ABSTRACT
Background
Early‐onset neonatal infections are among the most common neonatal diseases. However, the long‐term outcomes of the infections are not well understood.
Objective
To study the association between early‐onset neonatal infection and attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD).
Methods
A nationwide register‐based cohort study was conducted, including near‐term and term children born between 1997 and 2013 with follow‐up until 2021. An early‐onset infection was defined as an invasive bacterial infection occurring within the first week of life, including both physician‐assigned diagnoses and positive bacterial cultures. ADHD and ASD were defined by diagnoses or prescriptions of relevant medication. Associations between sepsis and the neurodevelopmental disorders were investigated using multivariable Cox regression to estimate adjusted hazard ratios (HR), whereas associations with meningitis were examined using person‐time incidence rate ratios (IRR). Sibling‐matched analyses were also conducted for associations with sepsis.
Results
A total of 981,869 children were included, with 8154 defined as having sepsis and 152 defined as having meningitis. Among these, only 257 children had culture‐positive sepsis, whereas 32 had culture‐positive meningitis. The incidence rate of ADHD and ASD for children with sepsis was 4.5 per 1000 and 3.3 per 1000 person‐years, respectively. Sepsis was associated with an increased adjusted likelihood of both ADHD (HR 1.28, 95% CI 1.17, 1.39) and ASD (HR 1.43, 95% CI 1.30, 1.58). However, sibling‐matched analyses especially attenuated the association with ADHD (HR 1.12, 95% CI 0.93, 1.34). Point estimates suggested that children with meningitis also had an increased likelihood of both ADHD (IRR 1.77, 95% CI 0.88, 3.17) and ASD (IRR 2.05, 95% CI 0.89, 4.04).
Conclusions
Early‐onset sepsis was associated with an increased likelihood of ASD, whereas the majority of the association with ADHD could be explained by unmeasured shared familial confounding.
Keywords: attention deficit disorder with hyperactivity, autism spectrum disorder, bacterial infections, bacterial meningitis, neonatal sepsis, neurodevelopmental disorders
1. Introduction
Early‐onset neonatal infection may be defined as an invasive bacterial infection, such as sepsis or meningitis, occurring within the first week of life [1]. The bacteria are typically transmitted from the mother before or during birth, with Group B Streptococcus (GBS) and Escherichia coli being the most frequently isolated pathogens [2, 3]. However, the majority of children are diagnosed by use of clinical signs and biomarkers, without the identification of a specific pathogen [4]. Although early‐onset infections are relatively common in newborns and their association with short‐term morbidities is well established, the long‐term effects remain poorly understood [5, 6, 7, 8]. The infections may induce systemic inflammation, including inflammation of the central nervous system, which could affect the neurodevelopment of the child [8, 9].
Attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) are common neurodevelopmental disorders that typically emerge in childhood [10]. ADHD is characterised by traits such as inattention, hyperactivity, and impulsivity, whereas ASD is characterised by impaired communication and social interaction as well as restricted interests and repetitive behaviours [11]. Both conditions may significantly impact quality of life and contribute to economic challenges for families and society [12, 13, 14]. Although the heritability of ADHD and ASD has been well documented, various foetal and newborn exposures are also believed to influence the overall likelihood and severity of the disorders [15, 16, 17].
Therefore, we aimed to investigate the association between early‐onset neonatal infection in near‐term and term children, defined both by diagnoses and bacterial cultures and neurodevelopmental disorders, including ADHD and ASD up until young adulthood.
2. Methods
2.1. Cohort Selection
This nationwide register‐based cohort study included all Danish liveborn singletons born at ≥ 35 completed gestational weeks from January 1997, to December 2013. Children with any major congenital anomaly were excluded [6]. The Danish Medical Birth Register was used to identify the population [18]. All Danish citizens receive a unique personal identifier at birth, which allowed for linkage of individual‐level information between the various Danish national registers [19]. A description of the data sources is available in the Table S1.
3. Exposures
An early‐onset infection was defined as an invasive bacterial infection within the first week of life, separated into sepsis and meningitis. A child with both sepsis and meningitis was classified as having meningitis. Both probable and culture‐positive infections were included in the definition, as early‐onset infections may be diagnosed based on clinical signs and biomarkers alone or through identification of the specific pathogen (Table 1) [1].
TABLE 1.
Study definitions of early‐onset sepsis and meningitis as well as attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD).
| Exposure | Probable infection (ICD10) | Culture‐positive infection |
|---|---|---|
| Sepsis |
Streptococcal sepsis (A40.0‐A40.9) Other sepsis (A41.0‐A41.9) Bacterial pneumonia (J15.0‐J15.9) Congenital pneumonia (P23.1‐P23.9) Bacterial sepsis of newborn (P36.0‐P36.9) |
Bacterial pathogen isolated from the blood |
| Meningitis |
Bacterial meningitis (G00.0‐G00.9) Meningitis, unspecified (G03.9) |
Bacterial pathogen isolated from the cerebrospinal fluid |
| Outcome | Diagnoses (ICD10) | Prescriptions of medication (ATC) |
|---|---|---|
| ADHD |
Disturbance of activity and attention (F90.0) Hyperkinetic conduct disorder (F90.1) Other hyperkinetic disorders (F90.8) Unspecified hyperkinetic disorder (F90.9) |
Dexamfetamine (N06BA02) Methylphenidate (N06BA04) Modafinil (N06BA07) Atomoxetin (N06BA09) Lisdexamfetamine (N06BA12) |
| ASD |
Childhood autism (F84.0) Atypical autism (F84.1) Asperger syndrome (F84.5) Other pervasive developmental disorders (F84.9) Unspecified pervasive developmental disorders (F84.9) |
NA |
Note: Bacterial pathogens included Enterobacterales (e.g., Escherichia coli ), Enterococcus faecalis , Enterococcus faecium , Haemophilus influenzae , Listeria monocytogenes , Pseudomonas aeruginosa , Staphylococcus aureus , Group B Streptococcus, Streptococcus dysgalactiae , Streptococcus pneumoniae , and Streptococcus pyogenes .
