Red blood cell subpopulation ratios identify samples at risk of dipstick–flow cytometry hematuria discordance in Sysmex UN-Series analyzers Authors: Jordi Tortosa-Carreres, Nahúm Martí-Montoro, Miguel Rodríguez-Manotas, Begoña Laiz-Marro Published: 2026 ABSTRACT AbstractObjectivesTo evaluate whether erythrocyte subpopulation structure, particularly the small-to-total RBC ratio (SRR), explains dipstick–flow cytometry hematuria discordance (DSD).MethodsA retrospective study including 3399 urine samples from unique primary care patients analyzed using Sysmex UN-series analyzers. Three discrepancy metrics were defined: dipstick–flow cytometry discordance (DSD), negative dipstick hematuria with elevated flow-cytometric RBC counts (≥15 cells/μL in men and ≥25 cells/μL in women); quantitative discordance (qDSD), the difference between flow-cytometric RBC counts and dipstick categories expressed on a standardized scale; and extreme discordance (eDSD), the highest decile of qDSD. Multivariable regression models evaluated associations between SRR, the lysed-to-total RBC ratio (LyRR), their interaction term, age, sex and urinary matrix parameters (white blood cell count, bacteria, osmolality, pH, yeast, and urinary crystals), combined to minimize collinearity. Additional analyses assessed whether these variables and clinical covariates were associated with SRR, and complementary models included alternative interaction terms, whereas sensitivity analyses included adjustment for total RBC.ResultsSRR showed the strongest association with increased odds of DSD, whereas LyRR was inversely associated, with a significant positive SRR×LyRR interaction. Debris and urinary pH were positively associated with DSD, while bacterial and leukocyte counts showed inverse associations. SRR was not associated with qDSD but remained consistently associated with eDSD. Associations remained robust across sensitivity analyses and were independent of absolute RBC counts; however, in eDSD models, total RBC burden emerged as the strongest determinant.ConclusionsDSD reflects a structured phenomenon largely driven by erythrocyte subpopulation heterogeneity rather than random analytical variability.Keywords: Urinalysis, Flow cytometry, Hematuria, Red blood cell subpopulations, Dipstick–sediment discordance Introduction Discrepancies between hematuria detected by urine dipstick testing and urinary red blood cell (RBC) counts obtained by flow cytometry are common in routine clinical laboratory practice, regardless of the analytical platform used. In everyday settings, dipstick-negative but cytometry-positive hematuria represents a frequent source of interpretative uncertainty. Urinalysis is highly susceptible to preanalytical variability, which may substantially influence analytical outcomes and result interpretation [1]. In many laboratories, such discordant results systematically trigger manual urine sediment review despite the higher analytical sensitivity and reproducibility of flow cytometry for RBC detection. This strategy increases workload and turnaround time without consistently providing clinically meaningful additional information. Identifying the determinants of this discordance is therefore essential to improve result interpretation and rationalize the use of manual microscopy [2,3]. Modern flow cytometry–based urine analyzers, such as the UF-5000 (Sysmex, Kobe, Japan), allow characterization of RBC subpopulations, including small, large, and lysed erythrocytes [4]. Because dipstick detection depends on the pseudoperoxidase activity of heme released after erythrocyte lysis [3,5], whereas flow cytometry directly quantifies intact RBCs or cellular particles [2], samples enriched in small or lysis-resistant erythrocytes may generate discordant results between both methods [2,5,6]. The aim of this study was to evaluate whether parameters derived from urinary RBC subpopulations obtained by automated urine cytometry—particularly the small-to-total RBC ratio (SRR)—are associated with dipstick–flow cytometry discordance (DSD) independently of absolute RBC counts, while simultaneously accounting for urinary matrix characteristics and the degree of erythrocyte lysis. Methods A retrospective study was conducted at Hospital Universitari i Politècnic La Fe (València, Spain) using consecutive primary care urine samples. Samples were collected, transported, and processed shortly after arrival according to standard institutional procedures using automated urinalysis systems (UC-3500/UC-1000 for dipstick analysis and UF-5000 for sediment analysis; Sysmex). The UF-5000 generates urinary RBC subpopulations using multiparametric flow cytometry after staining with dedicated reagents. Each particle is characterized according to forward-scattered light, side-scattered light, side fluorescence, and depolarized side-scattered light, providing information on particle size and permeability, internal structure, stainability, and birefringence, respectively. Among the particles classified as RBCs, large and small RBCs are distinguished by relatively high and