The cohort included children aged 1 month to 13 years who were admitted from January 2020 through December 2025. Researchers screened 864 medical records for suspected or documented pertussis and excluded 142: 52 for lack of laboratory confirmation or another diagnosis, 24 because the child was outside the eligible age range, 18 because they represented duplicate admissions for the same episode, 30 because essential clinical, laboratory or severity data were missing, and 18 because documentation was insufficient to confirm eligibility. The final sample comprised 722 children with documented microbiological evidence of Bordetella pertussis; polymerase chain reaction from a respiratory specimen was the principal confirmation method during the study period, although the analysis dataset did not retain the specific method used for each child.

Severe pertussis was defined as the occurrence during hospitalization of at least one major adverse clinical event. The composite included intensive-care admission, significant respiratory distress, clinically significant apnea or cyanosis, respiratory failure, persistent hypoxemia, noninvasive or invasive mechanical ventilation, hemodynamic instability, severe pulmonary or neurological complications, or death. Thresholds were more stringent than the broader descriptive definitions of some signs: clinically significant apnea meant recurrent apnea, defined as at least two episodes in 24 hours, or any apnea requiring stimulation, bag-mask ventilation or escalation of respiratory support. Respiratory distress counted when associated with oxygen saturation below 92% or requiring intervention, while persistent hypoxemia was defined as oxygen saturation below 92% sustained for at least five minutes or recorded on at least two consecutive measurements despite supplemental oxygen.

The severe and non-severe groups differed markedly in age and baseline history. Median age was 4.0 months among severe cases and 12.0 months among non-severe cases. Children younger than 6 months accounted for 64.3% of severe cases but 29.6% of non-severe cases. Prematurity was reported in 21.0% of the severe group and 8.3% of the non-severe group, while low birth weight was reported in 14.6% and 7.1%, respectively. Only 32.5% of severe cases had age-appropriate vaccination, compared with 60.7% of non-severe cases; incomplete-for-age or absent vaccination was correspondingly more frequent among severe cases. Underlying comorbidity was also more common, at 24.2% versus 12.2%. Male sex was not significantly associated with severity, with 57.3% of severe cases and 53.5% of non-severe cases being male.

Clinical differences were strongest for respiratory compromise, although several comparisons were partly circular because components of the severe-case definition also appeared in the descriptive clinical assessment. Apnea was recorded in 51.6% of severe cases and 11.9% of non-severe cases, while respiratory distress was present in 70.1% and 17.0%, respectively. Cyanosis was recorded in 38.2% versus 6.4%, feeding difficulty in 55.4% versus 22.1%, and oxygen saturation below 92% in 51.0% versus 4.2%. Age-adjusted tachypnea and tachycardia were also more frequent in the severe group. These were broader recorded clinical variables, and some non-severe children had signs such as apnea or respiratory distress without meeting the stricter thresholds required for the composite outcome. Paroxysmal cough, inspiratory whoop, post-tussive vomiting and wheezing did not differ significantly between the groups.

Severe cases also had distinct hematological, biochemical and immunological profiles. Median white blood cell count was 24.8 × 10⁹/L in severe cases and 15.7 × 10⁹/L in non-severe cases, while median absolute lymphocyte count was 12.4 versus 8.2 × 10⁹/L. Marked leukocytosis occurred in 43.3% of severe cases compared with 8.1% of non-severe cases, and lymphocytosis in 54.1% compared with 21.1%. Mean serum albumin was 34.8 g/L in severe cases and 39.1 g/L in non-severe cases; hypoalbuminemia was present in 42.0% and 11.7%, respectively. Median immunoglobulin G, or IgG, was 5.8 versus 7.1 g/L, while age-specific reduced IgG occurred in 28.0% of severe cases and 11.0% of non-severe cases. Severe cases also had higher median neutrophil counts, C-reactive protein, procalcitonin and alanine aminotransferase, along with lower haemoglobin and higher platelet counts in the descriptive cohort comparisons.

Imaging and hospital-course data showed additional differences. Radiological pneumonia was identified in 57.3% of severe cases and 23.0% of non-severe cases; pulmonary consolidation occurred in 42.7% and 8.7%, atelectasis in 27.4% and 4.2%, and multilobar involvement in 22.9% and 2.3%. Among the 157 severe cases, 80.3% required pediatric intensive care, 35.0% received invasive mechanical ventilation, 30.6% received noninvasive respiratory support and 8.3% died in hospital. These events were absent from the non-severe group by definition. Median hospital stay was 15 days for severe cases, compared with 7 days for non-severe cases. The descriptive comparisons were exploratory and were not adjusted for multiple testing.

