Delirium is a disturbance in attention and cognition from baseline that develops over a short period of time, tends to fluctuate in severity, and is not explained by another neurocognitive disorder (9). In neonates, identifying disturbances in cognition is challenging due to their changing cognitive baseline and high risk for other neurodevelopmental disturbances related to prematurity and complex comorbidities. It is unclear if premature infants have the level of brain development to experience the neurobiological processes thought to underly the pathogenesis of delirium (10). Despite these challenges, there are several factors to suggest that a pathology on the spectrum of delirium can occur in NICU patients.

The pathogenesis is not fully understood; several hypotheses have been proposed, all resulting in dysregulation of neuronal activity including neurotransmitter imbalance, oxidative stress with subsequent free radical-mediated injury, and neuroinflammation leading to elevated glucocorticoids and increased cytokine expression (1, 4). It is possible that there is increased susceptibility to delirium in neonates due to their developing nervous system and vulnerability to inflammatory conditions and oxygen toxicity; the lack of standardized diagnostic criteria and rigorously validated assessment tools for this unique population prevent accurate recognition.

Recognized delirium risk factors in adult and pediatric patients include prolonged hospitalization, baseline cognitive dysfunction, need for invasive mechanical ventilation, suboptimal pain or agitation management, and exposure to corticosteroids and neurosedative medications, all of which are pervasive in the NICU population (1, 3, 5).

Most cases of refractory agitation in the NICU are reported to occur at or around TEA, as seen in our patient population. By this age, patients should have a determined cognitive and behavioral baseline in which a change can be recognized. Although limited to case-reports and case-series they have demonstrated response to conventional delirium treatment with atypical antipsychotics (2, 6–8).

Based on the characteristics identified in our study population and existing literature, several risk factors for the development of refractory agitation were identified. These include extreme prematurity, ELBW status, severe bronchopulmonary dysplasia [BPD] with the need for prolonged mechanical ventilation, multiple steroid courses, pulmonary hypertension, recurrent infections, and feeding intolerance. Interestingly, severe intracranial hemorrhage or PVL did not appear to be a significant risk factor.

Pulmonary hypertension is the most significant risk factor. BPD is a major complication of prematurity with increasing incidence due to the improved survival of the most premature neonates. About 25% of infants with moderate to severe BPD will develop secondary pulmonary hypertension (11). Continuous infusions of multiple neurosedative medications are often required in the management of pulmonary hypertension (12). Even after resolution of PH crisis, weaning off these medications presents a challenge and can predispose infants to repeat PH crises. While the risk factors are largely non-modifiable, recognizing these features can help identify those patients at greatest risk allowing for earlier recognition and intervention.

One challenge in early recognition of refractory agitation is the lack of validated diagnostic tools. Available tools are based on subjective evaluation of behavioral responses and were primarily developed to assess acute responses to procedural pain or stress from handling. While there are several assessment tools available, our NICU uses the Neonatal Pain, Agitation and Sedation Scale (N-PASS) and Cornell Assessment of Pediatric Delirium (CAPD) scale to evaluate patients at increased risk for refractory agitation.

The N-PASS utilizes 5 domains: crying/irritability, behavior/state, facial expression, extremities/tone, and vital signs to assess pain/agitation and sedation with an assigned score between 0 and 10 for pain/agitation and 0 to −10 for sedation with additional points assigned based on corrected gestational age (13). It has been validated for evaluating pain and agitation in infants born as early as 23 weeks gestation, mechanically ventilated patients, and in cases of prolonged pain and agitation (13–15). The impact of neurologic abnormalities on validity and reliability of the N-PASS, however, have not been clearly determined (13). Additionally, it can be difficult to differentiate pain and agitation in neonates as they present with similar behavioral and physiologic manifestations. For these reasons, specific target N-PASS scores may need to be adjusted for individual patients based on their own baseline and interpreted within the context of their clinical status.

