Identifying Laboratory Parameters Profiles of COVID-19 and Influenza in Children: A Decision Tree Model

Abstract

Background: The COVID-19 pandemic has put other infectious diseases, especially in children, into a new perspective. Our study focuses on two important viral infections: COVID-19 and influenza, which often present with similar clinical symptoms. Taking into consideration that the pathophysiology and systemic impact of the two viruses are distinct, which can lead to measurable differences in laboratory values, this study aimed to analyze laboratory features that differentiate between COVID-19 and influenza virus infections in pediatric patients. Methods: We statistically analyzed the routinely available laboratory data of 98 patients with influenza virus and 78 patients with COVID-19. Afterwards, the classification and regression tree (CART) method was performed to identify specific clinical scenarios, based on multilevel interactions of different features that could assist clinicians in evidence-based differentiation. Results: Significant differences between the two groups were observed in ALT, eosinophils, hemoglobin, and creatinine. Influenza-infected infants presented significantly higher leukocyte, neutrophil, and basophil counts compared to infants infected with COVID-19. Regarding children (over 12 months), significantly lower levels of ALT and eosinophil counts were observed in those with influenza compared to those with COVID-19. Furthermore, the CART decision tree model identified distinct profiles based on a combination of features such as age, leukocytes, lymphocytes, platelets, and neutrophils. Conclusions: After further refinement and application, such machine learning-based, evidence-driven models, considering the large scale of clinical and laboratory variables, might help to improve, support, and sustain healthcare practices. The differential decision tree may contribute to enhanced clinical risk assessment and decision making.

1. Introduction

Respiratory infections are illnesses affecting the airways and lungs, from the nose and throat down to the deepest parts of the respiratory tract. The vast majority of them are caused by viruses (influenza, RSV—respiratory syncytial virus, coronaviruses, rhinoviruses), but they can also be caused by bacteria or, less commonly, by fungi. Respiratory viral infections significantly impact the health of the global population. They vary depending on the season and have a significant impact on global health, from pandemics such as the Spanish flu, Severe Acute Respiratory Syndrome coronavirus (SARS-CoV), Middle East Respiratory Syndrome (MERS-CoV), the pandemic caused by the coronavirus of 2019 (COVID-19), seasonal influenza (FLU), and other types of viruses [1].

The emergence of COVID-19 (caused by the SARS-CoV-2 virus) has created a significant diagnostic challenge due to the considerable overlap in its initial clinical presentation with seasonal influenza (caused by influenza A and B viruses) [2,3,4,5]. Both are acute respiratory infections presenting with common symptoms such as fever, cough, and fatigue.

Although the transmission pathways and clinical symptoms are similar, especially during periods of co-circulation, the pathophysiology and systemic impact of the two viruses are distinct, which could lead to measurable differences in laboratory values. Studies analyzing blood parameters in children with COVID-19 and influenza showed distinct, mixed, or conflicting blood parameter profiles/results [6,7,8,9]. Further analysis of these parameters is needed to develop practical, reliable, rapid methods for identifying and quantifying laboratory distinctions that can assist practitioners in ensuring appropriate patient diagnosis, management, and treatment and public health measures.

The current study aimed to analyze routinely available laboratory features to differentiate between COVID-19 and influenza virus infections in pediatric patients. Classification and regression tree (CART) analysis was also used to identify specific clinical scenarios based on the multilevel interactions of different features that can assist clinicians in evidence-based differentiation.

2. Materials and Methods

We conducted a retrospective study including demographic, clinical, and laboratory characteristics of 176 children (aged under 18 years) hospitalized at the Pediatric Clinical Hospital of Sibiu, from July 2022 to May 2023. A total of 98 patients were diagnosed with COVID-19 and 78 children with seasonal influenza A and/or B confirmed by multiplex real-time PCR assay (Allplex™, Seoul, Republic of Korea) that enables simultaneous amplification and differentiation of target nucleic acids of S gene, RdRP gene, and N gene of SARS-CoV-2, influenza A virus (Flu A), and influenza B virus (Flu B) and/or by rapid qualitative detection (using antigen tests developed to detect SARS-CoV-2/influenza A and/or B) by collecting nasal/pharyngeal swab samples (Xiamen Boson Biotech Co., Ltd., Xiamen, China). Patients above 18 years, as well as patients whose clinical and laboratory data could not be accessed, were excluded from the analysis. Informed consent was obtained from guardians of all patients, for all subjects included in this study, in accordance with the Declaration of Helsinki. The study was approved by the Institutional Review Board of Pediatric Clinical Hospital of Sibiu (No. 2218). All demographic parameters, clinical symptoms, and hematological and biochemical parameters of all patients were retrospectively reviewed from the digital register.

