Most children infected with SARS-CoV-2 experience only mild illness, but a small fraction develop severe or critical disease that requires hospital-level intervention. Identifying which hospitalized children are most likely to deteriorate has remained a persistent challenge for pediatricians, particularly as the virus continues to evolve and the clinical picture shifts with each successive variant. A new retrospective study from China, published in BMC Pediatrics, offers one of the most detailed looks yet at severe and critical COVID-19 among children admitted during and after the period when the Omicron BA.5 subvariant dominated circulation, and its findings carry a practical message for clinicians: two simple laboratory measurements, serum sodium and neutrophil count, may be the most reliable signals of impending severity.
The research team, led by Shengbo Cai of the Fifth Hospital of Xiamen and including colleagues from Peking University Shenzhen Hospital, reviewed the clinical records of 519 children hospitalized with symptomatic COVID-19 between December 2022 and December 2025. This three-year window allowed the investigators to compare two distinct epidemiological eras: a calendar period in which Omicron BA.5 and its close descendants predominated, and a subsequent post-BA.5 period characterized by later viral variants. The cohort was strikingly young. Infants and toddlers accounted for 345 of the 519 admissions, or 66.4 percent, underscoring that the youngest children continue to bear a disproportionate share of pediatric COVID-19 hospitalizations even years into the pandemic.
When the researchers classified disease severity, they found that 192 patients, or 37.0 percent, had mild disease, 227 patients, or 43.7 percent, had moderate disease, and 100 patients, or 19.3 percent, fell into the severe or critical category. Notably, every child in the cohort improved or achieved clinical recovery, a reassuring outcome that reflects both the inherent resilience of young patients and the quality of supportive care available at the study center. Nevertheless, nearly one in five hospitalized children required classification as severe or critical, a proportion high enough to justify a careful search for early warning signs that could help clinicians triage and monitor admissions.
The comparison between the two calendar periods revealed a consistent pattern of attenuation over time. Children admitted during the post-BA.5 period experienced a lower proportion of severe and critical disease, needed supplemental oxygen less often, spent fewer days in the hospital as measured by the median length of stay, and suffered fewer complications overall. All of these differences reached statistical significance, with P values below 0.05. The picture was not uniformly favorable in the later period, however. Neurological symptoms were actually more frequent during the BA.5-predominant calendar period, while the rate of viral co-detection, meaning the simultaneous presence of other respiratory pathogens alongside SARS-CoV-2, was lower during that same period. These shifting patterns illustrate how the clinical phenotype of pediatric COVID-19 is not static but evolves in tandem with the viral population.
To identify factors independently associated with severe and critical disease, the team turned to multivariable logistic regression, the standard statistical tool for isolating the effect of each candidate predictor while accounting for the others. In the primary analysis, three factors emerged: the presence of neurological symptoms, admission during the BA.5-predominant calendar period, and young age. On its face, this result seemed to suggest that children presenting with seizures, encephalopathy, or other neurological manifestations were at markedly elevated risk of progressing to severe or critical illness, and that infants and toddlers faced greater danger than older children.
But the investigators did not stop there, and it is their methodological rigor that elevates this study above many similar analyses. Recognizing two subtle statistical pitfalls, definitional circularity and quasi-complete separation, they repeated the analysis using Firth penalized regression, a technique that applies a penalty term to stabilize estimates when data are sparse or when a predictor almost perfectly separates the outcome groups. Definitional circularity arises when a symptom is itself part of the definition of severe disease, guaranteeing a spurious association; quasi-complete separation occurs when every patient with a certain feature shares the same outcome, producing extreme and unreliable odds ratios. The sensitivity analysis exposed exactly these problems. The extreme estimate for neurological symptoms was driven by circularity, and the associations involving the calendar periods and age groups proved unstable across different model specifications, fluctuating depending on how the variables were defined and adjusted.
What survived this gauntlet of sensitivity testing were two laboratory parameters: hyponatremia, an abnormally low concentration of sodium in the blood, and an elevated neutrophil count. Unlike the clinical and demographic variables, these two markers remained consistently and independently associated with severe and critical disease in both the primary multivariable regression and the Firth penalized sensitivity analyses. This robustness matters enormously for clinical practice, because a predictor that holds up across statistical approaches is far more likely to be genuine and reproducible than one that appears and disappears depending on model assumptions.
The biological plausibility of both markers lends further weight to the finding. Hyponatremia is a well-recognized complication of respiratory infections in children and can arise through several mechanisms, including the syndrome of inappropriate secretion of antidiuretic hormone, in which excess ADH causes water retention and dilution of serum sodium. Inflammatory cytokines such as interleukin-6, which surge in severe COVID-19, can stimulate ADH release, linking low sodium to the intensity of the host immune response. An elevated neutrophil count, meanwhile, is a classic hallmark of bacterial coinfection, intense systemic inflammation, or tissue damage, all of which correlate with worse outcomes in respiratory viral disease. In this sense, the two surviving predictors may be capturing the same underlying phenomenon: a hyperinflammatory host response that drives both electrolyte disturbance and granulocyte production.
The authors draw a careful and clinically useful distinction in their conclusions. Neurological symptoms, they argue, should be regarded as clinical red flags rather than independent statistical predictors. A child who presents with a seizure or altered consciousness warrants close observation and urgent evaluation regardless of what a regression model says, but the association between such symptoms and severe disease is partly an artifact of how severity is defined. By contrast, routine monitoring of serum sodium levels may be considered a practical addition to the care of hospitalized children with COVID-19, since a falling sodium concentration could serve as an early, objective warning of deterioration. Neutrophil counts are already measured routinely in most hospitals, making both of the validated markers essentially free to track once a child is admitted.
The study has limitations inherent to its design. As a single-center retrospective review conducted at the Fifth Hospital of Xiamen, its findings may not generalize to populations with different demographic profiles, vaccination coverage, or access to care, and retrospective designs cannot exclude residual confounding. The stability of the two laboratory associations across sensitivity analyses, however, gives the results a credibility that many risk-factor studies lack, and the honest reporting of unstable estimates for age, calendar period, and neurological symptoms is a model of statistical transparency. As SARS-CoV-2 continues to mutate and pediatric COVID-19 continues to evolve, studies like this one, which combine large observational cohorts with rigorous sensitivity analysis, will remain essential for helping clinicians distinguish the children who can be safely managed with routine supportive care from those who need intensified monitoring before their condition spirals. For now, the message for pediatricians is refreshingly concrete: watch the sodium, watch the neutrophils, and treat neurological symptoms as urgent warning signs rather than statistical curiosities.
Subject of Research: Risk factors for severe and critical COVID-19 in hospitalized children during the Omicron BA.5-predominant and post-BA.5 periods
Article Title: Severe and critical COVID-19 in children during the Omicron BA.5-predominant and post-BA.5 calendar periods: clinical characteristics and risk factors
Article References: Severe and critical COVID-19 in children during the Omicron BA.5-predominant and post-BA.5 calendar periods: clinical characteristics and risk factors. (n.d.). https://doi.org/10.1186/s12887-026-07751-0
Image Credits: AI Generated
DOI: 10.1186/s12887-026-07751-0
Keywords: COVID-19, children, SARS-CoV-2, Omicron BA.5, hyponatremia, neutrophils, pediatrics, risk factors, neurological symptoms, hospitalization, logistic regression, BMC Pediatrics
News Source: Ophelia Keating. (October 10, 2026). Low Sodium and High Neutrophils Mark Severe COVID-19 in Children, Study Finds. Scienmag.