low forward-scattered light intensity, respectively, whereas lysed RBCs are reported as a separate subpopulation of morphologically unstable erythrocytes [7]. Inclusion criteria required paired dipstick and flow cytometry results. Analytical data were extracted from the urinalysis middleware (UWAM, Sysmex). Clinical variables, including diabetes mellitus (DM), hypertension, and nephropathy, were recorded (Supplementary Table 1). Three discordance outcomes were defined: DSD was characterized by a negative dipstick hematuria result with an elevated flow-cytometric RBC count (≥25 cells/μL in women or ≥15 cells/μL in men) [8]. Quantitative discordance (qDSD) was defined as the difference between flow cytometric RBC counts and dipstick hematuria categories after z-standardization of both variables. This standardized approach was selected to place both measurements on a common scale despite their different measurement formats, allowing quantitative assessment of discordance beyond binary classification. Finally, extreme discordance (eDSD) corresponded to qDSD values at or above the 90th percentile. Baseline urinalysis-derived and clinical variables were compared according to DSD and eDSD status to characterize the study population. The multivariable statistical analysis followed a predefined, hypothesis-driven framework based on the expected contribution of erythrocyte size distribution and lysis susceptibility to dipstick–flow cytometry discordance. The primary analysis evaluated the association between SRR and DSD using multivariable logistic regression, including the lysed-to-total RBC ratio (LyRR), their interaction (SRR×LyRR), and relevant urinary and clinical covariates (bacterial counts [Bact], debris, osmolality, pH, white blood cell count [WBC], sex, and age). The SRR×LyRR interaction term was included a priori to evaluate whether the association between erythrocyte size distribution and dipstick–flow cytometry discordance varied according to erythrocyte lysis susceptibility. Predicted probabilities of DSD were plotted across SRR values at fixed LyRR levels (−1, 0, +1 SD), holding remaining standardized covariates at their mean values to visualize the interaction effect. Moreover, determinants of SRR were evaluated using multivariable linear regression including urinary matrix variables (Bact, debris, Osm, pH and WBC) and clinical covariates (DM, hypertension, and nephropathy). Sensitivity analyses were performed to assess the robustness of the findings. First, alternative DSD and eDSD models including the interaction term SRR×Debris instead of SRR×LyRR were fitted to evaluate whether urinary concentration modified the association between SRR and discordance. Second, the main model was additionally adjusted for total RBC count to assess whether the association between SRR and discordance was independent of erythrocyte burden. Third, Bact and WBC were replaced by yeast and urinary crystals (XTAL) to reduce collinearity between inflammatory sediment components (Supplementary Fig. 1). Within this framework, an additional model including the SRR×XTAL interaction term instead of SRR×LyRR was fitted to evaluate whether urinary crystals could account for the observed effect modification. Secondary analyses included modeling qDSD using multivariable linear regression and eDSD using multivariable logistic regression, using the same covariates as in the primary DSD model. Finally, as a complementary analysis, all models were repeated replacing SRR with the large-to-total RBC ratio (LRR) using the same modeling framework. All analyses were performed using R (R Foundation for Statistical Computing, Vienna, Austria). Continuous predictors were z-standardized prior to multivariable modeling to allow comparison of effect sizes across variables measured on different scales and to improve numerical stability of regression estimates. Collinearity was assessed using variance inflation factors (VIF; car package) [9], data were handled with dplyr, model outputs were summarized with broom, and figures were generated using ggplot2 and patchwork. The study was approved by the Ethics Committee for Research of the Health Research Institute La Fe (approval number: 2026-0210-1) and conducted in accordance with the Declaration of Helsinki. Results A total of 3399 urine samples from unique patients were included (median age 63 years [IQR 26], 66.3% female). DSD was identified in 369 samples (10.9%) and eDSD in 340 (10.0%). DM was present in 29.8%, hypertension in 41.7%, and nephropathy in 1.3%. Baseline distributions according to DSD and eDSD status are shown in Supplementary Table 2. Most urinalysis-derived parameters were higher in discrepant samples, whereas LRR and age were lower; clinical covariates showed comparatively modest differences. Higher SRR showed the strongest association with increased odds of DSD, whereas LyRR was inversely associated (Fig. 1A). A significant SRR×LyRR interaction was observed, with predicted probabilities of DSD differing across LyRR strata over the SRR range (Fig. 1B). Debris and pH were