The Impact of Pathologist Review on Peripheral Blood Smears: A College of American Pathologists Q-Probes Study of 22 Laboratories Open Access

This study investigated the impact of PB smear review by the pathologist, when requested by a technologist or provider, by measuring the rate of detection of clinically relevant findings reported by the pathologist. The study provided a measure of the rate of pathologist review based on technologist request and the ratio of manual PB smear differential count performed per 1000 complete blood cell counts (CBCs). Lastly, this study attempted to identify factors that were correlated with an increased or decreased incidence of reporting clinically significant findings.

During a 4-week study period, participants prospectively collected information on PB smears requested for pathologist review. Smears sent in consultation or for second opinion were excluded. Participants were asked to specify the dates of the study period and the total numbers of smears that underwent pathologist review, total number of manual differentials performed, total numbers of CBCs performed, and total number of provider-ordered smear reviews performed during that period. Laboratories were also asked to complete a form with demographic data on the hospital such as institution type, occupied bed size, and whether the laboratory trained pathology residents.

A data collection form (Supplemental Figure; see supplemental digital content at https://meridian.allenpress.com/aplm in the July 2025 table of contents, containing 1 figure and 3 tables) was completed for each smear in the study, reporting whether the smear was from an inpatient or outpatient and the review request source (technologist, physician provider, nonphysician provider, or both technician and nonphysician/physician provider). Based on the requestor category, different options for reason for pathologist review request were provided, with responses such as “blasts” and “dysplastic changes” for technologist requests, and options such as “platelet abnormality” and “red blood cell [RBC] abnormality” for nonphysician/physician providers. Physician requests for pathologist review asked for their subspecialty.

The reviewing pathologists were asked to respond yes/no to “Did the pathologist’s review result in a clinically relevant morphologic finding?” and were also asked to select the “primary clinically relevant finding.” For the purposes of the study, the following findings were considered clinically significant and were listed as options: “atypical (possibly neoplastic) lymphocytes,” “blasts,” “dysplastic changes,” “immature cells/left shift in myeloid cells or monocytes,” “microangiopathic changes/schistocytes,” “microorganisms,” “nucleated red blood cells (nRBC),” and “sickle cells.” Pathologists could also choose “other, specify:____” to write in other findings. For cases where review was requested by a technologist, the pathologists were asked if they agreed or disagreed with the technologist’s interpretation. Institutions were required to submit at least 7 provider-ordered cases and/or 7 technologist-requested cases to be eligible to receive each respective rate of clinically relevant findings and to be included in all institution analyses.

Participating laboratories were asked to provide information about general PB smear review practices in their institutions using a general practices questionnaire, including the criteria that would trigger technologist request for pathology review and whether pathologists had access to patient data when reviewing the PB smears.

Univariate differences were tested for using t tests for normally distributed variables and nonparametric Kruskal-Wallis tests for nonnormally distributed variables. For technologist-requested reviews, specific criteria for pathologist review were selected a priori to be tested for association with the rate of clinically relevant findings. The criteria selected for analysis included atypical lymphocytes more than numeric upper limit, suspected myeloproliferative disease, nRBCs, other RBC morphologic abnormalities (eg, thalassemia, sickle cell, anisocytosis, hypo/hyperchromasia, Howell-Jolly bodies), and bacterial sepsis (eg, bacteria on smear). Practice characteristic variables were tested for univariate association with the rate of clinically relevant findings disclosed by pathologists during technologist-requested review. Those associated at a significance level of 0.10 were then evaluated using a forward stepwise multiple regression model. The combination of variables used in the final model was selected based on optimal model fit, the results from a forward stepwise regression analysis, and a statistical significance level of .05 for each variable in the final model. All analyses were performed with SAS 9.4 (SAS Institute, Cary, North Carolina).

A total of 22 institutions submitted data, with 21 from the United States and 1 from Saudi Arabia. All had received a College of American Pathologists inspection in the past 2 years, and 3 (13.6%) had also had an inspection by the Joint Commission. Nine of 21 laboratories (42.9%) were affiliated with a hospital with an occupied bed size of 150 or fewer. Additional demographic characteristics are summarized in Table 1.

Data for 835 eligible PB smear cases were submitted. Eleven of the 22 participating laboratories submitted sufficient data to compute a rate of clinically relevant findings detected on provider-ordered review, and 18 submitted sufficient data for technologist-requested PB smear review. Only PB smears with data on both the pathology review request source and whether the pathologist’s review resulted in a clinically relevant morphologic finding were included in the analysis. There were 426 of 829 smears (51.4%) from inpatients and 403 smears (48.6%) from outpatients.

