Development and Impact of a Computerized Pediatric Antiinfective

Decision Support Program

Charles J. Mullett, MD, PhD*‡; R. Scott Evans, PhD*§储; John C. Christenson, MD‡; and
J. Michael Dean, MD, MBA‡

E
ABSTRACT. Objective. Computerized medical deci- rrors in prescription, distribution, and admin-
sion support tools have been shown to improve the qual- istration of pharmaceutical therapy are a sig-
ity of care and have been cited by the Institute of Med- nificant cause of injury to hospitalized patients.
icine as one method to reduce pharmaceutical errors. We Almost 2% of admissions in an adult, tertiary care
evaluated the impact of an antiinfective decision support university facility experienced preventable adverse
tool in a pediatric intensive care unit (PICU).
Methods. We enhanced an existing adult antiinfec-
drug events, costing an average of $4700 per epi-
tive management tool by adding and changing medical sode.1 Other investigators have published medica-
logic to make it appropriate for pediatric patients. Pro- tion prescription error rates of 3.99 errors per 1000
cess and outcomes measures were monitored prospec- orders in an adult hospital.2 Factors associated with
tively during a 6-month control and a 6-month interven- errors included a decline in renal or hepatic function
tion period. Mandatory use of the decision support tool (13.9%); a history of allergy to the same medication
was initiated for all antiinfective orders in a 26-bed PICU class (12.1%); the use of the wrong drug name, dos-
during the intervention period. Clinician opinions of the age form, or abbreviation (11.4%); incorrect dosage
decision support tool were surveyed via questionnaire. calculation (11.1%); and atypical or unusual dosage
Results. The rate of pharmacy interventions for erro-
neous drug doses declined by 59%. The rate of anti-
frequency (10.8%). Studies in children show similar
infective subtherapeutic patient days decreased by 36%, findings.3,4 Folli et al3 found that pediatric patients
and the rate of excessive-dose days declined by 28%. The who are younger than 2 years or require treatment in
number of orders placed per antiinfective course de- a pediatric intensive care unit (PICU) were at the
creased 11.5%, and the robust estimate of the antiinfec- greatest risk. In their study, dosage errors were the
tive costs per patient decreased 9%. The type of anti- most common medication error, with overdosage
infectives ordered and the number of antiinfective doses exceeding underdosage in frequency. Antibiotics
per patient remained similar, as did the rates of adverse were the most common class of pharmacotherapeu-
drug events and antibiotic-bacterial susceptibility mis- tics with errant orders.
matches. The surveyed clinicians reported that use of the
Technological solutions have been shown to have
program improved their antiinfective agent choices as
well as their awareness of impairments in renal function an impact on error rates. Computerized physician
and reduced the likelihood of adverse drug events. order entry systems for prescription of medications
Conclusions. Use of the pediatric antiinfective deci- enable the presentation of standard dosages, auto-
sion support tool in a PICU was considered beneficial to mated dosage calculations, and presentation of clin-
patient care by the clinicians and reduced the rates of ically important drug– drug, drug–allergy, and
erroneous drug orders, improved therapeutic dosage tar- drug–laboratory interactions at the time of the phy-
gets, and was associated with a decreased robust estimate sician’s decisions.5–7 In addition, these systems elim-
of antiinfective costs per patient. Pediatrics 2001;108(4). inate the inherent problems associated with hand-
URL: http://www.pediatrics.org/cgi/content/full/108/4/
writing interpretations and retranscription of orders.
e75; antiinfective agents, decision support systems, drug
therapy, medication errors, child, infant. Not surprising, the rate of serious medication errors
dropped 55% in one analysis of the impact of the
implementation of a physician order entry system in
ABBREVIATIONS. PICU, pediatric intensive care unit; IHC, In- a large tertiary care center.8 Crude estimates by those
termountain Health Care; PCMC, Primary Children’s Medical
Center; HELP, Health Evaluations through Logical Processing;
authors suggested that the net savings to the hospital
STICU, shock-trauma intensive care unit; LOS, length of stay. through fewer adverse events and their sequelae
amount to $5 million to $10 million per year.
Clinicians must strike a balance when choosing the
initial antimicrobial care. Appropriate empiric anti-
From the Departments of *Medical Informatics and ‡Pediatrics, University infective therapy improves outcomes, whereas un-
of Utah; §Department of Clinical Epidemiology, LDS Hospital, and 储De-
partment of Medical Informatics, Intermountain Health Care, Salt Lake
necessarily broad therapy puts the patient at risk for
City, Utah. the development of resistant organisms, adverse
Received for publication Mar 15, 2001; accepted May 22, 2001. drug events, and increased cost. Investigators have
Reprint requests to (C.J.M.) Robert C. Byrd Health Sciences Center, West sought to develop computer-based tools to facilitate
Virginia University, Box 9214, Morgantown, WV 26506. E-mail: cmullett@
hsc.wvu.edu
the clinician’s decision-making process. The first
PEDIATRICS (ISSN 0031 4005). Copyright © 2001 by the American Acad- work in this field was by Shortliffe and col-
emy of Pediatrics. leagues9 –11 with the MYCIN rule-based infectious

