Risk of Neuropsychiatric Disorders After Pediatric Delirium in Children Under 12: A Cohort Study
Article information
Abstract
Objective
Delirium is known to be related to neuropsychiatric comorbidities as long-term consequences in adult patients. However, the risk of major neuropsychiatric disorders after pediatric delirium remains largely unexplored.
Methods
We analyzed de-identified electronic health records (2013 to 2023) from the TriNetX Research network, a network with more than 111 million patients. Patients under age 12 years with delirium and a control group without delirium were identified and matched by age, sex, race, ethnicity, and physical comorbidities at a one-to-four ratio. We applied Cox regression and Kaplan–Meier analysis to assess the risk of neuropsychiatric disorders.
Results
A total of 618 pediatric patients with delirium were included, demonstrating a 2.15-fold higher risk of neuropsychiatric disorders than controls without delirium (hazard ratio [HR] with 95% confidence interval [CI]: 2.15, 1.82–2.56). The 5-year freedom from these disorders was 68.5% (95% CI, 65.5–71.8) in the study cohort, whereas it was 49.8% (95% CI, 44.1–56.2) in the control cohort. Subgroup analysis showed that children aged 6 years or younger were more likely to be diagnosed with externalizing disorder, including substance use disorder (HR=4.34; 95% CI, 2.00–9.44; p<0.001), intellectual disability (HR=1.71; 95% CI, 1.27–2.30; p<0.001), and attention-deficit/hyperactivity disorder (HR=2.25; 95% CI, 1.16–4.34; p=0.02).
Conclusion
Pediatric patients with delirium are at increased risk of major neuropsychiatric disorders compared to their control counterparts without delirium. Clinicians need to be aware of early symptoms and signs suggesting major neuropsychiatric disorders during the long-term follow-up period.
INTRODUCTION
Delirium is a serious neuropsychiatric condition characterized by global encephalopathy with cognition and perception disturbance, which occurs in more than 50% of patients admitted to the intensive care unit (ICU) [1-3]. Delirium is not a primary disorder, but a result of severe physical conditions eISSN 1976-3026 OPEN ACCESS such as trauma, infection, toxin, metabolic dysfunction, postoperation condition, and more [1,2]. Symptoms of delirium include disorientation, impaired attention, sleep-wake cycle disturbance, psychomotor change, speech or mood change, and psychosis [1]. Delirium is associated with longer lengths of hospitalization and higher rates of morbidity and mortality.
Following discharge, adult patients with a delirium diagnosis during the hospitalization were at risk of cognitive impairment, depression, anxiety, posttraumatic stress disorder (PTSD), and sleep disorders [2,3]. Although delirium is known as a temporary phenomenon, neuropsychiatric symptoms following the acute phase of delirium could become a year-long condition [3]. The prolonged course following delirium is documented, called as post-intensive care syndrome. The burden of post-delirium sequelae further underscores the importance of preventing and mitigating the condition’s effects.
Nevertheless, little remains known about post-delirium prognosis in children. Within a short-term post-discharge follow-up period between 1 to 3 months, children with delirium experienced a lower quality of life with an elevated risk of sleep disturbance and overall mortality [4,5]. However, the presence of long-term consequences from pediatric delirium remained controversial as studies presented mixed results. Some studies indicated that pediatric delirium is associated with a greater risk for lasting ailments, including cognitive impairment and psychiatric illness [2,3]. In contrast, others revealed no difference in global cognitive function and behavior [6].
In addition to the lack of investigation into the overall pediatric population, existing literary on pediatric delirium has traditionally focused on delirium in very young children, primarily due to the relatively high prevalence and severity of the condition in children under 5 years of age [7]. However, few studies have investigated children between 6 to 12 years old. The comparison of patient prognosis based on age group is an area that requires further exploration, as vast differences exist between the brains of infants and those of pre-adolescent children. Due to the limited information, we conducted a large sample size retrospective cohort study to investigate the long-term neuropsychiatric outcomes of delirium in individuals aged 12 years and below. We selected 6 years as a cutoff point to distinguish early childhood from later developmental stages, given the substantial changes in prefrontal cortex development, functional connectivity, and the onset reliability of psychiatric diagnoses such as attention-deficit/hyperactivity disorder (ADHD) and intellectual disability around this age [8,9].
