Reprinted from CLINICAL CHEMISTRY, 39, 1069 (1993).
Copyright 1993 by the American Association of Clinical Chemistry and reprinted by permission of the copyright owner
CLIN.CHEM. 39/6, 1069-1074 (1993)
Analytical Evaluation of i-STAT Portable Clinical Analyzer and Use by Nonlaboratory
Health-Care Professionals
Ellis Jacobs, 1,2,4 Elizabeth Vadasdi, 2 Laszlo Sarkozi, 1,2 and Neville Coman
We evaluated the performance of the i-STAT Portable
Clinical Analyzer, a hand-held instrument that, with its
current cartridge, analyzes for electrolytes, urea nitrogen,
glucose, and hematocrit in ~60 µL of whole blood in ~90
s. Accuracy, imprecision, and linearity studies were per-
formed with aqueous controls and standards and by
split-sample analysis. Intrarun imprecision (CV) ranged
from 0.34% to 3.97%. Total imprecision over a 2-month
period ranged from 0.42% to 4.83%, with urea nitrogen
and glucose analyses generating the higher values. Pa-
tients’ results from the Portable Clinical Analyzer corre-
lated well with those obtained for whole blood or plasma
by the Nova Stat Profile 5, the Beckman Synchron CX3,
or the Technicon H1 Hematology Analyzer, with Sy x
values <0.2 mmol/L for potassium; 1.5 mmol/L for
sodium, glucose, and urea nitrogen; <2.4 mmol/L for
chloride; and <2.4% for hematocrit. We also ascertained
imprecision and accuracy of the system placed in a
cardiothoracic intensive-care unit and operated by
nurses. There were no significant differences in either the
imprecision or accuracy of the system in this setting. We
conclude that operator technique is not a factor in the
analytical performance of the system and that it can be
used by nonlaboratorians with a high degree of confi-
dence that reliable results will be obtained.
Indexing Terms: point of care testing • biosensors • electro-
chemical sensors • emergency analyses • critical-care medi-
cine • electroyles • glucose • urea nitrogen
Changes in medical practice have intensified institu-
tional pressures to achieve clinical efficacy (1, 2). Thus
hospitals are decreasing the admission of patients with
nonacute conditions and increasing the proportion of
patients admitted for major therapeutic interventions,
e.g, transplants or open heart surgery, that require
more-intensive support systems (3–5). Additionally,
there is a push to decrease patients’ overall length of
stay as well as the use of high-cost resources such as
intensive-care units (ICUs).5
In the laboratory, these various pressures translate
into requests to improve service by decreasing test
1 Department of Pathology, Mount Sinai & School of Medicine;
2 Center for Clinical Laboratories, The Mount Sinai Hospital, New
York, NY 10029.
3 Veterans Administration Medical Center, Bronx, NY 10468.
4 Address for correspondence: The Mount Sinai Medical Center
STAT Laboratories, Box 1519, One Gustave L. Levy Place, New
York, NY 10029-6574.
5 Nonstandard abbreviations: PCA, Portable Clinical Analyzer
I C U, i n t e n s ive - c a re unit; QC, quality control; and NCCLS, N a-
tional Committee for Clinical Laboratory Standards.
Received September 23,1992; accepted December 11, 1992.
1–3
turnaround time (1, 6). Various mechanisms have been
used to meet these clinical needs, e.g., pneumatic tube
systems to improve specimen transport to centralized
laboratories, establishment of specialized “stat” labora-
tories, and use of point-of-care testing devices (glucose
meters, blood gas/electrolyte analyzers) (7–9).
One of the major impediments to the implementation
of point-of-care testing, in addition to concerns about
analytical accuracy and precision, are considerations of
the reliability of these systems in the hands of nontra-
ditionally trained health-care professionals. Until re-