Abbreviations: ATC, anatomical therapeutic chemical code; ICD10, International Classification of Diseases 10th revision; NA, not applicable.
Probable infection was defined as a physician‐assigned diagnosis of sepsis or meningitis by the International Classification of Diseases 10th revision (ICD10). The ICD10 codes were acquired from the Danish National Patient Register, with onset of infection defined as the time of admission [20]. The selected ICD10 codes complied with the Neonatal Sepsis and Meningitis guideline from the Danish Paediatric Society [21]. Hospitalisation had to last at least 5 days unless the child died prior to this. This was to ensure that the children could have received at least 5 days of intravenous antibiotics, which was the minimum required treatment duration of suspected infection during the inclusion period. Culture‐positive infection was present if a bacterial pathogen was cultured from blood or cerebrospinal fluid. Information on cultures was retrieved from all Danish Departments of Clinical Microbiology. Each department provided information on all children born from 2000 to 2013, except for two departments that only provided information from 2002 to 2013.
3.1. Outcomes
ADHD was defined as a physician‐assigned diagnosis of ADHD obtained from the Danish Psychiatric Central Register or a redeemed prescription of ADHD medication obtained from the Danish National Prescription Register (Table 1) [22, 23]. We included prescriptions of ADHD medication as Danish children may be diagnosed with ADHD by private psychiatrists, who are not required to report ICD10 codes to the health authorities [19]. ICD10 codes of ADHD included disturbance of activity and attention (F90.0), hyperkinetic conduct disorder (F90.1), other hyperkinetic disorders (F90.8), and unspecified hyperkinetic disorder (F90.9). Prescriptions of ADHD medication were identified by their anatomical therapeutic chemical code including Dexamfetamine (N06BA02), Methylphenidate (N06BA04), Modafinil (N06BA07), Atomoxetin (N06BA09), and Lisdexamfetamine (N06BA12). ASD was solely defined by diagnoses obtained from the Psychiatric Central Register including ICD10 codes of childhood autism (F84.0), atypical autism (F84.1), Asperger syndrome (F84.5), other pervasive developmental disorders (F84.9), and unspecified pervasive developmental disorders (F84.9) (Table 1) [22]. The diagnoses of ADHD and ASD from the Psychiatric Central Register have previously shown high positive predictive values when compared to information from medical records [24, 25]. The positive predictive value was estimated at 0.87 for ADHD diagnoses and 0.94 for ASD diagnoses.
3.2. Confounding Variables
Potential confounding variables were identified by the literature and depicted by directed acyclic graphs (Figure 1) [26]. These included child sex (male/female) and birth year (1997–2013); maternal age (continuous), parity (primipara/multipara), smoking during pregnancy (yes/no), diabetes (yes/no), and psychiatric disease (yes/no); and socioeconomic variables such as ethnicity (western/non‐western), highest parental education (three categories), family disposable income (continuous), and parental cohabitation (yes/no). Since gestational and birth weight may be considered intermediates between the exposure and outcome, we did not adjust for these variables in the present study. Maternal diabetes was defined as a diagnosis of diabetes mellitus or gestational diabetes or at least two prescriptions of antidiabetic medication. Maternal psychiatric disease was defined as any diagnosis of psychiatric disease or any prescription of antipsychotics or antidepressants. This information was obtained from the National Patient Register, the Psychiatric Central Register, and the National Prescription Register [20, 22, 23]. Other maternal and newborn characteristics were obtained from the Medical Birth Register, whereas the socioeconomic variables were obtained from Statistics Denmark and the Danish Civil Registration System [18, 27, 28]. Ethnicity was based on maternal birthplace and separated into western and non‐western according to the definition used by Statistics Denmark. Information on censoring events such as the dates of child death or emigration was also obtained from the Civil Registration System [28].
FIGURE 1.
Directed acyclic graph depicting potential confounding variables on the association between early‐onset infection and NDD. The directed acyclic graph was drawn in www.daggity.net (last accessed May 10, 2025). No arrows were drawn between confounding variables for simplicity. GBS, Group B Streptococcus; NDD, neurodevelopmental disorder; U, unmeasured confounding.
3.3. Statistical Analysis
Based on whichever occurred first, the children were followed from birth until being registered with ADHD or ASD, death, emigration or end of follow‐up (June 30, 2021). Incidence rates of the neurodevelopmental disorders were estimated in children with and without early‐onset infection. The associations between sepsis and the neurodevelopmental disorders were investigated using multivariable Cox regression analyses to estimate adjusted hazard ratios (HR). These were considered the primary analyses. As a child may have both ADHD and ASD, the association between sepsis and co‐occurring disorders was also conducted [29, 30]. Sibling‐matched analyses for early‐onset sepsis were also performed using stratified Cox regression comparing children with the same mother, where one was exposed to sepsis and the other was not. These analyses were also adjusted for the potential confounders specified above to account for non‐shared confounding between the siblings. Due to violation of the proportional hazards assumption and the sparse number of outcomes in children with meningitis, these associations were instead estimated using unadjusted person‐time incidence rate ratios (IRR). The small number of children also made sibling‐matched analyses unfeasible for this exposure. Subgroup analyses were also conducted only considering children with culture‐positive sepsis and meningitis with the reference population restricted to children born from 2000 to 2013. For the same reasons as above, these associations were also only estimated using unadjusted IRRs. All estimates were provided with 95% confidence intervals (CI). Statistical analyses were performed using STATA 18 (StataCorp 2023. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC).