METHODS
This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. The patient consents are waived because we did not utilize identifiable patient information in the analysis. This study has been deemed exempt from Carilion Clinic Institutional Review Board approval and is defined as a non-human subject study (Carilion Clinic IRB-23-1827). No identifiable participant was involved in this project. Thereby we did not obtain consent/assent.
Study design and data sources
This study employed a retrospective cohort design, utilizing patient information from the TriNetX Research Network from May 2013 to May 2023. The TriNetX platform is a dynamic online database based on electronic healthcare records (EHRs) from prominent healthcare organizations (HCOs). At the time of data collection, 99% of the patients included in this study were in the United States, with records obtained from 80 HCOs constituting a total of 113 million patients. These HCOs, including satellite hospitals, outpatient clinics, and affiliated academic medical centers, participate in the TriNetX platform and synchronize their de-identified EHRs approximately every 2 to 4 weeks. Accessible information encompasses demographic details, diagnoses, lab data, procedure codes, medications, visit types, and genomic information.
Cohort definitions
We defined our study cohort as children aged 0 to 12 with at least one emergency visit or inpatient visit as the index event who were diagnosed with delirium (ICD-10 Clinical Modification [ICD-10-CM]: F05) within 30 days after the index event. The delirium in the context of emergency/inpatient visits in the study cohort must be the first instance. The control cohort comprised patients with no delirium diagnosis during the study period (i.e., between May 2013 and May 2023). We excluded patients with neuropsychiatric disorder diagnoses prior to the index event. We excluded patients newly diagnosed with a mental illness in the first year of the study period to ensure the outcome is not a result of a pre-existed underlying condition and patients newly diagnosed with delirium in the last 1 year of the study period to provide sufficient time for the outcomes. The follow-up period started one day after the index event until the presence of outcome events, death, or patient lost follow-up.
Outcome measurement
The primary outcome was an incident of any psychiatric disorder (ICD-10-CM: F00-F99) documented in the EHR system with an ICD code. Secondary outcomes were the specific psychiatric disorders, including mood disorders (ICD-10-CM: F30-F39, F93.8, F93.9), psychotic disorders (ICD-10-CM: F20-F29), anxiety disorders (ICD-10-CM: F40-F41, F43-F45, F48.1.0, F93-F94), PTSD (ICD-10-CM: F43.1), somatoform and related disorders (ICD-10-CM: F44-F45), trauma and stress-related conditions (ICD-10-CM: F43), pervasive disorders and obsessive-compulsive disorders (OCD) (ICD-10-CM: F42, F84, F95), substance use disorders (ICD-10-CM: F10-F19), intellectual disability (ICD-10-CM: F70-F79, F80-F82), ADHD and related disorders (ICD-10-CM: F34.81, F63, F90-F91), depressive disorder (ICD-10-CM: F32, F33, F34.1), and bipolar disorders (F30, F31, F34.0). In this study, depressive disorders and bipolar disorders were analyzed both as components of the broader category of mood disorders (ICD-10: F30-F39) and as separate diagnostic categories to facilitate clinical interpretability. This dual classification allowed us to examine whether specific mood disorder subtypes carried distinct risk profiles following delirium, while still respecting the hierarchical structure of psychiatric diagnoses. Although depressive disorder is indeed a subset of mood disorders, we chose to report its hazard ratio (HR) independently because of its high clinical prevalence and distinct developmental trajectory in pediatric populations.
