicrobial
CBAM doesn’t create data. Rather, it integrates data from multiple databases
into physician- and pharmacist-derived algorithms to help guide antimicrobial
therapy. CBAM would detect, for example, excessive IV antimicrobial costs, suboptimal
therapy for serious infection, treatment of serious infection without adequate
cultures having been obtained, inadequate monitoring of nephrotoxic antimicrobials,
and situations where an oral antimicrobial might be substituted for an IV antibiotic.
"If there is a clinical situation where antimicrobial usage or management
of infection is suboptimal, the patient and algorithm are identified for clinical
review," Dr. Wilson says. Recommendations are then communicated to the physician
caring for the patient.
Pharmacy antimicrobial specialists commonly review, for example, these algorithms:
antimicrobial drug-to-drug interactions, conversions from IV to oral therapy
for enterically well-absorbed antimicrobials, and inappropriate or antagonistic
antimicrobial combinations. ID physicians commonly review the following: suboptimal
antimicrobial therapy, dosing infractions, inadequate monitoring of potentially
nephrotoxic drugs, inappropriate antimicrobial coverage or expense, and suboptimal
treatment of serious infections.
Both the ID physician and pharmacist communicate recommendations to the primary
physician caring for the patient. But only the ID physician formulates therapeutic
recommendations and decides whether a formal ID consult might be needed in the
more complex and difficult cases. "Computer-generated data is no substitute
for clinical patient evaluation at the bedside," notes Dr. Wilson. "We don’t
want to place our pharmacists in the uncomfortable position of making difficult
Not every patient condition identified by CBAM requires that an ID physician
or pharmacist intervene. For example, a blood culture with coagulase-negative
staphylococci growth from a patient without a prosthetic device or indwelling
intravenous catheter may not necessarily represent a bloodstream infection.
"Often this organism represents a skin contaminant and not infection," Dr. Wilson
says. "This is where clinical decision-making and medical judgment based on
the patient’s condition are important."
The software is developed in-house by a Mayo Clinic team, and the user interface
is Web-based and written using Microsoft Active Server Pages. The database is
Microsoft SQL Server. Microsoft Visual Basic is used to write the business logic,
CBAM has been used since 1994, and since then use of intravenous vancomycin
at the two hospitals has dropped by more than 50 percent. The cost of the top
20 antibiotics at the two hospitals declined by an average of $430,000 a year,
saving $2.15 million in cumulative costs between 1993 and 1998.
Baylor All Saints Medical Center, which operates two hospitals in Fort Worth,
Tex., is a more recent convert to linking microbiology and pharmacy data. It
purchased TheraTrac about a year ago.
"We wanted to be able to get microbiology rapid reporting into the pharmacist’s
hands as quickly as possible," says Angela Pinckard, PharmD, assistant director
for clinical pharmacy services. The pharmacist then suggests changes to physicians
Physicians would not change prescribing patterns solely on the basis of cost,
Baylor knew, but they do respond well to laboratory data. For instance, it has
customized TheraTrac to produce an alert for serum creatinine levels above 1.4
mg/dL. "We tell the physicians that the dose might be too high because of poor
renal function," Pinckard says.
Using TheraTrac, Baylor set up an interface between the microbiology and blood
culture labs, which use the Misys system, and the pharmacy information system.
"We had to work closely with the lab to make sure we understand the clinical
subtleties of what they’re sending us," she says. For example, while an initial
result on a culture may signal resistance, a confirmation test for methicillin-resistant
Staphylococcus aureus is often required before suggesting an antibiotic
Using the tool has improved collaboration between the pharmacy and the lab,
Pinckard says. "If something doesn’t fit our assumptions, we call microbiology
and talk to them before we make a suggestion for change," she says. "This has
opened up new communication between the pharmacist and the lab technologist."
Gayle Sondecker, MT(ASCP), SM, associate laboratory director and former microbiology
supervisor, agrees. In fact, she says, it was the microbiology laboratory staff
who recommended that the pharmacy buy the TheraTrac system. "The goal is to
take patients off expensive antibiotics when they don’t need them, and to get
them on antibiotics when they do," Sondecker says. The laboratory’s workflow
hasn’t changed, she says, and laboratory data are now better used.
