Predictors of Neurologic Outcomes and Mortality in Physically Abused and Unintentionally Injured Children Admitted to Intensive Care Units


 Background: This study aimed to identify the risk factors and predictors of neurologic outcomes and mortality for children with physical abuse and unintentional injury admitted to intensive care unit (ICU).Methods: We retrospectively evaluated the data for children with maltreatments admitted to the pediatric, neurosurgery and trauma ICUs between 2001 and 2019. Clinical factors were analyzed and compared between the physical abuse and the unintentional injury groups, including age, gender, season of admission, identifying settings, injury severity score, etiologies, length of ICU stay, neurologic outcomes, and mortality. Neurologic assessments were conducted with the Pediatric Cerebral Performance Category scale.Results: A total of 2,481 forensic children were collected, and there were 480 victims (19.3%) admitted to the ICus, including 156 physical abused and 324 unintentional injured. The significant difference between the two groups included age, histories of prematurity, clinical outcomes, head injury, neurosurgical interventions, clinical manifestations, brain computed tomography findings, and laboratory findings (all p<0.05). Traumatic brain injury (TBI) accounted for the majority of ICU patients. Abusive head trauma (AHT) was 87.1% in the abused group. In the abused group, only 46 (29.4%) cases achieved a favorable neurologic outcome. While268 (82.7%) achieved a favorable neurologic outcome in the unintentional injured group, Shock within 24 hours and spontaneous hypothermia (body temperature <35°C) presented the strong risk factors for poor neurologic outcomes and mortality in both group. Post-traumatic seizure was also associated with poor neurologic outcomes in both groups. Conclusions: In children with physical abuse and unintentional injuries on ICU admission, initial presentation with shock and spontaneous hypothermia may serve as the powerful predictors for mortality; and shock and convulsion may serve as the powerful predictors for poor neurologic outcomes. In addition, retinal hemorrhage could be a risk factor for poor neurologic outcomes in critical victims with physical abuse, and spontaneous hypothermia could be a risk factor for poor neurologic outcomes in unintentional injuries.


Introduction
Child maltreatment and the associated morbidity and mortality are a critical and complex public health issue worldwide. It is generally recognized as physical abuse, neglect, emotional abuse, and sexual abuse. Physical abuse and neglect are the two most common types of child maltreatment and are considered serious causes of child mortality (1). These cases are de ned as events that occur as a result of injury caused by in icted or careless consequences. According to the World Health Organization, physical abuse yields the highest mortality rate among the types of child maltreatment (2,3). In other studies, the mortality rate was reported to be approximately 9% in physically abused victims and 2% in neglected victims (4)(5)(6). Cases of neglect may include tra c accidents, falls, near drowning, burns, substance poisoning, and asphyxia, which are considered unintentional injuries (3). Victims of physical abuse were reported to have higher proportions of injuries in the head, thoracic region, and skin than those of neglect (1,2). Physical abuse involving the head is called abusive head trauma (AHT); it is the most critical form of child maltreatment with a mortality rate of 20% and permanent neurologic damage rate of up to 50% (6,7).
Because of the poor prognosis associated with child maltreatment, medical or surgical treatments should be performed immediately in critically ill victims, and the potential prognostic factors should be identi ed to avoid poor clinical outcomes. However, to date, only a few studies have focused on physically abused and unintentionally injured children admitted to the intensive care unit (ICU) (1,2,4,5,8,9). Therefore, this study aimed to evaluate victims of physical abuse and unintentional injury admitted to the ICU and to analyze the related predictors of clinical outcomes.

Study Population
From 2001 to 2019, we retrospectively evaluated physically abused and unintentionally injured victims aged 1 month to 18 years admitted to the pediatric, neurosurgical, and trauma ICUs from the emergency department (ED), outpatient department (OPD), and wards. A total of 2,481 children reported from the social welfare reporting system were included in this study. The study was approved by the Institutional Review Board of Chang Gung Memorial Hospital (reference number: 104-8307B). All procedures were performed in accordance with the relevant guidelines and regulations. Data were collected, reviewed, deidenti ed, and anonymized before analysis, and the Ethics Committee waived the requirement for informed consent because of the anonymized nature of the data and scienti c purpose of the study.

Study Setting and Patient Selection
Chang Gung Memorial Hospital is a tertiary medical center that receives cases transferred from local clinics, regional hospitals, and social and political institutions. The pediatric ICU has 29 beds, wherein patients aged 1 month to 18 years are hospitalized. Meanwhile, the neurosurgical and trauma ICUs have 30 and 28 beds, respectively, wherein patients with severe brain injury or physical trauma requiring medical or surgical treatments are managed. In the study, all maltreated children admitted to the ICUs were divided into two major groups based on the etiologies, i.e., physical abuse and unintentional injury.
Physical abuse is de ned as acts of commission by a caregiver that cause actual physical harm or have the potential for harm. Neglect is de ned as a failure by a child's caregiver to meet the physical, emotional, educational, or medical needs of the child. Unintentional injury is a part of neglect that causes physical damage (tra c accidents, falls, near drowning, burns, substance poisoning, and asphyxia). All victims of physical abuse and unintentional injury were evaluated by a child protection team.