Statistical analysis
The data source was the TriNetX Analytics Network. The study cohort comprised patients from 51 HCOs and the control cohort of 68 HCOs. The data were loaded into LUCID, a data bricks hosted computational platform associated with the TriNetX database and prepared for analysis. Patients in the study cohort were propensity-matched to patients in the control cohort at a one-to-four ratio by sex, age at index event, and diagnosis of major organ systems. These variables were selected based on their clinical relevance and known associations with both the development of delirium and psychiatric disorders. Age and sex are key developmental and diagnostic factors in pediatric psychiatry. Race and ethnicity were included to address potential disparities in access to care, diagnostic practices, and comorbidity profiles. Diagnoses across major organ systems were used as a proxy for medical complexity, which may increase the likelihood of both experiencing delirium and developing neuropsychiatric outcomes. In the multivariable regression model, we further adjusted for ICU admission as an indicator of illness severity and age group to reflect developmental stage, both of which have established relationships with long-term cognitive and emotional outcomes in children.
We performed multivariable Cox regression to obtain HRs with 95% confidence intervals (CIs) with adjustment for age group and critical care status as covariates. We generated Kaplan– Meier event-free curves with log-rank tests to compare the two groups.
We performed a subgroup analysis to investigate whether the subsequent neuropsychiatric outcomes following a delirium diagnosis were affected by the patient’s age and underlying disease severity. We estimated the incidence separately in patients equal to or under 6 years of age, greater than 6 years of age, patients with ICU admission, and patients without ICU admission within 30 days of the delirium diagnosis. A twotailed p-value<0.05 was considered statistically significant in all analyses. Statistical analysis was performed using R4.2.2 (https://cran.r-project.org/). We conducted a sensitivity analysis using E-values to evaluate the potential impact of unmeasured confounding (Supplementary Table 1).
RESULTS
We identified and matched 618 patients in the study cohort and 2,472 patients in the control cohort. The demographic data and comorbidities are summarized in Table 1. Demographic data before matching was displayed in Supplementary Table 2. During the 5-year follow-up period, we found 359 patients in the study cohort and 200 patients in the control cohort with at least one newly diagnosed neuropsychiatric disorder. Patients in the study cohort were more likely to be diagnosed with a neuropsychiatric disorder (HR=2.15; 95% CI, 1.82–2.56; p<0.001). Kaplan–Meier survival curves demonstrated a similar result (Figure 1). With a 5-year freedom from overall neuropsychiatric disorders of 68.5% (95% CI, 65.5–71.8) in the study cohort, compared to 49.8% (95% CI, 44.1–56.2) in the control cohort at the end of the follow-up period.
Baseline characteristics for the study and control cohorts after matching with propensity score
Kaplan–Meier survival curves for mental health outcomes. OCD, obsessive-compulsive disorder; ADHD, attention-deficit/hyperactivity disorder.
Among all the neuropsychiatric outcomes, patients in the study cohort were at a higher risk of receiving the following diagnosis: mood disorders (HR=1.82; 95% CI, 1.07–3.09; p=0.026), anxiety disorders (HR=1.69; 95% CI, 1.20–2.39; p=0.003), intellectual disability (HR=1.56; 95% CI, 1.19–2.05; p<0.001), depressive disorders (HR=1.74; 95% CI, 1.00–3.02; p=0.049), and bipolar disorders (HR=10.00; 95% CI, 1.15– 98.62; p=0.037) (Figure 2). Test for proportional hazard suggests valid results for overall psychiatric disorders (p<0.001), mood disorders (p=0.012), trauma and stress-related conditions (p=0.024), substance use disorders (p=0.005), and depressive disorder (p=0.035).
Forest plot of the outcomes hazard ratio of the primary outcomes. *indicates statistical significance (95% CI does not include 1). OCD, obsessive-compulsive disorder; ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval.
