The data for this study were obtained from the Korean National Health Insurance Service (NHIS), the entity responsible for administering the health insurance coverage mandated by the Korean government. In Korea, each individual is assigned a unique identification number, ensuring that over 98% of the population is covered by a single-payer healthcare system. Medical claims data were collected from all people who visited the hospital (all 1st, 2nd, and 3rd medical institutions).

The NHIS compiles medical claims data, demographic details, and outcomes from the National Health Screening Program for Infants and Children (NHSPIC). The NHSPIC provides annual health screening tests for children of preschool age. Collected medical claims data were cataloged using the ICD-10 codes. This study was performed according to the principles of the Declaration of Helsinki and received an exemption from review by the Institutional Review Board of Hanyang University Guri Hospital (approval no. GURI 2022–04-017). Additionally, this study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Online Resource 1).

Figure 1 depicts the flow diagram of the study population. We identified newborns born between 2008 and 2014 and classified them according to the ICD-10 codes for preterm birth (P07) and full-term birth (Z38), finding a total of 2,330,289 individuals. These individuals were monitored up to the age of 6 years. The cohort was divided into three groups based on gestational age (GA): (1) extremely preterm (EP) infants born at < 28 weeks of GA; (2) other preterm (OP) infants born between 28 and 36 weeks of GA; and (3) full-term (FT) infants born at ≥ 37 weeks of GA. Given the strong correlation between birth weight (BW) and GA, both indicate the level of prematurity in preterm infants.

Flow diagram of the study population. Abbreviations: IPTW, inverse probability of treatment-weighted; SES, socioeconomic status

Individuals who (1) died during the follow-up after birth (n = 5971), (2) were diagnosed with chromosomal anomalies (n = 5080), (3) had any malignancies (n = 5475), or (4) possessed incomplete medical records (n = 106,056) were excluded. Finally, a total of 2,224,476 children were included in the final analysis.

Asthma and AD were the primary outcomes of interest. Asthma was defined as having > 3 outpatient or inpatient visits coded as J45 or J46 according to the ICD-10 codes within a year, with at least 6 months separating the first and last visits. Early-onset asthma was categorized as a clinical diagnosis made before the child reached 2 years of age. Severe asthma was determined by at least 1 hospital visit of status asthmaticus (coded as J46 in ICD-10) or > 1 emergency department visit primarily for asthma.

AD was characterized by > 3 hospital visits for AD coded as J20.8 and J20.9 within a year, using the ICD-10 codes. An early AD diagnosis was defined as that made before the child reached 2 years of age. Severe AD was identified in patients who, upon diagnosis, were prescribed high-potency topical steroids, methotrexate, or cyclosporine.

Demographic data, including the child’s sex, birth year, residence, and socioeconomic status (SES), were sourced from the NHIS database. Residence was classified into metropolitan or rural, with metropolitan areas including Seoul and the six major Korean cities, exceeding a population of 1 million. SES was inferred from health insurance premium quartiles, which are indicative of household income levels.

Maternal and perinatal conditions, such as delivery type, maternal gestational diabetes mellitus (GDM), and pregnancy-induced hypertension (PIH), were linked to each child through maternal data in the NHIS database. Neonatal comorbidities, including respiratory distress syndrome (RDS), bronchopulmonary dysplasia (BPD), pulmonary hypertension, and allergic proctocolitis, were considered potential confounders and included in the analysis. However, given their prevalence in preterm infants, these comorbidities were excluded from the inverse probability of treatment-weighted (IPTW) analysis.

Furthermore, neonatal comorbidities, such as RDS, BPD, patent ductus arteriosus (PDA), pulmonary hypertension, necrotizing enterocolitis (NEC), intraventricular hemorrhage (IVH), hypothyroidism, neonatal sepsis, neonatal jaundice, and retinopathy of prematurity (ROP), were investigated as risk factors for asthma and AD. Online Resource 2 provides comprehensive details of variables.

We employed the IPTW based on the propensity score (PS) to mitigate selection bias for assessing the incidence of asthma and AD among the three population subgroups. This approach involved creating a pseudo-dataset, in which each participant was weighted according to the IPTW. Then, the dataset was subjected to regression analysis. The PS for each subgroup was determined using a logistic regression model incorporating baseline demographic information (sex, birth year, residence, and SES) and maternal factors (maternal asthma, maternal AD, GDM, cesarean section, PIH, and intrauterine growth restriction (IUGR)).

Baseline demographic characteristics and comorbidities across groups were compared using a one-way analysis of variance for continuous variables and the chi-squared test for categorical variables. We calculated the maximum absolute standardized difference (ASD) for each variable to assess the balance of characteristics among the groups, considering the maximum ASD of < 0.1 indicative of well-balanced variables between the groups.

Adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) for the risk of developing asthma and AD were derived from Cox proportional hazards models. A multivariate Cox regression analysis was conducted to adjust for potential confounders specific to each disease. For asthma, variables such as RDS, BPD, and pulmonary hypertension were adjusted. Regarding AD, adjustments were made for allergic proctocolitis. The assumption of proportional hazards was verified using log-minus-log plots.

The incidence of asthma and AD by age was depicted using Kaplan–Meier survival curves, with differences among subgroups assessed using the log-rank test. Additionally, logistic regression analysis was employed to identify major risk factors for diseases. A p-value < 0.01 indicated statistical significance. Statistical analyses were conducted using the SAS Enterprise Guide (version 7.1, SAS Institute Inc., Cary, NC, USA).

An additional sensitivity analysis was conducted to validate our statistical analysis. We divided the study population into three groups according to BW: extremely low birth weight (ELBW) including those with BW < 1 kg, low birth weight (LBW) including those with BW < 2.5 kg, and normal birth weight (NBW). We selected children who were assigned birth codes regarding BW. The IPTW with covariates, such as sex, birth year, residence, SES, maternal allergic disorders, GDM, cesarean section, PIH, and IUGR, was included in the analysis. Then, the association between BW and diseases (asthma and AD) was validated and expressed as HR with 95% CI using the Cox proportional hazards model.