positively associated with DSD, whereas Bact and WBC showed inverse associations. Female sex was associated with lower odds of DSD, while age showed no consistent effect (Fig. 1A). Multivariate association between SRR and DSD. A) Multivariable logistic regression model evaluating the association between SRR and DSD. (B) Predicted probability of DSD across standardized SRR values at low (−1 SD), mean (0 SD), and high (+1 SD) levels of LyRR, illustrating the SRR×LyRR interaction. Predicted probabilities were derived from the multivariable logistic regression model with the remaining covariates held at their mean values. (C) Multiple linear regression model evaluating the association between SRR and qDSD. (D) Multiple linear regression model evaluating determinants of SRR, including urinary matrix parameters and clinical covariates. Further stratification of nephropathy into primary glomerular disease and other nephropathies showed no significant associations with SRR in either univariable or fully adjusted multivariable analyses. Continuous predictors were z-standardized. Odds ratios (OR) represent the relative change in the odds of discordance associated with a one standard deviation increase in each predictor. β coefficients represent the expected change in the outcome variable per one standard deviation increase in each predictor. Dashed vertical lines indicate OR = 1 (A) or β = 0 (C–D). Error bars represent 95% confidence intervals. Bact, bacterial counts; Debris, urinary debris; DM, diabetes mellitus; DSD, dipstick–flow cytometry discordance; LyRR, lysed-to-total red blood cell ratio; Osm, urinary osmolality; qDSD, quantitative dipstick–flow cytometry discordance; RBC, red blood cells; SRR, small-to-total red blood cell ratio; WBC, white blood cell count. *p < 0.05; **p < 0.01; ***p < 0.001. For qDSD, neither SRR nor the SRR×LyRR interaction was significantly associated. In contrast, LyRR, Bact, WBC, and osmolality were inversely associated with qDSD, whereas age showed a positive association (Fig. 1C). Determinants of SRR included positive associations with debris, pH, and female sex, and inverse associations with WBC, Bact, and age, with DM showing the strongest negative effect (Fig. 1D). Sensitivity analyses confirmed the robustness of the primary findings. No significant interaction was observed for SRR×Debris (Fig. 2A). Additional adjustment for total RBC count did not materially modify the association between SRR and DSD (Fig. 2B). Substituting bacterial and leukocyte counts with yeast and XTAL yielded no significant associations for these variables (Fig. 2C), and no significant SRR×XTAL interaction was observed (Fig. 2D). Complementary and sensitivity analyses for DSD. (A) Multivariable logistic regression model including the SRR×Debris interaction instead of SRR×LyRR. (B) Sensitivity analysis additionally adjusting for total RBC count. (C) Alternative model substituting Bact and WBC with yeast and XTAL. (D) Alternative model including yeast and XTAL, additionally testing the SRR×XTAL interaction. Continuous predictors were z-standardized. Odds ratios (OR) represent the relative change in the odds of discordance associated with a one standard deviation increase in each predictor. Dashed vertical lines indicate OR = 1. Error bars represent 95% confidence intervals. Bact, bacterial counts; Debris, urinary debris; DSD, dipstick–flow cytometry discordance; LyRR, lysed-to-total red blood cell ratio; RBC, red blood cells; SRR, small-to-total red blood cell ratio; WBC, white blood cell count; XTAL, urinary crystal count. *p < 0.05; **p < 0.01; ***p < 0.001. In the eDSD analysis, SRR showed associations consistent with those observed in the primary DSD model, together with LyRR and the SRR×LyRR interaction. Female sex exhibited a strong positive association, in contrast to its inverse association observed for DSD (Fig. 3A). Sensitivity analyses yielded consistent results for eDSD; however, although SRR remained associated with this outcome, total RBC count emerged as the strongest independent determinant, in contrast to its lack of association with DSD (Fig. 3B). Multivariate association between SRR and eDSD. (A) Multivariable logistic regression model evaluating the association between SRR and eDSD, including the SRR×LyRR interaction term. (B) Sensitivity analysis additionally adjusting for total RBC count. Continuous predictors were z-standardized. Odds ratios (OR) represent the relative change in the odds of eDSD associated with a one standard deviation increase in each predictor. Dashed vertical lines indicate OR = 1. Error bars represent 95% confidence intervals. Bact, bacterial counts; Debris, urinary debris; eDSD, extreme dipstick–flow cytometry discordance; LyRR, lysed-to-total red blood cell ratio; RBC, total red blood cell count; SRR, small-to-total red blood cell ratio; WBC, white blood cell count. *p < 0.05; **p < 0.01; ***p < 0.001. Replacing SRR with LRR revealed consistent inverse associations across all outcomes, together with a significant LRR×LyRR interaction, supporting an