Results of laboratory tests included the following data: complete blood count (leukocytes [10³/μL], hemoglobin [g/dL], absolute neutrophils [10³/μL], absolute lymphocytes [10³/μL], absolute monocytes [10³/μL], absolute eosinophils [10³/μL], absolute basophils [10³/μL], platelet counts [10³/μL], NLR—neutrophil–lymphocyte ratio, PLR—platelet–lymphocyte ratio, LMR—lymphocyte–monocyte ratio, liver enzymes parameters (AST—aspartate aminotransferase [U/L], ALT—alanine aminotransferase [U/L]), renal function parameters (urea [mg/dL], creatinine [mg/dL]), and other parameters: RBC [10⁶/μL], C-reactive protein [mg/L]. Hematological parameters were measured using an automated hematological analyzer Sysmex XN (Sysmex Corporation, Kobe, Japan) and serum biomarkers were measured using an automated biochemical analyzer, Abbott, C 4000 (Abbott Park, IL, USA) with the specific kits in accordance with the manufacturer’s recommendations. We considered normal range references, adjusted to age and gender, according to the laboratory guidelines where the study took place [10].

The categorical data are presented as frequency and percentage and continuous data are presented as medians and IQR (interquartile range). For comparison of characteristics between the two patient groups, COVID-19 vs. influenza, the Mann–Whitney test, Chi-square test, or Fisher’s test were carried out. We performed these comparisons for all included patients and separately for two age groups: children under 1 year old (age ≤ 12 months) and children over 1 year old (age > 12 months). Classification and regression tree (CART), a supervised machine learning classification method, was used to identify important predictors, their corresponding cutoff points, and (homogenous) groups of patients, based on rules representing combinations of predictors. Decision tree algorithms are non-parametric methods, used for both classification and regression tasks, in order to create a model that predicts values for a target variable by learning decision rules inferred from data features. The learning process operates recursively, dividing the data based on the best features and optimal split point(s) in order to minimize prediction error. The resulting model is presented using a tree structure which includes nodes (root node, intermediate nodes, and terminal nodes) and edges between them. The tree can be translated into interpretable and understandable/explainable sets of “if-then” rules that mimic human decision making, which could be valuable to clinicians, researchers, and policymakers. In our study, a CART decision tree was constructed for the supervised classification of pediatric patients with COVID-19 and influenza. Also, the logistic regression method, a frequently used classification method in previous studies, was used. The model’s performance was assessed by performance metrics: accuracy, sensitivity, specificity, and ROC curve. The training (90%)/test (10%) method was used for (internal) validation. Data analyses were performed using R software v.4.0.5 (R Foundation for Statistical Computing, Vienna, Austria) and SPSS (IBM, Armonk, NY, USA, Statistical Package for the Social Science, v. 20).

3. Results

The study included 98 patients with influenza virus and 78 patients with COVID-19. The median age was higher in the influenza group (28, IQR: (11–57) months) than the COVID-19 group (10, IQR: (6–27) months); more than half of the patients with COVID-19 were infants while in the influenza group one-third of the children (34.2%) were between 2 and 6 years old, followed by infants (26.8%). The two groups were similar in terms of gender distribution, with males comprising 52.56% in the COVID-19 group and 50% in the influenza group (Appendix A,

Table A1. Age and gender distribution of patients.

	COVID-19 (n = 78)	Influenza (n = 98)	p
Gender
Male	41 (52.56)	49 (50.00)	0.537
Female	37 (47.44)	49 (50.00)
Age
Median (IQR)	10.00 (6–27)	28.00 (11–57)
<1 year	45 (57.69)	26 (26.80)	0.001
1–2 year	10 (12.82)	19 (19.59)
2–6 year	15 (19.23)	33 (34.02)
6–10 year	2 (2.56)	8 (8.25)
>10 year	6 (7.69)	11 (11.34)

). The main encountered symptoms in the influenza group were as follows: fever (89.80%), nasal congestion (71.43%), cough (70.41%), loss of appetite (55.10%), and rhinorrhea (47.96%), while in the COVID-19 group the hierarchy of symptoms was as follows: fever (87.18%), loss of appetite (57.69%), nasal congestion (50.00%), and cough (47.44%). Diarrhea and rash were more frequent in the COVID-19 patients compared to those with influenza (diarrhea: 29.49% vs. 17.35%, p = 0.056, rash: 24.36% vs. 13.27%, p = 0.058).

