Evaluation of Discrepancies in P

					CLIN. CHEM. 21/1, 87-92 (1975)
                                                                                                                                           R1
                                                                                                                                                    j’.-!     I




Evaluation of Discrepancies in Patients’ Results-
An Aspect of Computer-Assisted Quality Control

Philip Whitehurst,1 Thomas V. Di Silvio,2 and Gaydzag Boyadjian1



A computer program has been devised to select those                                      It has been        advocated      that error detection        routines
clinical chemistry results that have a high probability of                            be included         in clinical laboratory           computer     systems
error for inclusion on a discrepancy report, which is                                 (1-5),     and estimated           that computer-assisted             error
printed on demand throughout the day. Each report                                     detection       could result in a savings               of $3500-$5000
entry is evaluated by a supervisor, who decides whether                               yearly in technologist             and physician         time in a 350-
to accept the result or to re-assay. With this program,                               bed hospital         (2). General       recommendations          for such
8.4% of all results were included on the report, 1.9%
                                                                                      programs        have been published              (1, 5), as has a de-
were re-assayed, and 0.83% were judged to be in error
and corrected. Checking results at the time of their re-                              scription      of an algorithm         for checking SMA 12 nor-
lease to the computer has led to earlier report delivery                              mal-abnormal           patterns     (3).
and more convenient timing of re-assays without com-                                      Our program          was devised with the main objective
promise of patient safety.                                                            of detecting        potential     discrepancies        in results on pa-
                                                                                      tients soon after analysis, so that re-assay and needed
AddItIonal       Keyphrases:        screening     of data on patients                  correction       could be performed            in most cases before
                                                                                       report     printout.      Additional       objectives      were to alert
    After implementation              of a laboratory            computer             laboratory        supervisory       personnel       to the presence        of
system in the clinical chemistry                section of our labora-                extremely        ill patients,      and to provide           a means for
tory, it became possible to enhance                  quality control by               finding      patients       for special projects           from time to
visual report checking.            As the numerous           separate       re-       time.
ports formerly          issued on a single patient                had now
been collated         into three daily nursing-unit                 reports           Method
followed      by a cumulative             six-day     summary          report         Hardware
each evening,         convenient      formats        were available        for
                                                                                          The present system consists of an IBM S/1800-S/7
comparing         results     on patients,          and discrepancies
                                                                                      combination    (IBM Corp., White Plains, N. Y. 10604).
were found that would have previously                        escaped       no-
                                                                                      The S/7 acquires       data on-line       from an SMA 6, an
tice.
                                                                                      SMA 12 (Technicon          Instruments        Corp., Tarrytown,
    However,      there were several obvious drawbacks                       to
                                                                                      N. Y. 10591), seven channels          of AutoAnalyzers,         and a
the visual checking          system. First, the amount               of time
                                                                                      Model S hematological          system    (Coulter      Electronics,
required     to check more than 2900 results daily led to
                                                                                      Inc., Hialeah,    Fla. 33014). Test requests             and input
a delay in report delivery. Second, discrepancies                        went
                                                                                      from off-line instruments          are performed       via cathode
unnoticed       by even the most experienced,                  well-moti-
                                                                                      ray tube displays     and keyboards         (Trivex,    Inc., Costa
vated     observers.        Third,     most       discrepancies          were
                                                                                      Mesa, Calif. 92627), as are most routine system func-
found after the afternoon               report printout.         As a con-
                                                                                      tions. The system       also utilizes     four IBM 2311 disc
sequence,      the re-assays       created      a burden      for evening
                                                                                      drives,   two high-speed        line printers       (350 and 600
personnel,      and, in some instances,             had to be deferred
                                                                                      lines per minute),    and various data and error loggers.
until the following          day, with resulting           danger to the
patient    and inconvenience           to the attending         staff.                Software
                                                                                         The basic system              software is the IBM Clinical Lab-
        Management      Services,   Thomas      Jefferson   University,   Philadel-
                                                                                      oratory  Management                  System   (CLMS)  and Multi-
phia,     Pa. 19107.
   2    Clinical Laboratories,      Thomas      Jefferson   University    Hospital,
and Department     of Pathology,     Jefferson    Medical College         of Thom-
as Jefferson University,    Philadelphia,      Pa. 19107.                                  Nonstandard       abbreviations     used: SMA and AA, Sequential
    Received July 31, 1974; accepted        Oct. 17, 1974.                            Multiple   Analyzer     and AutoAnalyzer      (Technicon Corp.).


