Hospital Mortality and Functional Outcomes in Pediatric Neurocritical Care.

OBJECTIVES
Pediatric neurocritical care (PNCC) outcomes research is scarce. We aimed to expand knowledge about outcomes in PNCC by evaluating death and changes in Functional Status Scale (FSS) from baseline among PNCC diagnoses.


METHODS
We conducted a 2-year observational study of children aged 0 to 18 years admitted to the ICU with a primary neurologic diagnosis (N = 325). Primary outcomes were death and change in FSS from preadmission baseline to discharge. New disability was defined as an FSS change of ≥1 from baseline, and severe disability was defined as an FSS change of ≥3. Categorical results are reported as relative risk (RR) with 95% confidence interval (CI).


RESULTS
Thirty (9%) patients died. New disability (n = 103; 35%) and severe disability (n = 37; 13%) were common in PNCC survivors. New disability (range 14%-54%) and severe disability (range 3%-33%) outcomes varied significantly among primary diagnoses (lowest in status epilepticus; highest in infectious and/or inflammatory and stroke cohorts). Disability occurred in all FSS domains: mental status (15%), sensory (52%), communication (38%), motor (48%), feeding (40%), and respiratory (12%). Most (64%) patients with severe disability had changes in ≥3 domains. Requiring critical care interventions (RR 2.1; 95% CI 1.5-3.1) and having seizures (RR 1.5; 95% CI 1.1-2.0) during hospitalization were associated with new disability.


CONCLUSIONS
PNCC patients have high rates of death and new disability at discharge, varying significantly between PNCC diagnoses. Multiple domains of disability are affected, underscoring the ongoing multidisciplinary health care needs of survivors. Our study quantified hospital outcomes of PNCC patients that can be used to advance future research in this vulnerable population.

Patients with pediatric neurocritical care (PNCC) conditions account for 20% of all pediatric critical care admissions in the United States. 1-3 PNCC conditions include traumatic brain injury (TBI), status epilepticus, neuroinfectious and inflammatory conditions, stroke, and cardiac arrest. These patients frequently require critical care interventions, account for billions of dollars in annual hospital costs, and suffer high rates of mortality and morbidity. 4 However, quantification of detailed outcomes among many PNCC diagnoses is scarce and limits intervention research to improve outcomes for these vulnerable children.
Previous research has focused on death and gross outcomes, such as the Pediatric Overall Performance Category or Glasgow Outcome Scale-Extended. [5][6][7][8][9][10] Although easy to administer, these scales do not provide detail about the type or domain of the disability found and therefore may lack the sensitivity needed for intervention research. The Functional Status Scale (FSS) is a recently validated measure specifically developed to overcome these challenges in pediatric critical care outcomes research by measuring 6 domains of function: mental status, sensory, communication, motor, feeding, and respiratory. 11 Previous work using the FSS shows that 7% of pediatric critical care survivors with a neurologic diagnosis and 11% of patients with traumatic injuries have a change in FSS from baseline of $3. 12,13 Another study focused on children with TBI requiring critical care intervention showed that 37% had a change in FSS of $3. This suggests variability in outcomes between neurologic diagnoses, 14 but previous studies have not evaluated FSS among many PNCC diagnoses. Additionally, changes of $3 in FSS represent substantial impairments in function and a cutoff correlating with moderate or severe impairments in adaptive function 11,15 and underestimate the overall burden of new functional deficits that require ongoing care.
We evaluated hospital outcomes among PNCC admissions by quantifying death and changes in FSS from preadmission baseline both overall and among different PNCC diagnoses. We quantified changes by domain of FSS to provide greater detail on outcomes and explored risk factors for worse outcomes. We hypothesized that rates of disability measured by changes in FSS would vary significantly by primary diagnosis.

METHODS Study Design and Setting
We conducted an observational study of consecutive children ages 0 to 18 years admitted with a primary neurologic diagnosis to the ICU at an academic tertiary children' s hospital and accredited level 1 pediatric trauma center from August 2016 to August 2018. This study was approved by the institutional review board with a waiver of informed consent. Deidentified data will be made available on request.

