Streamline your automated hematology laboratory.

How can the hematology laboratory maintain high quality in this era of severe cost containment? Only by improving efficiency through redesigning the organization and delivery of laboratory services. This series, presented in a roundtable format, tells the tale of three laboratories describing their efforts to reengineer the automated hematology laboratory to improve efficiency.

The following transcript is excerpted from their discussion:

* MLO:

How can reorgnization of laboratory workflow improve efficiency?

* CORNBLEET:

Our laboratory has very open architecture, facilitating communication and movement between workstations. Thus, we can float technologists between workstations to respond to workload demand. We placed the instruments near related functions that could be performed by the operator on flagged samples, such as slide preparation and phase platelet counting.

In addition, the instruments are near the front of the laboratory, closest to the accessioning area.

* FERNANDES:

The automation of manual and repetitive tasks has provided savings of technologist time and also increased the speed of functions. Thus, technologists are free to perform functions requiring intellectual activity and decision making.

* MIERS:

Historically, we had a dedicated Star section within hematology with separate instruments and related workstations. Some reconstruction allowed us to dismantle the dedicated Stat area and locate all instruments close to each other. Stat and expedite samples now can be prioritized on all three instruments, minimizing the Stat queue time.

To meet the turnaround time (TAT) demands of the Hem/Onc Clinic, we began "pre-accessioning" those clinic samples into our laboratory information system (LIS) based on the clinic appointment schedule. This allowed the samples to be drawn and bar coded in the clinic using preprinted labels. Ultimately, we installed a pneumatic tube directly from the Hem/Onc Clinic into the hematology laboratory, bypassing specimen receiving altogether. An LIS terminal has been installed in the clinic for accessioning samples from patients not on the appointment schedule.

Other workflow improvements include:

1) An automatic printout by the instruments every 30 minutes to inform the technologist of incomplete adult and pediatric hem/onc samples

2) Repositioning the microscopes near the instruments

3) Automation of reticulocytes on the primary hematology instrumentation, decreasing the queues at the microscopes

4) Scheduling technologists to "teams" rather than "workstations" so the team members can float within a work area of related functions to expedite the workload.

* MLO:

What algorithms or rides do you use to identify spurious automated hematology result? Which manual procedures do you use for verification, and what is the impact of these flagging rules on workload processing?

* CORNBLEET:

Although automated hematology instruments are precise and accurate, some samples have spurious results and need to be analyzed by manual methods. Each flag, though, should detect an inaccuracy that is potentially clinically significant, and the specific action to be taken must be defined.

I have listed the instrument flags that we use in our laboratory in Table 1. We attempt to verify flagged white blood cell counts and platelet counts by slide estimate; if there is disagreement, then a hemocytometer count must be performed. Hemiglobinazide hemoglobin and centrifuged hematocrit analyses also are available.

We estimate approximately 13% of our automated hematology workload is flagged, in most cases because of a low platelet count ([less than]50 x [10.sup.9]/L), low white count ([less than]1.0 x [10.sup.9]/L), or a leukocyte differential-related flag.

* FERNANDES:

Instrument flags and/or messages are used in conjunction with reference ranges to identify abnormal or spurious results. Abnormal specimens are examined microscopically and, if necessary, macroscopically (e.g., for clots/hemolysis). Manual methods such as chamber counts and hematocrit are available to verify results when a discrepancy exists between the CBC and blood film appearances. Specimens above the linearity limits of the instrument are handled by appropriate dilution of the sample.

In addition, because of the high incidence of instrument flags and/or messages in some population groups (e.g. Oncology Clinic), the rules for blood film review are modified for different groups.

* MIERS:

The need for review of hemogram parameters is limited primarily to the platelet (PLT) count, which we resolve by smear estimate. Platelet counts [less than]10,000/[[micro]liter] are reported as "less than 10,000/[[micro]liter], confirmed by smear estimate." We occasionally perform phase PLT counting as a confirmatory procedure.

A significant feature of our instrumentation for the pediatric population is the dual white blood cell (WBC) count performed in two different cell counting channels. Nucleated red blood cells (NRBCs), clumped PLTs, etc., do not interfere in the second WBC count. The two counts are checked internally against each other. When interferences in the first count are present, the second count is substituted by the instrument.

Likewise, the instrument's unique direct measure of the cellular hemoglobin concentration mean (CHCM) allows for back-calculation of hemoglobin (HGB) when interference occurs in the photometric hemoglobin determination [CHCM = HGB/(RBC x MCV) x 1000]. Because of these instrument features, we have been able to virtually eliminate back-up WBC counting and HGB measurement. We do perform a rare spun microhematocrit (HCT) when the instrument HCT and the back-calculated HGB do not match.