When comparing laboratory parameters in influenza and COVID-19 patients, significant differences were noted in ALT, eosinophils, hemoglobin, and creatinine (

Table 1. Laboratory characteristics of patients.

Laboratory Characteristics	COVID-19 (n = 78)	Influenza (n = 98)	p-Value
urea [mg/dL]	19.00 (15.00–23.00)	20.00 (16.00–24.00)	0.500
creatinine [mg/dL]	0.45 (0.39–0.49)	0.46 (0.41–0.53)	0.046
AST [U/L]	48.00 (37.00–58.00)	41.00 (31.00–54.50)	0.065
ALT [U/L]	25.00 (17.00–35.00)	19.50 (13.50–25.00)	0.000
CRP [mg/L]	4.59 (2.00–15.00)	5.26 (2.00–24.07)	0.564
RBC [106/μL]	4.32 (3.99–4.59)	4.44 (4.23–4.70)	0.059
hemoglobin [g/dL]	11.30 (10.30–12.40)	11.95 (11.05–12.70)	0.013
leukocyte [103/μL]	6.94 (4.92–9.70)	6.85 (4.68–9.84)	0.602
neutrophils [103/μL]	2.93 (1.58–4.47)	3.38 (2.16–5.19)	0.212
lymphocytes [103/μL]	2.29 (1.34–4.09)	2.16 (1.41–3.23)	0.458
monocytes [103/μL]	0.82 (0.46–1.18)	0.61 (0.43–0.98)	0.065
eosinophils [103/μL]	0.02 (0.00–0.08)	0.00 (0.00–0.02)	0.000
basophils [103/μL]	0.02 (0.01–0.02)	0.01 (0.01–0.02)	0.254
platelet [103/μL]	285.50 (230.00–366.00)	261.50 (189.50–345.00)	0.078
NLR	1.30 (0.64–2.50)	1.73 (0.65–2.65)	0.195
PLR	114.50 (72.61–200.23)	123.12 (73.43–181.39)	0.700
LMR	2.72 (1.81–4.75)	3.17 (2.42–4.82)	0.105

AST—aspartate aminotransferase, ALT—alanine aminotransferase, CRP—C-reactive protein, RBC—red blood count, NLR—neutrophil–lymphocyte ratio, PLR—platelet–lymphocytes ratio, LMR—lymphocytes–monocytes ratio.

). Among children under 1 year old (≤12 months), influenza-infected children presented significantly higher levels of leukocyte, neutrophil, and basophil counts compared to the COVID-19 group. Among children over 1 year old (>12 months), significant lower levels of ALT and eosinophil counts were observed in children with influenza compared to the COVID-19 group (

Table 2. Laboratory characteristics of patients under and over 1 year old (12 months).