antagonistic pattern relative to that observed for SRR (Supplementary Fig. 2). No model showed evidence of relevant multicollinearity (all VIF <1.8). Representative examples of discordant and concordant cases are shown in Fig. 4. Representative cases illustrating DSD (A–C), concordant negative cases (D–E), concordant positive hematuria (F), and a proposed laboratory interpretation framework (G). Panels A–C illustrate representative DSD cases with elevated cytometric RBC counts selected to allow reliable morphological visualization of erythrocytes. For each panel, phase-contrast microscopy images are shown together with the corresponding UF-5000 RBC scattergram from the same urine sample. RBC events identified in the cytometric RBC channel are highlighted in red. Dipstick result, RBC count, and small-to-total RBC ratio (SRR) for each panel were as follows: (A) Neg, 3630 cells/μL, 89.94%; (B) Neg, 132 cells/μL, 56.34%; (C) Neg, 135 cells/μL, 42.96%; (D) Neg, 9 cells/μL, 62.76%; (E) Neg, 9 cells/μL, 44.08%; (F) Pos, 464 cells/μL, 6.68%. Panel G illustrates a proposed laboratory interpretation framework for dipstick-negative, flow cytometry-positive urine samples based on the present findings. The scheme is intended for illustrative purposes only and does not represent a validated clinical algorithm. DSD, dipstick–flow cytometry discordance; Neg, negative; Pos, positive; RBC, red blood cell; SRR, small-to-total red blood cell ratio. Discussion Dipstick–flow cytometry discordance appears to be largely influenced by erythrocyte subpopulations and urinary matrix characteristics, particularly in low-to-moderate discrepancy settings, where SRR emerged as the main independent determinant. In contrast, increased LyRR showed a protective effect, which may be mediated by extracellular hemoglobin release enhancing dipstick peroxidase activity and reducing false-negative results [1,10]. A plausible explanation is that smaller erythrocytes may be less susceptible to lysis [11] and contain lower hemoglobin content per cell, thereby attenuating dipstick signal and promoting analytical discordance. In routine practice, this population is likely to correspond to red blood cells that remain undetected by conventional microscopy, consistent with the previously described concept of erythrocyte cryptodetection [5,11]. The positive interaction between SRR and LyRR may reflect, at least in part, a cytometric classification effect related to erythrocyte fragmentation during lysis, whereby erythrocyte-derived subcellular particles are detected within the small erythrocyte region, thereby amplifying the observed association between SRR and discordance despite the protective biochemical effect of extracellular hemoglobin on dipstick reactivity. In addition, debris-related parameters showed an independent effect on discordance but no significant interaction with SRR, suggesting that these particles may partially contribute to cytometric misclassification; however, the magnitude and consistency of the SRR effect across all multivariable models in which these variables coexisted are consistent with the hypothesis that erythrocyte structural heterogeneity is the primary determinant of discordance. Several urinary matrix parameters also influenced discordance. Higher urinary pH may reduce dipstick peroxidase activity [3,6,12], whereas increased estimated osmolality [derived from conductivity measurements) may modify dipstick hemoglobin detection independently of erythrocyte lysis or volume-related subpopulation shifts [6,12,13]. The inverse association observed between WBC and bacteria and discordance could be explained by enhanced erythrocyte lysis in inflammatory or infectious contexts [14] and, in the case of bacteria, by additional non-hematic peroxidase activity and a higher likelihood of overt hematuria, favoring larger erythrocyte populations [5]. In contrast, XTAL and yeast showed no significant effect in multivariable analysis, in agreement with manual sediment review findings, suggesting a limited role as practical DSD drivers. In extreme discordance settings, total erythrocyte burden appears to emerge as a major determinant, reflecting high RBC concentrations in which the semiquantitative and saturable nature of dipstick testing contrasts with the wider dynamic range of flow cytometry, resulting in amplified discrepancies at higher hematuria levels. The opposite direction of the association observed for female sex between DSD and eDSD is likely explained by the sex-specific cytometric thresholds used to define DSD. Because these thresholds are higher in women than in men, female sex appears protective at lower discrepancy levels. This definition-dependent protective effect disappears when a sex-independent percentile-based definition is applied for eDSD, indicating that the observed association is at least partly attributable to outcome definition rather than to a true biological effect. Taken together, these findings support the hypothesis of a structured, multifactorial model of dipstick–flow cytometry discordance driven by