3.4. Missing Data
Maternal smoking was the only variable with missing values (3.4%), which was accounted for by multiple imputation using chained prediction equations. A total of five imputations were used including several auxiliary variables related to both the parents, pregnancy, and child (Table S2). When information on parental education and family income was missing for the children at the year of birth, registrations within 3 years of birth were included instead.
3.5. Sensitivity Analyses
Several pre‐planned sensitivity analyses were conducted to evaluate the robustness of the estimates from the primary analyses. Complete‐case analyses were conducted as well as cluster analyses considering dependence between children of the same mother with robust standard errors. Alternative definitions of ADHD and ASD were employed to examine any potential misclassification of the outcomes. Analyses were performed considering only primary diagnoses of the neurodevelopmental disorders. Additionally, an analysis was carried out using a stricter definition of ADHD, requiring both a diagnosis and a prescription of ADHD medication. An analysis was also conducted examining the association with childhood autism (F84.0) only, as this may be considered the most severe presentation among the ASDs [24]. Under the assumption of non‐differential misclassification of the outcomes, a quantitative bias analysis was also conducted by estimating the risk ratio for each neurodevelopmental disorder with and without correction for the previously reported positive predictive values of the diagnoses [24, 25]. To evaluate the potential influence of selection bias due to censoring, analyses were also conducted applying the inverse probability of censoring weights. The stabilised weights were estimated by multivariable logistic regression using the same variables as in the primary analyses.
4. Ethics approval
This study was approved by the Central Denmark Region (No. 1‐16‐02‐144‐22) and Statistics Denmark (No. 704920) and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology guideline [31].
5. Results
5.1. Characteristics of the Cohort
A total of 981,869 children were included from 1997 to 2013, with 8154 (0.8%) children defined as having sepsis and 152 (< 0.1%) defined as having meningitis. Among these, 257 (< 0.1%) children had culture‐positive sepsis, whereas 32 (< 0.1%) had culture‐positive meningitis. The most common pathogens were GBS (54%), Staphylococcus aureus (18%) and Escherichia coli (12%). A total of 47,767 (4.9%) children met the criteria for ADHD and 29,699 (3.0%) for ASD, whereas 11,312 (1.2%) had both ADHD and ASD. Median follow‐up time for both ADHD and ASD was 15 years (IQR 11, 20). The median age of being registered with ADHD was 10 years (IQR 8, 14) for children with an infection and 11 years (IQR 8, 15) for children without an infection. The median age of being registered with ASD was 10 years (IQR 6, 14) for children with an infection and 11 years (IQR 7, 14) for children without an infection. The age distribution at the time of diagnosis is presented in Table S3. A total of 62,486 (6.4%) children were censored before being registered with ADHD or ASD, including 2281 (0.2%) who died and 60,205 (6.1%) who emigrated. The mortality rate for children with early‐onset infection was 1.6%, increasing to 6.9% for the subgroup of children with culture‐positive infection, compared with 0.2% in the reference population. Figure 2 shows a flow chart of the study population, whereas a flow chart of the sibling cohort is shown in Figure S1. Table 2 shows baseline characteristics of children with and without early‐onset infection, whereas baseline characteristics of children solely with culture‐positive infection are shown in Table S4.
FIGURE 2.
Flow chart describing the population in the Danish nationwide cohort study on the association between early‐onset infection and NDD. NDD, neurodevelopmental disorder.
TABLE 2.
Characteristics of Danish near‐term and term children with and without early‐onset infection (sepsis and meningitis) born from 1997 to 2013.
| Reference (n = 973,563) | Early‐onset sepsis (n = 8154) | Early‐onset meningitis (n = 152) | |
|---|---|---|---|
| Child characteristics | |||
| Male | 496,377 (51%) | 3008 (37%) | 60 (39%) |
| Female | 477,186 (49%) | 5146 (63%) | 92 (61%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Gestational age, weeks | 40 (39–41) | 40 (39–41) | 40 (38–41) |
| Missing | 8548 (1%) | 123 (2%) | NR |
| Birthweight, g | 3560 (510) | 3591 (663) | 3505 (615) |
| Missing | 9321 (1%) | 145 (2%) | NR |
| Maternal characteristics | |||
| Age, years | 30 (4.9) | 30 (5.1) | 30 (5.4) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Primipara | 416,995 (43%) | 5245 (64%) | 92 (61%) |
| Multipara | 556,568 (57%) | 2909 (36%) | 60 (39%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Smoking during pregnancy | 141,524 (15%) | 1292 (17%) | 22 (15%) |
| Non‐smoking | 799,095 (82%) | 6500 (80%) | 125 (82%) |
| Missing | 32,944 (3%) | 362 (4%) | 5 (3%) |
| Diabetes | 30,165 (3%) | 374 (5%) | 12 (8%) |
| No diabetes | 943,398 (97%) | 7780 (95%) | 140 (92%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Psychiatric disease | 120,267 (12%) | 1111 (14%) | 33 (22%) |
| No psychiatric disease | 853,296 (88%) | 7043 (86%) | 119 (78%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Socioeconomics | |||
| Non‐western origin | 116,036 (12%) | 963 (12%) | 17 (11%) |
| Western origin | 857,527 (88%) | 7191 (88%) | 135 (89%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Parental education | |||
| Low | 463,407 (48%) | 4096 (50%) | 74 (49%) |
| Medium | 295,651 (30%) | 2421 (30%) | 51 (34%) |
| High | 214,505 (22%) | 1637 (20%) | 27 (18%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Yearly family income, 1000 DKK | 188 (145–234) | 188 (143–231) | 188 (134–240) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
| Non‐cohabiting parents | 120,978 (12%) | 1277 (16%) | 34 (22%) |
| Cohabiting parents | 852,585 (88%) | 6877 (84%) | 118 (78%) |
| Missing | 0 (0%) | 0 (0%) | 0 (0%) |
Note: Continuous variables are presented as mean values with standard deviations (normally distributed) or medians with interquartile ranges (non‐normally distributed), whereas categorical variables are presented as numbers with percentages. NR, not reported (values are not reported in accordance with Statistics Denmark's regulations on microdata).