http://www.pediatrics.org/cgi/content/full/108/4/e75 PEDIATRICS Vol. 108 No. 4 October 2001 1 of 7

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 disease expert system. Chung and colleagues12,13 Development of the Pediatric Antiinfectives
used a statistical approach to publish a monthly list Management Program
of most likely organisms and effective therapies by Using the adult version from IHC as a template facilitated the
culture site. Evans et al14 at Intermountain Health task of developing the pediatric antiinfectives management pro-
Care (IHC) developed the first tool that automated gram. Although the adult edition could be run at PCMC on
pediatric patients, its advice usually was inappropriate and some-
the data-gathering process and calculated the most times harmful. Thus, 2 of the authors (C.J.M. and J.C.C.) reviewed
probable effective therapy at the time the physician each adult rule governing recommendations for infectious ill-
was choosing the antibiotic. Later enhancements to nesses and pathogen culture results and identified which would
this tool included drug dosage selection assistance, be safe and beneficial to keep in the pediatric edition. New pedi-
atric-specific empiric antiinfective therapy logic was developed
renal function and microbiology results monitoring, and incorporated into the program while maintaining the same
and reviews of antibiotic costs and bacterial sensitiv- framework and overall “look and feel” of the adult tool. For the
ity patterns. In adult patients, this computerized an- antiinfective doses, pediatric pharmaceutical texts20,21 were con-
tiinfective decision support program reduced antibi- sulted first and the list of candidate dosages then was reviewed by
otic/bacterial susceptibility mismatches, orders for the infectious disease specialist (J.C.C.) with modifications made
for local experience (eg,, the recommended dose of cefuroxime
drugs to which the patient had reported allergies, was increased because of the risk of resistant Streptococcus pneu-
and alerts of excessive dosage of antiinfective moniae). Special doses were developed when indicated for severe
agents.15 Benefits to the patients were noted in fewer disease, such as meningitis or bacteremia, or for atypical patient
adverse drug reactions, fewer total antiinfective populations, such as those with cystic fibrosis. The neonatal dos-
age recommendations were constructed using standard tables that
doses, and lower antiinfective costs for the hospital. take into account the postconceptional age (estimated gestational
Other, more recent empiric antibiotic decision sup- age at birth plus age in weeks since birth) and age in days since
port systems have been associated with potential birth (postnatal age).22 Dosage adjustments for impairments in
improvements in clinical care.16 –18 renal function also were standardized for patients who are older
Given this experience with an adult decision sup- than 6 months. Younger patients were excluded because the
Schwartz formula for creatinine clearance estimation has been
port tool, we anticipated a similar benefit in pediat- cited as less accurate in neonates and infants who are younger
rics. We hypothesized that a clinical decision support than 6 months.23
system designed to account for the therapeutic indi- Copies of the pediatric antiinfectives medical logic and dosing
cation, the age and weight of the patient, the renal recommendations were distributed to the other pediatric infec-
tious diseases faculty of the University of Utah. Their comments
function, and the level of prematurity would im- and suggestions were incorporated into the final logic submitted
prove antiinfective choices and dosage selections, for use in the pediatric antiinfective management program.
reduce the rate of adverse drug events, and reduce The pediatric antiinfective management program subsequently
the cost of antiinfectives used in the care of critically was loaded onto the HELP system as a restricted-access program
ill infants and children. This article describes the for testing by the authors. During the summer and fall months of