The investigators developed two analytically distinct multivariable logistic-regression models. A concurrent clinical-laboratory model included features that overlapped with the severe-disease definition. In that analysis, age below 6 months, prematurity, incomplete or absent vaccination, radiological pneumonia, marked leukocytosis, hypoalbuminemia and reduced IgG were associated with severe disease. Apnea and respiratory distress had large adjusted associations, but were markers of established severe disease rather than independent risk factors. The integrated model’s apparent AUC was 0.89, with sensitivity of 83.4%, specificity of 82.1% and accuracy of 82.4%. Because outcome-defining information entered the model, this estimate may be inflated by outcome-information leakage and cannot be interpreted as prognostic performance.

The admission-time model excluded apnea, respiratory distress, respiratory failure, intensive-care admission, respiratory support and other in-hospital outcome variables. It used demographic, perinatal, vaccination, comorbidity and laboratory information available at the first admission blood sample. In this model, age below 6 months was associated with higher adjusted odds of severe disease, with an adjusted odds ratio of 3.17 and 95% confidence interval of 1.75–5.74. Incomplete-for-age or absent vaccination had an adjusted odds ratio of 2.59, with a 95% confidence interval of 1.42–4.70. Each 5 × 10⁹/L increase in white blood cell count was associated with higher odds, with an adjusted odds ratio of 3.10 and 95% confidence interval of 2.34–4.12; the corresponding value for absolute lymphocyte count was 7.09, with a 95% confidence interval of 4.33–11.60. Higher albumin and IgG concentrations were associated with lower odds. Sex, prematurity and comorbidity were not independently associated with severe disease in this model, although the prematurity confidence interval was wide.

At a probability threshold of 0.50, the full admission-time model had sensitivity of 75.2%, specificity of 95.8%, accuracy of 91.3%, positive predictive value of 83.1% and negative predictive value of 93.3%. Its apparent AUC was 0.961, with a 95% confidence interval of 0.947–0.975, and the optimism-corrected AUC was 0.957. Calibration did not indicate poor fit, and sensitivity analyses showed no evidence that the continuous predictors were non-linear on the logit scale. A simplified model excluding IgG retained an apparent AUC of 0.956, with a 95% confidence interval of 0.941–0.972 and an optimism-corrected AUC of 0.952. Its sensitivity was 71.3%, specificity 96.5% and accuracy 91.0% at the 0.50 threshold. These estimates describe classification within the development cohort and require independent validation before clinical application.

The timing analysis explains why the admission-time AUC must not be presented as evidence of future-deterioration prediction. Of the 157 children classified as severe, 104, or 66.2%, had already met at least one severity criterion at or before the index blood sample; 41, or 26.1%, first met a criterion afterward; and timing could not be reliably established for 12, or 7.6%. The model therefore separated children ultimately classified as severe from those classified as non-severe using information available at admission. It did not demonstrate that admission variables can identify children initially considered non-severe who will later worsen. Other limitations included the retrospective single-center design, missing-data exclusions that could introduce selection bias, no non-pertussis control group, incomplete information on nutrition, fluid balance and organ dysfunction, inability to reconstruct each case’s laboratory confirmation method, and absence of external validation. The associations also do not show causation: IgG varies with age, low albumin may reflect several illness-related processes, and vaccination opportunity is closely linked to age. Prospective multicenter studies would be needed to determine whether a clearly timed model predicts new severe events and generalizes to other pediatric populations.

Subject of Research: Admission-time clinical and laboratory classification of severe pertussis in hospitalized children

Article Title: Integrated clinical and laboratory assessment for severity classification in pediatric pertussis

Article References: Wang, X., & Zhu, X. (2026). Integrated clinical and laboratory assessment for severity classification in pediatric pertussis. Frontiers in Medicine, 13, 1961021. https://doi.org/10.3389/fmed.2026.1961021

Image Credits: AI Generated

DOI: 10.3389/fmed.2026.1961021

Keywords: pediatric pertussis, severity classification, admission biomarkers, leukocytosis, lymphocytosis, hypoalbuminemia, immunoglobulin G, vaccination status, retrospective cohort, hospitalized children, Integrated, clinical