Other error-limiting protocols for our laboratory include:

1) Peripheral smear review for new patients if significant distribution abnormalities are indicated, including WBC count [greater than]20,000/[[micro]liter] or PLT count [less than]100,000/[[micro]liter]

2) Delta checks, requiring confirmation if significant change occurs in the patient's HCT or PLT count

3) Panic or crisis values, requiring confirmation and immediate call-back of abnormal results.

Table 1

Suspect (distributional) flags

WBC                RBC                           Platelets

Blasts             RBC agglutination             Large platelets
Immature GRAN      Turbidity/Hb interference     Small platelets
Left shift         Iron deficiency               Platelet clump
Atypical lymph     Hb defect                     Micro RBC/platelet
                                                  interference
NRBC               Fragments

* MLO:

How can data management systems streamline identification of samples that need further verification or manual procedures?

* CORNBLEET:

We continually process the flagged samples in the IIS and place the results in the LIS, even if they are unverified. This helps to pass result verification to technologists quickly in other parts of the laboratory.

* FERNANDES:

Identification of samples requiring further verification is streamlined by embedding rule based algorithms in the information system. This is further facilitated by presentation of data to simplify decision making. Thus, previous results are shown in reverse chronological order and degrees of change may be computed automatically to facilitate delta checking. Color coding of results to represent abnormal high, abnormal low, or abnormal delta checks also makes it easier to identify samples requiring verification.

* MIERS:

Automatic verification, or result transmission to the host computer with validation by exception based on user-defined rules, is available using our primary instrumentation. We have chosen not to use this option, however. Using our LIS instead, delta flags are generated when current data are significantly different from previous results, even if several weeks or months have elapsed since the patient's last clinic visit. Panic or crisis values also are flagged.

* MLO:

What is your laboratory's philosophy regarding performance of leukocyte differentials? How did you gain clinician acceptance of the automated differential?

* CORNBLEET:

The slide differential is a separate test, uncoupled from the CBC. When a CBC with differential is ordered, we provide the automated differential, which is a more precise quantitative measurement of the major leukocyte sub-populations.

We perform a slide review or slide differential using flagging criteria that have a high probability of detecting clinically significant abnormalities. We do not use rules with poor specificity because such flags would result in a low yield of true positive samples in a tertiary care hospital. Thus, we do not guarantee that our review criteria will detect all samples with abnormal cells, particularly when these cells are few.

Ultimately, it is the physician's responsibility to order a slide differential when hematologic malignancy is suspected. The clinical history may be the most important factor in making this decision.

When the five-part differential was introduced three years ago, we distributed written material to all physicians showing the performance of the automated differential and sensitivity and specificity of the flagging algorithms we would be using. We encouraged them to consider the clinical utility of the slide differential.

Finally, we discussed problems in the use of band enumeration to diagnose bacterial infection, citing the imprecision of the 100-cell differential, the lack of reproducible identification, and the clinical non-specificity of band elevation.

Clinician acceptance was not hard to gain because the managed-care market was significant in Northern California at that time. Clinical pathways were being written that included decreased laboratory testing, and the slide differential appeared to be a reasonable target for reduction. Changes in the requisition slip helped to reinforce that the slide differential is not a routine part of the CBC.

* FERNANDES:

When a CBC is requested, an automated white cell differential count is provided to maximize the value of the CBC. However, I believe that in an era of increasing cost containment, discrete testing will become more prevalent when only limited information (e.g., hemoglobin or platelet count) is required.

The literature indicates in samples with normal populations of white cells, the automated differential is more precise and accurate than the microscopic differential. In the presence of abnormal populations, instruments generate flags and messages that can be used to select samples for microscopic review.

Since these flags and messages are highly sensitive, but lack specificity, they should be tailored to the requirements of the laboratory and used in conjunction with other parameters for sample identification.

Physicians' acceptance of the automated differential was obtained by convincing them of the improved accuracy and precision for normal populations and the existence of laboratory protocols to ensure that abnormal populations are accurately enumerated and reported. If a physician wanted a microscopic differential specifically, however, this had to be written out and justified.

* MIERS:

Leukocyte differentials are provided from our laboratory only as an ordered test. The WBC differential is performed as an automated differential and both the relative and absolute numbers are reported.

If the cells must be enumerated manually, only relative numbers are reported. We do not perform differentials if the WBC count is [less than]1,000/[[micro]liter].

Our instrument actually counts a 6-part differential (neutrophils, lymphocytes, monocytes, eosinophils, basophils, and large unstained cells, or LUCs). The LUC count is significant in that it separates abnormal mononuclear cells from other cell populations, thus allowing smear scans with appended comments versus manual enumeration of specific cell types (i.e., atypical lymphocytes).