1998, rigorous daily trials of the pediatric logic were performed on
development and clinical evaluation of a pediatric sample populations of patients from the wards and the pediatric
antiinfective decision support tool founded on the and neonatal intensive care units of the children’s hospital. Once
adult antiinfective management program developed the logic and underlying code were judged to be reliable and
at IHC. accurate, plans were made for the installation and evaluation of
the effects in the PICU. PICU personnel were readied for initial
use through a series of demonstrations and tutorials.
METHODS
Study Design
Setting
A study to measure the impact of the pediatric antiinfectives
Primary Children’s Medical Center (PCMC) is a 232-bed facility decision support tool was planned, using 6-month pre- versus
set on the University of Utah medical campus and owned and postimplementation comparison periods. With an average of 1700
operated by IHC. It is the primary pediatric teaching facility for admissions per year and an estimation that 75% would be treated
the University of Utah School of Medicine. The hospital serves as with antibiotics, we anticipated capturing more than 600 patients
the tertiary referral center for all of Utah and significant portions in each study period. The time periods chosen for the study placed
of Nevada, Arizona, Montana, Wyoming, and Idaho. PCMC approximately one half of the summer season (with many trauma
houses a PICU comprising 26 beds and averaging 1700 admissions patients) and one half of the winter season (with many bronchi-
per year of a broad array of critically ill medical and surgical olitis patients) in each phase of the evaluation. The institutional
patients. Pediatric critical care specialists, working together with review boards of the University of Utah and PCMC approved the
critical care fellows-in-training, pediatric residents, and nurse study protocol.
practitioners, staff the unit. This team is responsible primarily for For the PICU team of physicians and nurse practitioners, man-
the medical patients and co-manages the surgical admissions. datory use of the pediatric antiinfective management tool for
Bedside computer terminals that run the Health Evaluations ordering antiinfectives was initiated on January 22, 1999. Al-
through Logical Processing (HELP) hospital information system though use of the tool was obligatory, clinicians chose whether to
facilitate patient care.19 HELP is a fully integrated hospital infor- accept its recommendations on antiinfective agents and/or doses.
mation system that provides data collection and reporting for a During the last week of each 4-week resident rotation, resident
broad range of clinical arenas, including laboratory results, phar- physicians were surveyed using a questionnaire composed of
macy, radiology, and pathology. The PICU physicians and nurse 5-point Likert-type scales. The PICU pediatric nurse practitioners
practitioners use the HELP system primarily for laboratory results were surveyed once, at the end of the 6-month experimental
review and the generation of a morning summary report of patient period.
vital signs, labs, pharmacology, radiology, and ventilator data.
Before the implementation of the pediatric antiinfective decision
support tool, all patient care orders from the physicians were Analysis
handwritten. Antibiotic and other medication orders typically The pharmacy staff monitored and recorded adverse drug
were interpreted by the clerk and rewritten onto the bedside events and kept a log of their interventions on drugs and drug
medication administration record. Carbon copies of the handwrit- dosages. A computer alerting program reported mismatches of
ten order were physically sent to the pharmacy and read by a bacterial culture sensitivities and patient antibiotic therapy. These
pharmacist, who entered the order via the keyboard into the HELP mismatches were recorded and investigated as appropriate. A
system’s pharmacy module. computer program was developed to review the PCMC patient