Of 835 cases, review was requested by technologists for 520 (62.3%) and ordered by physicians for 283 (33.9%) and by nonphysician providers for 20 (2.4%). Review was both ordered by a provider and requested by a technologist for only 12 of 835 cases (1.4%); these were included in both technologist-requested and provider-ordered analyses.

Table 2 shows the rate of pathologist detection of clinically relevant findings for technologist-requested and provider-ordered smears. Pathologists identified clinically relevant findings for a median of 53.4% of technologist-requested PB smear reviews and a median of 14.3% of provider-ordered PB smear reviews.

Overall, a pathologist’s review resulted in a clinically relevant morphologic finding in 382 of 835 cases (45.7%). Table 3 and the Figure summarize the specific clinically relevant findings for all cases together, technologist-requested only cases, and provider-ordered only cases. Blasts were the most reported finding for the technologist-requested and overall cases in 91 of 532 (17.1%) and 97 of 835 smears (11.6%), respectively, followed by atypical (possibly neoplastic) lymphocytes in 74 of 532 (13.9%) and 82 of 835 smears (9.8%) and immature cells/left shift in myeloid cells or monocytes in 63 of 532 (11.8%) and 74 of 835 smears (8.9%). For provider-requested cases, other was the most frequent, reported in 55 of 315 smears (17.5%), followed by immature cells/left shift in myeloid cells or monocytes in 12 of 315 smears (3.8%). Blasts, atypical (possibly neoplastic) lymphocytes, and microangiopathic changes/schistocytes were each reported in 9 of 315 (2.9%) provider-requested PB smear reviews. Supplemental Tables 1 and 2 list the “other” responses specified for technologist-requested and physician-ordered PB smear review, respectively.

Table 4 summarizes the reasons for technologist-requested pathologist review selected by the technologist (multiple responses were allowed). The most frequent reasons for technologist-requested review were “other” for 156 of 531 smears (29.4%), “atypical (possibly neoplastic) lymphocytes” for 147 of 531 smears (27.7%) and “blasts” for 141 of 531 smears (26.6%). The pathologists indicated they agreed with the reason for review in 458 of 513 (89.3%) technologist-requested cases (data not shown).

Table 5 shows the reasons that providers gave for ordering PB smear review (multiple responses were allowed). Of 312 provider-based review orders, RBC abnormalities were the most frequent reason reported for ordering PB smear review (138; 44.2%) followed by white blood cell (WBC) abnormalities (104; 33.3%) and platelet (PLT) abnormalities (84; 26.9%). The other 2 options allowed for the question were “unknown” and “other.” The indication was listed as unknown for 40 (12.8%) smears and other for 33 (10.6%). Provider PB reviews requested for WBC abnormalities had clinically relevant morphologic findings for 45 of 104 examinations (43.3%) compared with 56 of 138 (40.6%) requested for RBC abnormalities and 27 of 84 (32.1%) for platelet abnormalities. The most commonly reported specialties of 307 ordering physicians were hematology/oncology (86; 28.0%) and hospital medicine (83; 27.0%), followed by primary care (53; 17.3%), emergency medicine (20; 6.5%), and critical care (11; 3.6%).

Table 6 shows the study period length and test volumes of participants. The median length of the study period was 28 days (6 and 39 days for the 10th and 90th percentiles, respectively). The median number of CBCs performed was 7556, with a median number of manual differential counts performed of 819 and a median of 43 pathologist-reviewed PB smears. Table 7 shows overall rates of pathologist reviews and manual differential counts during the study period adjusted by relevant test volumes. A median of 4.7% of all manual differential counts from 20 institutions underwent pathologist review during the study. The median number of manual differential counts performed per 1000 CBCs was 105.3 for 22 institutions.

All 22 responding laboratories indicated that their institution had a written policy defining the criteria for pathologist review of PB smears (question 1 of general practices questionnaire). The reported criteria for review are listed in Table 8. Most laboratories allowed repeat pathologist reviews of PB smears if requested by a clinician (20; 90.9%) or if the sample was flagged by the technologist or instrument (17; 77.3%). All laboratories indicated pathologists had access to detailed clinical information either always (16; 72.7%) or frequently (6; 27.3%). Full answers to the general practices questionnaire are presented in Supplemental Table 3.