TheraTrac can not only track the cost of an antibiotic but also the impact
on overall hospital costs. For six months, Pinckard has been tracking the cost
savings related to labor, drug, and total cost of care. For 2,500 interventions,
"we averaged about $21 in savings per intervention," she says.
TheraTrac is now being used for other high-cost therapies such as colony-stimulating
factors. By looking at laboratory data such as neutrophil counts, "we can get
that medication stopped once it’s no longer needed," she says.
Like the Mayo Clinic, Health Midwest, a 12-hospital integrated delivery system
in Kansas City, Mo., has developed its own approach to antibiotic prescribing,
led by L. Patrick James, MD, medical director of integrated services (lab, radiology,
pharmacy, respiratory services, case management, and access management).
The integration started, Dr. James says, when Health Midwest consolidated
its laboratories in 1995 and 1996 on Cerner PathNet, and used that step to look
at how best to integrate clinical information with other departments such as
pharmacy. Beginning in 1999, culture sensitivity results were collected from
all the hospitals and fed directly to clinical pharmacists.
The next step, in 2000, was developing an antibiotic formulary. "We looked
at best practices and sensitivity results and tried to identify the most cost-effective
agent given a particular sensitivity," says Dr. James.
Kathy Chase, PharmD, clinical pharmacy director at Health Midwest, says the
impetus for the formulary came from the microbiology lab. As the labs consolidated,
"they noticed there were different panels [of antibiotics] in every hospital.
The microbiologist said it would be good to have a standard formulary," she
Pharmacy and microbiology worked with the system’s eight infectious disease
physicians to come up with an antibiotic list, based on sensitivities, usage,
and cost. Clinicians resisted at first, she concedes. "Anytime you try to change
drugs, you take away someone’s favorite." Eventually, the pharmacy’s outcomes
data convinced them.
With a formulary in place, "we went to tiered reporting," she says. That is,
when microbiology identifies an organism and its sensitivity, the only report
the physician sees is for the acceptable antibiotic. "Within the LIS we built
a series of rules that said if it’s a gram-positive organism susceptible to
a first-generation cephalosporin, that’s the information we will provide, which
ends up in the patient’s chart," Chase explains. However, the attending physician
can override that if, for example, the patient is allergic to a particular drug.
Antibiotic costs at Health Midwest dropped from $3.1 million in 1999 to $2.36
million in 2001, according to Chase. "Not only have we been successful from
that standpoint," she says, "but we have not seen an increase in resistance."
At first, physicians feared that by limiting the number of available antibiotics,
resistance might increase. Instead, using narrow-spectrum (and less costly)
antibiotics wherever possible has kept resistance at bay. Limiting antibiotics
also meant Health Midwest could buy the approved ones in greater bulk and was
able, therefore, to negotiate better prices from vendors.
It is the clinical pharmacists who intervene on antibiotic therapy. "Since
we have the microbiology core lab across the system, all results are on the
same database," Dr. James says. Each morning, the pharmacist pulls up laboratory
findings on culture and sensitivity, and the same information goes to the attending
physician. "If we’re seeing any problems, the pharmacist can intervene with
the physician," he adds. For instance, the physician might be using an unnecessarily
expensive or inappropriate antibiotic.
The pharmacist does not overrule the initial order; he or she recommends that
the physician make a change if there’s a discrepancy with the formulary. However,
as Health Midwest moves to electronic order entry, "down the road the pharmacist
will have the ability to change those orders," after informing the physician,
"With a standard antibiotic formulary, we have cut down on what’s available,"
says Dr. James. Physicians have pushed back a bit, he says, but the data have
convinced most of them. "Our resistance patterns have gotten much better since
we initiated the formulary," he says.
A clinical pharmacist and PhD microbiologist make rounds together on antibiotic
therapy. "Having one infectious disease group for the whole system helps, too.
The microbiologist and his colleagues are very data driven," Dr. James says.
Health Midwest is now expanding the integration of lab and pharmacy data to
expensive biotech drugs and anticoagulants. "We want to better utilize these
agents based on lab data," Chase says. For instance, the immature reticulocyte
fraction can help nephrologists determine whether a patient is responsive to
epoetin alfa, an expensive agent used to treat patients with chronic renal failure
or on chemotherapy.