                                                                                                            CLINICAL    CHEMISTRY,    Vol. 21, No.1,        1975   87
Programming        Executive       (MPX),     the former      having               plier and an allowed range of deviation.     For instance,
been extensively         modified.    The basic S/7 software                       if an SMA 12 result for lactate dehydrogenase          (upper
consist    of an IBM field-developed              assembler      lan-              limit of normal, 225 U/liter)     is to be compared       to a
guage program,       also extensively     modified.                                kinetic result for lactate   dehydrogenase    on the same
    The program      that is the subject of this publication                       specimen     (upper limit of normal,      90 U/liter),     the
was written     by one of us (G.B.) in Fortran             IV com-                 SMA 12 value is multiplied      by 0.4. This converted re-
puter language.      Limits for the various checks are con-                        sult must agree with the kinetic result within 15 U!
tained in a single file, and can be changed             readily via                liter. If the units for the methods being compared are
punched-card       input. A second file, the discrepancy                           directly comparable    (for example, if an SMA 12 albu-
file, contains   the pertinent       information      on those re-                 min result is being compared with an albumin value
sults that have failed a check.                                                    obtained   by electrophoresis),    a multiplier of 1.0 is
                                                                                   used.
Program Concepts                                                                      The        previous     value    check      is used        to compare          a re-
    It is obviously       impossible       for either a laboratory                 sult with the previous      result for the same test on the
supervisor      or a computer         to assess the validity             of a      most recent previous      specimen,      within four days. For
single isolated laboratory              result.      Thus,     individuals         instance,  serum    protein      results    differing by more
who evaluate         laboratory      results       (in the absence          of     than    1.5 g/dl from a previous            value will be written
knowledge       of the patient’s       clinical problems)           will or-       into the discrepancy       file. For some tests, the previous
dinarily    attempt      to compare        a result with results of                value check is more complex.                 For instance,        if the
related    tests performed          on the same specimen                  and      limit of 1.5 g/dl is accepted as the maximal allowable
with results of the same test performed                    on prior spec-          day-to-day      change in serum protein                concentration,
imens. If the result cannot             be compared          to other re-          this limit seems as valid for a patient with a serum
sults, it may still be questioned               if it is at an extreme             protein concentration         of 4.0 g/dl as for a patient with
of the range of the total patient                population.       The ex-         a serum protein concentration               of 9.0/dl. However, a
perience     and knowledge           of the individual            checking         patient whose serum urea nitrogen is 75 mg/dl would
will be brought        to bear on the decision              as to the va-          be expected to change more from day-to-day                          than
lid ity of the result.                                                             one with a concentration          of 15 mg/dl. Therefore, for
    This program        has been designed            to select those re-           some tests, the limits file contains three entries for
sults most likely to be in error, and present                      them in         previous      value     checks:    the allowable            low-range
an organized       format to the supervisor for further                    ac-     change (in absolute units), the allowable high-range
tion. The program          provides     four types of checks, des-                 change (in percent),          and the point of division               be-
ignated     as arithmetic,        discordance,          previous      value,       tween high and low range. For example, if a given
and alert. For each check, we set limits (Table 1). No                             urea nitrogen concentration            is less than 25 mg/dl, it
attempt was made to establish these limits by rigor-                               will be written       to the discrepancy           file if it differs
ous statistical        means; rather,           the limits        were set         from the preceding           value by more than 5 mg/dl.
subjectively      by the clinical         chemistry        section     head,       However, if a given urea nitrogen              concentration        is 25
and were based on known intralaboratory                          analytical        mg/dl or above, it will be written to the discrepancy
variance and his clinical experience.                                              file only if it differs from the preceding                   result by
    The arithmetic         check is used to compare                     more       more than 20% of the result being checked.
than two results from the same specimen                       that have a              For most tests, the previous               value check limits
demonstrable         mathematical         relationship.        At present          have been set to be rather narrow, on the basis of an-
it is used for “balancing”            electrolytes       and for check-            ticipated    change in 24 h, even though this check is
ing acid-base        calculations.          For balancing         electrolytes,    made on results obtained as much as four days pre-
a term    “DELTA”           is defined       as                                    viously. Setting such narrow limits allows the labora-
                                                                                   tory supervisor to exercise his judgment                 to determine
     DELTA       =    [K]     +      [Na]     -   [HC03]      -     [Cl-I    (1)   whether or not re-assay is truly needed, based on the
                                                                                   time interval or other knowledge he may have about
DELTA       values of less than 10.0 mmol/liter       or greater                   the patient.      Although      many valid values are thus
                                                                                   called to the supervisor’s        attention,      the risk of miss-
than 21.0 mmol/liter        are written   into the discrepancy
                                                                                   ing a true change is lessened.
file for eventual printout.       For checking acid-base cal-
                                                                                       The alert check is designed to find very large ana-
culation,    actual bicarbonate       and total CO2 are calcu-
lated from pH and p co2 by means of the Henderson-                                 lytical, entry, or transcription          errors, to alert labora-
Hasselbalch       equation,    and the calculated    values are                    tory supervisors       to very ill patients,          or to find pa-
                                                                                   tients of special interest. For instance,               a urea nitro-
compared to the values entered by the technologist.
Differences      of more than 2.0 mmol/liter        are written                    gen of less than 3 mg/dl or greater than 50 mg/dl is
into   the discrepancy            file.
                                                                                   written into the discrepancy          file.
   The discordance     check is used to compare two re-
                                                                                   Program          Operation
sults from the same specimen that have a defined
mathematical     relationship.    For   the discordance                                  After    verifying     test   results     for     a given      worklist       via
check, the limits file contains two entries, a multi-                              the     screen      of the    cathode         ray     tube,    and     printing       a