For the technologist criteria for pathologist review used at each institution (question 7 on the general practices questionnaire, Supplemental Table 3), 5 of the 26 queried options were selected a priori to test for univariate association with the rate of technologist-requested clinically relevant findings: atypical lymphocytes more than numeric upper limit, suspected myeloproliferative disease, nRBCs, other red cell morphologic abnormalities (eg, thalassemia, sickle cell, anisocytosis, hypo/hyperchromasia, Howell-Jolly bodies), and bacterial sepsis (eg, bacteria on smear) in addition to relevant practice characteristics.

Among these queried characteristics, only 2 were found to be significantly associated with the rate of clinically relevant findings disclosed by pathologists during technologist-requested PB smear review using a multiple linear regression model at the 0.05 level. These significant associations were (1) the rate at which institutions had pathologists or designated staff conduct postanalytic patient follow-up on PB smears (eg, look at medical records to review patient diagnosis, treatment; question 6 of the general practices questionnaire, Supplemental Table 3), and (2) whether or not institutions used an upper numeric cutoff of atypical lymphocytes as an indication for review. Table 9 shows the rates of clinically relevant findings for these subgroups. Rate of detection of clinically relevant findings on technologist-requested PB smear review was estimated from a statistical model to be 31.6% higher for institutions that always, frequently, or sometimes conducted postanalytic patient follow-up compared with institutions that did this rarely or never (P = .02). Conversely, the rate of clinically relevant findings was estimated to be 27.2% lower for institutions that used atypical lymphocytes more than numeric upper limit as a criterion for review compared with institutions that did not (P = .03).

Variables that were not significantly associated with rate of disclosure of clinically significant findings (P > .05) included who was responsible for maintaining the list of criteria for PB smear reviews, whether or not there was a policy that specified a period of time between flagged PBs reviewed when a pathologist has recently reviewed a smear, how repeat pathologist reviews of a PB smear were handled, and how often pathologists had access to detailed clinical information. The other a priori selected criteria for pathologist review were also not found to be significantly associated with rate of clinically relevant findings: suspected myeloproliferative disease, nRBCs, other red cell morphologic abnormalities (eg, thalassemia, sickle cell, anisocytosis, hypo/hyperchromasia, Howell-Jolly bodies), and bacterial sepsis (eg, bacteria on smear).

The Impact of Pathologist Review on Peripheral Blood Smears study showed that pathologist review disclosed clinically relevant findings in a median of 53.4% of PB smear reviews requested by technologists and a median 14.3% of PB smears ordered by providers. It also showed that the rate at which institutions have pathologists or designated staff at their institution conduct postanalytic patient follow-up and numeric cutoff of atypical lymphocytes as an indication for review were associated with the rate of detection of clinically relevant findings among technologist-requested PB smear reviews.

Only 11 of the 22 participants reported data for provider-ordered PB smear review. This could indicate either that the other 11 institutions did not submit enough provider-requested cases during the study period or that they did not allow providers to order physician review. The study did not directly ask participants whether or not providers were able to order smear reviews. It is unclear why only 18 laboratories submitted data on technologist-requested smears, because 100% of laboratories stated that they had a policy defining criteria for pathologist review. This may be related to difficulties collecting these data, or that the policy only pertained to the pathologist review of PB smears when providers ordered them.

There was a wide distribution for rates of clinically relevant findings in both technologist-requested (9.1% at the 10th percentile, 100.0% at the 90th percentile) and provider-ordered (0% at the 10th percentile, 54.5% at the 90th percentile) PB smear reviews. This is likely indicative of differences in laboratory policies regarding what abnormalities trigger technologists to request review and how easily providers may order PB smear review. The most commonly reported clinically relevant findings were blasts, atypical (possibly neoplastic) lymphocytes, and immature cells/left shift in myeloid cells or monocytes. These were also the most common indications for technologists to request PB smear review. These 3 findings are all WBC abnormalities; however, interestingly, RBC abnormality was the most frequent indication for provider-ordered reviews.

Some of the free-text responses given for other clinically significant findings were English-language equivalents of numerical findings that are present in the CBC, such as “decreased platelets,” “microcytic anemia,” and “lymphocytosis.” Because these are restatements of data already provided, it is not clear if such comments add clinically relevant information. There were 25 of these interpretations in the provider-ordered PB smears, and if these were removed, the total number with clinically relevant morphologic findings would be decreased from 100 to 75 of 315 (23.8%). Fifteen technologist-requested reviews had similar interpretations; removing them would lower the number with clinically relevant findings to 274 of 532 (51.5%). Pathologists reviewing PB smears may feel compelled to include some sort of interpretation, either as a courtesy or to fulfill billing requirements, even if they do not add specific novel information. However, others included additional interpretation such as “favor reactive lymphocytosis,” “iron-deficiency anemia or anemia of chronic disease,” and “large to giant platelets concerning for ITP [idiopathic thrombocytopenic purpura],” which add morphologic information/interpretation beyond what can be gleaned from the CBC.