Beyond the formulary
Franciscan Health System, a three-hospital system in Tacoma, Wash., already
had an antibiotic formulary but wanted to make sure it was using the most cost-effective
drugs. "We were very concerned about using the best presumptive therapies, to
lessen resistance and cost," says pharmacy manager J. Kelly Martin Jr., RPh.
Franciscan selected TheraTrac because it incorporated patient demographic
data with pharmacy and microbiology information. "We’ve set two standard alarms
with TheraTrac," Martin says. The first is where a patient has a culture that’s
positive but therapy is not prescribed. This culture reading can come from routine
analysis of blood, urine, or wounds. "If we get a positive culture, the system
recognizes there’s no antibiotic covering it and sends an alarm." Typically,
the pharmacist gets the alarm and notifies the physician, and they decide whether
the positive culture is one that requires treatment. Generally, "we get about
six to 10 of these alarms every 24 hours," he says.
The second alarm is for a resistant isolate—the patient has been prescribed
an antibiotic but the bug is resistant. "We’ll suggest a change to more effective
therapy," Martin says. In some cases that will mean moving from a broad-spectrum
to a narrow-spectrum antibiotic, once the bacterium has been identified. The
system generates two to five of these alarms every 24 hours.
"All antibiotic orders take communication and interaction between the pharmacist
and clinician," Martin says. "Those discussions give us an opportunity to focus
on patients at risk for problems." In most cases, the physician accepts the
pharmacist’s recommendation, but patient allergies or other information the
pharmacist doesn’t have could change that.
In February alone—a typical month, Martin says—Franciscan saved
$13,400 on therapy changes recommended by pharmacists based on laboratory data,
either by shifting to a lower-cost antibiotic or eliminating a redundant one.
That doesn’t include the potential savings that came from starting the patient
on an antibiotic earlier and perhaps moving the patient out of the hospital
The microbiology lab helped select the system and develop parameters for how
it works. "Microbiology and pharmacy have been very cooperative," Martin says.
Microbiology supervisor Claudia Willis, MT(ASCP), agrees. "This is a good
system because it doesn’t impact our workflow, because the computer automatically
gets the results to pharmacy and alerts the pharmacist if there’s a problem."
Occasionally, an anomalous result will trigger a phone call to the pharmacists
telling them "to throw that one out, and we’ll rerun," she says.
Enlisting the medical record
Brigham and Women’s Hospital in Boston uses internal software to tie together
lab data and pharmacy through the patient’s medical record.
"Our lab operates on a paperless format," says Andrew Onderdonk, PhD, professor
of pathology and director of clinical microbiology at Brigham and Women’s. "As
soon as we have a culture result, it’s available in real time as part of the
patient’s medical record."
Pharmacists can examine the data and make a recommendation on an antimicrobial
agent. Generally, however, once the laboratory comes up with its susceptibility
data, it’s clear what agent should be tried first. That’s because data on agents
that aren’t included on the hospital’s formulary are suppressed, Dr. Onderdonk
"The susceptibility profile is run based on our formulary," he says. "We don’t
include information on other agents that might be used." But if a physician
requests the information, "we will make it available." This additional information
is rarely requested, he says, and the drug of choice is the formulary selection.
Occasionally, a patient allergy might result in an alternative.
The laboratory also makes its antibiograms available on the computer system,
so that physicians or pharmacists who want to know, for example, how sensitive
E. coli is to a particular antibiotic can find out. Specialists have
antibiograms tailored to their services, Dr. Onderdonk says. "If we’re trying
to track a resistant organism, we have software that allows us to do that by
site or by organism," and those data are also available through the internal
information system known as BICS.
Bill Churchill, MS, RPh, director of pharmacy services, says the clinical
pharmacists make rounds with physicians and "routinely check cultures and sensitivities
to see if anything more efficacious and less expensive" could be used.
But it isn’t just software that makes the system work. Much of the credit,
he says, goes to the "good relationship between pharmacy, the clinical lab,
and our infectious disease service." Thanks to a strict formulary and cooperation
between departments, Brigham and Women’s has kept its antibiotic costs below
20 percent of the drug budget, Churchill says. "That’s very reasonable, given
what we could be spending."
Karen Southwick is a writer in San Francisco.