Subgroup analysis by age showed that children over 6 years old were at a higher risk of mood disorder (HR=1.96; 95% CI, 1.14–3.36; p=0.02), anxiety disorder (HR=2.05; 95% CI, 1.37–3.06; p<0.001), trauma and stress-related conditions (HR=1.87; 95% CI, 1.04–3.38; p=0.04), and depressive disorders (HR=1.79; 95% CI, 1.02–3.14; p=0.04). On the other hand, children under or equal to 6 years old were at a higher risk of substance use disorder (HR=4.34; 95% CI, 2.00–9.44; p<0.001), intellectual disability (HR=1.71; 95% CI, 1.27–2.30; p<0.001), and ADHD and related disorders (HR=2.25; 95% CI, 1.16–4.34; p=0.02). Children in both age groups were at a higher risk of overall neuropsychiatric disorders (age >6 years: HR=2.29, 95% CI, 1.74–3.02, p<0.001; age ≤6 years: HR=2.29, 95% CI, 1.86–2.83, p<0.001) (Figure 3A).
Forest plot of the outcomes hazard ratio by subgroup analysis. *indicates statistical significance (95% CI does not include 1). OCD, obsessive-compulsive disorder; ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval.
Subgroup analysis by the presence of critical care showed that children with a history of critical care were at a higher risk of anxiety disorders. On the other hand, children without a history of critical care were at a higher risk of mood disorders, anxiety disorders, trauma and stress-related conditions, pervasive disorders and OCD, substance use disorders, intellectual disability, ADHD and related disorders, and bipolar disorder (Figure 3B).
DISCUSSION
To date, few studies have examined the long-term neuropsychiatric outcomes of pediatric delirium. During the 11-year follow-up period, we found that children with delirium were associated with a 2.15-fold increased risk of neuropsychiatric disorder compared with control without delirium. The risk was highest for mood disorders, including depressive disorders and bipolar disorders, followed by anxiety and intellectual disability. Notably, the subgroup analyses suggest age as a critical factor regarding neuropsychiatric outcomes. Children over 6 years old were at a higher risk of internaliz-ing disorders in which psychosocial factors play a major role, including mood disorders, anxiety disorders, trauma-related conditions, and depressive disorders. In contrast, children aged 6 and under were at risk of externalizing disorders characterized by impulsivity and maladaptive behaviors, including substance use disorders and ADHD and related disor-ders. Children aged 6 and under were also more likely to be diagnosed with a disorder with strong endogenous origin such as intellectual disability. It is noticed that children aged 6 and under were more prone to developmental disorder diagnoses as well. In brief, although delirium is a transient central nerve dysfunction, our study underscores the importance of long-term neuropsychiatric sequelae after pediatric delirium. However, we acknowledge that diagnoses such as ADHD and intellectual disability in early childhood must be interpreted with caution. These conditions may not be recognized until later developmental stages, and diagnostic practices vary widely. It is possible that underlying central nervous system vulnerabilities were present before the delirium episode but only became clinically apparent afterward. Therefore, the temporal association observed in our study does not imply causation and may reflect the delayed manifestation of pre-existing conditions. Future studies using longitudinal developmental data or prospective neurocognitive assessments may help clarify these possibilities.
Mood disorders, trauma-related conditions, and anxiety disorders were more prevalent among children older than 6 years in our study. According to the neuroinflammatory hypothesis of delirium, delirium suggests neuronal and synaptic dysfunction in the central nervous system with an inflammatory status [10]. Inflammatory conditions can lead to depressive symptoms [11,12]. Research has demonstrated elevated levels of inflammatory biomarkers in children exhibiting depressive symptoms [13]. The neuroinflammatory changes lead to not only temporary manifestations of delirium but also long-term neuropsychiatric consequences. We assume that children older than 6 exhibit a lower susceptibility to complications with longterm cognitive consequences because their brains are more developed than those of their younger counterparts. This finding aligns more closely with delirium studies conducted in adult populations. Studies in adults have similarly demonstrated long-term cognitive and psychiatric sequelae following delirium, providing a reference point for understanding pediatric trajectories [1].