cently, technology has not been able to produce a point-
of-care analytical system that was robust enough to
withstand the rigorous environment associated with
dispersed locations (10, 11). To maintain the desirable
accuracy and precision, systems have had demanding
operational and maintenance requirements. Thus, it
has been difficult for nurses, physicians, or other non-
laboratory health-care professionals to obtain and main-
tain the skills necessary for proper operations and
maintenance of the systems (12). Various technological
innovations have extended the lifetime of electrochem-
ical sensors, reduced maintenance, and led to the devel-
opment of single-use and unit-use disposable sensors
(13, 14). Additionally, in conjunction with reagent sta-
bilization through the use of biosensor technology, the
system must have built-in software capability for qual-
ity-control (QC) activities and detection of instrument
malfunctions.
Recently, a new point-of-care testing instrument, the
Portable Clinical Analyzer (PCA; i-STAT Corp., Prince-
ton, NJ), was introduced into the market (10, 15). This
system was designed for use by nonlaboratory personnel
in the critical-care environment of a hospital’s ICUs,
surgical units, and emergency department. In its cur-
rent configuration, this hand-held instrument analyzes
for sodium, potassium, chloride, urea nitrogen, glucose,
and hematocrit in ~60 µL of whole blood in ~90 s.
We report here the results of our study of the i-STAT
PCA. Correlation studies were performed by comparison
with established laboratory methodologies. A system was
placed in our Cardiothoracic ICU to evaluate its
reliability in the ICU, when operated and maintained
by nonlaboratory health-care professionals. Precision
and accuracy of the system were assessed for utilization
by both laboratory technologists and ICU nurses.
Materials and Methods
System Description
The PCA is a hand-held system that utilizes a single-
use disposable cartridge (2.7 x 4.5 x 0.5 cm, w x l x h)
containing microfab ri c ated biosensors. Whole blood
CLINICAL CHEMISTRY, Vol. 39, No. 6, 1993 1069
from a capillary fingerstick or a syringe is placed in the
c a rt ri d ge, the cart ri d ge is then inserted into the ana-
ly ze r, and operator and patient identification (numeri-
cal only) are entered into the system. Automatically a
calibration solution is released and a one-point calibra-
tion is performed just before sample analysis. The re-
sponses of the various biosensors are monitored during
the rehydration, calibration, and sample analysis pro-
cesses. If any of the sensor responses fall outside prede-
termined limits, then the controlling software sup-
presses the output from that sensor. Additionally,
mechanical and user operational errors are identified
through the controlling software and, if detected, cause
a suppression of result reporting.
Laboratory Study Protocol
Syringe specimens, anticoagulated with dry lithium
heparin (3-mL preheparinized syringes; Concord-Por-
tex, Keene, NH) to a final concentration of 30 kIU/L,
were selected without conscious bias from those rou-
tinely received in the stat laboratory for analysis.
These specimens were obtained from either the operat-
ing rooms, the ICU, or the emergency department. The
samples were transported to the stat laboratory either
by a high-speed pneumatic tube system or by hand.
Once in the laboratory, the samples were remixed by
hand. The whole-blood samples were tested without
delay on the PCA, the Stat Profile 5 (Nova Biomedical,
Waltham, MA), and the H-1 Hematology analyzer
(Miles, Tarrytown, NY). Plasma was obtained by cen-
trifuging an aliquot of the whole-blood specimen for 2
min at 2536 x g in a Microfuge 12 (Beckman Instru-