88     CLINICALCHEMISTRY,VoI.21,                  No.1,1975
                            Table 1. Limits for Discordance, Previous Value, and Alert Checksa
                                                                                                       Previous    value   limits
                                                     Discordance   check     limits             -                                           Alert   limits
                                                                                                     Low     Division          High
               Test                         Tested    against              Multiplier   Range       range      point          range   Lower            Upper

Glucose, SMA65                       Glucose, SMA12                            1.0      ±10         ±100          300         ±50%     50                500
Glucose, SMA12b                      Glucose, SMA6                             1.0      ±10         ±100          300         ±50%     50                500
Urea nitrogenb                       Creatinine, AAI                          13.0      ±10         ±5             25         ±20%      3                     50
                                     Creatinine, SMA12                        13.0      ±10
Creatinine, AAIb                      Creatinine, SMA12                        1.0      ±0.3        ±0.5            2         ±25%    0.3                7.5
                                      Urea nitrogen                            0.07     ±1.0
Creatinine,    SMA12b                Creatinine, AAI                          1.0       ±0.3        ±0.5            2         ±25%    0.3                7.5
                                     Urea nitrogen                            0.07      ±1.0
Uric acid, AAP                       Uric acid, SMA12                         1.0       ±1.0        ±1.5           8.0        ±25%    1.0               12.0
Uric acid, SMA12b                    Uric acid, AAI                           1.0       ±1.0        ±1.5           8.0        ±25%    1.0               12.0
Albumin, electrophoresisc            Albumin, SMA12                           1.0       ±0.6        ±1.5          N/A         ±1.5    1.5                6.0
Albumin, SMA12C                      Albumin, electroph.                      1.0       ±0.6        ±1.5          N/A         ±1.5    1.5                6.0
Protein, manual”                     Protein, SMA12                           1.0       ±0.8        ±1.5          N/A         ±1.5    4.0                9.0
Protein, SMA12”                      Protein, manual                          1.0       ±0.8        ±1.5          N/A         ±1.5    4.0                9.0
Sodium4                              N/At                                    N/A        N/A         ±6            N/A         ±6      120                150
Potassium4                           N/A                                     N/A        N/A         ±1.2          N/A         ±1.2    3.0                6.0
Chlorides4                           N/A                                     N/A        N/A         ±8            N/A         ±8       85                115
CO2 content,SMA6d                    CO2 content,calc.                       1.0        ±4.0        ±6            N/A         ±6       10                 40
Osmolality, serume                   N/A                                     N/A        N/A         ±15           N/A         ±15     265                320
Osmolality, urinea                   N/A                                     N/A        N/A         N/A           N/A         N/A     150               1200
Calcium, AASb                        Calcium, SMA12                           1.0       ±1.0        ±1.5          N/A         ±1.5    6.5               13.0
Calcium, SMA12b                      Calcium, AAS                             1.0       ±1.0        ±1.5          N/A         ±1.5    6.5               13.0
Phosphorus, AAI5                     Phosphorus, SMA12                        1.0       ±1.0        ±2.0          N/A         ±2.0    1.0                8.0
Phosphorus, SMA12b                   Phosphorus, AAI                          1.0       ±1.0        ±2.0          N/A         ±2.0    1.0                8.0
Iron’                                N/A                                     N/A        N/A         ±12           100         ±20%     20                250
Iron-binding   cap.’                 N/A                                     N/A        N/A         ±50           350         ±20%    200                600
Cholesterol, AAl’                    Cholesterol, SMA12                       1.0       ±40         ±50           300         ±30%    100                500