The CAPD is an observation tool which evaluates eye contact, purposeful actions, awareness of surroundings, communication, restlessness, inconsolability, activity level in awake states, and timing and response to interactions. It has been validated as a screening tool for delirium in the pediatric ICU, including infants and newborns (validation included only 25 infants, with 7 < 1 month of age) (16). Despite its increasing use, with an estimated 22% of TEA infants in the NICU screen positive for delirium, the CAPD has not been validated in the NICU setting (3). The impact of prematurity, neurodevelopmental impairment, and invasive ventilation on the predictive value of these scores is not well understood. Higher rates of false positive screens have been demonstrated in each of these groups (16, 17). While developmental anchor points can be used to improve accuracy of screening in children under 2, they are difficult to apply in the NICU given the atypical neurodevelopmental trajectories.

Kaur et al (18), recognizing the challenges of using CAPD scores in patients with developmental delay, demonstrated that a combined CAPD score >9 with a fluctuation in Richmond agitation sedation scale score of >2 was both sensitive and specific for detecting delirium in developmentally delayed PICU patients. Despite their individual limitations, using the N-PASS and CAPD scores together and considering each patient's individual cognitive and behavioral baseline can help improve the usefulness of these tools in identifying refractory agitation and response to treatment. It is important to note that these are screening tools and cannot be used to diagnose refractory agitation or delirium.

Management of pain and agitation remains a complex challenge in the NICU. The known negative neurodevelopmental outcomes of untreated pain and stress must be weighed against the detrimental effects of opioids and benzodiazepines and the relatively unknown long-term safety and efficacy of emerging therapies including alpha-2 receptor agonists and atypical antipsychotics (19, 20). These challenges are magnified in the extremely premature neonate who will likely require prolonged exposure to any number of these drugs.

Analgesics and sedatives are used frequently in the NICU to manage pain and agitation, promote comfort and safety (21). Administration of opioids and benzodiazepines has been associated with altered brain development and poor neurodevelopmental outcomes ⁽19, 22). Prolonged use (>7 days) of opioids and benzodiazepines has been associated with lower Bayley scales of infant development (BSID-III) cognitive, motor, and language scores for premature infants examined at 2 years (23). It is also associated with analgesic and sedative tolerance and physical dependence, necessitating escalating doses and/or addition of adjunctive agents to achieve adequate pain and sedation control (24).

Antipsychotics have become a therapy of interest for the management of refractory agitation and delirium. Antipsychotics regulate neurotransmitter levels essential in affecting mood and behavior. First-generation, or typical, antipsychotics (haloperidol, chlorpromazine) are dopamine receptor antagonists. Second-generation, or atypical, antipsychotics (quetiapine, risperidone) are dopamine and serotonin receptor antagonists. By blocking these receptors, antipsychotics decrease the rate of elimination of these neurotransmitters released by electrical stimulation of neurons (25). Although antipsychotics have not been FDA approved for pharmacological delirium treatment in pediatric patients and the PANDEM guidelines recommend that antipsychotics should not be routinely used to prevent or decrease the duration of delirium in a critically ill patient. PANDEM guidelines do recommend consideration of antipsychotics for the management of severe delirium manifestations with caution of possible adverse drug effects (26). Further, there is no data available on long term neurodevelopmental impacts of these medications.

Quetiapine is an oral atypical antipsychotic that has been demonstrated to be tolerated in pediatric ICU patients as young as 2 months of age (27, 28). Common adverse events associated with quetiapine include dyslipidemia, extrapyramidal symptoms, hyperglycemia, temperature dysregulation, sedation, weight gain and QT prolongation. Other less common adverse effects include neuroleptic malignant syndrome, hypothyroidism, and hematologic abnormalities. Anticholinergic effects including constipation, urinary retention, xerostomia, and blurred vision are also possible although quetiapine's anticholinergic activity is low in comparison to other second-generation antipsychotics. Quetiapine results in symptom improvement within 24–72 h after initiation (29, 30).