ments, Brea, CA) with a 13.2 fixed-angle rotor. The
plasma obtained was then analyzed by the Synchron
CX3 analyzer (Beckman) and by the PCA. All testing
was done according to manufacturers’ recommended
procedures. All specimens were analyzed in duplicate
and testing was completed within 10 min of specimen
receipt. Five different PCA units were used during this
study, one on each day of the week, to assess intermeter
variability. Aqueous bi-level QC and linearity test
solutions were provided by i-STAT Corp.
ICU Study Protocol
In the Cardiothoracic ICU, eight nurses were given 30
min of instructions on the use and operations of the
PCA. The aqueous QC material was analyzed singly on
Table 1. PCA Within-Laboratory Precision Study: Extended NCCLS EP5 Protocol
Level 1
Mean S IR(CV, %)
Sodium, mmol/L 115.0 0.521 (0.45)
Potassium, mmol/L 5.73 0.079 (1.38)
Chloride, mmol/L 78.5 0.589 (0.75)
Urea Nitrogen, mmol/L 18.90 0.496 (2.63)
Glucose, mmol/L 10.63 0.421 (3.96)
Hematocrit, %PCV (–3.04)a 0.264 (–8.69)
SIR, intrarun standard deviation; S TOT, total standard deviation;PCV, packed cell volume.
a
These values represent artefacts from measuring aqueous control material rather than a biological specimen.
1070 CLINICAL CHEMISTRY, Vol.39, No. 6, 1993
the system each day. Specimens were obtained in lith-
ium heparinized syringes from arterial lines and ana-
lyzed by the PCA in the ICU before being sent, by
pneumatic tube, to the stat laboratory. Once in the
laboratory, the specimens were tested according to the
protocol previously described.
Statistics
The National Committee for Clinical Laboratory
Standards (NCCLS) EP5 & EP6 protocols for the assess-
ment of imprecision of an analytical system and for the
assessment of linearity were utilized (16, 17). However,
instead of the standard EP5 protocol, i.e., duplicate runs
twice a day for 20 days, we used an extended protocol of
65 days.
Student’s t-test and Demings linear-regression anal-
ysis (18) were used in the methodology and specimen-
type comparisons.
Results and Discussion
Imprecision and Linearity Studies
The precision, or imprecision, of the i-STAT system
was determined when the system was operated either by
traditionally trained laboratory personnel or by Car-
diothoracic ICU nurses who do not have training to
operate laboratory equipment. Within the laboratory,
rotating the use of five PCAs and multiple lots of
cartridges and using our extended NCCLS EP5 protocol
yielded a within-run imprecision (CV) of 0.34–3.96%;
the total imprecision ranged from 0.42% to 4.82% (Table
1). Sodium, potassium, and chloride had total CVs <2%,
whereas for urea nitrogen and glucose total imprecision
was between 3.8% and 4.8%. Because an aqueous con-
trol was used, the hematocrit data reflect the stability of
the conductivity electrode used in this measurement.
These values are comparable with those obtained by
established laboratory-based analytical systems, i.e.,
0.3–4.1%.
Even though five PCA analyzers were used through-
out the imprecision study, and >250 data points were
collected per analyte, there were no statistical outliers.
However, some data points were suppressed by the
analyzer because of some malfunction identified by the
controlling software. The rate of data suppression by
the PCA shown in Table 2 represents the results of
testing both QC material and patients’ specimens with
> 1800 test cartridges, from seven different lots. One or
Level 2
STOT(CV, %) Mean SIR(CV, %) S TOT(CV, %)
0.535 (0.47) 142.5 0.490 (0.34) 0.592 (0.42)
0.088 (1.53) 2.84 0.046 (1.62) 0.055 (1.93)
0.671 (0.85) 106.6 0.767 (0.72) 0.822 (0.77)
0.772 (4.08) 22.79 0.514 (2.26) 0.879 (3.85)
0.513 (4.82) 3.03 0.087 (2.87) 0.146 (4.82)
0.264 (–8.69) ( 3 . 0 3 )a 0.205 (–6.75) 0.205 (–6.75)