Cholesterol, SMA125                  Cholesterol, AAI                         1.0       ±40         ±50           300         ±30%    100                500
Triglyceridesb                       N/A                                     N/A        N/A         ±30           250         ±25%     40               1000
Bilirubin, AAIb                      Bilirubin, SMA12                         1.0       ±0.4        ±1.0           3.0        ±50%    0.2               10.0
                                     Bilirubin, for newborn                   1.0       ±1.0
Bilirubin, SMA12t                    Bilirubin, AAI                           1.0       ±0.4        ±1.0           3.0        ±50%    0.2               10.0
Bilirubin, for newbornb              Bilirubin, AAI                           1.0       ±1.0        ±2.5          10.0        ±25%    1.0               15.0
Lactate dehydr.,                     Lactate dehydr.,                         2.5       ±40         ±50           250         ±50%     50                600
                                        kinetic
Lactate dehydr.,       kinetics       Lactate dehydr.,                         0.4      ±25         ±25           150         ±40%     10               1000
                                        SMA12
Alk phosphatase,           SMA125     Alk phosphatase,                         2.4      ±20         ±50           125         ±50%     10                    300
                                         phenolphthalein
                                         monophosphate
Alk.phosphatase,#{176}                Alk.phosphatase,                         0.42      ±9         ±15            60         ±30%      5                    250
  phenolphthalein                       SMA12
  monophosp hate
Asp. aminotrans.,                     Asp. aminotrans.,                         1.8     ±20         ±25            75         ±50%      5                    300
                                         kinetic
Asp. aminotrans.,                     Asp. aminotrans.,                        0.55     ±10         ±20           100         ±40%      2                    750
  kinetics                              SMA12
Ala. aminotrans.                      N/A                                     N/A       N/A         ±20           100         ±40%      2                    750
Acid   phosphatase#{176}              N/A                                     N/A       N/A         ±2.0          10.0        ±25%     0.1              10.0
Amylase                               N/A                                     N/A       N/A         ±100          500         ±25%     20               1000
Creatine kinase5                      N/A                                     N/A       N/A         ±50           250         ±40%      5               1500
Acid-base group
pH                                    N/A                                     N/A       N/A         N/A           N/A         N/A     7.10                   7.65
                                      N/A                                     N/A       N/A         N/A           N/A         N/A     40                     135
pcop                                  N/A                                     N/A       N/A         N/A           N/A         N/A     10                      80
Actual bicarb.4                       N/A                                     N/A       N/A         ±8            N/A         ±8       9                      39
CO, content, calc.”                   CO, content,        SMA6                 1.0      ±40         ±9            N/A         ±9      10                      40
Base excess4                          N/A                                     N/A       N/A         ±6.0          N/A         ±6.0    -12.0             +12.0
Standard bicarb.4                     N/A                                     N/A       N/A         ±8            N/A         -8      10                      40
  “Arithmetic     check limits are contained       in text. For discordance       check, multiplier acts on result being tested, and prod uct is compared
with result listed as “tested        against,”   applying     respective    range. For previous value check, respective     low range or high range limits
will be selected depending         on relation of result being tested to division point. For alert check, values below respective          low or above re-
spective    upper limits fail check.
     mg/dl; g/dl;     “mmol/liter;     ‘mosm/kg;       fg/dl;      U/liter;   h mm Hg;    N/A, not applicable.
                                                                                                                  ‘.,