2 of 7 COMPUTERIZED ANTIINFECTIVE DECISION SUPPORT

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 files comparing all administered antiinfective agents to published However, the number of orders placed per antiinfec-
therapeutic ranges with modifications for age and renal function. tive course decreased 11.5% from an average of 1.56
This program identified all doses that fell outside the therapeutic
ranges and generated a file that contained subtherapeutic and to 1.38 orders/patient-antiinfective (P ⬍ .01). In ad-
excessive-dosage risk days. dition, application of Tukey’s biweight estimator,
which downweights extreme values in non-normal
Statistical Methods distributions, revealed an underlying 9% decrease in
Study data were stored and manipulated in Microsoft Access the costs of antiinfectives used for the average inter-
(Microsoft Corp, Redmond, WA). Between-group comparisons vention group PICU patient (Table 2). Logarithmic
were performed using Fisher’s exact test for equality of propor-
tions, ␹2 test for independence, and 2-tailed t tests for comparisons transformations of non-normal data and application
of means. During the latter analyses (t tests), consideration was of Tukey’s biweight estimator did not change the
given to logarithmic transformations and the use of Tukey’s bi- interpretations of the other baseline population or
weight estimator for skewed variances when appropriate for non- antiinfective use measurements between the 2
normally distributed data.24 All ␹2 analyses were performed using
Microsoft Excel. The regression analyses were performed using groups. Total antiinfective use, by a comparison of
SPSS (SPSS Inc, Chicago, IL). All other analyses were performed the count of patients treated with each antiinfective,
using Statit Custom QC (Statware Inc, Corvallis, OR). Statistical also was similar between the 2 groups by ␹2 analysis.
significance levels were set at P values of .05 a priori. Therapeutic mismatches between pathogens cul-
RESULTS tured in the microbiology laboratory and the antimi-
crobials being used to treat the patients were as-
During the 12-month study period, the PICU ad-
sessed by a time-driven computer alerting program
mitted 1758 patients: 809 patients during the prein-
at 1:00 pm on the day of published sensitivities. There
tervention period and 949 during the intervention
was no difference in the incidence of mismatches
period. The intervention group was more likely to be
treated with antimicrobials while in the PICU (66.5% between the 2 groups. Only 1 event was noted dur-
vs 60.2%; P ⬍ .05), but the rate of antimicrobial use ing the control period: a Staphylococcus epidermidis
during the total hospital stay did not differ signifi- blood culture that initially was perceived as a con-
cantly. taminant and therefore was not being treated. Dur-
Additional comparisons are limited to “study pa- ing the intervention period, only 1 event was noted
tients,” defined as those patients with antiinfectives as well: an Enterococcus species urinary tract infection
ordered while hospitalized in the intensive care unit that was being treated with but ultimately found to
during the 2 periods. The intervention group was be resistant to amoxicillin.
significantly younger (Table 1). However, the 2 The total number of adverse drug reactions re-
groups were similar with respect to gender, PICU corded in the PICU for the 12-month study period
length of stay (LOS), total hospital LOS, All Patient was 119, with 24 of those secondary to antiinfectives.
Refined Severity of Illness,25 All Patient Refined Risk Twelve events were recorded in each of the 2 study
of Mortality,25 percentage mortality, and total hospi- periods. A breakdown of the reactions into the cate-
tal costs. gories of mild (requiring no therapy change), mod-
Per-patient antiinfective use measurements also erate (requiring a change in therapy), and severe
were similar between the 2 groups, despite the im- (potentially life-threatening) found no significant dif-
plementation of the new management tool (Table 2). ference. In each group, only 1 of the 12 potentially
Specifically, there were no differences in the PICU or was preventable secondary to known allergic sensi-
total hospital count of antiinfectives or antiinfective tivities. During the intervention period, the poten-
doses used per patient. There also was no difference tially preventable allergic reaction occurred when a
in the PICU or total hospital costs of antiinfectives. surgery resident, not using the antiinfectives man-

TABLE 1. Population Statistics for Patients With a PICU Antiinfective Order

Variable Preintervention Intervention Significance
(⫾SEM) (⫾SEM)
Female (%) 41.5 43.5 NS*
Age (y) 6.2 (⫾0.302) 5.3 (⫾0.244) P ⬍ .05†
PICU LOS (d) 4.93 (⫾0.490) 4.90 (⫾0.313) NS†
Hospital LOS (d) 10.76 (⫾0.778) 10.76 (⫾0.521) NS†
APR-DRG Severity of Illness 1:98 1:120 NS‡
category (count) 2:128 2:153
3:142 3:177
4:124 4:178
APR-DRG Risk of Mortality 1:228 1:280 NS‡
category (count) 2:84 2:135
3:119 3:139
4:56 4:74
Mortality (count) 18 20 NS*
Hospital costs (1999 dollars) $28 257.67 (⫾2375.66) $25 032.11 (⫾1210.70) NS†
APR-DRG indicates All Patient Refined–Diagnosis Related Groups; SEM, standard error of the mean;
NS, not significant.
* Fisher’s exact test for equality of proportions.
† Two-tailed independent t test.
‡ ␹2 test.

http://www.pediatrics.org/cgi/content/full/108/4/e75 3 of 7

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 TABLE 2. Average Per-Patient Antiinfective Use Measurements
Variable Preintervention Intervention Significance
(⫾SEM) (⫾SEM)
Total antiinfective costs ($) 274.79 (⫾28.57) 289.60 (⫾23.47) NS*
Total number of antiinfectives used 2.18 (⫾0.066) 2.22 (⫾0.063) NS*
Total doses used 19.8 (⫾1.35) 22.0 (⫾1.33) NS*
PICU antiinfective costs ($) 177.03 (⫾18.02) 183.53 (⫾14.75) NS*
Robust estimate PICU antiinfective costs ($) 86.60 (⫾2.98) 78.43 (⫾2.29) P ⬍ .05†
PICU antiinfective doses 12.8 (⫾0.985) 13.4 (⫾0.777) NS*
PICU number of antiinfectives 1.85 (⫾0.056) 1.97 (0.052) NS*
PICU antiinfective orders per 1.56 (⫾0.060) 1.38 (⫾0.032) P ⬍ .01*
patient-antiinfective course
SEM indicates standard error of the mean; NS, not significant.
* Two-tailed independent t test.
† Two-tailed modified t test using Tukey’s biweight estimator to downweight extreme observations.