Table 2. Data Suppression Rage of PCA Cartridges
Results suppressed No. (%) of cartridges
All tests 24 (1.31)
All tests except glucose 26 (1.42)
Chloride 1 (0.05)
Urea nitrogen 6 (0.32)
Glucose 4 (0.22)
Hematocrit 11 (0.60)
more test results were suppressed in 3.94% of the
c a rt ri d ges used. For the majority (69%) of the car-
t ri d ges that had tests suppressed, the internal moni-
toring blocked the reporting of either all the analytes,
or all except glucose.
Lot variations. A significant concern with unit-based
testing systems, especially with biosensor technology as
in the PCA, is the va ri ability of calibration and re-
sponse, both within and across various manufacturing
lots of cartridges. By sorting and analyzing the QC data
by cartridge lots, one can assess calibration reproduc-
ibility. At least 15 data points were obtained from six
different manufacturing lots of cartridges. Of the with-
in-lot variations (Table 3), the highest values were
obtained for glucose and urea nitrogen. In general, the
lot-to-lot variation was excellent for the analytes tested
(Table 3). CVs at high concentrations of urea nitrogen
and glucose (CV of within-lot means >3.0%), were
substantially greater than with the other analytes
(<0.6%). Nevertheless, the lot-to-lot variations for all the
analytes were well within acceptable clinical limits
for overall imprecision and would have had no impact on
patient care.
Linearity. We assessed the linearity (or proportional-
ity of response) of the chemistry tests on the PCA by
analyzing aqueous standards according to the NCCLS
EP6 protocol (17). This method utilizes linear-regres-
sion analysis as its basis for determination of linearity.
However, in this protocol, the data are subjected to two
verifications before regression analysis is performed.
First is determination of any outliers. Then Cochran’s C
test is applied, to test the hypothesis of equality of
variance across dilutions. After these two determina-
tions, linear-regression analysis is performed and tested
for lack of fit to the linear model.
To test the linearity of the PCA, we used five aqueous
solutions with concentrations distributed equally across
the following ranges (mmol/L): sodium 102-167, potas-
sium 2.5-8.5, chloride 70-130, glucose 1.93-19.57, and
urea nitrogen 1.29-39.42 (Figure 1). There were no
outliers in the data sets. At a significance level of 0.05
and over the range being tested, there was equality of
variance for all the analytes except glucose. Even at P =
0.01, glucose failed Cochran’s test of equality—primar-
ily because of the increased imprecision of the glucose
sensor at concentrations >11 mmol/L. Even though
these variances are significant statistically, there is
sufficient precision in the data to continue with the
linearity study.
After performing the linear-regression calculations,
we compared the linear “significance” value (G) with
the critical F-value, at P = 0.1, for testing the hypoth-
esis of a linear fit. Sodium, potassium, and glucose tests
had G-values lower than the critical F-value, and thus
were determined to be linear over the ranges tested.
Over the entire range tested, urea nitrogen was linear
but at P = 0.01. Furthermore, visual inspection of the
data in Figure 1 reveals a definite break in urea
linearity at 23 mmol/L. Over the limited range of 1.3-23
mmol/L, urea nitrogen testing was linear at P =
0.1. Chloride had G-values significantly greater (6-13-
fold) than the critical F-values and thus failed the
hypothesis of linear fit. However, for very precise
assays, data with clinically acceptable linearity may be
d e cl a red statistically nonlinear. Furt h e rm o re, upon
inspecting the plot of data in Figure 1, it is obvious that
the system is linear over the range of values tested.
This further proves the need for a visual inspection of

