Differences in delirium evaluation and pharmacologic management in children with developmental delay: A retrospective case-control study

Abstract

Delirium is associated with increased mortality and cost, decreased neurocognition and decreased quality of life in the pediatric ICU population. The Cornell Assessment for Pediatric Delirium (CAPD) is used in pediatric ICUs for delirium screening but lacks specificity in children with developmental delay (DD). Within a cohort of children receiving pharmacologic treatment for ICU delirium, we compared delirium scoring and medication management between children with and without DD. We hypothesized that CAPD scores and treatment decisions would differ between DD and neurotypical (NT) patients.

In this retrospective case-control study, we queried the medical record for patients admitted to our PICU with respiratory failure June 2018 to March 2022 who received antipsychotics typically used for ICU delirium. antipsychotics prescribed for home use were excluded. Non-parametric statistics compared demographics, CAPD scores, medication choice, dosing (mg/kg), and medication continuation after discharge between those with and without DD based on ICD-10 codes.

Twenty-one DD admissions and 59 NT admissions were included. Groups did not significantly differ by demographics, LOS, drug, or initial dosage. DD patients had higher median CAPD scores at admission (17 v.13; p=0.02) and treatment initiation (18v.16.5; p=0.05). Providers more frequently escalated doses in DD patients (13/21v.21/59; p=0.04) and discharged them home on new antipsychotics (7/21v.5/59; p=0.01).

DD patients experience delirium screening and management differently than NT counterparts. Providers should be aware of baseline elevated scores in DD patients and carefully attend to indications for dosage escalation. Further work is needed to understand if prolonged duration, even after hospital discharge, benefits patients or represents potential disparity in care.

Delirium, defined as acute alterations in consciousness and cognition caused by systemic illness or treatment effect, occurs in over 20% of all pediatric critical care patients^1,2. Delirium in children is associated with increased mortality, prolonged length of intensive care unit (ICU) stay, increased costs, and declines in neurocognitive function and reduced quality of life following discharge. There are four recommended, validated, and reliable delirium monitoring tools in pediatric ICU patients: Sophia Observation withdrawal Symptoms-Pediatric Delirium scale (SOS-PD), Pediatric Confusion Assessment Method for the Intensive Care Unit (pCAM-ICU), Preschool Confusion Assessment Method for the Intensive Care Unit (psCAM-ICU), and Cornell Assessment for Pediatric Delirium (CAPD)^6,7. CAPD is increasingly used in pediatric ICUs to screen for delirium; it is based on observations of a child’s ability to interact with their environment. A CAPD of ≥9 at any timepoint is the threshold for a positive screen⁸. Although highly sensitive and specific for identification of delirium in the general pediatric ICU population, the CAPD lacks specificity for children with developmental delay (DD), as the CAPD assumes prior neurotypical communication and behavior. In the scoring system, the CAPD is unable to account for a baseline impaired ability to interaction with one’s environment, as may be seen in children with static encephalopathy or motor impairment rendering them unable to make purposeful movements, thus falsely elevating their scores for acute alterations⁹. Yet, DD is a significant risk factor for delirium, as children with DD are diagnosed with delirium almost three times as often as children without delay when evaluated using psychiatric assessment taking into account baseline behaviors⁸. As such, children with DD are at increased risk for both underrecognized delirium, if scores are assumed to elevated at baseline and not monitored for changes, and overtreatment, if the strict threshold is applied for treatment consideration, with subsequent risk of disparate care.

Both unrecognized delirium and overtreatment of delirium pose significant risks to children with DD. Unrecognized delirium increases risk of over-sedation and post-traumatic symptoms, prolongs length of stay, and contributes to decline in neurocognitive function^8,10. However, delirium overtreatment increases polypharmacy, risk of drug-drug interactions, and risk for adverse drug reactions, and may also increase the risk of inappropriately attributing symptoms to delirium leading to missed diagnoses.

Given the importance of recognizing and treating delirium, and the lack of specificity for scoring CAPD in patients with DD, we aimed to characterize delirium scoring and treatment in patients with DD compared to those without DD in our ICU. To assess delirium recognition and management, we investigated treatment patterns for delirium in our pediatric ICU amongst neurotypical (NT) children and those with DD. We utilized pharmacologic treatment as a surrogate marker for a clinical diagnosis of delirium and severity of symptoms. We hypothesized patients with DD would have higher initial baseline CAPD scores than their NT counterparts but similar adjusted scores at onset of delirium treatment, and may show treatment differences in drug choice, dosage, and duration.