5.2. Sepsis
The incidence rate of ADHD for children with sepsis was 4.5 per 1000 person‐years, compared with 3.2 per 1000 person‐years for children without infection. The incidence rate of ASD for children with sepsis was 3.3 per 1000 person‐years, compared with 2.0 per 1000 person‐years for those without. Table 3 shows the results from the primary analyses. Sepsis was associated with an increased likelihood of both ADHD (adjusted HR 1.28, 95% CI 1.17, 1.39) and ASD (adjusted HR 1.43, 95% CI 1.30, 1.58). Sepsis was also associated with an increased likelihood of co‐occurring disorders (adjusted HR 1.49, 95% CI 1.28, 1.74). The estimates from the sibling‐matched analyses were attenuated compared with the estimates from the primary analyses (Table 3). The association between sepsis and ADHD disappeared (adjusted HR: 1.12, 95% CI 0.93, 1.34), whereas the association with ASD remained (adjusted HR: 1.32, 95% CI: 1.05, 1.66).
TABLE 3.
Associations between early‐onset sepsis and neurodevelopmental disorders (NDD) including attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) in Danish near‐term and term children born from 1997 to 2013.
| NDD | Events (%) in children without sepsis | Events (%) in children with sepsis | Unadjusted HR (95% CI) | Adjusted HR (95% CI) |
|---|---|---|---|---|
| ADHD | ||||
| Unpaired | 47,194 (4.8%) | 562 (6.9%) | 1.41 (1.29, 1.53) | 1.28 (1.17, 1.39) |
| Siblings | 365 (5.3%) | 359 (7.0%) | 1.16 (0.97, 1.38) | 1.12 (0.93, 1.34) |
| ASD | ||||
| Unpaired | 29,278 (3.0%) | 413 (5.1%) | 1.66 (1.51, 1.83) | 1.43 (1.30, 1.58) |
| Siblings | 209 (3.1%) | 244 (4.8%) | 1.46 (1.18, 1.80) | 1.32 (1.05, 1.66) |
| Co‐occurring | ||||
| Unpaired | 11,152 (1.2%) | 166 (2.0%) | 1.79 (1.53, 2.09) | 1.49 (1.28, 1.74) |
Note: A total of 973,563 children were included in the reference population with 8154 children defined as having sepsis. A total of 5127 children had early‐onset sepsis with 6834 siblings who did not. Variables in the adjusted Cox regression included sex, birth year, maternal age, parity, maternal smoking, maternal diabetes, maternal psychiatric disease, ethnicity, parental education, family disposable income and parental cohabitation. Results are provided as hazard ratios (HR) with 95% confidence intervals (CI).
5.3. Meningitis and Culture‐Positive Infection
Table 4 shows the incidence rates and IRRs for the different definitions of infection. Point estimates indicated that children with meningitis had an increased likelihood of both ADHD (IRR 1.77, 95% CI 0.88, 3.17) and ASD (IRR 2.05, 95% CI 0.89, 4.04). Point estimates similarly indicated that children with culture‐positive infection had an increased likelihood of the neurodevelopmental disorders, with the exception of the association between culture‐positive sepsis and ADHD (IRR 0.98, 95% CI 0.47, 1.79).
TABLE 4.
Associations between different definitions of early‐onset infection and attention deficit hyperactivity disorder (ADHD) and autism spectrum disorder (ASD) in Danish near‐term and term children born from 1997 to 2013.
| Definition of infection | Total n | Events (%) | IR per 1000 years (95% CI) | Person‐time IRR (95% CI) |
|---|---|---|---|---|
| ADHD | ||||
| No infection (1997–2013) | 973,563 | 47,194 (4.8%) | 3.2 (3.2, 3.2) | 1.00 (reference) |
| No infection (2000–2013) | 800,188 | 36,348 (4.5%) | 3.3 (3.3, 3.3) | 1.00 (reference) |
| Sepsis | ||||
| All infections | 8154 | 562 (6.9%) | 4.5 (4.2, 4.9) | 1.42 (1.30, 1.54) |
| Culture‐positive | 257 | 10 (3.9%) | 3.2 (1.7, 5.9) | 0.98 (0.47, 1.79) |
| Meningitis | ||||
| All infections | 152 | 11 (7.2%) | 5.7 (3.1, 10.2) | 1.77 (0.88, 3.17) |
| Culture‐positive | 32 | 4 (12.5%) | 10.0 (3.7, 26.6) | 3.05 (0.83, 7.82) |
| ASD | ||||
| No infection (1997–2013) | 973,563 | 29,278 (3.0%) | 2.0 (1.9, 2.0) | 1.00 (reference) |
| No infection (2000–2013) | 800,188 | 23,844 (3.0%) | 2.2 (2.1, 2.2) | 1.00 (reference) |
| Sepsis | ||||
| All infections | 8154 | 413 (5.1%) | 3.3 (3.0, 3.6) | 1.67 (1.52, 1.85) |
| Culture‐positive | 257 | 12 (4.7%) | 3.8 (2.2, 6.7) | 1.77 (0.91, 3.09) |
| Meningitis | ||||
| All infections | 152 | 8 (5.3%) | 4.1 (2.0, 8.1) | 2.05 (0.89, 4.04) |
| Culture‐positive | 32 | 3 (9.4%) | 7.2 (2.3, 22.5) | 3.35 (0.69, 9.80) |
Note: The analyses of all infections included both probable and culture‐positive infections from 1997 to 2013. The subgroup analyses of culture‐positive infection only included children from 2000 to 2013. Results are provided as unadjusted person‐time incidence rate ratio (IRR) with 95% confidence intervals (CI).