Table 3. Variation of QC Results due to Different Lots

of PCA Test Cartridges
Mean of CV of
Range of lot lot
Control means, Range of means, means,
Analyte level mmol/L lot CVs, % mmol/L %
Sodium 1 114.9–115.4 0.28–0.74 115.1 0.16
2 142.1–142.8 0.28–0.42 142.4 0.22
Potassium 1 5.70–5.75 0.96–2.12 5.73 0.29
2 2.82–2.86 1.59–2.21 2.84 0.58
Chloride 1 78.2–78.9 0.55–1.04 78.6 0.36
2 106.2–107.0 0.50–0.93 106.6 0.27
Urea Nitrogen 1 18.05–19.72 1.71–4.31 18.97 3.08 Fig.1.Linearity of i-STAT Portable Clinical Analyzer for (▲) chloride,
2 21.85–23.75 1.68–3.78 22.94 3.18 (●) sodium, (●) potassium, (▲) urea nitrogen, and (■) glucose
Glucose 1 9.96–10.89 3.36–6.22 10.54 3.14 assays
All testing was performed as described in the test. Each solution was analyzed
2 2.95–3.16 2.55–5.26 3.06 2.15
four times, according to the NCCLS EP6 protocol, and the averages repre-
sented here

CLINICAL CHEMISTRY, Vol.39, No. 2, 1993 1071

the data in addition to the application of mathematical
formulas.
Method Comparison
The analytical performance of the i-STAT system was
compared with various established laboratory-based
systems. Initially, to ensure that there was no bias due
to sample type, both whole blood and plasma derived
from the same specimen were analyzed by the i-STAT
system. The data in Table 4 show the lack of significant
analytical bias due to sample type, with average differ-
ences ranging from 0.1 mmoVL for potassium to 0.44
mmoVL for glucose. Other than for glucose values >11
mmoVL, the standard deviations of the differences dem-
onstrate good precision. Demings regression analysis
reveals excellent correlations, with slopes between 1.01
and 1.06 and r >0.95. Only glucose showed significant
scatter about the regression line (Syx = 0.5216 mmol/L).
The correlation of results from whole blood analyzed
on the PCA with whole blood on the Stat Profile 5 or H1
analyzer or with plasma analyzed on the Synchron
CX3 is shown in Table 5. In general, there was good
agreement between the PCA and the CX3, with slopes
between 0.987 and 1.07. The lowest r value and the
largest Sy x value were obtained with chloride testing.
Chloride ion-selective electrode systems are sensitive
to protein effects (19). The CX3 system uses a diluted
sample, which minimizes the impact of specimen pro-
tein on the analysis. Glucose results by the PCA were
~7% greater than by the CX3 and showed greater
scatter (Sy x ) than for the other analytes. The hematocrit
measured by conductivity on the PCA correlated
well with the hematocrit calculated from erythrocyte
counts and indices obtained on the H1: slope = 1.143
and Sy x = 2.32%.
As would be expected, given the matrix problems
associated with whole-blood testing, the regression
data for the PCA and the Stat Profile 5 were not as
tight as the correlation with the CX3. Nevertheless,
except for sodium and chloride, the r values were >0.93
and, except for chloride, the slopes were >0.95. The
lower r value for sodium is due to the limitation of the
numerical range of the data, rather than a problem
with either testing system. This is seen from the
difference data, which show an average difference of 2.0
mmol/L in the range of 130-150 mmol/L, the standard
deviation of the differences being 2.16. Despite an r
value of 0.977, a slope of 1.162, and an intercept of
-1.102 mmol/L, glucose results on the PCA were
significantly higher than the Stat Profile 5 for samples
with glucose concentrations >11 mmol/L. This is sim-
ilar to the results of the sample-type comparison study,
which showed decreased precision of the glucose bio-
sensor at concentrations > 11 mmol/L. The PCA hemat-
ocrit methodology correlated well with the conductivity
methodology on the Stat Profile 5 system: slope =
1.119, S y x = 2.06%, and intercept = -3.759%.
Reliability in Point-of-Care Settings
There was no significant difference in the imprecision
of the system when it was operated by the Cardiotho-
racic ICU nurses (CV 0.56-4.76%) (Table 6). Addition-
ally, there was no significant difference in the rate of
data suppression by PCAs used by the nurses vs those
used by laboratory personnel. Thus operator technique
is not a factor in the analytical performance of the
i-STAT system.
When the same sample was first analyzed by the
nurses in the Cardiothoracic ICU and then transported
and analyzed within 5 min in the stat laboratory on the
i-STAT system, no significant difference was observed in
the difference and regression analysis (Table 6). The Syx
values ranged from 0.0792 to 1.2528 mmol/L for the
chemistry tests, and was 2.0671% for hematocrit. This

