Objective: Band neutrophil counts are helpful for the detection of neonatal sepsis. Our goal was to optimize our laboratory process to detect elevated band neutrophils whenever present, without performing unnecessary manual differential counts.
Methods: We retrospectively tracked consecutive complete blood count with differential (CBCD) orders from neonatal intensive care unit (NICU) specimens according to the level of manual review. We identified specimens for which manual differential counts were performed due to telephone request and categorized them according to band percentage.
Results: Our laboratory routinely performed manual scans on 96% of NICU specimens, but only 87% of specimens required manual differential counts. Abnormal band counts were present in 5 of 70 specimens for which manual differential counts were requested because of clinically suspected sepsis. We elected to perform universal manual scans on all NICU specimens.
Conclusion: The revised laboratory process could be useful for other hospitals treating neonatal patients at risk of sepsis.
Band neutrophil counts are used as part of sepsis scoring systems1–4 for the detection of neonatal sepsis. At our institution, the trend of the band count is used in conjunction with other clinical and laboratory parameters to evaluate the probability of culture negative sepsis in neonates. Modern automated hematology instruments provide an accurate and precise total neutrophil count but do not report band numbers. Instrument flagging for samples with increased bands has been characterized as unreliable in the literature.3,5,6 One way for the physician to obtain a band count is to request a manual differential count whenever a band count is desired. However, manual leukocyte tabulation is a time-consuming, labor-intensive process that adds a significant laboratory cost to the complete blood count3 and reduces laboratory productivity.7 An alternative to the manual differential count is a qualitative manual scan, which involves the examination of several microscopic fields without enumeration of cell types. The manual scan can be followed by a manual differential count when abnormalities are present on the scan. Most laboratories perform manual scans for specific purposes, which vary by laboratory.7
The importance of band counts for hospitalized neonates may lead to a “standing order” request for manual differential counts. Alternatively, the laboratory technologist can perform a manual scan of the blood film and then proceed to enumerate the cell types if significant band forms are present. We evaluated the effectiveness and efficiency of our laboratory process for blood film review of neonatal intensive care unit (NICU) specimens. We asked the following questions: (1) How frequently were we performing manual scans and manual differential counts on NICU specimens? (2) How frequently did our laboratory process miss elevated bands which were only picked up after a telephone request for a manual differential count? (3) How would “standing order” manual differential counts for NICU specimens impact our hematology lab workflow? (4) How would universal manual scans for NICU specimens impact our hematology lab workflow? Our objective was to optimize our laboratory process to detect elevated bands whenever present, without performing unnecessary manual differential counts.
Definitions of Terms
Complete blood count with differential (CBCD): A complete blood count (CBC) performed on an automated analyzer, in which the WBC differential count is determined either by the automated analyzer or by a manual differential count, if triggered by numerical abnormalities or instrument flags.
Automated differential count: A CBCD in which the WBC differential count is determined by the automated analyzer without manual review of a film.
Manual differential count: An examination of a film in which a technologist reviews the film microscopically and performs a formal leukocyte differential count.
Manual scan of film: An examination of a blood film in which a technologist reviews at least 10 microscopic fields to confirm the absence of significant abnormality but does not perform a differential count.
Manual review: A manual scan of a film and/or a manual differential count.
Automated blood counts were performed on the Beckman Coulter GenS or LH785 hematology analyzer (Beckman Coulter, Fullerton, CA).
Laboratory Process for Manual Review of CBCD Specimens
Prior to this study, our laboratory process for manual review of all CBCD specimens took place as follows. When a CBCD was ordered, an automated differential count was released except in the following circumstances. A manual differential count was required when the WBC count was above 18 × 109/L or below 3 × 109/L, or when the analyzer flagged the specimen as suspicious for blasts or variant lymphocytes (Table 1, Section A). A manual scan was triggered by any of the criteria listed in Table 1, Section B. Namely, a manual scan was triggered by instrument flags for immature granulocytes or nucleated red cells (NRBC), by elevated proportions of monocytes or eosinophils, and by certain erythrocyte and platelet measurands. If any of the abnormalities listed in Table 1, Section C were found on scan, the technologist would proceed to perform a manual differential count. Such abnormalities include bands elevated above the reference interval, immature granulocytes, variant lymphs, blasts, and NRBC. The reference interval for bands in our laboratory is 0% to 15%.
A – Criteria for which manual differentialcountis required
WBC >18 × 109/L or WBC< 3 × 109/L
“Variant lymphocyte” flag
B – Criteria for which manualscanis performed
Instrument flag for bands/immature granulocytes*
Instrument flag for nucleated red blood cells (NRBC)
Numerical abnormalities or instrument flags for erythrocytes and platelets
*accompanied by Neutrophils >80% or Neutrophil # >8.5 × 109/L
C – Criteria for which manual scan is followed by manual differential count
Any immature granulocytes – promyelocytes, myelocytes, or metamyelocytes
The automated differential count would be released without any manual review if the specimen had none of the abnormalities listed in Table 1, Sections A and B. Otherwise, a manual scan would be performed. Normal findings on a manual scan would permit the technologist to release the automated differential count. Specimens that had manual scans due to immature granulocyte flags but showed no elevation in bands would be reported with a comment indicating “bands/immatures within reference range.”