Risperidone is an atypical antipsychotic that is formulated in injectable (intramuscular, subcutaneous) and enteral formulations. It is commercially available in various tablet strengths as well as a 1 mg/ml oral solution, which allows for more precise tailoring and titration of pediatric dosing as well as easier administration in infant or pediatric patients unable to swallow tablets. Common side effects include somnolence, fatigue, hyperprolactinemia, weight gain, constipation, nausea, vomiting, dizziness, and extrapyramidal side effects (25, 31, 32). Given its atypical antipsychotic status, extrapyramidal side effects should theoretically be less than a typical antipsychotic agent. When compared with quetiapine, risperidone has been associated with more extrapyramidal symptoms (33). Data regarding dosing is limited in the infant and younger pediatric populations. Symptom improvement may be seen within 24 h of starting risperidone with resolution of delirium within 3.5–12 days (34).

Alpha-2 agonists, specifically dexmedetomidine and clonidine, provide analgesia, anxiolysis, and sedation without impacting respiratory drive or gastrointestinal motility (19). Pre-synaptic activation of alpha-2 adrenergic receptors in the central nervous system leads to the inhibition of norepinephrine release causing termination of pain signals. These medications also stimulate the release of substance P leading to analgesia, potentiating the effects of opioids (35–37). Clonidine is available as an oral liquid, tablet, or transdermal patch with an onset of action of 30–60 min (36, 37). Dexmedetomidine is administered as an IV infusion with an onset of action of 5–10 min and is 8 times more selective for alpha-2 adrenergic receptors than clonidine (38). Although limited, there is data to suggest that the use of alpha-2 agonists may be effective in reducing the need for adjunctive sedation or analgesia, decreasing the duration of mechanical ventilation, and accelerating the attainment of full enteral feeds (39). Additionally, the use of dexmedetomidine has been associated with a decrease in overall opioid and benzodiazepine exposure (40, 41). Post-synaptic activation of alpha-2 receptors inhibits symptomatic activity leading to decreases in blood pressure and heart rate (33).

Other medications such as gabapentin, with a better-known safety profile, could be considered. Gabapentin, a gamma-aminobutyric acid analog, is increasingly used in infants with neuro-irritability of various etiologies (42). Its use is associated with decreased N-PASS scores and reduced need for analgesic or sedative medications, though data regarding dosing, efficacy and long-term neurodevelopmental impact remains limited (22, 42, 43).

Our proposed clinical pathway focuses on timely intervention and limiting exposure to opioids and benzodiazepines by prioritizing non-pharmacologic interventions and alpha-2 agonist medications.

Several non-pharmacologic interventions have been shown to reduce physiologic and behavioral stress responses. The efficacy of these interventions is dependent upon neonatal maturity and must be tailored based on PMA (44–47). They should be routinely incorporated into the care of all patients admitted to the NICU to prevent stress and agitation. Standard measures should include multidisciplinary involvement with establishment of daily routines, controlling light and sound exposure, encouraging parental involvement, creating a familiar environment, promoting uninterrupted sleep, clustering patient care activities, and minimizing painful procedures.

If pharmacological intervention is needed, intermittent dosing should be trialed before escalating to continuous infusions and alpha-2 agonists should be used first-line, when able. Opioids and benzodiazepines may be necessary; however, their use should be limited based on their known detrimental neurodevelopmental effects. Should agitation persist despite the above interventions, there is not sufficient evidence to routinely recommend the use of atypical antipsychotics in the neonatal population. Consultation with pediatric psychiatry, neurology, and pain team, if available, is recommended prior to initiation of medications outside of the clinical pathway.

Although further understanding of the underlying etiology of refractory agitation, its potential association with delirium, and dedicated research on the safety and neurodevelopmental impacts of emerging treatment strategies is critical, we must begin with recognition of this phenomenon in our unique patient population. Identifying high-risk patients can enable the use of non-pharmacologic neuroprotective measures that may prevent refractory agitation. Improved surveillance of this high-risk population may allow for earlier intervention to minimize symptoms.

We acknowledge that this pathway is limited by the availability of existing data, but necessary given the pervasiveness of agitation in the neonatal population and its potential long-term detrimental effects. We aim to call attention to the need for further research into the efficacy of current management strategies for and long-term impacts of neonatal agitation.