agement program, ordered a cephalosporin on a nificant 36% decrease in the rate of subtherapeutic
PICU patient who was known to be allergic to pen- risk days was found for the intervention group when
icillins. (Note: Unlike pediatric residents in the PICU, compared with the control group (Table 3). Likewise,
the surgery residents were not involved in the study a significant 28% decrease was noted in the exces-
and were not required to use the pediatric anti- sive-dosage risk days. The combined effect is a 32%
infectives management program.) Had the decision decrease in the rate of antiinfective days that fall
support tool been used, it would have alerted the outside published recommended parameters.
physician to the allergy history. Questionnaires were returned by 28 of the 31 users
The pharmacists in the PICU serve as a human (26 pediatric residents, 5 nurse practitioners). Ques-
“safety net” for ordered pharmaceuticals, making tions were formatted as 5-point Likert-type scales,
interventions on erroneous drug doses and other and a favorable response was defined as 1 on the
therapeutic improvement opportunities. In this ca- “beneficial” or “positive effect” side of the neutral
pacity, they routinely keep a log of their interven- response. A majority of the users responded favor-
tions on the drugs ordered by the clinicians. During ably to the decision support tool. Specifically, they
the study period, the interventions for all pharma- reported improved overall antibiotic choices, in-
ceuticals numbered approximately 1800, with anti- creased awareness of renal function, beneficial dos-
infectives comprising approximately 30%. Analysis age calculation assistance, association with fewer ad-
of the relevant intervention categories revealed a verse drug events, and improved quality of care
59% decrease in the rate of intervention for errone- (Table 4). The median estimation of how often the
ous antiinfective doses and a 58% decrease in the rate users ordered the recommended antibiotic was 50%,
of clinician requests for antiinfective dosing help (Fig and the estimation of how often they ordered the
1). recommended dose was 75%. Most (79%) reported
An analysis of patient antiinfective doses com- that they learned something from the system, and
pared with published minimum and maximum rec- nearly all (93%) would recommend it to others.
ommendations for age, weight, and renal function Two post hoc analyses were performed to answer
was performed. Days of antiinfective therapy that questions raised by the initial analyses. Was the
fell outside the minimum and maximum recommen- younger age of the intervention patients responsible
dations were called subtherapeutic and excessive- for the 5% increase in antimicrobial usage? The an-
dosage risk days, respectively, and were determined swer seems to be “no.” Multiple linear regression
for each patient and analyzed by study day. A sig- showed that age did not explain PICU antiinfective
usage variability either alone or when controlled for
study group. The second question was whether the
decrease in PICU antiinfective costs found when an-
alyzed using Tukey’s biweight estimator also was
secondary to the younger age and presumably size of
the intervention patients. Again, multiple linear re-
gression did not find age to be a significant predictor
of PICU antiinfective costs when controlled by study
group or by the combination of study group, severity
of illness, and risk of mortality.

DISCUSSION
This study found that implementation of comput-
erized antiinfective decision support, provided at the
time antiinfectives are ordered, increased the likeli-
hood that the dose was on target for the given age,
Fig 1. Pharmacists’ rates of interventions by category on users’ weight, and renal function of the pediatric patient.
drug orders per 1000 antiinfective orders. The tool provided support to the clinician in a num-

4 of 7 COMPUTERIZED ANTIINFECTIVE DECISION SUPPORT

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 TABLE 3. Subtherapeutic and Excessive-Dosage Risk Days
Variable Preintervention Intervention Significance
(⫾SEM) (⫾SEM)
Subtherapeutic antiinfective days 7.350 (⫾0.610) 4.702 (⫾0.441) P ⬍ .001*
per 100 patient days
Excessive-dosage antiinfective days 8.454 (⫾0.460) 6.063 (⫾0.490) P ⬍ .001*
per 100 patient days
Total risk days per 100 patient days 15.80 (⫾0.734) 10.766 (⫾0.733) P ⬍ .0001*
SEM indicates standard error of the mean.
* Two-tailed independent t test.