We retrospectively tracked all CBCD orders originating in our NICU between July 2005 and July 2008, by searching our laboratory information system. We categorized the CBCD specimens according to the level of manual review performed. Our laboratory process, described above, was employed during the study period. From lowest to highest, the levels of manual review were none, manual scan only, and manual differential count. We further categorized specimens that had manual scans according to whether normal or abnormal findings were seen on scan. When normal findings were obtained on scan, the automated differential was released. Specimens with abnormal findings on scan had manual differential counts performed, as described above.
We sought to determine whether any specimens with elevated band counts were missed by the automated analyzer, escaped manual review, and were thus reported without the desired information. In order to estimate retrospectively whether this problem was occurring, we took the following strategy. We identified specimens for which the automated differential count was initially released, but a manual differential count was subsequently performed because of a clinician’s telephone request. We categorized these specimens according to whether the percentage of bands obtained was within or above the reference interval (0%–15%).
Use of Manual Review for NICU Specimens
We reviewed the outcomes for 14303 consecutive CBCD orders from the NICU which resulted during the study period (Figure 1). The specimens were obtained from 3039 different NICU patients. Of the 14303 orders, 10051 specimens (70%) directly required manual differential counts by the criteria listed in Table 1, Section A. Manual scans were performed on an additional 3623 specimens (25%), per the criteria listed in Table 1, Section B. The remaining 629 specimens (4.4%) required no manual review according to our laboratory process and were resulted with the automated differential count only.
Outcomes for all CBCD specimens received from the NICU during the study period.
Of the 3623 specimens that initially had manual scans, 2442 specimens (17% of total) required a subsequent manual differential count because of elevated bands or other abnormalities listed in Table 1, Section C that were seen on the scan. The remainder of the 3623 specimens that had manual scans, namely 1181 specimens (8.3% of total), did not require manual differential counts. Figure 2 depicts the relative workflow distribution for NICU CBCD specimens during the study period.
Detection of Elevated Band Counts After Telephone Request
Manual differential counts were routinely ordered by telephone, if not already reported, for newly admitted patients with suspected early onset sepsis. During the study period, out of 1810 specimens that did not have manual differential counts per our laboratory process, there were 70 specimens for which manual differential counts were requested by telephone. Five of these already had a scan performed, which was read as less than 15% bands, and this was confirmed by the manual differential count. The other 65 were requested from among the 629 specimens that had no manual review.
Of the 70 specimens for which manual differential counts were performed solely because of telephone requests, band counts were found to be elevated above the reference interval (0%–15%) in 5 specimens. Thus, abnormal band counts were missed in 5 (7.1%) of these 70 specimens failing to meet the threshold for a manual scan or differential. The band range was found to be between 16% and 34% on the manual count. In 4 of these 5 cases, the patients would have received a full course of antibiotics even if the band count was within the reference interval because of other factors including elevated C-reactive protein (CRP), hypotension, hypoglycemia, neutropenia, or respiratory distress. In 1 case, antibiotics were stopped after a negative blood culture (17% band count). Because telephone requests were made for a subgroup of patients with selected clinical characteristics, the false-negative rate of 7% cannot be extrapolated to the larger number of specimens not flagged by the instrument.
Relative workflow distribution for NICU CBCD specimens during the study period.
Use of Manual Review for NICU Specimens
Under our existing laboratory process, we were performing manual scans and/or manual differential counts on nearly 96% of NICU CBCD specimens, based on instrument flags. Therefore, universal “standing order” manual scans would not impact our workload significantly. On the other hand, we were performing manual differential counts on only 87% of NICU CBCDs. During the study period, there were 1810 specimens not requiring manual differential counts, including 629 specimens requiring no manual review and 1181 specimens requiring manual scans only. Of the specimens requiring manual scans, approximately one-third did not require manual differential counts. This suggested that “standing order” manual differential counts would add hundreds of manual differential counts to our laboratory workload annually.
These findings are specific to the NICU patient population. Neonatal intensive care unit specimens require manual review more frequently than other patient groups, partly because of features common to the blood of neonates. Morphologically immature lymphocytes, granulocytes, and NRBC are seen more frequently in neonates, and these cell types trigger instrument flags, which require manual review. Furthermore, the nature of the specimen and pre-analytical variables related to collection in Microtainer (BD, Franklin Lakes, NJ) tubes contribute to higher rates of instrument flagging.