Study Protocol
Information related to the cases of physical abuse and unintentional injury was obtained from the social welfare reporting system and the medical chart records; this included age, sex, personal and family histories, types of abuse, types of injury, initial clinical presentations on admission, physical examinations, laboratory tests, imaging examinations, length of stay (LOS) in hospitals, LOS in the ICU, time of presentation (morning shift, night shift, or graveyard shift), time of report (8 a.m. to 5 p.m., 5 p.m. to 0 a.m., and 0 a.m. to 8 a.m.), transferring unit (ED, OPD, wards, or other hospitals), morbidity, neurologic outcomes, and mortality rate. Admission to the ICU was considered to indicate an urgent and critical condition requiring emergency critical care and a high risk of mortality. The factors associated with both mortality and neurologic outcomes were analyzed between the physically abused and unintentionally injured groups. Moreover, the predictors of mortality were further analyzed in each group using a logistic regression analysis.

Outcome Variables
The in-hospital mortality and neurologic outcomes of the survivors were the main outcome variables of this study. Meanwhile, the general functional status at discharge was the secondary outcome variable, assessed using the Pediatric Cerebral Performance Category (PCPC) scale. The neurologic outcomes of the survivors with head injuries were recorded on the basis of their PCPC scale score and categorized as favorable (PCPC scale score, ≤ 2) or poor (PCPC scale score, ≥ 3). A survival analysis was conducted to compare the different survival rates between the physically abused and unintentionally injured groups according to the period of critical care in the ICU.

Statistical Analysis
The chi-square test, Fisher's exact test, Student's t-test, the Mann-Whitney U test, and a multivariate logistic regression analysis were used, where appropriate. In the descriptive analysis, values were presented as means ± standard deviations. Differences between the groups were presented as 95% con dence intervals (CIs). The chi-square test or Fisher's exact test was used to compare dichotomous variables and the Mann-Whitney U test to compare continuous variables between the groups. The predicted probabilities of mortality and 95% CIs were calculated using a logistic regression model. The log-rank test and Kaplan-Meier survival curves were applied to compare survival differences. A Cox proportional hazard model was used to test the effect of independent variables on hazards. Statistical signi cance was set at p-values of < 0.05, and all statistical analyses were conducted using the IBM SPSS Statistics software (version 22.0; SPSS Inc., Chicago, IL, USA).

Demographics
During the 19-year study period, 2,481 cases of child abuse were reported via the social welfare reporting system; there were 480 (19.3%) patients admitted to the ICU for critical care, including 304 (63.3%) boys and 176 (36.7%) girls ( Fig. 1). Among them, 156 (32.5%) were physically abused, and 324 (67.5%) were unintentionally injured. The demographics between the two groups are shown in Table 1. Age, history of prematurity, clinical outcomes, head injury, neurosurgical interventions, clinical manifestations, brain computed tomography (CT) ndings, and laboratory ndings signi cantly differed between them (all p < 0.05). Traumatic brain injury (TBI) was the major etiology for admission to the ICU, including AHT and accidental TBI. The incidence of AHT was 87.1% in the physically abused group, and that of accidental TBI was 57.4% in the unintentionally injured group (p < 0.01). In terms of brain CT ndings, subdural hemorrhage (SDH) and subarachnoid hemorrhage were more common in the physically abused group, while epidural hemorrhage was more predominant in the unintentionally injured group (both p < 0.05). Moreover, neurosurgery was more commonly performed in the physically abused group than in the unintentionally injured group (53.2% vs. 18.5%; p < 0.05). In terms of blood laboratory ndings, the physically abused group had lower hemoglobin level, platelet count, and serum sodium level but higher blood pH and serum potassium level than the unintentionally injured group (all p < 0.05). In terms of clinical presentation, shock, retinal hemorrhage, and convulsions were more common in the physically abused group than in the unintentionally injured group (all p < 0.05). In terms of clinical outcomes, the LOS in both the hospital and ICU was longer in the physically abused group than in the unintentionally injured group (both p < 0.05). In addition, the proportion of victims with worse neurologic outcomes (PCPC scale score, ≥ 3) was higher in the physically abused group than in the unintentionally injured group (p < 0.05).

Factors Associate with Poor Neurologic Outcomes
The results of the comparison of the favorable and poor neurologic outcomes between the two groups are listed in Table 2. A total of 110 (70.6%) and 56 (17.3%) children had poor neurologic outcomes in the physically abused group and unintentionally injured group, respectively. Moreover, the children with poor neurologic prognosis had signi cantly higher rates of shock, head injury, and convulsions and lower initial Glasgow Coma Scale (GCS) score and body temperature in both groups (all p < 0.05). In the multivariate logistic regression analysis, shock within 24 hours of ICU admission and initial presentation of convulsions were both found to be associated with poor neurologic outcomes in both groups (both p < 0.05) (Table 3). However, retinal hemorrhage was associated with poor neurologic outcomes in only the physically abused group and spontaneous hypothermia (body temperature, < 35°C) in only the unintentionally injured group. Among the related signi cant clinical factors, shock within 24 hours of ICU admission yielded the highest odds ratio for poor neurologic outcomes in both groups.