Table 2. Classification of Discrepancy File Entries
          (Representative 12-day Period)
                                                                    Percent
                                                         Percent      of die-
                                                         of total   crepancy
                                                           test         file
                                                          results    #{149}ntrl.s

Total test results                           30870
Total discrepancy file entries                 2579       8.35
Total errors found and                          255       0.93          9.9
     corrected (*)
Arithmetic          check    failures            580        1.9       22.5
     Electrolytes,          judged correct
     Electrolytes,          insufficient         132        0.4         5.1
       quantity        to reassay
     Electrolytes           correct     by      335         1.1       13.0
    reassay
 5Electrolytes in error by                       147        0.5         5.7
    reassay
  Acid-base calculations                          28        0.1         1.1
    judged correct
 5Acidbase    calculations in                     32        0.1         1.2
    error
Discordance          check failures
     Judged correct                             320         1.0       12.4
     Not enough to repeat                        16         0.1        0.6
 *Error      found          by reassay            53        0.2        2.1
Previous result check failure                                                       Fig. 1. Outline of program logic (Y = Yes, N = No). The
  Judged correct                                434         1.4       16.8          checking algorithm is automatically initiated immediately after
  Correct by reassay                              4         0.1        0.2          the technologist has visually reviewed the data on a CR1, and
 5Error found by reassay                          9         0.1        0.3          indicated approval
Alert check failures
  Judged correct                                438        1.4        17.0
 *Error found by reassay                          5        0.1         0.2
 5Entry errors                                    9        0.1         0.3          tient; the result; the type of check that has been
  Judgment not recorded                          37        0.1         1.4          failed; and the test code, date, time, and specimen
                                                                                    number     associated   with that result. If electrolytes
                                                                                    have not balanced, the DELTA             value is also printed,
                                                                                    together with the associated glucose and urea nitro-
                                                                                    gen values. In many instances, this allows rapid ex-
hard copy of those results, the technologist           releases                     planation    of elevated DELTA values. For failures of
the results into the master log file (Clinical Laborato-                            any check except discordance,         the most recent previ-
ry Management        System, IBM). It is at that time that                          ous result for that test (within four days) will appear.
the discrepancy      checks are automatically      performed.                       In case of discordance      check failure, the discordant
Results are released into the master log file, whether                              result appears, similarly       identified.    Ample spacing
or not they have been written            into the discrepancy                       between lines of the report allows it to be used as a
file, and are then available for reports or inquiry.                                log of action taken for each report entry.
    In addition to the limits already discussed, the lim-                              Once the report has been printed,            the laboratory
its file contains      a group of bits to identify        which                     supervisor    evaluates each entry. In many cases one
checks are to be performed            on a given result. The                        can decide immediately,       from the information       on the
checks are performed         in the order: arithmetic,      dis-                    report itself, whether      or not to re-assay. In other
cordance, previous value, alert (Figure 1). This order                              cases, before making      a decision, the supervisor        may
allows checks making the greatest number of com-                                    need to refer to other results for the patient. These
parisons     to be performed       first. If a result fails a                       are easily accessible via cathode ray tube screen. Fre-
check,    subsequent      checks are not performed          and                     quently, clinical information is obtained from the pa-
computer      time is saved.                                                        tient’s chart by the clinical pathologist or resident, or
    Results failing any check are written into the dis-                             the attending physician may be consulted.
crepancy file. At frequent intervals during the day,                                   After a re-assay, the supervisor       will update the
laboratory supervisors call for a discrepancy report,                               master log file containing     the result. If re-assay has
which prints out the contents of the discrepancy file                               shown the original     result to be correct, a two-digit
(as exemplified in Figure 2). This report contains the                              code will be-entered by the supervisor via cathode ray
name, identification  number, and location of the pa-                               tube,    causing    the     abbreviation      “CKD”    (for