Laboratory Characteristics	≤12 Months			>12 Months
	COVID-19 (n = 45)	Influenza (n = 26)	p-Value	COVID-19 (n = 33)	Influenza (n = 71)	p-Value
urea [mg/dL]	18.00 (13.00–21.00)	14.00 (11.00–20.00)	0.153	21.00 (17.00–27.00)	20.00 (18.00–27.00)	0.965
creatinine [mg/dL]	0.41 (0.37–0.45)	0.40 (0.36–0.43)	0.308	0.48 (0.46–0.58)	0.50 (0.44–0.58)	1.000
AST [U/L]	54.00 (42.00–63.00)	55.00 (47.00–79.00)	0.483	39.50 (24.00–51.50)	38.00 (29.00–50.00)	0.951
ALT [U/L]	28.00 (22.00–41.00)	27.00 (20.00–32.00)	0.504	19.50 (15.00–31.55)	16.00 (13.00–22.00)	0.054
CRP [mg/L]	3.69 (2.00–11.00)	4.03 (2.01–27.02)	0.343	9.46 (2.00–26.50)	6.35 (2.00–24.07)	0.567
RBC [106/μL]	4.09 (3.82–4.47)	4.15 (3.86–4.46)	0.793	4.48 (4.34–4.65)	4.51 (4.31–4.70)	0.805
hemoglobin [g/dL]	10.90 (9.90–11.30)	11.10 (10.70–12.10)	0.070	12.30 (11.70–13.00)	12.20 (11.30–12.90)	0.494
leukocyte [103/μL]	6.36 (4.69–9.80)	10.08 (7.06–12.71)	0.012	6.95 (5.17–8.71)	6.29 (4.47–8.09)	0.104
neutrophils [103/μL]	2.32 (1.46–3.27)	3.87 (2.10–4.60)	0.011	3.89 (2.88–6.80)	3.20 (2.25–5.21)	0.116
lymphocytes [103/μL]	3.20 (1.77–4.88)	3.75 (2.34–7.32)	0.166	1.66 (1.08–2.29)	1.85 (1.16–2.65)	0.570
monocytes [103/μL]	0.96 (0.64–1.71)	1.04 (0.69–1.88)	0.724	0.54 (0.38–0.85)	0.55 (0.40–0.79)	0.800
eosinophils [103/μL]	0.03 (0.00–0.12)	0.01 (0.00–0.05)	0.056	0.02 (0.00–0.05)	0.00 (0.00–0.01)	0.004
basophils [103/μL]	0.02 (0.01–0.03)	0.03 (0.02–0.04)	0.024	0.01 (0.01–0.02)	0.01 (0.01–0.02)	0.208
platelet [103/μL]	322.00 (238.00–382.00)	353.00 (281.00–448.00)	0.090	241.00 (216.00–318.00)	249.00 (182.00–292.00)	0.111
NLR	0.80 (0.39–1.36)	0.90 (0.34–1.85)	0.473	2.50 (1.44–4.86)	1.88 (1.18–3.09)	0.147
PLR	109.92 (56.82–172.74)	104.01 (55.44–138.29)	0.593	149.31 (104.95–233.04)	128.23 (85.39–188.20)	0.175
LMR	2.56 (1.86–4.55)	3.32 (2.59–6.08)	0.096	3.32 (1.81–5.07)	3.12 (2.39–4.66)	0.660

AST—aspartate aminotransferase, ALT—alanine aminotransferase, CRP—C-reactive protein, RBC—red blood count, NLR—neutrophil–lymphocyte ratio, PLR—platelet–lymphocytes ratio, LMR—lymphocytes–monocytes ratio.

, Figure 1).

The decision tree model, presented in Figure 2, identified age as the first predictor for discriminating between COVID-19 and influenza patients. In the case of patients under 12 months (indicated by the “age ≤ 12” branch of the tree), 63.4% (45/71) had COVID-19, while in the group of patients over 12 months (the “age > 12” branch), 68.6% (72/105) had influenza. Among infants (age ≤ 12 months), leukocytes were identified as the second splitting predictor (cutoff value: 6.295 × 10³/μL), while splitting predictors at the third level were platelets (cutoff value: 286.500 × 10³/μL) and lymphocytes (cutoff value: 3.625 × 10³/L). Among older children (age > 12 months), neutrophils were identified as the second splitting predictor (cutoff value: 10.037 × 10³/L), followed by platelet count (cutoff value: 170.000 × 10³/μL) at the third split. The normalized importance of predictors used to construct the decision tree model is presented in Figure 3.

The confusion matrix and prediction performance indicators of the two classification models are presented in

Table 3. Confusion matrix of prediction for the two models.

Decision Tree (CART)			Logistic Regression
	Predicted		Predicted
Observed	COVID-19	Influenza	COVID-19	Influenza
COVID-19	56	22	42	35
Influenza	10	88	21	74

and

Table 4. Evaluation metrics of classifiers for differential diagnosis between COVID-19 and influenza.

Algorithm	Acc (%)	Se (%)	Sp (%)	AUC (95%CI)
Decision tree (CART)	81.8	71.8	89.8	80.8 (73.9–87.7)
Logistic regression	67.4	54.5	77.9	66.2 (57.9–74.5)

Acc—accuracy, Se—sensitivity, Sp—specificity, AUC—area under the curve.

, while the ROC curve for the models is presented in Figure 4.