Toward the end of the study period, we began using a newer generation analyzer, the Coulter LH 785 (Beckman Coulter), which has improvements in false-positive flagging, thus allowing for even lower rates of manual review.8 We repeated the analysis on the subset of NICU specimens which were run on the LH785 analyzer and found that manual review was performed on 93% of 860 NICU CBCD specimens. Thus, the impact of universal manual scans on our workload would be comparable with either analyzer.
Adequacy of Detection of Elevated Band Counts
Our laboratory missed elevated bands in 5 of 70 specimens for which manual differential counts were performed by telephone request. The reason for the false negatives was not because of technologist error in performing manual scans. When scans had been performed, subsequent manual differential counts confirmed that the true band count was <15%. Rather, the 5 false-negative specimens were not flagged by the automated analyzer as abnormal, and they did not meet any of our criteria for manual review. Indeed, automated hematology analyzers generally are unreliable for flagging samples with increased bands.3,5,6
From our results, we inferred that performing a manual scan on every specimen would ensure the detection of elevated band counts when present. Therefore, we revised our manual review process to reduce our rate of false-negative errors in the detection of elevated band counts for NICU specimens. We tailored our manual review process for the needs of the NICU, without requiring manual differential counts to be performed on a “standing order” basis.
Revised Manual Review Process for NICU Specimens
Figure 3 illustrates our manual review process for NICU specimens including 2 changes to ensure the detection of elevated band counts. The first of these changes is to perform universal manual scans for all NICU specimens. Thus, all specimens are examined microscopically by a technologist. Second, the technologist proceeds to a manual differential count whenever the band count appears to exceed 5% on the manual scan. Previously, we used an estimate of 15% bands as our threshold for performing a manual differential count. Under the new process, the automated differential count is released only for specimens appearing to have a negligible (less than 5%) proportion of bands on manual scan. In this circumstance, a comment is appended stating that “Bands/Immature granulocytes appear less than 5% by scan.”
Based on the distribution of band counts in the study sample, only a minority of specimens had between 5% and 15% bands. Thus, only a small number of manual differential counts would need to be performed to meet this criteria change. The small increase in manual differentials resulting from this criteria change would not offset the improved efficiency from the overall process change.
Clinical Use of Band Counts for Neonates
The literature on the clinical use of band counts to screen for neonatal sepsis reflects variable sensitivity and specificity, in part because the definition of a band varies greatly from laboratory to laboratory. One report noted that 45% of febrile infants under 60 days of age seen in the emergency department of 1 institution had elevated band counts, compared with 6% at another institution. A much larger proportion of infants seen at the former institution would have been excluded from a low-risk category for serious bacterial infection solely on the basis of the band count, using the Rochester criteria.9 The discrepancy was attributed to differences in the morphologic definition of the band neutrophil between the 2 hospital laboratories. Definitions of a segmented neutrophil vary from “a neutrophil with nuclear constriction to less than one third the width of a lobe” to the strictest definition requiring nuclear lobes to be connected by “only a thin filament with no discernible chromatin.”3 Laboratories using a strict definition of segmented neutrophils will have a higher reference interval for band counts, and many more infants will be excluded from the low-risk group solely on the basis of elevated band counts. Indeed, our laboratory uses a strict definition of segmented neutrophils, and our reference interval for band counts is 0%–15%. Another limitation in the use of band counts is that manual counting procedures inherently have a high degree of variability.
The diagnosis of neonatal sepsis presents a clinical and laboratory challenge. The utility of CRP, interleukin-6, procalcitonin, and TNF-α have been intensively studied.10 Alternative markers for the detection of neonatal sepsis are currently under investigation, including neutrophil CD64.11
Manual review process adapted for NICU specimens as a result of this study. This decision tree depicts the manual review process adapted in our laboratory to meet the needs of the NICU. The automated differential count is not released without blood film examination. A manual differential count may be required per the criteria shown in Table 1. Otherwise, a manual scan is performed. If the band count appears to be higher than 5% on scan, a manual differential count is performed.
We note that our findings are limited to the hospitalized neonatal population. In other patient groups, the clinical use of band counts is not comparable to hospitalized neonates.3 Also, the rate of manual review of films obtained in this study is specific to this patient group.
Based on our workflow analysis and the need to eliminate false-negative errors for detecting elevated bands, we modified our manual review process for NICU specimens. We elected to perform universal manual scans to confirm the lack of significant abnormality but not universal manual differential counts on all NICU specimens. This change would be expected to increase the rate of manual scans from 96% to 100%, while maintaining the rate of manual differential counts around 87%. The revised laboratory process overcomes the limited ability of automated hematology analyzers to detect elevated bands. This tailored manual review process could be useful for other hospitals treating neonatal patients at risk of sepsis.
Laboratory productivity and the rate of manual peripheral blood smear review: A College of American Pathologists Q-Probes study of 95,141 complete blood count determinations performed in 263 institutions. Arch Pathol Lab Med. 2006;130:596–601.