Background: Iron deficiency (ID) and hemoglobinopathies are highly prevalent in Southeast Asia. Accurate estimation of reference intervals (RIs) for red cell parameters is complicated by the need to exclude individuals with these conditions from the reference population. Indirect RI estimations using machine learning could help overcome these challenges. Methods: We developed a binary classification model using eXtreme Gradient Boosting (XGB) to distinguish normal individuals from those with ID, hemoglobinopathies, or other anemias. The model was trained on an annotated dataset comprising 5,520 complete blood count (CBC) results and validated with a holdout dataset of 2,367 CBC results. An independent dataset of 64,100 CBC results was used to identify individuals predicted to be normal, from which RIs were estimated using the refineR algorithm. Results: The XGB model achieved an area under the ROC of 0.97 (95% confidence interval: 0.96–0.97) for distinguishing between individuals with normal versus abnormal values. Among individuals within the independent dataset, 40,300 (62.9%) were predicted to be normal. The refineR-based reference limits (RLs) derived from this subset approximated those obtained through a direct approach. Improvements in the accuracy of indirect RL estimates were most evident for hematocrit, hemoglobin, and red cell concentrations. Conclusions Combining XGB with refineR to indirectly derive RIs for red cell parameters improved the accuracy and yielded results comparable with those of directly derived RIs. A further benefit was the capacity to generate sex- and age-specific ranges, which has remained difficult to achieve through direct approaches.

Introduction: The objective of the study is to determine the level of agreement between measured total carbon dioxide (TCO2) and calculated bicarbonate (HCO3–) in our laboratory. Materials and Methods: TCO2 and HCO3– values of 1820 samples drawn at the same time from the patient were compared. TCO2 from venous samples was measured on Dimension RxL while HCO3 – was obtained from arterial blood gas samples analyzed on Radiometer ABL 700. Results: The TCO2 and HCO3– values correlated well (r = 0.977, p<0.001), with the correlation given by the equation, y = 0.986x – 0.5335. Using Bland-Altman analysis, the bias was 0.87 mmol/L (SD 1.42 mmol/L), and the limits of agreement (LOA) were -1.92 to 3.67 mmol/L. Story and Poustie’s criteria were applied to study the agreement between these two methods. Based on the fi rst criterion that the bias between TCO2 and HCO3– should be less than ±1 mmol/L, the results for the two methods appear to be in good agreement. The second criterion requires that the LOA between the two methods should range between a bias of ± 2 mmol/L or a total span of 4 mmol/L; the LOA was exceeded in our study. Using the total allowable error in the Bland Altman plot also showed that the two values cannot be used interchangeably especially at the lower values. Conclusions: TCO2 did not show good agreement with HCO3–. Clinicians should be aware of this discrepancy and hence should be cautious when using HCO3– for management of acid-base disorders.