Methodological comparison of capillary and venous blood parameters in pediatric Mycoplasma pneumoniae pneumonia: a...

Routine blood testing is one of the most common tests for patients. However, the clinical value of many parameters remains to be further explored. The aim of this study was to evaluate the differences in routine blood parameters between capillary blood (CB) and venous blood (VB), and to explore their clinical application in stratifying the severity of Mycoplasma pneumoniae pneumonia (MPP).

This prospective study enrolled 142 children with MPP from July 2023 to July 2024. The differences, associations, and consistency between CB and VB routine blood parameters were systematically evaluated by the t-test, Spearman correlation test, and Bland-Altman analysis. The Boruta algorithm was applied to identify features associated with disease severity, and eight machine learning models were used to assess the feasibility of using CB and VB routine blood parameters to stratify the severity of MPP.

Among the 24 routine blood parameters, 79.17% (19/24) showed significant differences between CB and VB (P<0.05). Eight parameters [mean corpuscular hemoglobin concentration (MCHC), platelet count (PLT), eosinophil count (EOS), basophil count (BASO), red cell distribution width coefficient of variation (RDW-CV), plateletcrit (PCT), platelet-to-lymphocyte ratio (PLR), and lymphocyte-to-monocyte ratio (LMR)] were significantly lower in CB than in VB, while 11 parameters [white blood cell count (WBC), red blood cell count (RBC), hemoglobin level (HGB), hematocrit level (HCT), mean corpuscular volume (MCV), neutrophil count (NEU), monocyte count (MONO), red cell distribution width standard deviation (RDW-SD), neutrophil-to-lymphocyte ratio (NLR), neutrophil-to-platelet ratio (NPR), and systemic inflammation response index (SIRI)] were significantly higher in CB. The association analysis showed that 91.67% (22/24) of the parameters in the total group exhibited varying degrees of positive association between CB and VB (correlation coefficients r range: 0.226 to 0.770). The consistency analysis showed 62.5% (15/24) had the prespecified statistical agreement criterion, whereas 37.5% (9/24) did not. Using the Boruta algorithm, four CB parameters [NLR, PLR, SIRI, platelet distribution width (PDW)] and six VB parameters [lymphocyte count (LYM), EOS, mean platelet volume (MPV), NLR, RBC, SIRI] were selected to independently build eight machine learning models for MPP severity stratification. All models exhibited poor performance in stratifying MPP severity (In CB models, accuracy ≤0.67, AUC ≤0.59; VB models, accuracy ≤0.62, AUC ≤0.61).

The blood routine parameters of CB and VB show significant differences, but they have good consistency. Simply relying on these parameters is not sufficient to accurately stratify the severity of MPP.