5.4. Results From Sensitivity Analyses
The sensitivity analyses showed estimates similar to those from the primary analyses (Tables S5). In the quantitative bias analysis, the risk ratio estimates for the association between early‐onset sepsis and the neurodevelopmental disorders increased only slightly after correction for the previously reported positive predictive values of the diagnoses (Table S6).
6. Comment
6.1. Principal Findings
In this nationwide register‐based cohort study of near‐term and term children, early‐onset sepsis was associated with an increased likelihood of ADHD and ASD, both separately and as co‐occurring disorders. However, sibling‐matched analyses indicated that the association with ADHD could be explained by unmeasured time‐stable maternal and familial confounding. Point estimates also indicated that children with meningitis had an increased likelihood of both disorders, but the small number of children with these infections resulted in a high degree of uncertainty.
6.2. Strengths of the Study
This study addresses several needs previously emphasised by other reviews and meta‐analyses on the association between early‐onset neonatal infection and long‐term outcomes [7, 8]. We included children born between 1997 and 2013, with follow‐up extending to 2021, providing a minimum of 8 years and up to 24 years of follow‐up. The large population size allowed us to investigate rare exposures such as meningitis and culture‐positive infection and to perform sibling‐matched analyses. Another strength was the access to high‐quality data from the Danish registres, which enabled adjustment for multiple potential confounders and the conduct of several sensitivity analyses. These analyses supported the robustness of our findings.
6.3. Limitations of the Data
Differential misclassification of the outcome is always a possibility in non‐blinded studies. In this study, an early‐onset infection could have increased the likelihood of receiving a diagnosis of the neurodevelopmental disorders. Parents may be traumatised by their child's admission to the neonatal intensive care unit right after birth, leading to persistent anxiety and lower threshold for seeking healthcare later in childhood. Additionally, a history of perinatal infection may influence the decision of the psychiatrist to evaluate the child for these disorders. However, Denmark's universal healthcare system and the presence of school psychologists who help identify and refer children with suspected neurodevelopmental disorder are likely to minimise this potential detection bias.
Non‐differential misclassification of either the exposure or outcome could also have affected our results. No consensus definition exists for early‐onset sepsis and meningitis [1]. Early‐onset infections are mostly diagnosed without a pathogen being identified [4]. This could be due to maternal or newborn antibiotics before sampling or small sample volumes compared with adults [32]. However, some children may also be misdiagnosed with infection due to other conditions with similar symptoms. Our definition of sepsis and meningitis included both probable and culture‐positive infections, ensuring that most children with an infection were captured, though some without an infection may have been included as well. That some children without an infection were included in our definition could be supported by the fact that the mortality was higher in children with culture‐positive infection. However, this subgroup may also represent more severe forms of infection (e.g., higher concentration of bacteria). To evaluate this limitation, we also performed subgroup analyses of children with culture‐positive infection only. Except for the association between early‐onset sepsis and ADHD, the point estimates for culture‐positive infection were higher than those for probable infection. However, the wide and overlapping confidence intervals hinder any definitive conclusions. This also meant that too few culture‐positive samples were available for any meaningful pathogen‐specific analyses. Regarding misclassification of the outcome, several steps were also taken to minimise and assess this risk. We only considered diagnoses obtained from psychiatric departments, which previously have shown high positive predictive values when compared with medical record data. The quantitative bias analysis also indicated that our results only slightly underestimated the associations. In addition, we also conducted several sensitivity analyses using stricter definitions of the neurodevelopmental disorders. These analyses showed results consistent with the primary analyses.
Selection bias and missing information are other factors that may affect the results of cohort studies. As we were able to include all children born in Denmark during the study period, baseline selection is not an issue in the present study. The dataset was also virtually complete, with maternal smoking as the only variable with missing values (3.4%). Missing values were handled by multiple imputation, including several auxiliary variables that explained the underlying mechanisms of the missing data. The assumption of missing at random was therefore considered fulfilled. Regarding loss to follow‐up, children with early‐onset infection had a higher mortality compared to those without infection. This may have caused an underestimation of the associations if those who died had a higher likelihood for the neurodevelopmental disorders had they survived. Nonetheless, the results from the primary analyses were similar with and without applying the inverse probability of censoring weights.
Residual confounding is also a possible limitation of observational studies, but we included several potential confounders in the adjusted model related to both the child and parents. We also conducted sibling‐matched analyses to assess the potential effect of any unmeasured shared familial confounding, potentially including both genetic, intrauterine, environmental, and socioeconomic factors. However, adjusted and sibling‐matched analyses were not conducted for associations regarding meningitis and culture‐positive infection, as the number of outcomes within these exposures was believed to be too small to yield any meaningful results. In addition, the sibling analyses should still be interpreted with some caution [33]. Non‐shared confounders between siblings may still bias the estimates, whereas misclassification of the exposure may result in more substantial underestimations compared with unpaired analyses. The statistical power of the sibling analyses was also limited by the reduced number of included children.
