Abstract Background Urinalysis is important in screening for and monitoring nephrological and urological conditions. Urinalysis tests include chemistry and sediment analysis. When performed manually, urinalysis is time-consuming and associated with extensive analytical and clerical errors. Current automated technologies and informatics have greatly reduced the labor intensity of urinalysis and have allowed workflow improvements while optimizing accuracy of results reporting. These instruments, when compared with manual microscopy, achieve acceptable results for major cell types and formed elements. Despite the advantages offered by automated urinalysis, there are concerns that automated urinalysis systems do not perform in the real-world equivalently. These concerns make thorough assessment of new automated urinalysis systems – especially the formed elements flagging and characterization – a necessary element of good laboratory practice. We compared two urinalysis work cells during routine urinalysis and evaluated the differential recognition of abnormal findings. Methods Our study involved prospective analysis of specimens submitted for urinalysis at Corewell Health William Beaumont University Hospital, Royal Oak, MI (Sysmex) with repeat testing at Henry Ford Hospital, MI, Detroit (Beckman Coulter). Sysmex system includes UF-5000 fluorescent flow cytometry analyzer and UD-10 digital imager. Beckman system includes DxU Iris Workcell with digital flow morphology technology and proprietary auto-particle recognition. The study population was adult emergency and inpatients at Corewell. Abnormal findings from biochemical testing proceeded to sediment testing following each site’s usual protocol. Following testing at Corewell Health, specimens were transferred to tubes containing preservative and couriered to Henry Ford for testing the following morning. Results 272 samples were tested with only RBC, WBC and bacteria yielding sufficient abnormal finding for analysis. Since our goal was to compare reflex rates by each system, we analyzed initial screening results, not final reports. Results were first categorized as normal or abnormal, with limits of RBC =3, WBC =10, bacteria 1. Overall agreement was RBC 0.864; WBC 0.917; bacteria 0.773. Sysmex had higher rates of abnormal findings compared to DxU: RBC 19.7% v. 12.5%; WBC 18.2% v. 12.5%; bacteria 31.1% v. 13.2%. We calculated the probability that a normal specimen by Sysmex would yield normal by DxU: RBC 0.958; WBC 0.981; bacteria 0.962. Similar calculations for abnormal screens yielded: RBC 0.519; WBC 0.375; bacteria 0.646. In previous work with manual microscopy as predicate, we found DxU yielded 91.70% sensitivity, 94.44% specificity, overall accuracy 93.43%, in gratifying agreement with reports by others. Conclusion This real-world study of two urinalysis systems found very high overall agreement between individual specimens but lower agreement for abnormal findings. Falsely abnormal screens follow up is by visualizing the captured images or manual microscopy, with added operator time. We found a significant number of positive screens were not confirmed with visualized image reviews increasing the analysis time for those specimens. A separate assessment of the DxU was consistent with published reports. Our findings could have been influenced by one or more factors: inherent differences in the image recognition devices and flagging rules, differences in how the specimens were handled, biased sample collection and differences between testing protocols and staff.
Maiorana et al. (Wed,) studied this question.