4. Discussion

Differentiating between COVID-19 and influenza is a critical challenge in clinical practice. In this study, we assessed the clinical and laboratory characteristics of SARS-CoV2 and influenza infection in hospitalized pediatric patients.

The most prevalent encountered symptom in both groups was fever, affecting 87.7% of patients in the COVID-19 group and 89.8% of those with influenza. Among COVID-19 patients, loss of appetite was the second most prevalent symptom. Cough and nasal congestion were found mostly in children with influenza. About one-fifth of patients in both groups reported breathing difficulties. Headache was the least frequently reported symptom, given the fact that it is a difficult-to-objectify symptom, especially for younger children. Numerous studies, ours included, indicate that patients with influenza have fever, nasal congestion, and cough more frequently than those diagnosed with COVID-19, while diarrhea is most prevalent in SARS-CoV-2 infection [4,6,9,11].

The higher prevalence of these symptoms can be attributed to differences in viral tropism, replication dynamics, transmission pathways, and the specific immune responses triggered by each infection. The influenza virus primarily targets the respiratory epithelial cells [6,12]. The virus has a surface glycoprotein called hemagglutinin, capable of binding to sialic acid residues from the surface of the target cells. The distribution of sialic acid along the respiratory tract is not uniform, with higher concentrations being found in the upper respiratory tract. This explains why the inflammation is predominantly localized in this region [6,13]. On the other hand, SARS-CoV-2 especially affects the lower respiratory tract via angiotensin-converting enzyme 2 (ACE2) receptors. Recent studies suggest that infection of the lower respiratory tract can occur as a primary event or as a result of the aspiration of infectious particles. Differences between these types of infection are also observed in the host immune response. The influenza virus is considered a respiratory virus, so the inflammatory response tends to remain localized, and extrapulmonary dissemination is rare. In contrast, SARS-CoV-2 can infect a wide range of cell types and tissues, and the infection tends to be more generalized [14].

Various biochemical and hematological parameters were evaluated for comparison between groups.

The mean serum levels of creatinine and urea were lower in the COVID-19 group compared to the influenza group. The influenza group had fewer infants, a factor that could have significantly influenced the serum creatinine values. Serum creatinine levels are dependent on age and muscular mass, so younger children have physiologically lower levels [15,16].

In our study, the mean levels of ALT and AST were higher in the COVID-19 group, but there was a statistically significant difference (p = 0.000) only for alanine aminotransferase (ALT). Age-stratified analysis revealed that this difference remained significant only in children older than 12 months. The current literature presents conflicting data. Most studies indicate that liver enzyme levels are lower in SARS-CoV-2 infections compared to influenza. However, some research has shown higher AST levels in COVID-19 patients [3,6,17]. The mechanisms involved in hepatic injury in COVID-19 are multifactorial. The SARS-CoV-2 virus was proven to have a direct cytopathic effect on hepatocytes by binding to ACE2 receptors, which are expressed not only in hepatocytes but also in cholangiocytes and endothelial cells [18,19]. Histopathological examination of liver biopsy samples has revealed findings such as macrovesicular steatosis and an increased number of mitotic figures, suggestive of hepatocyte apoptosis [19,20]. These alterations collectively contribute to a reduced regenerative capacity of hepatocytes [20]. In a study conducted by Sanyaolu A. et al. [19], it was observed that SARS-CoV-2 infection is associated with abnormal liver function tests, regardless of whether patients had pre-existing liver conditions or not. This suggests a direct viral role in hepatocyte injury. In the same study, liver injury was reported in up to 53% of SARS-CoV-2-infected patients, and those rates were higher (69%) among those requiring hospitalization. The influenza virus is considered non-hepatotropic, so influenza infection is associated with a lower rate of liver injury compared to SARS-CoV-2 infection.

Certain influenza virus strains can affect hepatocytes and can alter liver function tests. The most commonly implicated mechanism is immune-mediated inflammation, often associated with drug-induced hepatotoxicity. As shown in the study by Shafran Noa et al. [21], a key distinguishing feature between the two infections is the moment at which hepatic abnormalities occur. In patients with SARS-CoV-2 infection, liver enzyme alterations tend to occur later in the course of the disease, whereas in influenza infection, such abnormalities typically emerge earlier.