Additionally, our analyzer has a low rate of suppression of automated data. Whether an instrument presents or suppresses "suspect" data seems to be based on the philosophical view of the manufacturer. The suppression of automated data forces a manual differential count, whereas the release of data with appropriate flagging man, dates the need for careful data scrutiny and possible blood film review.

When we switched to the automated differential, we presented the technology to multiple physician groups. We also distributed a one,page written summary of the technology and our policy changes to all physicians. This educational process was key in gaining physician acceptance.

Added value with the automated differential includes:

1) The ability to release the ANC as a preliminary differential to the hem/onc clinicians. This allows the clinician to make treatment decisions about the patient early before the complete differential results are available. The ANC is accurate and can be verified in most cases even if other differential flags warrant a smear review. Exceptions include a WBC count [less than]1,000/[[micro]liter] or a "no fit" instrument flag.

2) Faster TAT because of the labor-intensive nature of the manual differential count.

3) Generation of a smear scan if a "blast" flag is triggered, even if a differential count is not ordered. This adds fewer than five additional scans per day to our workload and allows us to detect occasional abnormalities that may have been missed otherwise. It's possible to use the "blast" flag in this way because of the relative specificity of this flag.

* MLO:

Do you have data documenting the sensitivity and specificity of the rules you use for flagging samples for slide review and/or slide differential?

* CORNBLEET:

We studied the performance of the instrument in flagging "abnormal" samples and in flagging samples with immature granulocytes (IG = metamyelocytes, myelocytes, or promyelocytes). Abnormal samples were ones that contained any nucleated red cells, blasts, dysplastic neutrophils, malignant lymphoid cells, [greater than]5% reactive lymphocytes, megakaryocyte nuclear fragments, or platelet clumps, as noted on a 400-cell slide differential. Instrument flagging criteria for abnormal samples included the differential-related suspect flags (except for flags related to immature granulocytes or bands), monocytes [greater than or equal to]40%, basophils [greater than or equal to]3%, or absolute lymphocytes [greater than]5 x [10.sup.9]/L.

As a separate study, we looked at samples without abnormal cells, but varying numbers of IG, assessing the performance of the instrument immature granulocyte flag. The control ("negative") group for specificity contained a wide range of quantitative abnormalities. The number of bands was not considered in either of these studies.

Specificity of the flagging criteria was excellent: 91% for the flags relating to abnormal cells, and 96% for the immature granulocyte flag. Sensitivity was 77.1% for flagging abnormal samples, and increased to 81.4% when only samples with [greater than or equal to]1% abnormal cells on the 400-cell differential were considered. About half the abnormal samples missed by our flagging algorithm consisted of low density abnormalities (abnormal cells [less than or equal to]5%).

Other unflagged abnormal samples with larger numbers of abnormal cells included one sample with nRBC, two samples with blasts (one acute lymphocytic leukemia, one acute myelocytic leukemia), four samples with malignant lymphoid cells, and six samples with 5%-10% reactive lymphocytes. Sensitivity of the immature granulocyte flag was 38.5% for IG[greater than]0%; 84.6% for IG[greater than or equal to]1%; 96.0% for IG[greater than or equal to]2%, and 100% for IG[greater than or equal to]5%.

* FERNANDES:

The efficiency of the algorithm for blood film review using instrument messages was evaluated on a random group of 220 patients. When reference intervals were used as action limits, the sensitivity was 93% and the specificity 47% for detection of any morphological abnormality. Using expanded action limits, the sensitivity remained at 93%, but the specificity increased to 73%. The positive predictive value was 73% and the negative predictive value was 92%. Morphology scans are performed on approximately 30%-40% of specimens, and differential counts are performed on 8%-12% of specimens.

* MIERS:

Using the rules we've established for review of instrument flags, the sensitivity and specificity were 92.9% and 86.1%, respectively, when 100 abnormal samples were tested. Two samples were morphologically false negative: The first had a slight left shift with 20.3% bands (ANC, 12.8 x 109/L); the second had 4.8% immature granulocytes (2% metas, 2.8% myelos) and one NRBC/100 WBCs (ANC, 14.1 x [10.sup.9]/L). In both cases, the elevated ANC was an indicator of neutrophil abnormality.

* MLO:

When is assessment of red cell morphology clinically important? Can specific instrument values or flags identify the need for evaluation of red cell morphology?

* CORNBLEET:

We decided to screen slides for very specific red cell abnormalities and report a complete description only when we see findings that could lead to an unrecognized diagnosis (see Table 2).