90     CLINICALCHEMISTRY,Vol.21,             No.1,1975
PRINTEO        06/20/74

                                                                     THOMAS        JEFFERSON      UNIVERSITY             HOSPITAL                                                          XXX-XXXX-XXX
       TIMF    1753      HRS                                                       CLINICAL     LABORATORIES
                                                                                      DISCREPANCY       REPORT                                                                                     PAGF           S
(.UMPL F T F                                                                                     CHEMI   STRY

               -         CURRENT       RELEASED      RESULTS        FAILING         INDICATED        CHECK                            COMPARISON             RESULTS             OTHER      INFORMATION

CHECK          itO       CTC       X-NUMBER       SPEC      NO    LOCAT       AGE       SEX       TIME       RESULT           RESULT          lCD      CTC       DATE   TIME     PATIENT       NAME

PREy           CPK       CPK        0714618       0620366         451H        48Y            N    1039          2030            1624          CPK      CPK       0619   0825                 ROBERT           J

                                                                              0Y             M    0900           820                 93       CPK      CPK       0619   0900                        CARNAN
PREy           CPK       CPK        0708339       0620015         551C


915CR              RUN   SN6        0052704       0620097        USiA         bOY            M    0900            49                 1.7      CRT      CR1       0620   0900                   JOSFPH

AR ITH         CL        SM6        0054270       0620170        N6488        85Y            N    0900           106                                                                                JACOB
AR I TN        C02                                                                                0900            29
AR ITH         K                                                                                  0900            3.8
ARIIH          NA                                                                                 0900           141
AR ITH             BuN                                                                            0900            34
API IN         GLF                                                                                0900           181
DEL TA                                                                                                            9.7

ALERT          CL        SM6        0054205       0620071        1057*        49Y            F    0900             83                                                                         AL ICE      0

PREy           NA        SM6        0053298       0620030         811A        Sly            F    0900           143                136       NA       SM6       0619   1122      t         ELOISE

PREy               BUN   SM6        0053199       0620038         8258        21Y            F    0900             58                76       BUN      SM6       0619   0900                        PAuLETT

PREy               BUN   SM6        0703199       0620150        M222A        34Y            N    0900             20                14       BUN      SM6       0618   0900                   JAMES

0!SCR          CRT       SMA        0712034       0620339        11578        385’           N    0900           10.4                57       BUN      SN6       0620   0900                       REV RALP
ALERT          ALP                                                                                0900           354
AL FRI         GOT                                                                      ..        0900           335


015CR          CLF       SMA        0051292       0620010         5518        64Y            F                   342                359       GLU      5Mb       0620   0900
UISCR          CRT                                                                                0700            5.3                88       BUN      5Mb       0620   0900
PREy           LDH                                                                                0700           524                606       lOU      SNA       0619   0915
PREy           GOT                                                                                0715             33                64       GOT      SMA       0619   0915
ALERT          PHD                                                                                0715            8.7                6.9      PHI)     SNA       0619   0915
ALERT          IIRC                                                                               0715           12.1               11.5      URC      SMA       0619   0915


r)ISCR         CR1       SMA        0052670       0620126        0*248        69Y            F                     1.0                 .7     CR1      CR1       0620   0900                          LILY            0
ALFRT          ALP                                                                                0700           339


015CR          CRT       SMA        0706457       0620133        0*54*        41Y            F                     3.4               3.1      CRT      CRT       0620   0900                       JACKSON