Although CRP levels were higher on average in the influenza patients, this difference did not reach statistical significance. When adjusting for age, we found that infants in both groups had lower values compared to older children. Our study supports findings from previous research, indicating that inflammatory markers like C-reactive protein (CRP) and procalcitonin generally tend to have lower levels during SARS-CoV-2 infection [3,6,17,22,23]. This can be explained by the rapid action of the influenza virus, which promptly infects the epithelial cells lining the respiratory tract, thereby activating innate immune mechanisms and triggering acute inflammation and subsequent elevations in CRP levels. On the other hand, in their study, Levinson T. et al. [24] demonstrated that “higher CRP level upon admission is approximately twice as common among SARS-CoV-2 patients compared to other widespread respiratory viruses, which may demonstrate the higher intensity of inflammation caused by SARS-CoV-2”.

Several hematological parameters were assessed within the study, including hemoglobin concentration, total leukocyte count, leukocyte differential, and platelet count, as well as derived indices such as the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and lymphocyte-to-monocyte ratio (LMR).

In most subjects of the study, red blood cells (RBCs) had normal values for age. No differences were observed based on the type of infection. In the infant group, where iron-deficiency anemia is more common, the lowest RBC values were found: 15.6% for infants infected with SARS-CoV-2 and 11.54% for those infected with influenza. One-third of the patients included in the study had low hemoglobin levels. There were no statistically significant differences identified regarding the type of infection. In the infant group, more than half of those infected with SARS-CoV-2 (51.11%) had hemoglobin values below the normal range. In the influenza cohort, one-third of the children presented with anemia upon hospital admission (34.62%) among those under 12 months of age, and 34.78% among those over 12 months. The majority of studies comparing the hematological parameters between influenza and COVID-19 have reported significantly higher red blood cell counts and hemoglobin concentrations in patients with COVID-19 [7,25,26]. One explanation for this finding is the agglutination of red blood cells, which occurs due to the influenza virus binding to sialic acid receptors on their surface [25,27].

Most of the patients enrolled in the study had normal leukocyte counts, regardless of the type of infection. This result was consistent between age categories (infants and children over 12 months). Among infants, the median leukocyte count was found to be significantly higher in influenza infection compared to COVID-19 (p = 0.012). Most studies indicate that leukocyte counts are lower in SARS-CoV-2 infection compared to influenza. Our study also supports this conclusion for children under 12 months of age [3,11,22,28]. In infants, leukopenia is more severe in SARS-CoV-2 infection than in influenza, primarily due to cytokine-mediated lymphocyte depletion, bone marrow suppression, and systemic immune activation. The physiological immaturity of the infant’s immune system has an additional effect.

Examining the rest of the hematological parameters, specifically, lymphocyte, monocyte, eosinophil, and basophil counts, we found no statistically significant differences between patients with COVID-19 and those with influenza, except for eosinophil counts in individuals over 12 months of age (p = 0.004) and basophil counts in infants (p = 0.024). Eosinophil counts were normal in almost all patients included in the study, except for 6.06% of children over 12 months of age with COVID-19. Lymphopenia was observed in one-third of patients with COVID-19 and influenza. When analyzed by age groups, lymphopenia was more common in COVID-19 patients than in those with influenza. These findings were supported by other studies [29]. On the other hand, in their studies, Lee H.Y. (Taiwan) and Siddiqui (Turkey) highlighted that lymphopenia is more characteristic of influenza infection than of COVID-19 [26,30]. In our study, lymphocytosis was less common in the studied group, but it occurs more frequently in patients infected with COVID-19, except for infants where it is more prevalent among those with influenza. Despite these observations, no statistically significant differences were found. Median lymphocyte counts were lower in older children compared to infants and in COVID-19 compared with influenza for both age groups. These findings were supported by the study of Akkoç et al. [29]. Studies report conflicting data regarding this parameter. Dulkadir et al. [7] showed that lymphocyte counts were higher in the COVID-19 group. In his systematic review and meta-analysis, Bang Yu et al. [25] found “no significant difference in terms of lymphocyte count” between these infections. Recent studies confirm that patients with influenza may have low lymphocyte counts because of hematopoietic suppression and virus-induced apoptosis. Recently, lymphopenia has also been identified as an important marker of SARS-CoV-2 infection. A possible explanation might be an increased production of interleukin-6 (IL-6), which leads to the excessive activation of monocytes and a decrease in lymphocyte levels [5].