The rate of “clinically relevant” findings on PB smear reviews in the literature varies from 1% to 85%; however, the few studies available had widely variable practices regarding availability of ordering pathologist review of PB smear, as well as how clinical impact of the results were measured. In a study⁴  from Wake Forest School of Medicine, clinicians had to free text an indication to order a PB smear review. The study authors did chart reviews to see if the results influenced clinical decision-making: “results considered clinically influential were those that led to a diagnosis of a novel disease or process or those that resulted in a change in therapy.” The authors performed 277 reviews during 3 months in 2015, but only 3 (1% of 277) “appeared to influence clinical decision making” using this relatively strict definition of impact.⁴  An older study⁵  from University Hospitals of Cleveland showed a high rate of clinically relevant findings, but those authors looked only at a subset of their flagged blood smears and body fluids, which was enriched to focus on abnormalities likely to have clinical significance (eg, “findings suggestive of previously undiagnosed myeloproliferative disorder, myelodysplastic syndrome, or acute leukemia”), biasing the outcome. Using these criteria, 96 samples were included during 2 months of the study in 2005. Clinical impact was measured by residents calling the provider after the results to ask, “Did the pathologist review contribute to the clinical diagnosis?” The result was that 85% said yes; however, there is potentially an element of confirmation bias and/or politeness contributing to this high rate. Interestingly, 51% of the ordering providers had not seen the review result before they were contacted by the resident up to 48 hours later.⁵  A recent multi-institutional study from Beckman et al⁶  used methodology more similar to the current study. The authors defined a smear review as adding clinical value “if the pathologist’s interpretation included a morphologic abnormality that had the potential to impact patient management, and that could not be diagnosed by an automated complete blood count with white blood cell differential or automated iron studies alone.” Of 515 consecutive provider-ordered smears at the authors’ 3 institutions, 23% of interpretations were considered to have added clinical value. The relative frequency of indications for ordering were also similar to ours: RBC 50%, WBC 24%, and PLT 29%; and the rates of clinically relevant findings for each indication were 19%, 25%, and 13%, respectively.⁶  Our study similarly showed PB smear reviews requested for WBC abnormalities most frequently yielded clinically relevant findings, followed by RBC and PLT, albeit at higher frequencies (43.3%, 40.6%, and 32.1%, respectively).

There was wide variability in the ratio of PB manual differential counts performed per 1000 CBCs during the study period across the 22 laboratories, ranging from 44.7 to 192.3 for the 10th and 90th percentiles, respectively (Table 7). This likely reflects the varying practices among institutions for technologist review based on the CBC and automated differential count. Laboratories at the higher end of either of these distributions may benefit from revisiting policies for such manual differential counts as an opportunity for decreasing technologist workload and burnout with improved efficiency.

There was less variety in the percentage of manual differential counts that underwent pathologist review, with 1.3% at the 10th percentile, 27.5% at the 90th percentile, and a median of 4.7% among 20 responding institutions. However, this can be a time-consuming task for a pathologist, regardless if they are flagged for review by a technologist or review is ordered by a provider. Reimbursement at the University of Minnesota Medical Center in 2013 and 2018 showed an average professional fee reimbursement of $30 (range, $0–$75).⁶ 

Interestingly, pathologists agreed with findings that prompted the technologists’ requests for PB smear reviews in 458 of the 513 cases with pathologist agreement reported (89.3%), but reported clinically relevant findings in only 289 of the 532 total cases (54.3%). This suggests there may be a disconnect between what pathologists consider is worth reviewing and what they find essential to report. There may also be some instances where certain findings are flagged for review to use in education or by technologists and pathologists in training. Finally, there may be an emotional desire to agree with one’s laboratorians. This also calls into question the value added by a pathologist smear review in these cases. For certain findings that may indicate a new malignant diagnosis, such as blasts or lymphoma cells, it is appropriate for a board-certified pathologist to provide confirmation of the morphologic assessment. However, subsequent review of that same morphology is unlikely to add clinically relevant information; such reviews could be performed by the technologists, or potentially by increasingly sophisticated artificial neural networks used in digital imaging systems. Review by a pathologist may also incorporate more clinical information, such as symptoms of viral infection, which aids in the overall interpretation; this is another area in which artificial intelligence/machine learning may have a potential role for integrating data from across the medical record. Pathologist review also provides an opportunity to suggest or order additional testing, such as flow cytometry in lymphocytosis or BCR::ABL1 in left-shifted neutrophilia.