TABLE 4. User Survey Results subtherapeutic risk days by increasing antibiotic

doses, it would not have been surprising if we had
Question Topic % Positive Results
found that the average antiinfective cost per patient
Improved overall antibiotic choices 81 was appropriately higher with the use of the tool.
Increased awareness of renal function 79 It is instructive to compare the results of this pe-
Beneficial dosage calculation assistance 96
Associated with fewer adverse drug events 89
diatric trial with the findings from the evaluation
Improved quality of care 81 performed in the adult shock-trauma intensive care
unit (STICU) at LDS Hospital. As shown in Table 5,
in the STICU evaluation, a marked impact was noted
ber of ways that can account for this improvement. in the number of mismatches between the sensitivity
The renal function was estimated and updated auto- patterns of the cultured bacterial pathogens and the
matically daily, and suggested doses were calculated antibiotics used for therapy. An improvement was
with adjustments for evidence of impairment. Age not noted in the current study, but sensitivity mis-
and prematurity considerations were factored auto- matches occur far more frequently in the adult ICU.
matically, and doses were calculated without arith- Thus, the opportunity for improvement in the PICU
metic errors. Order legibility also was rendered a was diminished. The measures of orders for drugs to
nonissue. These mechanisms explain the decrease in which the patient was known to have a history of
pharmacy interventions for erroneous doses and also allergy was improved in the adult study but without
the decreased number of days of therapy that fall change in the pediatric study. The frequency of drug
outside recommended therapeutic ranges. allergy is much higher in adult patients; young chil-
The last step in the system’s antiinfective recom- dren have not had as much time to have a known
mendations algorithm is to consider costs. The anti- history of drug allergy and have had fewer expo-
infective management program recommends the less sures to antibiotics than adult patients. We anticipate
expensive agent when 2 or more drugs are found to that our decision support tool would decrease sensi-
be therapeutically equal. We therefore anticipated a tivity mismatches and administration of drugs to
cost benefit, as was seen in the adult study. In this allergic children significantly with a much larger
pediatric study, the average cost of antiinfectives was study population.
no different between the 2 groups. However, appli- A large impact was noted in the rates of pharma-
cation of Tukey’s biweight estimator, which down- cists’ intervention for erroneous drug orders in the
weights extreme observations in non-normal distri- pediatric study. This information was not recorded
butions, identified a 9% decrease in the robust by pharmacists in the STICU evaluation. In the 2
estimate of the cost of antiinfectives used in the studies, a similar benefit was noted in the rate of
intervention group. Therefore, a longer study may reduction of excessive drug dosage days. In addition,
have documented cost savings using more conven- a benefit was seen in underdosage days in the PICU
tional analytical methods. Moreover, given that one study, but this is less of an issue in adults and was
of the tool’s beneficial effects is to increase the doses not measured in that study. No change was noted in
administered when clinically indicated and 1 of its the rate of adverse drug events in the PICU, but a
documented effects is to minimize the number of large benefit was found in the adult study. It is

TABLE 5. Comparison of Antiinfective Management Program Impact in Adult and Pediatric Studies
Measurement PICU Impact STICU Impact Baseline Rates Comparison
Susceptibility-mismatch alerts No change Large reduction 18/100 admissions in STICU vs
0.2/100 admissions in PICU
Drug allergy alerts No change Large reduction 12/100 admissions in STICU vs
0.4/100 admissions in PICU
Excessive days of antiinfective dose Reduction Reduction
Adverse drug events attributable to No change Large reduction 2.4/100 admission in both STICU
antiinfectives and PICU
Pharmacists’ interventions Large reduction Not measured
Antiinfective costs 9% reduction 20% reduction $412/patient STICU vs
$177/patient in PICU
Length of stay No change No change 6.3 d in STICU vs 4.9 d in PICU
Mortality No change No change 22% in STICU vs 3.7% in PICU