Participants and Data Collection
The primary exposure was admission diagnosis: TBI, status epilepticus, neuroinfectious and inflammatory diseases (meningitis, encephalitis, abscess or empyema, or demyelinating), stroke (hemorrhagic, ischemic, or cerebral sinus venous thrombosis [CSVT]), and hypoxicischemic encephalopathy (HIE) due to out-ofhospital cardiac arrest requiring chest compressions for .2 minutes. In patients with multiple diagnoses, the primary diagnosis was used (eg, patients with seizures due to meningitis were classified as infectious). All patients meeting inclusion criteria during daily census review were included in the study. Only characteristics of the first PNCC admission during the study period for a particular patient were included in the analysis.
Demographic and clinical characteristics were collected from electronic medical records and entered into Research Electronic Data Capture prospectively during admission. 16 Preadmission chronic conditions were grouped into system categories (Supplemental Table 4). Pediatric Index of Mortality 2 scores and critical care interventions evaluated illness severity. Interventions included intubation, noninvasive ventilation, central venous catheterization, arterial catheterization, intracranial pressure monitoring, continuous antiepileptic infusion, neurosurgical intervention (eg, decompressive craniectomy and hematoma evacuation), hemodynamic resuscitation or vasopressor use, and in-hospital cardiopulmonary resuscitation (CPR). Interventions were not included if they were used during operative management only (eg, intubation for operation). Seizures were diagnosed clinically and on EEG monitoring. TBI severity was measured by the initial Glasgow Coma Scale (mild complicated 13-15, moderate 9-12, and severe 3-8) recorded in the emergency department. Location and type of TBI were identified from radiology reports, and concurrent nonbrain traumatic injuries were identified from radiology and clinical reports.

Outcomes
The primary outcomes were in-hospital death and change in FSS from baseline. Manner of death was recorded from provider notes. The change from baseline FSS at hospital discharge was used to define disability outcomes and collected for all eligible patients. The FSS is validated for use in pediatric critical care outcomes in children of all ages, has excellent interrater reliability (intraclass correlation coefficient 5 0.95), 11 and correlates with the Adaptive Behavior Assessment System-Second Edition. 11 Baseline FSS scores, reflecting patients' preillness function, were documented in the medical record by the treating attending PICU physician within 24 hours of admission per our institution' s standard. Discharge FSS scores were inputted into the medical record by a treating attending PICU or neurology physician as part of the PNCC program. 17 Discharge FSS scores reflected the patient' s status on the day of discharge. Physicians prospectively inputted FSS on the basis of the 5-category rubric designed by Pollack et al 12 in a tool embedded in the electronic medical record. For this study, we defined new disability as an increase from baseline to discharge FSS total score of $1 point and new severe disability as an increase of $3. No patients had improvement from baseline FSS. Baseline FSS, median (IQR); maximum value 6 (6-6); 18 6 (6-6); 14 6 (6-11); 18 6 (6-6); 8 6 (6-6); 15 6 (6-6); 9 Discharge FSS, c median (IQR); maximum value 6 (6-8); 27 6 (6-7); 22 7 (6-11); 21 7 (6-8); 15 6.5 (6-8); 26 7.5 (6-9); 27

Statistical Analysis
Descriptive statistics were used, including percentage for categorical variables and median with interquartile range (IQR) for continuous variables, because data were not normally distributed. Demographic and clinical characteristics were compared between survival and disability groups. x 2 tests for categorical variables (with Fisher' s exact correction for expected cell counts ,10) and Mann-Whitney U tests for continuous variables were used. Bivariate results for categorical variables were reported as relative risk (RR) with 95% confidence interval (CI). The primary analysis compared new disability groups among the overall PNCC cohort. The hypothesized variability in new disability rate by diagnosis was tested by using x 2 tests. Secondary analyses were conducted to compare the overall cohort by new severe disability and to explore outcomes among individual diagnoses (Supplemental Tables 5-9). All tests were 2 tailed, and significance was defined as P , .05. Because secondary analyses were exploratory within limited populations, no adjustment was made for multiple comparisons. All analyses were conducted by using SPSS (version 24.0; IBM SPSS Statistics, IBM Corporation).