Table 4. PCA Sample-Type Comparison: Whole Blood vs Plasma

Differenceb Regression statistics

Analytea Conc range n Mean SD Slope y-Intercept Sy|x r

Sodium 100–130 5 –0.200 0.447 1.0311 –3.8546 0.6419 0.9723
130–150 198 0.424 0.908
Potassium 2.0–3.0 4 0.125 0.126 1.0304 –0.0617 0.0758 0.9837
3.0–5.0 189 0.057 0.096
5.0–9.0 10 0.100 0.094
Chloride 70–90 1 –1.000 0.0— 1.0620 –5.9309 1.0306 0.9532
90–110 184 0.576 1.462
110–140 18 –0.056 1.259
Urea Nitrogen 0–8.9 145 –0.185 0.388 1.0061 –0.2029 0.4737 0.9971
8.9–21.4 42 0.187 0.789
21.4–35.7 11 0.844 1.154
35.7–53.6 5 –1.142 1.053
Glucose 1.1–5.6 16 0.132 0.156 1.0247 0.0239 0.5216 0.9880
5.6–11.1 122 0.225 0.374
11.1–16.7 49 0.432 1.012
16.7–25.0 18 0.293 1.487
a
Units as in Table 1.
b
Plasma value minus wholeblood value.

1072 CLINICAL CHEMISTRY, Vol.39, No. 6, 1993

 Table 5. Method Comparison of the PCA
PCA vs Synchron CX3 PCA vs Stat Profile 5

Differenceb Regression statistics Differenceb Regression statistics

Analytea Conc range n Mean SD Slope y-Inter. S y|x r n Mean SD Slope y-Inter. Sy|x r
Sodium 100–130 4 –0.000 0.816 0.9969 –0.1089 1.1695 0.8650 2 0.500 3.536 0.9489 5.2559 1.5263 0.8377
130–150 185 –0.546 2.032 192 –1.948 2.163
Potassium 2.0–3.0 4 0.025 0.150 0.9695 0.0217 0.0764 0.9780 3 –0.100 0.000 0.9880 –0.0137 0.1124 0.9475
3.0–5.0 169 –0.100 0.095 177 –0.059 0.169
5.0–9.0 10 0.100 0.094 13 –0.108 0.246
Chloride 70–90 1 5.000 0.0— 0.9874 –0.8202 1.6538 0.8170 0 .— .— 0.9265 4.2629 2.3272 0.7516
90–110 145 –1.697 2.822 131 –2.740 2.929
110–140 18 –0.056 1.259 61 –5.541 3.264
Urea Nitrogen 0–8.9 137 –0.122 0.407 0.9756 0.0370 0.5258 0.9966 . —c .— .— .— .— .— .—
8.9–21.4 39 0.073 0.729
21.4–35.7 9 –1.349 1.208
35.7–53.6 4 –0.893 1.220
Glucose 1.1–5.6 14 0.135 0.144 1.0698 –0.3719 0.6206 0.9886 10 –0.039 0.200 1.1615 –1.1018 0.8934 0.9770
5.6–11.1 114 0.163 0.357 125 0.150 0.629
11.1–16.7 50 0.493 0.839 49 1.034 1.086
16.7–25.0 13 1.218 1.820 10 2.222 2.993
Hematocritd 15–30 55 1.418 2.291 1.1432 –2.597 2.3170 0.9102 34 1.418 2.291 1.1189 –3.7590 2.0634 0.9315
30–50 127 2.228 2.658 157 2.228 2.658
50–70 0 .— .— 1 2.000 .—
a
Units as in Table 1.
b
PCA value minus wholeblood value.
c
Urea nitrogen is not available on the Stat Profile 5
d
The comparison method was the H1 analyzer.