Recent studies have shown that pediatric delirium is related to neurocognitive impairment in attention, memory, and executive function [14]. In our study, children with a delirium diagnosis were at a higher risk of a subsequent intellectual disability diagnosis, which is consistent with the previous finding. We assume the correlation could be deemed as the consequence of brain injury in early childhood. The human brain develops rapidly and dynamically in the first 2 years of life, and the cortical surface area continues to expand during preschool and early school-age years [14,15]. Brain imaging studies further suggest significant signal intensity differences before and after 3 years of age, while regional and functional development continues, and cortical thickness changes could continue until 20 years of age [16,17]. Physiological distress during the critical developmental period could have detrimental effects on the brain. The presence of delirium could represent such a type of injury. These age-specific nuances highlight the need for tailored monitoring strategies while following up with children recovered from delirium. On the other hand, preexisting cognitive impairment or developmental delay are known to be risk factors for pediatric delirium [2]. Thereby, the casualty of intellectual disability and delirium is still deferred for further investigation. Beyond cognitive delay, broader central nervous system vulnerabilities may also predispose children to both delirium and later psychiatric disorders. Additional factors including neurotransmitter imbalances such as dopamine dysregulation, disruptions in the sleep-wake cycle, and psychological trauma related to illness may also contribute to the development of psychiatric disorders after delirium [18-22].
Our study results suggested a potential correlation between delirium and an increased risk of substance use disorder diagnosis later in life. Subgroup analysis showed that the correlation is only significant in children under 6 years and children without a history of ICU admission, along with a delirium diagnosis. To the best of our knowledge, there is no existing literature supporting a direct correlation between pediatric delirium and substance use disorders in very young children. However, substance use disorders are known to be related to a variety of neuropsychiatric disorders, including mood disorders, intellectual disability, ADHD, and more [16,17,21]. We believe the substance use disorder diagnosis might be a result secondary to a neuropsychiatric condition followed by delirium instead of a consequence of delirium. On the other hand, we could also consider substance use disorders as a result of biological vulnerability, evident by shared genetic liability regardless the drug of choice [23]. Genetic factors contribute to dopamine dysregulation which put individuals with an increased risk of substance use disorders [23]. Similarly, brain injury resulting from delirium can also lead to dopamine dysregulation, thereby increasing susceptibility to substance use disorders [20].
However, on the other hand, the unexpected finding of the elevated risk of substance use disorder diagnoses among children aged 6 years or younger may not be an actual prognosis in our study population. From a clinical standpoint, the plausibility of a genuine substance use disorder diagnosis in this age group is extremely low, as DSM-based diagnostic criteria cannot reliably be met in early childhood. These cases may instead reflect several possibilities: 1) diagnostic or coding errors within electronic health records; 2) misclassification of related clinical scenarios such as accidental ingestion, adverse medication reactions, or severe behavioral dysregulation under substance use disorder codes; or 3) the use of these diagnostic codes as proxies for significant environmental or psychosocial adversity, such as maltreatment or parental substance use. Therefore, rather than representing genuine early-onset substance use disorder, this association may serve as a marker of heightened psychosocial vulnerability or as a limitation of EHR-based research. Clinicians and researchers should interpret these findings with caution, and future studies incorporating prospective developmental assessments or chart reviews are needed to clarify this observation.
The study also highlights the association between critical care history and subsequent neuropsychiatric outcomes of pediatric delirium. We found that children with a history of critical care were at a heightened risk of anxiety disorders. In contrast, those without such a history faced increased risks across a broader spectrum of conditions, including mood disorders, trauma-related conditions, substance use disorders, and more. ICU admission is related to long-term neuropsychiatric consequences in pediatric populations, including PTSD, lower IQ scores, and general emotional and behavioral issues [3,22,24]. Our results did not align with previous studies. We believe the effect of mortality rate on the cohort study design should be taken into consideration. Children admitted to the ICUs were more likely to have a life-threatening medical condition with a mortality rate of 20% [25]. The high mortality rate in ICU patients could lead to shorter follow-up periods, thereby leading to the long-term consequences being underestimated in our study. Alternatively, there is still room for discussion, as most existing studies have not considered the impact of illness severity, sedation strategies, and common ICU interventions [3]. Furthermore, disease severity may act as a confounding factor in the relationship between delirium and psychiatric outcomes, as children with more complex or severe medical presentations may inherently be at higher risk for both conditions. This limitation is intrinsic to observational cohort designs despite matching procedures [26].