015CR          PRO       SNA        0054239       0620134        0*56*        79Y            F                    5.5                6.5      T P      PEL       0620                               JULIA             N


015CR          CR1       SMA        0243188       0620092        11518        67Y            N                     1.2               30       BUN      SM6       0620   0900     #{149}    LOUIS


015CR          C,LF      SNA        0054262       0620165        M633R        37Y            F                   117                106       GLU      5Mb       0620   0900                          DORIS


I)ISCR         CR1       SMA        0054205       0620071        1057*        49Y            F                     1.2               29       RUN      SN6       0620   0900                  ALICE       0

                                                                         Fig. 2. Example of discrepancy report




“checked”)    to appear with the result on all subse-                                                        sis and cardiac                patients         comprise          a large segment                    of
quent reports. If an error has been found, entry of a                                                        our patient population. As expected, patients with
different  two-digit   code will cause the abbreviation                                                      renal failure, diabetic acidosis, cardiac failure, and
“LAB COR” (for “laboratory        correction”) to appear                                                     anoxia    accounted     for most high DELTA          values.
with the result. If the erroneous result has already                                                         Many of the patients with low DELTA            values were
appeared on a report, the nursing unit is additionally                                                       hypoproteinemic      or on intravenous  hydration.    Errors
notified by telephone.                                                                                       in sodium and chloride were about equal in number.
                                                                                                             An alternative      approach     to the arithmetic     check
Results                                                                                                      would have been to calculate the DELTA             for each
   The results from 12 consecutive days in May 1974,                                                         specimen as soon as it had been analyzed by the SMA
have been tabulated       (Table 2). A total of 255 errors                                                   6. That approach would have given the advantage of
were found and corrected, a rate of 0.83%. If the spec-                                                      immediate    checking. However, by delaying the check
imens judged in need of re-assay, but not available in                                                       for a few minutes, as in our system, the responsibility
sufficient quantity   for the re-assay, had an error rate                                                    for a decision not to re-analyze is placed with the su-
comparable    to that for those that were re-assayed,                                                        pervisor, rather than the operator. Throughput     is not
there would have been an estimated          329 errors, or                                                   adversely affected, since the specimen would be re-
1.07% of all the results.                                                                                    analyzed at the end of the worklist in either case.
                                                                                                                Most of the discordance check failures judged by a
Discussion                                                                                                   supervisor   to be correct were creatinine-urea    nitro-
   Most of the discrepancies were found by the arith-                                                        gen discrepancies     in patients  who were on dialysis,
metic check of electrolytes,       not surprising,    since                                                  pregnant patients, and patients with liver disease or
about 25% of all the assays performed      in the clinical                                                   pre-renal   azotemia. Next most common were differ-
chemistry  section are for electrolytes and since dialy-                                                     ences for cholesterol    as measured by the direct SMA