Table 6. Use of PCA by Cardiothoracic ICU Nurses and Laboratory Personnel Compared
Imprecision analysis

Level 1 Level 2 Differenceb Regression statisticsc

Conc
Analytea Mean S TOT (CV, %) Mean S TOT (CV,%) range n Mean SD Slope y-Inter. Sy|x r
Sodium 115.0 0.556 (0.48) 142.7 0.795 (0.56) 100–130 3 0.333 0.577 0.9483 7.2761 0.7087 0.9671
130–150 96 0.125 1.059
Potassium 5.69 0.065 (1.14) 2.82 0.047 (1.65) 2.0–3.0 1 0.000 .— 0.9651 0.1365 0.0792 0.9772
3.0–5.0 93 –0.005 0.103
5.0–9.0 93 –0.020 0.130
Chloride 78.5 0.504 (0.64) 106.7 0.779 (0.73) 70–90 1 0.000 .— 1.0080 –1.1825 1.2528 0.9431
90–110 88 –0.205 1.627
110–140 10 –1.700 1.829
Urea Nitrogen 18.34 0.708 (3.86) 22.25 1.059 (4.76) 0–8.9 71 –0.055 0.411 1.0098 –0.1031 0.4023 0.9983
8.9–21.4 21 0.068 0.645
21.4–35.7 2 0.068 0.645
35.7–53.6 4 –0.357 1.237
Glucose 10.54 0.440 (4.17) 2.93 0.111 (3.77) 1.1–5.6 4 0.194 0.147 1.0274 0.2400 0.5223 0.9836
5.6–11.1 60 –0.029 0.392
11.1–16.7 31 0.208 0.947
16.7–25.0 4 –0.444 1.807
Hematocrit (–3.0) 0.000 (0.00) (–3.0) 0.000 (0.00) 15–30 11 0.818 1.328 0.9719 1.1637 2.0671 0.8754
30–50 86 0.081 2.879
STOT, total standard deviation.
a
Units as in Table 1.
b
ICU PCA value minus stat lab PCA value.
c
x = stat lab PCA value, y = ICU PCA value.

further proves that the system is fairly technique- Quality Control

independent and can be used by nonlaboratorians in
expectations, with a high degree of confidence, that Quality assurance deals with ensuring the reliability of
reliable results will be obtained. a system via various checks and balances. In our