http://www.pediatrics.org/cgi/content/full/108/4/e75 5 of 7

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 interesting that the baseline rate of events per 100 dication, age, weight, and renal function rendered by
admissions was similar in both units. One would the antiinfective management program.
surmise that fewer of the events in children are pre- After the completion of the study, by joint decision
ventable, as they are less likely to be secondary to of the members of the medical, pharmacy, and nurs-
failing drug metabolism due to hepatic or renal dys- ing staffs, use of the pediatric antiinfectives manage-
function. The study data support this notion as only ment program remained mandatory for ordering an-
1 of the 12 pediatric events at baseline could be tiinfectives within the PICU. Usage also spread to
judged as preventable by retrospective evaluation other areas of the children’s hospital. Although the
and none of these events could be attributed to failed pediatric antiinfectives management program is not
organs or otherwise poor drug metabolism. transferrable without the HELP hospital information
A 20% reduction in the costs of antiinfectives used system, the implementation of a clinical computer-
at baseline was found in the adult study, whereas the ized physician order entry system combined with a
pediatric drug costs are, at best, 9% improved. The comprehensive system of antiinfective decision sup-
baseline analysis shows that antiinfective costs aver- port rules likely would provide similar benefit in
aged $412 (1995 dollars) per patient in the adult other pediatric institutions.
study and only $177 (1999 dollars) per patient in the
pediatric evaluation. Once again, it should be noted CONCLUSION
that there was less opportunity for improvement in
the pediatric case. Last, although the measures of Implementation of a pediatric antiinfective deci-
severity of illness used in the 2 studies are different sion support tool had a positive impact on the anti-
and are not comparable directly, it is instructive to infective therapeutic milieu of a PICU through better
note the 5-fold difference in incidence of death be- dosage selection, as documented by fewer pharmacy
tween the 2 units. The adult patients were far more interventions on antiinfective orders and fewer anti-
likely to have terminal disease. One may conclude infective subtherapeutic and excessive-dosage risk
that the adults have more end-organ dysfunction days. These findings are supported by the survey of
affecting their responses to and internal metabolism users who reported that use of the tool would result
of the antiinfective therapy used in the study. There- in fewer adverse drug events and improved quality
fore, it should not be a surprise that the impact of the of care.
antiinfective management program differs between
the 2 units in the process and outcome measures ACKNOWLEDGMENTS
evaluated. Direct and indirect support was provided by the University of
With the improvements in the rate of pharmacy Utah, Intermountain Health Care Corporation, and the National
interventions and the number of days outside thera- Library of Medicine (NLM Grant T15-LM07-124, University of
Utah Medical Informatics Training Grant to C.J.M.). This project
peutic ranges, one could anticipate a patient benefit would not have been possible without the assistance of Reed M.
in outcomes. Unfortunately, we were not able to Gardner, PhD, Chairman of the Department of Medical Informat-
document this, given the insensitivity of our patient ics at the University of Utah; Jared Cash, RPh, and partners in the
outcome measures: adverse drug events, LOS, and PCMC PICU pharmacy; and Mary B. Price, MBA, PCMC CQI
Coordinator.
mortality. However, one can conclude that although
the incidence of adverse sequelae from medical er-
rors is low, it is not 0, and minimization of outright REFERENCES
errors and improvements in therapeutic dosing tar- 1. Bates DW, Spell N, Cullen DJ, et al. The costs of adverse drug events in
gets should affect both adverse events and the qual- hospitalized patients. Adverse Drug Events Prevention Study Group.
ity of care, given enough time. It is through consid- JAMA. 1997;277:307–311
2. Lesar TS, Briceland L, Stein DS. Factors related to errors in medication
erations such as these that a majority of pediatric prescribing. JAMA. 1997;277:312–317
residents and nurse practitioners reported that they 3. Folli HL, Poole RL, Benitz WE, Russo JC. Medication error prevention
believed that the use of this clinical decision support by clinical pharmacists in two children’s hospitals. Pediatrics. 1987;79:
tool should have a beneficial impact on adverse drug 718 –722
4. Wilson DG, McArtney RG, Newcombe RG, et al. Medication errors in
events and the quality of care. This is consistent with paediatric practice: insights from a continuous quality improvement
the adage that humans and computers working in approach. Eur J Pediatr. 1998;157:769 –774
tandem are better than either one alone. 5. Schiff GD, Rucker TD. Computerized prescribing: building the elec-
The findings of this study are likely to be a con- tronic infrastructure for better medication usage. JAMA. 1998;279:
servative assessment of the benefit that would be 1024 –1029
6. Nightingale PG, Adu D, Richards NT, Peters M. Implementation of
anticipated from wide-scale implementation of this rules based computerised bedside prescribing and administration: in-
computerized decision support tool throughout the tervention study. BMJ. 2000;320:750 –753
children’s hospital. A significant portion of the anti- 7. Wyatt J, Walton R. Computer based prescribing. BMJ. 1995;311:
microbial therapy measured in this study was or- 1181–1182
8. Bates DW, Leape LL, Cullen DJ, et al. Effect of computerized physician
dered without the use of the tool— either by pediat- order entry and a team intervention on prevention of serious medica-
ric residents on the floor before transfer of the patient tion errors. JAMA. 1998;280:1311–1316
to the PICU or by surgery and surgery subspecialty 9. Shortliffe EH, Axline SG, Buchanan BG, Merigan TC, Cohen SN. An
residents transferring patients from either the floor artificial intelligence program to advise physicians regarding antimicro-
or the operating room. Therefore, once these groups bial therapy. Comput Biomed Res. 1973;6:544 –560
10. Shortliffe EH, Davis R, Axline SG, Buchanan BG, Green CC, Cohen SN.
become users of the system, the doses ordered also Computer-based consultations in clinical therapeutics: explanation and
would have the benefit of the elimination of calcula- rule acquisition capabilities of the MYCIN system. Comput Biomed Res.
tion errors, plus the careful consideration of the in- 1975;8:303–320