                                                                                                                                     CLINICAL        CHEMISTRY,           Vol. 21, No.1,1975                      91
12 method and the AutoAnalyzer                          I (isopropanol-ex-    consistent           in the subjective            clinical    judgment       of the
tracted) method in patients with                     bilirubinemia   or tn-   examiner.
glyceridemia. The discordance                       errors substantiated        Because of these residual apparent errors, one can-
by re-assay were usually glucose                    assays performed on        not omit visual checking of reports. However, the vi-
both the SMA 6 and SMA 12.                                                     sual check does not now interfere           with report deliv-
   At first glance, it would appear               that very few previ-         ery or laboratory    operation. The nursing unit reports,
ous result-check            failures     were re-assayed. In our sys-          printed at 5:00 a.m., 1:00 p.m., and 4:00 p.m. are no
tem, previous results are made available on worklists,                         longer    checked     visually,     because no correlations
and in most cases re-assays will have been initiated                           could be made visually that have not already been
and completed              by the bench technologist. Thus, the                checked by the program.             Because of the very low
previous result            check serves only as a double check                yield of additional       errors, the summary       reports are
here.                                                                          not checked before delivery; however, the laboratory
     The alert check          is also    expected     to   yield only a low   copy of these reports is being checked by the evening
return      of   errors,    because         the main function of this         supervisor in chemistry          while the report is being de-
check      is to discover      very     large entry errors. It will also      livered, saving nearly an hour in delivery time. This
alert    the supervisors to very ill patients.                                copy is then routed to the head of the clinical chemis-
     Although    for the program under discussion,                   limits   try section, who rechecks it early the next day. Be-
were set subjectively,             the program can use limits set             cause the routine checking has been done, the visual
by any method, as these limits are contained              in an               checkers can apply their expertise to discovering            re-
easily accessible file. It is possible that limits set by                     lationships    among results that are inconsistent         with
sophisticated    statistical techniques      and then opti-                   their clinical judgment.
mized could reduce the number of results selected for                             The program described has made it possible to dis-
visual checking. We have elected instead to use the                           cover many errors that otherwise             would have been
more direct, but admittedly       subjective     approach, in                 missed. It has benefited the operation           of the labora-
the interest   of a more immediate           solution   to the                tory, allowing errors to be found with minimal effort,
problem of checking a truly overwhelming            number of                 at a time convenient for re-assay. In this manner, dis-
results.                                                                      ruption     of laboratory      routine   and salary costs for
     While it
            is quite easy to document   the number of                         overtime are avoided. Most important,            errors are de-
errors found and corrected by re-assay, an accurate                           tected and corrected before a clinical decision can be
count of the errors missed by the program is impossi-                         made that might be deleterious             to the patient. We
ble. Published estimates of laboratory error rates are                        conclude that this program is an effective use of the
 understandably       uncommon        (6, 7). We attempted         to         computer in laboratory        quality control.
approach this problem by making an admittedly                    sub-
jective survey of in-patient         six-day summary reports                  References
 during the same 12-day period. These reports were
 checked by one of us (T.V.D.)            for residual apparent               1. Council         of the Association       of Clinical   Pathologists,   Data    Pro’
 errors. Fifty-eight     possible errors were found (0.2%).                   cessing     in Clinical   Pathology.      J. Clin. Pathol.    21, 299 (1968).
                                                                              2. Krieg,      A. F., Johnson,          T. J., McDonald,     C., and Cotlove,  E.,
 Of these, 28 represented        electrolytes   (seven sets) the              Clinical Laborator           Computerization.       University    Park Press, Bal-
 DELTA      of which had been questioned              by the pro-             timore,     Md.,    1971, pp 86-87.
gram, but which had not been re-assayed, either be-                           3. Poletti,   I. J., Zack, J. F., and Mueller, T. J., Computer     control in
 cause of lack of sample or of a supervisor’s              decision.          the clinical laboratory.     Amer. J. Clin. Pathol 53, 731 (1970).
                                                                              4. Walter, A. R., What to look for in a computerized         laboratory    in-
 The remaining       30 had not been detected by the pro-
                                                                              formation    system, Part I. Lab. Med. 3/1, 32 (Dec. 1972).
 gram. Thirteen        involved    discordances      on different             5. Walter,      A. R., What to look for in a computerized            laboratory      in-
 days, such as total protein by SMA 12 of 8.0 g/dl, fol-                      formation      system, Part 11. Lab. Med. 4/1, 32 (Jan.            1973).
 lowed a day later by a manually-obtained                total pro-           6. Reed, A. H., and Henry, R. J., Accuracy,     precision and control
 tein value of 5.6 g/dl. Ten represented            irregularities            charts.   In Clinical Chemistry,   Principles and Techniques,      2nd
                                                                              ed., R. J. Henry, D. C. Cannon,   and J. W. Winkelman,    Eds. Harper
 in trends, without enough day-to-day             change to have              & Row, Hagerstown,     Md., 1974, p 298.
 exceeded the previous value limits. The remaining                            7. McSwiney,       R. P., and Woodrow,   D. A., Types of error              within       a
seven were multitest             patterns that appeared to be in-             clinical laboratory.    J. Med. Lab. Technol. 26, 340 (1969).




92       CLINICALCHEMISTRY,             Vol. 21, No. 1. 1975

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:29
posted:5/20/2010
language:English
pages:6