Some possible interventions to maximize clinical utility of provider-ordered PB smear reviews include restricting ordering to certain providers or care settings. Although hematology/oncology was the most frequently reported specialty among providers ordering PB smear review at 28.0%, a similar percentage (27.0%) were hospital medicine providers, and a significant portion (17.3%) were primary care providers (Table 5); restricting ordering to specialists only may increase the rate of referral to hematology/oncology consultation.

Electronic medical record interventions may also be helpful. Beckman et al⁶  performed chart reviews on 57 smears with clinically relevant findings reported and found that 46% (26) already had a PB smear previously reviewed by a pathologist based on laboratory protocols. Alerts could be used to notify providers if a PB smear review was recently reviewed by a pathologist. Having a mechanism for the provider to indicate the reason for a smear review may also be helpful, especially in specific cases such as looking for spherocytes in suspected extravascular hemolytic anemia or acanthocytes for neuroacanthocytosis. An initiative at UCLA Medical Center looked at overall PB smear reviews (not necessarily pathologist-performed) from April 2016 to January 2017; adding a pop-up showing the laboratory indications for smear review (inpatient) or requiring a prior approval (outpatient) reduced inpatient and outpatient orders by 35.56% and 66%, respectively.⁷  Additional clinician education may be helpful; traditional ordering practices promulgated over time, such as always ordering a smear review for thrombocytopenia, even if mild and isolated, may have been established when hematology analyzers were less accurate and precise, with fewer capabilities for flagging abnormalities.

The rate of clinically relevant findings disclosed by pathologists during this technologist-requested PB smear review is estimated from a statistical model to be 31.6% higher for institutions that always, frequently, or sometimes had pathologists or designated staff at their institution conduct postanalytic patient follow-up on PB smears previously reviewed by a pathologist compared with institutions that did this rarely or never (P = .02). Based on this result, incorporating an ongoing clinical feedback mechanism that requires pathologists and technologists to meet periodically and review select PB smear cases could improve efforts to ensure accuracy of PB smear review, which in turn would arguably improve patient care. There are several ways in which this could be accomplished, such as reviewing results of a set percentage of smear reviews during quality meetings or having the pathologist prospectively flag cases for follow-up while performing the smear review. The criteria for a technologist request for pathologist review should be reviewed and updated if necessary and be readily available to the laboratory staff.

The rate of clinically relevant findings disclosed by pathologists during technologist-requested PB smear review is estimated from a statistical model to be 27.2% lower for institutions that use “atypical lymphocytes more than numeric upper limit” as a technologist criterion for pathology review compared with institutions that do not (P = .03). Paradoxical as this association sounds, elevated lymphocytes may be seen in younger patients, including pediatric populations, and in infectious mononucleosis, each in the benign setting, frequently with morphologically atypical-looking lymphoid cells. Reactive/virus-associated lymphocytes were not included in the provided list of clinically relevant findings. However, in practice, recognition and correct reporting of these cells may help to guide additional testing, such as testing for Epstein-Barr virus or other viral infections rather than bone marrow biopsy.

This study has several limitations. The input form and general practices questionnaire (Supplemental Figure; Supplemental Table 3) did not ask what hematology analyzers were used, whether digital imaging and artificial neural network preclassification systems were used, or if the pathologist reviewing the smears had hematopathology fellowship training. The input form did not have an area to report details on whether or not the pathologist review required a change in differential count. Subset analysis of smears received from inpatients versus outpatients was not performed, though we recognize that smear reviews may be ordered for different indications in these 2 populations.

Because relevant findings were frequently identified in referred PB slides, pathologist review of slides is recommended as a diagnostic aid to clinicians. Technologist request frequently resulted in clinically relevant findings; however, there were still some cases where pathologist review did not add information. Eliminating repeated pathologist review for the same morphologic finding could potentially reduce unnecessary work. Pathologists can ensure receipt of relevant referrals by incorporating an ongoing clinical feedback mechanism that requires pathologists and technologists to meet periodically to review select PB smear cases to ensure accuracy of PB smear review. Pathologist review also provides an opportunity to order or suggest appropriate follow-up testing. A median of 14.3% of provider-ordered PB smear reviews resulted in clinically relevant findings. Potential methods to maximize the clinical impact of these reviews could include electronic medical record interventions to indicate if a review had already been done, or for the provider to indicate the reason for the review. Other opportunities include assessing ordering/use patterns and increasing education regarding the accuracy and precision of contemporary hematology analyzers.