Methods:

This retrospective single-site case-control study took place in a quaternary care pediatric ICU at an urban academic medical center. All children admitted between June 2018 and March 2022 with a diagnosis of acute respiratory failure (inclusive of acute respiratory failure, acute hypoxic and/or hypercarbic respiratory failure, and acute on chronic respiratory failure) and who received any of four antipsychotics traditionally used for ICU delirium treatment (olanzapine, haloperidol, risperidone, or quetiapine) were included in the initial data extraction. We limited the diagnosis to acute respiratory failure to minimize confounding risk factors of direct neurologic injury or involvement that may influence initial CAPD score in either group. Patients who had antipsychotics listed in home medications prior to admission were excluded. The Washington University in St. Louis Institutional Review Board (IRB) approved this minimal risk study (IRB# 202205134) with a waiver of informed consent.

We queried our institutional electronic medical record to identify the population of interest. Manual chart review confirmed inclusion criteria that all patients received antipsychotics expressly for the treatment of documented delirium and were not on antipsychotics in the outpatient setting. Delirium diagnosis was confirmed with manual chart review of provider notes and/or problem list denotation. Patients with repeat admissions were included in the study as a distinct encounter so long as the patient was not on an antipsychotic upon readmission. Demographic data including age, sex, weight, need for mechanical ventilation, primary diagnosis, and mortality were collected.

Patients were separated into cases (DD) and controls (NT) based on ICD-10 and SNOMED CT codes relating to DD, cognitive impairment, or intellectual disability (Appendix 1). Patient problem lists were also cross-referenced to assure no diagnoses of developmental delay were missed. Data including length of ICU and hospital stay, initial CAPD score, subsequent CAPD scores, CAPD at diagnosis of delirium, CAPD at initiation of pharmacologic delirium treatment, medication choice, dose, and duration were collected.

Non-parametric statistics compared demographics, length of stay, diagnosis codes, initial and subsequent CAPD scores, medication choice, dose, and duration between those with and without DD. Demographic and clinical characteristics of patients were described as n (%) or mean (SD) and median (IQR). Chi-square/Fisher exact tests or Wilcoxon tests were used as appropriate to compare cases and controls with regards to CAPD scores, drug dosages, and continuation of antipsychotic at discharge. Analyses were performed in Prism9 version 9.4.1 (GraphPad Software, LLC). A STROBE checklist was utilized per the Equator network guidelines (Appendix 2).

Results:

Electronic medical record query resulted 144 discrete admissions for respiratory failure who received antipsychotics (Appendix 3). Sixty-four admissions were excluded: 38 received antipsychotics as a home medication, 24 received antipsychotics for indications other than delirium, and two were excluded as the medication was ordered, but never given. Eighty discrete admissions (76 patients) were included in the study, 20 patients with DD (21 discrete admissions) and 56 patients with NT development (59 discrete admissions). There was no significant difference in readmission between the two groups (4.8% vs 5.1%, p=0.99). Demographic information is shown in Table 1. Patients in each group did not significantly differ by age, sex, weight, ICU length of stay, or hospital length of stay. Of note, 9 of the control group had co-existing mental health diagnoses including major depressive disorder (MDD), generalized anxiety disorder, bipolar disorder, adjustment disorder, and attention deficit hyperactivity disorder (ADHD). In the group with DD, 1 patient carried a diagnosis of MDD and 1 ADHD.

Table 1.

Demographics of overall cohort and patients with DD and NT.

Characteristic	Overall cohort (n=80), n (%) mean (SD)	DD (n=21), n (%) mean (SD)	NT (n=59), n (%) mean (SD)	p
Sex
Male	46 (57.5)	9 (42.9)	37 (62.7)	0.11
Female	34 (42.5)	12 (57.1)	22 (37.3)
Age at admission (yr)	8y6m (6y1m)	7y3m (5y11m)	8y11m (6y2m)	0.29
Dosing weight at admission (kg)	36.1 (27.8)	34.6 (33.6)	36.6 (25.7)	0.78
ICU LOS (days)	30.9 (55.8)	41.9 (66.7)	35.88 (45.7)	0.71
Hospital LOS	49.99 (58.7)	48.9 (65.5)	50.3 (56.7)	0.92

DD (developmental delay), NT (neurotypical)

In comparing first recorded CAPD scores between the two groups, Median time from ICU admission to first CAPD was 26.3 hours for the overall cohort and without significant difference between the two groups (p=0.92). Patients with DD had significantly higher initial CAPD scores (Figure 1 (A)): DD median 17 [14–19], NT median 13 [6–17], p=0.02. Patients with DD had only slightly higher CAPD scores at initiation of pharmacologic treatment for delirium (Figure 1 (B)), with median score of 18 [17–21] than their NT counterparts (median 16.5 [12–19], (p=0.047). Time from first CAPD score to initiation of therapy did not differ significantly between the groups, with a median of 5.8 days [1.3–10.8 days] in the DD group and 4.3 days [2.3–11.0 days] in the NT group (p=0.97). Although patients with DD had a smaller change between their initial CAPD and their CAPD at initiation of treatment, (DD median 3, [1–8], NT median 5, [3–11], p=0.06), this did not reach statistical significance in this small study (Figure 1 (C)).