We chose to focus on near‐term and term children, as studies in this population are limited, as also noted in a recent meta‐analysis [7]. We believe that preterm and term children represent distinct populations that would require separate analyses, making the scope too large for one single article. This means that our findings can only be generalised to near‐term and term children and that further studies are needed to study the association for preterm populations.
6.4. Interpretation
The biological mechanisms through which early‐onset infections may influence neurodevelopment are likely complex. Early‐onset sepsis may trigger systemic and local inflammatory responses, which could disrupt brain development during a critical stage of rapid growth [8]. This disruption could be worsened by a bacterial invasion of the central nervous system, exacerbating the cerebral inflammation and cytotoxicity [8]. Our point estimates may support this theory as meningitis showed a higher likelihood of the neurodevelopmental disorders compared to sepsis.
Beyond this acute damage, it has been hypothesised that the early activation of the immune system may lead to sustained inflammatory changes, which also could affect neurodevelopment [34]. Evidence suggests that especially individuals with ASD may have an altered inflammatory state, including deviations in cytokine levels and microglial activation [35, 36, 37]. However, it remains unclear whether these inflammatory changes are consequences of the neurodevelopmental disorders rather than the underlying cause.
Another potential explanation could relate to the antibiotic treatment, which may affect brain development through the gut‐brain axis [38, 39]. It has been hypothesised that dysbiosis might impact neurodevelopment through both endocrine, neuronal and inflammatory pathways [39]. Studies have found that children with ASD have an increased prevalence of symptoms from the gastrointestinal tract [40]. A recent meta‐analysis investigated this theory, showing that antibiotic exposure during childhood was associated with an increased likelihood of both ADHD and ASD [41]. However, both associations disappeared when only studies with sibling‐matched designs were considered, suggesting that unmeasured shared familial confounding might be responsible for the findings.
This points to another potential explanation for an association between early‐onset infection and the neurodevelopmental disorders, i.e., that certain genes may be associated with susceptibility to both infection and impaired neurodevelopment. Our sibling‐matched analyses also attenuated the association between sepsis and the neurodevelopmental disorders; notably, the association with ADHD disappeared. However, the association with ASD remained, further suggesting that early‐life inflammation may play a role in the pathophysiology of this disorder.
7. Conclusions
Early‐onset neonatal sepsis was associated with an increased likelihood of ASD in near‐term and term children, whereas the majority of the association with ADHD could be explained by unmeasured shared familial confounding. Preventive measures directed towards bacterial infections in the first week of life may therefore not only reduce the immediate morbidity and mortality related to the infection, but also the long‐term likelihood of ASD. However, such preventive strategies may not reduce the likelihood of ADHD.
Author Contributions
M.A., N.B.M., S.Y.N., and T.B.H. have made substantial contributions to the concept and design. All authors have made substantial contributions to acquisition, analysis, or interpretation of data. M.A. drafted the manuscript, while all authors have made critical review of the manuscript for important intellectual content. All authors have approved the final manuscript as submitted and agree to be accountable for all aspects of the work. M.A. is responsible for the overall content (as guarantor).
Consent
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Data S1
Funding: This study was funded by Graduate School of Health at Aarhus University (MA), Elsass Foundation (No. 21‐3‐0256) (MA), Helsefonden (No. 21‐B‐0186) (MA), and Beckett Foundation (No. 21‐2‐6912) (MA).
Editors note: A commentary based on this article appears on pages 598‐600.
Data Availability Statement
Individual‐level data are not publicly available according to Danish law. Aggregated data may be made available upon request to the corresponding author.
References
- 1. McGovern M., Giannoni E., Kuester H., et al., “Challenges in Developing a Consensus Definition of Neonatal Sepsis,” Pediatric Research 88, no. 1 (2020): 14–26, 10.1038/s41390-020-0785-x. [DOI] [PubMed] [Google Scholar]
- 2. Cailes B., Kortsalioudaki C., Buttery J., et al., “Epidemiology of UK Neonatal Infections: The neonIN Infection Surveillance Network,” Archives of Disease in Childhood. Fetal and Neonatal Edition 103, no. 6 (2018): F547–F553, 10.1136/archdischild-2017-313203. [DOI] [PubMed] [Google Scholar]
- 3. Stoll B. J., Puopolo K. M., Hansen N. I., et al., “Early‐Onset Neonatal Sepsis 2015 to 2017, the Rise of Escherichia coli, and the Need for Novel Prevention Strategies,” JAMA Pediatrics 174, no. 7 (2020): e200593, 10.1001/jamapediatrics.2020.0593. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Fjalstad J. W., Stensvold H. J., Bergseng H., et al., “Early‐Onset Sepsis and Antibiotic Exposure in Term Infants: A Nationwide Population‐Based Study in Norway,” Pediatric Infectious Disease Journal 35, no. 1 (2016): 1–6, 10.1097/INF.0000000000000906. [DOI] [PubMed] [Google Scholar]