RESULTS
Over 2 years, 325 patients had a primary PNCC diagnosis, accounting for 16% of all ICU admissions. Diagnoses included TBI (n 5 154; 47%), status epilepticus (n 5 71; 22%), infectious or inflammatory disease (n 5 40; 12%), HIE (n 5 37; 11%), and stroke (n 5 23; 7%). One-third (n 5 96) had preadmission chronic conditions, which varied by primary diagnosis (Supplemental Table 4). Most (84%) patients had normal baseline FSS (FSS 5 6). Severity of illness varied, but 56% received at least 1 critical care intervention, and 25% received 3 or more.   Among patients with new disability, fewer changes (#15%) were found in mental status and respiratory domains compared with sensory (52%), motor (48%), feeding (40%), and communication (38%). Table 3 shows the prevalence of individual FSS domain changes, which varied by diagnosis. Within most FSS domains, changes of 1 were most prevalent Supplemental Fig 3). We found that nearly 1 in 10 PNCC admissions result in death, confirming results from previous administrative data and point-prevalence studies in PNCC 3,4 and showing that PNCC mortality is substantially higher than rates for other PICU cohorts. 12,19 We also found high rates of morbidity that were consistent with previous reports. Pollack et al 12   Similar to previous studies, most children requiring PNCC admission had normal functional baselines in our study. 13,14 Therefore, we also evaluated disability as defined by smaller changes in FSS than previous studies to better quantify the breadth of morbidities these children suffer. Using our definition of FSS increases of $1 from baseline, the prevalence of new disability after PNCC increased substantially to 35% (range 14%-53% by diagnosis). We believe these small FSS changes are important at the individual patient level when considering risk for ongoing health care needs and reduced quality of life [20][21][22][23][24] and are important considerations for clinicians. Changes of 1 point among previously healthy children are substantial and correspond to suspected hearing or vision loss, need for oxygen or suctioning, need for age-inappropriate help with feeding, or functional impairment of 1 limb depending on the domain affected. 11 These data can be used in future studies aimed at reducing disability after PNCC.
Our study adds granularity to outcomes data by using the FSS. The 6 domains of the FSS offer approximations of activities of daily living and correlate with results of comprehensive adaptive behavior testing. 11 Evaluating outcomes at the FSS domain level, we found that all domains are affected after PNCC, and patients often have disabilities in multiple domains. The wide breadth of morbidities found by using FSS, coupled with knowledge about cognitive, emotional, and psychosocial morbidities after ICU hospitalization, [24][25][26][27][28] suggests that a multidisciplinary approach to patient followup is needed to adequately care for these children. Despite the high rate of new disability in multiple domains, ,10% of our patients were discharged to inpatient rehabilitation facilities. Currently few institutions offer systematic follow-up care in PNCC survivors that can provide this type of comprehensive care outside of a rehabilitation setting, although increasing recognition of long-term sequelae is driving the development of longitudinal programs. 3,17,29,30 Our study underscores the need for developing systematic support for patients and families after discharge.
Few interventions have been shown to improve outcomes in PNCC, and intervention studies in this population are hindered by limited knowledge of outcomes and risk factors. 1 With our study, we explored risk factors that could be used to predict outcomes or targeted in future intervention studies. Not surprisingly, markers of illness severity, such as need for critical care interventions, lower Glasgow Coma Scale score in TBI, and increased CPR duration in cardiac arrest, were associated with worse outcomes, similar to previous studies. 14,31-34 However, we also identified seizures during admission as a risk factor for worse outcomes in the overall cohort as well as among individual diagnoses. Seizures may represent worse primary brain injury, but seizures can also lead to secondary brain injury and may serve as a modifiable target to improve outcomes. Although our institution has guidelines for seizure prophylaxis and EEG monitoring in patients with TBI, we did not systematically evaluate EEG for other diagnoses and, notably, may have underdiagnosed subclinical seizure activity. Seizure burden is linked to worse outcomes in the PICU population, 35 and our results further support guidelines for systematic EEG monitoring in PNCC. 36 It remains unknown if interventions targeting seizures can improve outcomes.
Our study has several limitations to consider, including the single-center setting when admission characteristics, treatments, and outcomes in PNCC patients may vary by region and center. 3  our patients had normal baseline FSS, but larger studies could explore the development of new disabilities and severity of decline relative to baseline among children with preadmission neurologic impairments. This study was also limited to hospital outcomes, and little data exists on longitudinal outcomes after PNCC, particularly in nontraumatic diagnoses. It is likely that FSS improves or declines in the months after discharge, 40 and future work should evaluate postdischarge FSS to determine potential for recovery and need for ongoing medical intervention after PNCC. Additionally, the FSS does not capture important morbidities, such as those of detailed neuropsychological assessments of cognitive and mental health, or assess impairments in quality of life. Future studies should assess the relationship between FSS and these other important outcomes.

CONCLUSIONS
More than one-third (35%) of PNCC survivors have new disability when evaluating changes in FSS from baseline, and outcomes vary significantly with primary diagnosis. All functional domains measured by the FSS can be affected after PNCC, and patients with new disability often have multiple domains affected concurrently, underscoring the ongoing need for multidisciplinary care after discharge. Future studies should determine the longitudinal course of recovery in FSS after discharge and evaluate risk factors for worse outcomes in larger multicenter populations to identify targets for future intervention.