Figure 1.

CAPD score comparisons between patients with developmental delay and those with neurotypical development. (A) Intial CAPD scores; (B) CAPD at treatment initiation; (C) Difference between initial CAPD and treatment CAPD.

Specific drug choice did not differ significantly between the two groups, with providers prescribing quetiapine most frequently in both groups (57.14% in DD and 69.49% in NT, respectively, p=0.55). The second most prevalent in both groups was risperidone. Three children received haloperidol (2NT, 1DD) and olanzapine once overall (1 NT), (p=0.49). Patients in both groups had similar dosing weights on admission: 34.58 kg (SD 33.61) and 36.60 kg (SD 25.71), for DD and NT (p=0.78). Both groups had similar initial medication dosages per kilogram of bodyweight (p=0.11). Median initial quetiapine dosage in both groups was 12.5 mg. In the NT cohort, initial quetiapine dosage ranged from 3mg to 25mg, with maximum dosage of 50mg. In the DD cohort, initial quetiapine dosage ranged from 6.25mg to 25mg, with maximum dosage of 50mg. Median initial risperidone dosage in the NT cohort was 0.1mg, ranging from 0.05mg to 1mg, with maximum dosage of 2mg. In the DD cohort, median initial risperidone dosage was 0.18mg, ranging from 0.1mg to 0.5mg, with a maximum dosage of 1mg. Haloperidol dosage in the NT cohort (n=2) ranged from 0.4mg to 2.2mg with no escalation. In the DD cohort, one patient received haloperidol at a starting dose of 2mg escalated to 3mg maximum. Despite similar initial doses, as seen in Figure 2 (A), patients with DD were more likely to have their drug dosages escalated (61.9% vs 35.6%, p=0.04). Similarly, patients with DD were also more likely to discharge home on an antipsychotic (33.0% vs 8.5%, p=0.01, Figure 2 (B)).

Figure 2.

Comparisons between patients with developmental delay and those with neurotypical development who (A) had their drug dosages escalated and (B) went home on newly prescribed antipsychotic.

Discussion:

In this single-center retrospective study of children with acute respiratory failure receiving pharmacologic treatment for delirium in the intensive care unit, we found that children with DD were likely to have higher CAPD [or delirium screening] scores triggering pharmacologic therapy and more escalation of medications after initiation compared to their neurotypical peers.

Providers using antipsychotic to treat delirium prescribed quetiapine most frequently in our cohort, with equal use between the two groups. The first study demonstrating use of quetiapine for delirium in pediatric patients was a case report published in 2013 by Traube et al describing four patients who received quetiapine offering it as a possible therapeutic option in critically-ill children with delirium¹¹. In 2015, Joyce et al offered a review demonstrating its safety profile, specifically noting a lack of dysrhythmias, extrapyramidal symptoms, and neuroleptic malignant syndrome¹². However, both studies noted the need for prospective, randomized trials of quetiapine for treatment of pediatric critical care delirium. Quetiapine use in pediatric delirium patients quickly rose, and our study is consistent a recent report of quetiapine being the favored choice for treatment of delirium in critically-ill pediatric patients¹³. However, despite increased use and assumed increased comfort with quetiapine’s safety profile and efficacy, it has only been shown to decrease opioid and benzodiazepine requirements in pediatric patients with congenital heart disease¹⁴. It has not been shown to improve delirium screening scores when compared to untreated, matched patients¹⁵. As such, despite the increased use of quetiapine, as described by recent publications and consistent with our findings, randomized trials of quetiapine usage for pharmacologic delirium continue to be indicated to demonstrate efficacy and safety.

Despite quetiapine being the preferred drug in both groups, we found how providers managed antipsychotic drugs differed between the two groups. Providers initiated any antipsychotic at a higher CAPD score in children with DD compared to NT, with treatment differences persisting after ICU and hospital discharge. The change in CAPD score did not differ significantly between the two groups, so this may reflect providers accounting for the elevated baseline CAPD. The elevated baseline CAPD scores in the DD group is consistent with Traube et al’s findings that CAPD has decreased specificity in children with developmental delay. Traube’s study defined “significant clinical developmental delay” based on clinical assessment and/or parental report of developmental problems impairing ability to communicate⁸. Though we used ICD-10 codes for developmental delay or cognitive impairment, this would likely encompass a very similar patient population, though we may have missed some patients with parental concerns not delineated in the medical record. To counteract this decreased specificity, newer data recommends combining Richmond Agitation-Sedation Scale (RASS) score with CAPD in patients with DD to attain both sensitivity and specificity in this at-risk, but challenging to assess population⁹.