CLINICAL CHEMISTRY, Vol.39, No. 6, 1993 1073

operations of the PCA system over 6 months and utiliz-
ing > 1800 cartridges from various production lots, there
were no analytical outliers, neither in the precision data
nor in the method-comparison data. All data obtained
were included in our analysis. However, 3.94% of the
cartridges had one or more tests suppressed by the
analyzer. All the tests, or all except glucose, were
suppressed in 2.73% of the cartridges. Thus, in our
experience, the software in the system that monitors
and analyzes the rehydration and calibration responses
of the electrodes is apparently able to adequately detect
and suppress analytical malfunctions. If this observa-
tion holds true after data from multiple studies are
combined (n in the thousands), it will have a significant
impact on how quality control is performed on this
system.
An effective QC program tests both the mechanical/
electronic component of the system, i.e., the analyzer,
and the analytical component (the disposable test car-
tridge). With the PCA system, an electronic cartridge
s i mu l ator is used to ensure that electrical ch a ra c t e ri s-
tics of the analyzer are within the manufacturer’s spec-
ifications. This test is done on each analyzer every 24 h.
Mechanical characteristics of the analyzer are checked
on each cartridge by monitoring fluid flow. However, as
a departure from traditional laboratory QC concepts, it
is not necessary to assay liquid QC material with each
analyzer each day. Instead, a randomized sampling, in
accordance with various manufacturing QC standards
(20, 21), of the test cartridges in stock should be ana-
lyzed each day to ensure that they have been stored
correctly and are still functioning properly.
In conclusion, the i-STAT PCA is an analytical sys-
tem for the testing of whole blood, as well as serum or
heparinized plasma, that has accuracy and precision
equivalent to, if not better than, established laboratory-
based systems. It is simple to operate and gives reliable
results when used by either laboratory personnel or ICU
nurses, i.e., nonlaboratory-trained individuals. Because
of the high level of sophistication in the controlling
software, a cost-effective QC program can be based on
s t a n d a rd industrial manu fa c t u ring sampling tech-
niques, rather than on the traditional laboratory QC
concept of testing each analyzer/cartridge combination
on each day of use.
We acknowledge and thank i-STAT Corp. for supporting this
study through a grant. In particular, we thank Robert Martin, from
i-STAT, for his aid in the data reduction. We are grateful to
1074 CLINICAL CHEMISTRY, Vol.39, No. 6, 1993
Carolyn Lo, Nursing Education Specialist, and the nurses of the
Cardiothoracic ICU for their cooperation with this study.
References
1. Rock RC. Why testing is being moved to the site of patient care.
Med Lab Observ 1991,23:2 5.
2. Civetta JM, Hudson-Civeta JA. Maintaining quality of care
while reducing charges in the lCU. Ann Surg 1985,202:523-32.
3. Jacobs P, Noseworthy TW. National estimates of intensive care
utilization and costs: Canada and the United States. Crit Care Med
1990;18:1282–6.
4. Hines R, Mylott L, Barash PG. The evolving needs of critical
care. Med Lab Observ 1991,23:6-9.
5. Zaloga GP, Hill TR, Strickland RA, Kennedy D, Visser M, Ford
K, et al. Bedside blood gas and electrolyte monitoring in critically
ill patients. Crit Care Med 1989,17:920-5.
6. Hall JR. Critical care medicine and the acute care laboratory.
Clin Chem 1990;36(Suppl):1552B-6B.
7. Hardwick DF, Morrison JI, Cassidy PA. Managing lab costs and
utilization: balance supply and demand. Dimens Health Serv
1985;62: 18-22.
8. Fleisher M, S ch wa rtz MK Strat egies of orga n i z ation and ser-
vice for the critical care laboratory. Clin Chem 1990;36(Suppl):
1557B-61B.
9. Jacobs E, Sarkozi L, Colman N. A centralized critical care (stat)
laboratory: the Mount Sinai experience. Crit Care Rep 1991;2:397–
405.
10. Wilding P. New approaches to point-of-care testing. Adv Hosp
Technol 1990;2:38-42.
11. Claremont DJ. Biosensors: clinical requirements and scientific
promise. J Med Eng Technol 1987;11:51.
12. Nani AA Poon R, Hinberg I. Quality of laboratory test results
obtained by non-technucal personnel in a decentralized setting. Am
J Clin Pathol 1988;89:797-801.
13. Burritt MF. Current analytical approaches to measuring blood
analytes. Clin Chem 1990;36(Suppl):1562B-6B.
14. Meyerhoff ME. New in vitro analytical approaches for clinical
chemistry measurements in critical care. Clin Chem 1990
36(Suppl): 1567B-72B.
15. Wilding P, E ri ckson K Eva l u ation of a hand-held micro fab-
ricated sensor system, the i-STAT lOOTM, for rapid assay of
electrolytes, glucose, urea nitrogen and hematocrit [Abstract].
Clin Chem 1990;36:1206.
16. NCCLS Te n t at ive Guideline EP5-T. User eva l u ation of pre c i-
sion performance of clinical chemistry devices. Villanova, PA:
National Committee for Clinical Laboratory Standards, June
1984.
17. NCCLS Proposed Guideline EP6-P. Evaluation of the linearity
of quantitative analytical methods. Villanova, PA: National Com-
mittee for Clinical Laboratory Standards, September 1986.
18. We i s b rot IM. Statistics for the clinical lab o rat o ry. Philadel-
phia: JB Lippincott Co., 1985:36-7.
19. Corimer A, Vitiello J. Chloride ion concentration by ISE’s. In:
Aizawa M, Kuwa K eds. Methodology and clinical applications of
blood gas, pH, electrolyte and sensor technology. Vol. 13, Proceed-
ings of an international symposium. Copenhagen, Denmark: Int
Fed Clin Chem, 1992:85-93.
20. American national standard for sampling procedures and
tables for inspection by variablefi for percent nonconforming.
ANSVASQC Zl.9-1980. Milwaukee, Wl: American Society for Quality
Control, 1980.
21. American national standard for sampling procedures and
tables for inspection by attributes. ANSVASQC Z1.4-1981. Milwaukee,
W I :A m e rican Society for Quality Control, 1981.