Neutrophilia was more prevalent among children with influenza and neutropenia in both age groups with COVID-19. COVID-19 infection was associated with a higher proportion of neutrophilia among children aged 12 months or older, and influenza among infants. Neutropenia was more frequent in SARS-CoV-2 infection across all age groups, without a statistically significant difference. Neutrophil count was significantly lower in infants with COVID-19 than with influenza (p = 0.011). This conclusion is also supported by other studies, such as that of Yilmaz K. et al. [3], which reported a significantly lower neutrophil count in patients with COVID-19. Similar results were reported by studies conducted by Liang F. et al., Li Y. et al., and Săsăran O. et al. [22,31,32]. However, in a meta-analysis, Bang Yu et al. [25] reported that neutrophil counts were similar in both infections, whereas Dulkadir R. [7] found that neutrophil counts were significantly higher in SARS-CoV-2 infection compared to influenza.

The vast majority of the children enrolled in our study had normal monocyte counts. Monocytosis was more prevalent in COVID-19 compared to influenza in both age groups, but no statistically significant difference was reached. The median monocyte count was lower in the COVID-19 group compared with the influenza group. This conclusion is supported by the study of Jiangnan Chen, which showed that in the influenza group, both monocyte percent and count are significantly higher regardless of the severity of COVID-19 [33]. Similar results were reported by Mudd PA et al. who identified that “COVID-19 subjects exhibited significantly reduced numbers of circulating monocytes” [34]. On the other hand, Curtolo A. et al. showed that monocyte count was higher in the COVID-19 group than in the influenza group, and the monocyte count is predictive for the SARS-CoV-2 infection [35]. Macrophages and monocytes are an important part of the innate immune system and among the first line of defense, especially against viruses [36]. Their roles are complex in inflammation and inflammatory disease and in tissue homeostasis, and, as recent studies show, they can be reprogrammed to anti-inflammatory cells [37]. Recent studies show that up to 10% of the blood monocytes can be infected by SARS-CoV-2 and become a source of excessive inflammation and severe COVID-19. The cytokine and chemokine pattern changes contribute to a severe inflammatory response and higher mortality [36,38,39].

In this study, we found that a substantial majority of patients exhibited platelet counts within the normal range, with 89.89% of individuals diagnosed with COVID-19 and 83.33% of those with influenza falling into this category. The platelet count was lower in COVID-19 patients across both age groups. Platelets play important roles in the inflammatory response, especially due to vascular endothelial injury. However, they can also become activated in the absence of injury due to proinflammatory or infectious factors [40]. Similar results were reported by Akkoç G. et al. [29], while Bang Yu [25] showed that no significant difference in platelet count could be found in the studies he included in his meta-analysis. Yilmaz K. et al. [3] found that in the COVID-19 group the platelet count was higher compared to the influenza group. In SARS-CoV-2 infection, platelets are involved in thrombotic events and lower counts are associated with severe disease course, disease progression, and poor outcomes [40,41,42,43].

The resulting decision tree model delineates patient profiles associated with COVID-19 and influenza, based on a combination of laboratory parameters. The model segments patients by age, leukocyte count, neutrophil count, lymphocyte count, and platelet count, revealing characteristic patterns for each viral infection (across different branches of the decision tree). Infant patients (age ≤ 12 months) predominantly classified with COVID-19 are characterized by a combination of relatively lower leukocyte counts (≤6.296 × 10³/μL, further refined to <5.000 × 10³/Μl), and relatively lower platelet counts (≤286.500 × 10³/μL). Normal leukocyte values (>6.296 × 10³/μL) combined with a predisposition for lymphopenia (<3.625 × 10³/μL) are predominantly encountered in cases of influenza infection. When both leukocyte and lymphocyte counts remain within normal ranges, platelet values above 363.000 × 10³/μL characterize the COVID-19 infection. When platelet counts are below this level, the percentage of cases with influenza infection lowers to 33%, whereas the percentage of cases with SARS-CoV-2 infection rises to 66%. Regarding children with age > 12 months, a neutrophil count over 10.000 × 10³/μL is encountered in cases of patients with COVID-19 infection. Below this neutrophil threshold, when associated with a tendency towards thrombocytopenia (≤170.000 × 10³/μL), the findings indicate influenza infection. These patterns are most likely related to the fact that, with increasing age, the immune system becomes more robust and more capable of responding effectively to viral stimuli. SARS-CoV-2 infection is more prone to induce systemic inflammation compared to the influenza virus, which often has a suppressive effect on the bone marrow, resulting in lower platelet counts.