- 5. Fleischmann C., Reichert F., Cassini A., et al., “Global Incidence and Mortality of Neonatal Sepsis: A Systematic Review and Meta‐Analysis,” Archives of Disease in Childhood 106, no. 8 (2021): 745–752, 10.1136/archdischild-2020-320217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Andersen M., Matthiesen N. B., Murra M., Nielsen S. Y., and Henriksen T. B., “Early‐Onset Neonatal Infection and School Performance: A Danish Nationwide Population‐Based Cohort Study,” Clinical Microbiology and Infection 31 (2025): 389–395, 10.1016/j.cmi.2024.12.029. [DOI] [PubMed] [Google Scholar]
- 7. Ong W. J., Seng J. J. B., Yap B., et al., “Impact of Neonatal Sepsis on Neurocognitive Outcomes: A Systematic Review and Meta‐Analysis,” BMC Pediatrics 24, no. 1 (2024): 505, 10.1186/s12887-024-04977-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Thomas R., Bijlsma M. W., Gonçalves B. P., Nakwa F. L., Velaphi S., and Heath P. T., “Long‐Term Impact of Serious Neonatal Bacterial Infections on Neurodevelopment,” Clinical Microbiology and Infection 30, no. 1 (2024): 28–37, 10.1016/j.cmi.2023.04.017. [DOI] [PubMed] [Google Scholar]
- 9. Han V. X., Patel S., Jones H. F., et al., “Maternal Acute and Chronic Inflammation in Pregnancy Is Associated With Common Neurodevelopmental Disorders: A Systematic Review. Transl ,” Psychiatry 11, no. 1 (2021): 71, 10.1038/s41398-021-01198-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. GBD 2019 Mental Disorders Collaborators , “Global, Regional, and National Burden of 12 Mental Disorders in 204 Countries and Territories, 1990–2019: A Systematic Analysis for the Global Burden of Disease Study 2019,” Lancet Psychiatry 9, no. 2 (2012): 137–150, 10.1016/s2215-0366(21)00395-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. American Psychiatric Association , ed., Diagnostic and Statistical Manual of Mental Disorders, 5th ed. (American Psychiatric Association, 2013). [Google Scholar]
- 12. Jonsson U., Alaie I., Löfgren Wilteus A., et al., “Annual Research Review: Quality of Life and Childhood Mental and Behavioural Disorders ‐ a Critical Review of the Research,” Journal of Child Psychology and Psychiatry 58, no. 4 (2017): 439–469, 10.1111/jcpp.12645. [DOI] [PubMed] [Google Scholar]
- 13. Catalá‐López F., Hutton B., Page M. J., et al., “Mortality in Persons With Autism Spectrum Disorder or Attention‐Deficit/Hyperactivity Disorder: A Systematic Review and Meta‐Analysis,” JAMA Pediatrics 176, no. 4 (2022): e216401, 10.1001/jamapediatrics.2021.6401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Sampaio F., Feldman I., Lavelle T. A., and Skokauskas N., “The Cost‐Effectiveness of Treatments for Attention Deficit‐Hyperactivity Disorder and Autism Spectrum Disorder in Children and Adolescents: A Systematic Review,” European Child & Adolescent Psychiatry 31, no. 11 (2022): 1655–1670, 10.1007/s00787-021-01748-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Kim J. Y., Son M. J., Son C. Y., et al., “Environmental Risk Factors and Biomarkers for Autism Spectrum Disorder: An Umbrella Review of the Evidence,” Lancet Psychiatry 6, no. 7 (2019): 590–600, 10.1016/s2215-0366(19)30181-6. [DOI] [PubMed] [Google Scholar]
- 16. Kim J. H., Kim J. Y., Lee J., et al., “Environmental Risk Factors, Protective Factors, and Peripheral Biomarkers for ADHD: An Umbrella Review,” Lancet Psychiatry 7, no. 11 (2020): 955–970, 10.1016/s2215-0366(20)30312-6. [DOI] [PubMed] [Google Scholar]
- 17. Gidziela A., Ahmadzadeh Y. I., Michelini G., et al., “A Meta‐Analysis of Genetic Effects Associated With Neurodevelopmental Disorders and Co‐Occurring Conditions,” Nature Human Behaviour 7, no. 4 (2023): 642–656, 10.1038/s41562-023-01530-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Bliddal M., Broe A., Pottegard A., Olsen J., and Langhoff‐Roos J., “The Danish Medical Birth Register,” European Journal of Epidemiology 33, no. 1 (2018): 27–36, 10.1007/s10654-018-0356-1. [DOI] [PubMed] [Google Scholar]
- 19. Schmidt M., Schmidt S. A. J., Adelborg K., et al., “The Danish Health Care System and Epidemiological Research: From Health Care Contacts to Database Records,” Clinical Epidemiology 11 (2019): 563–591, 10.2147/clep.S179083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Schmidt M., Schmidt S. A., Sandegaard J. L., Ehrenstein V., Pedersen L., and Sorensen H. T., “The Danish National Patient Registry: A Review of Content, Data Quality, and Research Potential,” Clinical Epidemiology 7 (2015): 449–490, 10.2147/CLEP.S91125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Stanchev H., Fenger‐Grøn J., Simonsgaard M., Malchau E., and Nygaard U., “Neonatal Sepsis Og Meningitis.” (2018), http://paediatri.dk/images/dokumenter/vejledninger_2018/DPS_sepsis_og_meningitis_11.11.2018.pdf.
- 22. Mors O., Perto G. P., and Mortensen P. B., “The Danish Psychiatric Central Research Register,” Scandinavian Journal of Public Health 39, no. 7 Suppl (2011): 54–57, 10.1177/1403494810395825. [DOI] [PubMed] [Google Scholar]
- 23. Kildemoes H. W., Sorensen H. T., and Hallas J., “The Danish National Prescription Registry,” Scandinavian Journal of Public Health 39, no. 7 Suppl (2011): 38–41, 10.1177/1403494810394717. [DOI] [PubMed] [Google Scholar]
- 24. Lauritsen M. B., Jorgensen M., Madsen K. M., et al., “Validity of Childhood Autism in the Danish Psychiatric Central Register: Findings From a Cohort Sample Born 1990‐1999,” Journal of Autism and Developmental Disorders 40, no. 2 (2010): 139–148, 10.1007/s10803-009-0818-0. [DOI] [PubMed] [Google Scholar]
- 25. Mohr‐Jensen C., Vinkel Koch S., Briciet Lauritsen M., and Steinhausen H. C., “The Validity and Reliability of the Diagnosis of Hyperkinetic Disorders in the Danish Psychiatric Central Research Registry,” European Psychiatry 35 (2016): 16–24, 10.1016/j.eurpsy.2016.01.2427. [DOI] [PubMed] [Google Scholar]
- 26. Williams T. C., Bach C. C., Matthiesen N. B., Henriksen T. B., and Gagliardi L., “Directed Acyclic Graphs: A Tool for Causal Studies in Paediatrics,” Pediatric Research 84, no. 4 (2018): 487–493, 10.1038/s41390-018-0071-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. “Statistics Denmark,” https://www.dst.dk/en.