erythrocyte structural heterogeneity and modulated by urinary matrix characteristics. To our knowledge, this is the first study to systematically evaluate the contribution of urinary RBC subpopulation ratios to this phenomenon, with consistent effect sizes across models and sensitivity analyses underscoring the robustness of the observed associations beyond statistical significance alone. From a practical laboratory perspective, our findings suggest that dipstick–flow cytometry discordance should not be interpreted as random analytical variability but as a predictable phenomenon associated with urinary erythrocyte subpopulation composition. In routine practice, SRR may provide additional interpretative information in dipstick-negative, flow cytometry-positive urine samples and may help identify cases in which dipstick testing underestimates erythrocyte burden despite positive cytometric detection. Consequently, these findings suggest that it may be appropriate to prioritize flow cytometry-derived erythrocyte counts over reagent strip results in appropriately selected discordant cases, thereby supporting greater confidence in cytometric RBC findings. If prospectively validated, this approach could help optimize reflex microscopy strategies and reduce unnecessary manual microscopy, particularly in samples without significant sediment abnormalities or high debris burden (Fig. 4G) [15]. However, marked discrepancies that are clearly disproportionate to the overall urinalysis findings should still prompt evaluation for potential analytical interference or other atypical causes before relying on this interpretative approach. However, these findings should be interpreted in light of several limitations. First, the observational single-center design using a single analytical platform precludes causal inference and may limit the generalizability of our findings to other laboratory settings and urinary flow cytometry platforms. Accordingly, external validation in independent laboratories and across different urinary flow cytometry platforms will be required before these findings can be extrapolated to routine clinical practice. Second, the analysis relied primarily on analyzer-derived parameters rather than systematic morphological characterization of erythrocyte size distribution across all DSD cases. Nevertheless, the UF-5000 classifies urinary erythrocytes based on light-scatter properties directly related to cellular size [16], and its analytical performance for urinary RBC detection has been well established through comparisons with manual microscopy in previous studies [14,[17], [18], [19]]. Moreover, in a small exploratory subset of DSD-positive samples (n = 50), erythrocytes were morphologically confirmed in 94% of cases by phase-contrast microscopy performed by two experienced observers. Although this exploratory analysis supports the biological plausibility of the proposed mechanism, it was not intended to provide definitive validation of the proposed cytometric classification hypothesis. Third, although SRR showed consistent associations with DSD and eDSD, no independent association was observed with the continuous qDSD metric, suggesting that SRR may primarily identify biologically meaningful discordant phenotypes rather than the full quantitative spectrum of discordance. Finally, although the practical implications of these findings have been further discussed and illustrated through a proposed laboratory interpretation framework (Fig. 4G), this approach has not yet undergone prospective validation in terms of clinical outcomes or laboratory workflow. Prospective implementation studies will be required to determine the impact of this approach on reflex microscopy utilization and routine laboratory practice, whereas further studies incorporating systematic morphological characterization of erythrocyte subpopulations will help clarify the structural basis of the observed associations. Conclusion DSD is strongly associated with erythrocyte subpopulation heterogeneity, particularly an increased proportion of small urinary erythrocytes. Integration of parameters such as SRR may improve result interpretation and reduce unnecessary manual microscopy in routine laboratory practice. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used ChatGPT (OpenAI) in order to review the English grammar. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. Funding The article processing charges (APC) were supported by Sysmex. The company had no role in study design, data analysis, interpretation of results, or manuscript preparation. CRediT authorship contribution statement Jordi Tortosa-Carreres: Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft. Nahúm Martí-Montoro: Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Writing – review & editing. Miguel Rodríguez-Manotas: Conceptualization, Funding acquisition, Investigation, Validation, Writing – review & editing. Begoña Laiz-Marro: Project administration, Resources, Supervision, Writing – review & editing. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.