Otherwise, we enter standard comments such as "red cells appear normal." Physicians may specifically request a "complete red cell morphology" if warranted by the clinical problem. Only about two to three requests for evaluation of red cell morphology occur each month.

In the tertiary care setting, we haven't found any instrument flags to be of value in identifying which samples should have the red cells examined manually.

* FERNANDES:

We attempt to document any abnormalities of red cells at least once on the patient's records. We have found the hematology analyzer instrument flags of some use to identify samples for evaluation of red cell morphology when used in conjunction with numerical action limits. Some red cell abnormalities (e.g., hereditary elliptocytosis) may not be detected by instrument flags or messages. If they are clinically significant, however, such as having associated anemia, they will be picked up by instrument flags and messages.

* MIERS:

The instrument's measurements of RBC size and hemoglobin content are much more accurate than a technologist's visual impression of these parameters. What the instrument cannot do is flag for RBC inclusions. Therefore, we use combinations of instrument flags for RBC size and hemoglobin content to indicate the necessity for RBC smear review. This is a reasonable approach since inclusions typically are seen in conjunction with abnormally sized/shaped RBCs.

Our concern, however, is that this algorithm is very non-specific and less than optimally productive in identifying clinically important RBC abnormalities. We are beginning to look at our current guidelines for RBC review with intent to modify/streamline them.

Probably the most specific and clinically significant RBC flag is a "small red cells" flag when seen in conjunction with a "low PLT count" flag. This combination indicates that the technologist should confirm a low PLT count and the possible presence of schistocytes on the peripheral smear.

Table 2 Stanford Laboratories: Policy for red cell morphologic evaluation

A complete description of RBC morphology is performed when specifically ordered by the physician. However, when slides are examined because either a slide review or slide differential is ordered, a complete description of RBC morphology is performed only if an abnormality is found that could lead to an unrecognized diagnosis, such as:

1 Hemolytic anemias ([greater than]5% polychromatophilic red cells)

A. Microangiopathic hemolysis ([greater than or equal to]1% schistocytes plus helmet cells and the presence of any number of spherocytes)

B. Oxidative-drug-induced hemolysis ([greater than or equal to]2% bite cells and the presence of any number of spherocytes)

C. Autoimmune hemolysis ([greater than or equal to]5% spherocytes, except newborn and hyposplenic)

2. Hereditary anemias

A. Hemoglobinopathies (any sickle cells or Hgb "C" crystals)

B. RBC membrane defects ([greater than]10% of any individual poikilocyte not listed above; e.g., spherocytes, elliptocytes, stomatocytes)

C. RBC enzyme deficiencies ([greater than or equal to]5% spherocytes, except newborn and hyposplenic; [greater than or equal to]2% bite cells; [greater than or equal to]5% blister cells)

3. Myelodysplasia

(MCV [greater than or equal to]110 fL, dysplastic nRBCs, macrocytes with hypochromia)

4. Myeloma 

([greater than or equal to]3+ rouleaux)

5. Severe iron-deficiency anemia

(moderate to marked hypochromia)

6. Heavy metal poisoning 

(microcytosis with [greater than or equal to]1% coarse basophilic stippling)

7.Bone marrow fibrosis with myeloid metaplasia

([greater than or equal to]5% teardrop cells)

8. Hemoglobin [less than or equal to]7 g/dL

(Severe anemia possibly with unknown etiology)

Standardized red cell morphology comments include the following:

1. "Red cells appear normal." (Use when insignificant numbers of abnormal cells are seen.)

2. "Marked poikilocytosis or polychromatophilia is not present." (Use when small numbers of abnormal cells are seen but findings are not significant to trigger a complete RBC morphologic description.)

3. "No significant change from previously reported RBC morphology." (Use when abnormal findings trigger a complete RBC morphologic description, but current findings are nearly identical to those previously reported.)

4. "Red cell morphology consistent with hyposplenic state."

A. Use as concise description when Howell-Jolly bodies are present, accompanied by small numbers (e.g., [less than]5%-10%) target cells, acanthocytes, spherocytes, and Pappenheimer bodies. An occasional ([less than]1%) schistocyte or helmet cell also may be seen. B. If any of the cell types above are increased beyond 5%-10%, describe RBC morphology completely rather than assuming all of the findings are due solely to splenectomy.

Joanne Cornbleet is associate professor and medical director of the hematology laboratory at Stanford University Medical Center in Stanford, Calif. Bernard J. Fernandes is head of Hematopathology and Blood Transfusion Services, Mount Sinai Hospital, as well as an associate professor at the University of Toronto in Toronto, Ontario. Martha Miers is executive director of the diagnostic laboratories at the Vanderbilt University Medical Center Hematology Laboratory in Nashville, Tenn.