Limitations
We were unable to identify the underlying medical condition causing delirium, as we could only assess patient information through the TriNetX EHR. Considering the nature of retrospective cohort study analysis based on HER record, the causality of delirium and the subsequent neuropsychiatric diagnosis was undetermined. Due to the above, this approach introduces the possibility of information bias, particularly misclassification of neuropsychiatric outcomes and underdiagnosis of delirium. Psychiatric diagnoses in children may be inconsistently documented across health systems or delayed due to variable access to mental health services [27,28]. Similarly, pediatric delirium remains under-recognized, especially in non-ICU settings, due to limited screening implementation and clinical awareness [2].
Selection bias is another concern in retrospective EHR-based studies, particularly regarding which children are evaluated and documented for delirium. Although our propensity score matching accounted for demographic and comorbidity profiles, we could not adjust for unmeasured variables such as illness severity scoring or social determinants of health. To mitigate this as much as possible, we perform subgroup analysis in ICU admission status in the model and exclude patients diagnosed with delirium in the last year of the study period to allow for adequate follow-up.
A critical limitation of our study concerns the increased risk of substance use disorder diagnoses among children aged 6 years or younger. Given the developmental implausibility of an actual substance use disorder diagnosis in this age group, this finding most likely reflects diagnostic misclassification, variable coding practices across institutions, or underlying family-level and psychosocial risk factors rather than genuine clinical cases. Because TriNetX relies on administrative data, we were unable to validate individual cases through chart review or to differentiate between coding errors and early psychosocial risk indicators. This limitation reflects the challenges of relying solely on retrospective EHR data to capture complex developmental phenomena and emphasizes the importance of cautious interpretation. Future research should validate such findings with prospective longitudinal data or detailed clinical assessments.
Last, we were not able to obtain information on unquantifiable risk factors of neuropsychiatric disorders, such as adverse childhood experiences and family history. Despite the diverse population in the US in terms of race and ethnicity, whether the results can be generalized globally is in question.
Conclusions
In summary, this study highlights the correlation between pediatric delirium and long-term neuropsychiatric outcomes. Recognizing the potential sequelae of delirium is paramount for optimizing the mental health trajectories of children. Clinicians should provide information to parents regarding potential long-term neuropsychiatric outcomes in children with delirium and closely monitor those children after recovery.
Supplementary Materials
The Supplement is available with this article at https://doi.org/10.30773/pi.2025.0034.
Sensitivity analysis
Demographic data of subgroup analysis
Notes
Availability of Data and Material
Researchers obtained data directly from TriNetX platform. Due to the contract with TriNetX, we cannot make the dataset publicly available.
Conflicts of Interest
The authors have no conflict of interest to report. One of the lead authors (CF Sun) was an American Psychiatric Association fellow in the year of publication. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the American Psychiatric Association or American Psychiatric Association Foundation.
Author Contributions
Conceptualization: Ching-Fang Sun, Chih-Sung Liang. Formal analysis: Yingxing Wu. Investigation: Ching-Fang Sun, Chih-Sung Liang, Yingxing Wu. Methodology: Ching-Fang Sun, Chih-Sung Liang, Yingxing Wu. Project administration: Ching-Fang Sun. Resources: Anita S. Kablinger. Supervision: Anita S. Kablinger. Writing—original draft: Ching-Fang Sun, Rebecca E. Mardis. Writing—review & editing: Kiran Khalid, Chih-Sung Liang, Anita S. Kablinger.
Funding Statement
None
Acknowledgments
None