- 28. Pedersen C. B., “The Danish Civil Registration System,” Scandinavian Journal of Public Health 39, no. 7 Suppl (2011): 22–25, 10.1177/1403494810387965. [DOI] [PubMed] [Google Scholar]
- 29. Zablotsky B., Bramlett M. D., and Blumberg S. J., “The co‐Occurrence of Autism Spectrum Disorder in Children With ADHD,” Journal of Attention Disorders 24, no. 1 (2020): 94–103, 10.1177/1087054717713638. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Rong Y., Yang C.‐J., Jin Y., and Wang Y., “Prevalence of Attention‐Deficit/Hyperactivity Disorder in Individuals With Autism Spectrum Disorder: A Meta‐Analysis,” Research in Autism Spectrum Disorders 83 (2021): 101759, 10.1016/j.rasd.2021.101759. [DOI] [Google Scholar]
- 31. Fitchett E. J. A., Seale A. C., Vergnano S., et al., “Strengthening the Reporting of Observational Studies in Epidemiology for Newborn Infection (STROBE‐NI): An Extension of the STROBE Statement for Neonatal Infection Research,” Lancet Infectious Diseases 16, no. 10 (2016): e202–e213, 10.1016/S1473-3099(16)30082-2. [DOI] [PubMed] [Google Scholar]
- 32. Klingenberg C., Kornelisse R. F., Buonocore G., Maier R. F., and Stocker M., “Culture‐Negative Early‐Onset Neonatal Sepsis—At the Crossroad Between Efficient Sepsis Care and Antimicrobial Stewardship,” Frontiers in Pediatrics 6 (2018): 285, 10.3389/fped.2018.00285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Frisell T., Öberg S., Kuja‐Halkola R., and Sjölander A., “Sibling Comparison Designs: Bias From Non‐Shared Confounders and Measurement Error,” Epidemiology 23, no. 5 (2012): 713–720, 10.1097/EDE.0b013e31825fa230. [DOI] [PubMed] [Google Scholar]
- 34. Hagberg H., Gressens P., and Mallard C., “Inflammation During Fetal and Neonatal Life: Implications for Neurologic and Neuropsychiatric Disease in Children and Adults,” Annals of Neurology 71, no. 4 (2012): 444–457, 10.1002/ana.22620. [DOI] [PubMed] [Google Scholar]
- 35. Mitchell R. H. and Goldstein B. I., “Inflammation in Children and Adolescents With Neuropsychiatric Disorders: A Systematic Review,” Journal of the American Academy of Child and Adolescent Psychiatry 53, no. 3 (2014): 274–296, 10.1016/j.jaac.2013.11.013. [DOI] [PubMed] [Google Scholar]
- 36. Liao X., Yang J., Wang H., and Li Y., “Microglia Mediated Neuroinflammation in Autism Spectrum Disorder,” Journal of Psychiatric Research 130 (2020): 167–176, 10.1016/j.jpsychires.2020.07.013. [DOI] [PubMed] [Google Scholar]
- 37. Zhao H., Zhang H., Liu S., Luo W., Jiang Y., and Gao J., “Association of Peripheral Blood Levels of Cytokines With Autism Spectrum Disorder: A Meta‐Analysis,” Frontiers in Psychiatry 12 (2021): 670200, 10.3389/fpsyt.2021.670200. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Fjalstad J. W., Esaiassen E., Juvet L. K., van den Anker J. N., and Klingenberg C., “Antibiotic Therapy in Neonates and Impact on Gut Microbiota and Antibiotic Resistance Development: A Systematic Review,” Journal of Antimicrobial Chemotherapy 73 (2018): 569–580, 10.1093/jac/dkx426. [DOI] [PubMed] [Google Scholar]
- 39. Góralczyk‐Bińkowska A., Szmajda‐Krygier D., and Kozłowska E., “The Microbiota‐Gut‐Brain Axis in Psychiatric Disorders,” International Journal of Molecular Sciences 23, no. 19 (2022): 11245, 10.3390/ijms231911245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Leader G., Abberton C., Cunningham S., et al., “Gastrointestinal Symptoms in Autism Spectrum Disorder: A Systematic Review,” Nutrients 14, no. 7 (2022): 1471, 10.3390/nu14071471. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Yu H.‐y., Zhou Y.‐y., Pan L.‐y., Zhang X., and Jiang H.‐y., “Early Life Antibiotic Exposure and the Subsequent Risk of Autism Spectrum Disorder and Attention Deficit Hyperactivity Disorder: A Systematic Review and Meta‐Analysis,” Journal of Autism and Developmental Disorders 52, no. 5 (2022): 2236–2246, 10.1007/s10803-021-05121-6. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1
Data Availability Statement
Individual‐level data are not publicly available according to Danish law. Aggregated data may be made available upon request to the corresponding author.