Consistent with the higher baseline scores, providers initiated pharmacologic treatment for delirium at higher CAPD scores for patients with DD. Both groups’ median score at initiation of pharmacologic treatment was significantly higher than the recommended treatment score of 9. It is unknown if delirium is being recognized at lower CAPD scores and non-pharmacologic methods (e.g., strict day/night lighting) or other pharmacologic methods (e.g., cycling of dexmedetomidine) are implemented prior to antipsychotics, if delirium is recognized but untreated, or if more education is needed about interpreting CAPD scores. Further, as CAPD can be falsely elevated in the setting of pain, trauma, and anxiety, it is possible that with these concerns, providers waited longer or desired more time and data to determine a delirium diagnosis. However, as late onset delirium may be more refractory to pharmacologic intervention than early-onset delirium, it is critical to appropriately identify delirium and initiate treatment as indicated¹⁶.

In those who were identified and treated for delirium, providers were significantly more likely to escalate antipsychotic dosage and prescribe antipsychotics as a new home medication for DD patients. Polypharmacy is a significant challenge in pediatric patients with critical illness, and even more so in patients with DD, who are already often on more medications than NT patients¹⁷. Polypharmacy leads to increased adverse drug reactions, drug-drug interactions, increased patient and institutional costs, and mortality¹⁸. Further, polypharmacy is associated with decreased adherence at home^18,19. Providers should give careful attention to escalating drug dosages in the DD population and ensure doses are increased for intended indications. Throughout the hospital course and at discharge, providers should evaluate whether there is a true need for continuing antipsychotics to avoid worsening polypharmacy in this already at-risk population. Follow-up providers should be made aware of new antipsychotics, and the delirium-specific indication, so these medications are not continued indefinitely.

It is currently unknown if delirium superimposed on an incompletely reached neurocognitive state in DD may portend similar poorer outcomes as adults with dementia or cognitive impairment; it is also unknown if pediatric delirium is associated with increased readmission as shown in the adult delirium population. Like the CAPD’s poor specificity in pediatric DD patients, adult data proves identifying delirium in the presence of dementia to be challenging. Studies show delirium superimposed on dementia is associated with poorer clinical outcomes and delirium imposed on cognitive impairment is associated with higher incidence of postoperative complications, and longer hospital LOS^20,22,23.

Patients with DD are already at increased risk of disparate care. Children with special needs are significantly less likely to receive care in a well-organized system, face more barriers accessing care, and experience disparities in quality of healthcare services; those with the most severe disabilities are the most impacted^24–27. Given patients with DD are already at risk of disparate care, our findings demonstrating possible overtreatment and extended treatment of delirium increase the risk of unequal care in this already vulnerable population. Providers should maintain a high index of suspicion for delirium evaluation and pharmacologic treatment in patients with DD to minimize disparities in care.

Our study had several limitations. First, our retrospective study was dependent on EMR accuracy and consistent provider documentation. Approximately a third of the cohort had their first CAPD documented after 48 hours of ICU admission, even though CAPD scoring is ordered every 12 hours in our unit on every patient. Yet, as both groups had equal amounts of missing data, this is not expected to change the overall findings. The non-ideal documentation in our study is likely not unique and reflects educational opportunities: It reinforces previous studies noting inadequate knowledge among providers about delirium screening and prevention and recommending increased standardization and education^13,28. Second, in our ICU, day-night dexmedetomidine cycling is often used as first-line pharmacologic treatment for delirium prior to antipsychotic use. As such, some patients who received non-antipsychotic pharmacologic intervention for delirium were likely excluded from the analysis. The lack of capturing alternative therapies in the data set may have contributed to higher CAPD scores at antipsychotic initiation but would not explain the difference in dose escalation between NT and DD. Our small sample size is also a limitation, and the study was likely underpowered to evaluate magnitude of delta CAPD between the two groups.

In conclusion, future studies investigating antipsychotic efficacy and safety in pediatric patients, CAPD scoring in all pediatric patients, and larger cohorts of patients with and without DD are needed. These studies will further enhance our screening, prevention, and treatment of delirium in the pediatric DD population and minimize risk of disparate care.

Supplementary Material

Appendix 1

ICD-10 codes

Appendix 2

STROBE EQUATOR checklist

Appendix 3

Inclusion/exclusion flow chart

Abbreviations

CAPD

Cornell Assessment for Pediatric Delirium

DD

developmental delay

EMR

electronic medical record

NT

neurotypical

PICU

pediatric intensive care unit