We analyzed ratios of various immune-related cell types. Our findings show that the neutrophil-to-lymphocyte ratio (NLR) was higher in patients with influenza compared to those with COVID-19; however, this difference was not statistically significant (p = 0.195). The NLR values are sex- and age-dependent regardless of the condition (disease vs. health) [44,45,46]. Prozan L. [47] found that the NLR median is significantly higher in influenza compared to COVID-19. In SARS-CoV-2 infection, NLR is an independent prognostic factor for progression to pneumonia [48], and it is proven to be better than PLR for predicting the severity of the disease [47].

Our study showed that the lymphocyte-to-monocyte ratio (LMR) was higher in influenza infection compared to SARS-CoV-2 infection (3.17 vs. 2.72, p = 0.105). A study conducted by Săsăran O. et al. supports these findings [32]. The lymphocyte-to-monocyte ratio (LMR) indicates the involvement of immune cells in both localized and generalized inflammatory processes [49]. In the scientific literature, LMR has been associated with various conditions, including bacterial and fungal infections, sepsis, different types of cancers, trauma, and vascular diseases such as spontaneous intracerebral hemorrhage, ischemic stroke, and atherosclerosis [49,50,51,52,53]. The lymphocyte-to-monocyte ratio seems to effectively differentiate between patients with influenza type A and those without this infection [54]. Furthermore, an LMR below 2 is significantly useful in distinguishing influenza type A infections from other viral infections [55]. The lower LMR value is associated with a more severe infection because of the decreased lymphocyte counts that delay viral clearance. Additionally, the LMR appears to be a valuable dynamic biomarker for monitoring hospitalized patients [49,56,57,58].

In our study, we found that the platelet-to-lymphocyte ratio (PLR) was elevated in SARS-CoV-2 infections across all age groups. The same results, as reported in a study by Yan Zhao conducted in China, showed that PLR levels were lower in influenza than in COVID-19 [59].

The platelet-to-lymphocyte ratio (PLR) reflects the complex interaction between platelets and lymphocytes in inflammation. An increase in this ratio is associated with both acute and chronic inflammatory processes [40]. The PLR increases constantly during childhood as a result of the decrease in lymphocytes and platelets. In SARS-CoV-2 infection, the PLR is higher in severe and critical cases and predicts the severity of the disease [40,60,61,62].

This study represents our preliminary research in utilizing machine learning-based models to understand the clinical and laboratory markers associated with COVID-19 and influenza. The results are promising and there are several directions for future research that could further refine and expand upon these findings. The study was a retrospective, single-center study with a small sample size. To enhance the robustness and generalizability of the model, prospective, multi-center studies involving larger sample sizes from diverse regions are needed. These studies could provide more reliable results and better applicability to broader patient populations. Moreover, while this study focused on a specific set of laboratory features, future work could explore the inclusion of additional variables—such as genetic, imaging, and environmental factors—that may individually influence disease outcomes. Incorporating these factors could offer a more comprehensive understanding of the pathophysiology and systemic impacts of both viruses. The data-driven approach presented here lays the groundwork for future multidimensional, data-driven frameworks. Such frameworks could integrate diverse clinical data sources and enhance decision making by supporting clinicians in more personalized and effective patient care.

5. Conclusions

Machine learning-based, evidence-driven models that consider a large scale of clinical and laboratory variables can help to improve, support, and sustain healthcare practices. The developed decision tree model is an exploratory one providing an interpretable framework that can be useful for understanding the differential clinical presentations of COVID-19 and influenza, highlighting the specific combinations of laboratory markers that contribute to their discrimination. COVID-19 frequently involves relatively low leukocyte and platelet counts in younger patients. Influenza often shows a combination of moderate leukocytes and moderate-to-lower lymphocytes in the cases of younger patients. In older patients, distinguishing features involve more complex patterns, combining neutrophil and platelet variations. After further refinement and application, the decision tree model differentiating between COVID-19 and influenza has the potential to improve clinical risk assessment and decision making in practice.