6 of 7 COMPUTERIZED ANTIINFECTIVE DECISION SUPPORT

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user
 11. Yu VL, Fagan LM, Wraith SM, et al. Antimicrobial selection by a 18. Warner H Jr, Reimer L, Suvinier D, Li L, Nelson M. Modeling empiric
computer. A blinded evaluation by infectious diseases experts. JAMA. antibiotic therapy evaluation of QID. Proc AMIA Symp. 1999;440 – 444
1979;242:1279 –1282 19. Gardner RM, Pryor TA, Warner HR. The HELP hospital information
12. Chung SJ, Chung RY. Choice of antimicrobial drugs by the site of system: update 1998. Int J Med Inf. 1999;54:169 –182
infection. Comput Methods Programs Biomed. 1987;24:47– 62 20. Nelson JD. Pocket Book of Pediatric Antimicrobial Therapy. 12th ed. Balti-
13. Chung SJ, Gosnell WO, Purkey MS. Choice of antibiotics by the site of more, MD: Williams & Wilkins; 1996
infection in a community hospital. Int J Biomed Comput. 1995;40:31–39 21. Taketomo CK, Hodding JH, Kraus DM. Pediatric Dosage Handbook. 5th
14. Evans RS, Classen DC, Pestotnik SL, Lundsgaarde HP, Burke JP. Im- ed. Cleveland, OH: Lexi-Comp Inc.; 1998
proving empiric antibiotic selection using computer decision support. 22. Young TE, Mangum OB. Neofax ’95: A Manual of Drugs Used in Neonatal
Arch Intern Med. 1994;154:878 – 884 Care. 8th ed. Columbus, OH: Ross Products Division, Abbott
15. Evans RS, Pestotnik SL, Classen DC, et al. A computer-assisted man- Laboratories; 1995
agement program for antibiotics and other antiinfective agents. N Engl 23. Schwartz GJ, Brion LP, Spitzer A. The use of plasma creatinine concen-
J Med. 1998;338:232–238 tration for estimating glomerular filtration rate in infants, children, and
16. Leibovici L, Gitelman V, Yehezkelli Y, et al. Improving empirical anti- adolescents. Pediatr Clin North Am. 1987;34:571–590
biotic treatment: prospective, nonintervention testing of a decision sup- 24. Hoaglin DC, Mosteller F, Tukey JW. Understanding Robust and Explor-
port system. J Intern Med. 1997;242:395– 400 atory Data Analysis. New York: John Wiley & Sons Inc.; 1983
17. Warner H Jr, Blue SR, Sorenson D, et al. New computer-based tools for 25. Iezzoni LI, Shwartz M, Ash AS, Mackiernan YD. Using severity mea-
empiric antibiotic decision support. Proc AMIA Annu Fall Symp. 1997; sures to predict the likelihood of death for pneumonia inpatients. J Gen
238 –242 Intern Med. 1996;11:23–31

http://www.pediatrics.org/cgi/content/full/108/4/e75 7 of 7

Downloaded from http://publications.aap.org/pediatrics/article-pdf/108/4/e75/894228/pe1001000jx1.pdf

by Dalhousie University-Dal 11762 user