Factors Associated With Mortality
The overall mortality rate among the study population was 9.8%; speci cally, the mortality rate was 12.8% in the physically abused group and 8.3% in the unintentionally injured group. Table 4 shows the Cox proportional hazards for predicting 28-day mortality. Shock within 24 hours of ICU admission was the strongest predictor of 28-day mortality, followed by spontaneous hypothermia in both groups. The rate of poor neurologic prognosis (PCPC scale score, ≥ 3) was also signi cantly higher in the children with spontaneous hypothermia on admission to the ICU (93.5%) than in those without (28.1%, p < 0.05). The Kaplan-Meier survival curves in the two groups are shown in Figs. 2 and 3. The occurrence of initial shock within 24 hours of ICU admission was analyzed using the Kaplan-Meier survival curves in the physically abused ( Fig. 2A) and unintentionally injured groups (Fig. 2B). The proportion of victims who initially presented with shock and then expired was higher, and those who initially presented with shock expired earlier than did those who did not in both groups. Meanwhile, the proportion of victims who initially presented with spontaneous hypothermia and then expired was higher, and those who initially presented with spontaneous hypothermia expired earlier than did those who did not in the physically abused (Fig. 3A) and unintentionally injured groups (Fig. 3B).

Discussion
Approximately 80% of mortality caused by AHT may be survived with earlier interventions (10). Primary clinicians should be highly alert to recognize the possibility of child maltreatment. However, timely diagnosis of AHT is di cult because the initial presentations often involve nonspeci c complaints, such as di culty feeding and breathing, poor activity, apnea, convulsions, or drowsiness; this consequently causes confusion on the diagnostic process among doctors (11). Differentiating between AHT and accidental TBI may be equivocal; this is especially true, as histories provided are usually ctitious to conceal abuse.
This study is the rst to analyze the risk factors associated with the clinical prognosis of maltreated children admitted to the ICU. Herein, we included and investigated up to 480 victims admitted to the ICU within a relatively long study period (19 years). The ICU admission rate was 19.3% in the total study population of 2,481 children, which indicates that 1 in every 5 cases required critical care owing to child maltreatment. The mortality rate was 12.8% in the physically abused group and 8.3% in the unintentionally injured group, which was both higher than previous rates (5,6). We think that this may be attributed to the rarely reported cases from the medical and social systems in the past. Only very critically ill victims were admitted to the ICU and reported in the past.
In our study, we identi ed that initial presentation with shock within 24 hours was the strongest predictor of poor prognosis in both groups. The major type of shock observed was neurogenic shock, mainly caused by TBI, which accounted for approximately 51% of the mortality herein. TBI is a leading cause of morbidity and mortality in child maltreatment cases worldwide, and previous studies have demonstrated several risk factors associated with poor outcomes in children with TBI, including low GCS score, posttraumatic seizure, retinal hemorrhage, young age, and SDH (12)(13)(14)(15)(16)(17). However, in our study, we found that shock, head injury, convulsions, retinal hemorrhage, lower initial GCS score, and lower body temperature were all risk factors of poor neurologic outcomes. Severe brain swelling and mass effect caused by TBI may lead to neurogenic shock. Shock may cause ischemic injuries and reperfusion impairment, which may increase the rates of cell death and multi-organ failure, resulting in high mortality rates (18)(19)(20)(21). In our study, the children with shock expired earlier and easily and had poor neurologic outcomes. Therefore, the occurrence of shock within 24 hours of ICU admission may indicate a high mortality and a poor neurologic outcome in physically abused or unintentionally injured children. Hence, closely monitoring the early signs of shock, such as tachycardia, poor perfusion, and low blood pressure, and emergent uid resuscitation with vasopressors could be important for children with TBI.
The second strongest risk factor associated with mortality in our study was spontaneous hypothermia following injuries. Previous studies have demonstrated that spontaneous hypothermia was a risk factor for mortality in adult patients with severe trauma and hemorrhage and identi ed it as a new risk factor after cardiac arrest in adults (22)(23)(24). Under normal conditions, body temperature is regulated by the hypothalamus. Spontaneous hypothermia after severe injuries may indicate a harmful mechanism of thermoregulation by the cerebrum resulting from ischemia and reperfusion damage. Therefore, spontaneous hypothermia may be translated as a sign of cerebral injury (24). In addition, hypothermia may disrupt systemic homeostasis, deplete energy stores, and nally cause multi-organ failure (23). This is the rst study to analyze the impact of body temperature changes on ICU admission in abused children. In physically abused and unintentionally injured children, critical conditions with spontaneous hypothermia may indicate a high mutability rate. Moreover, spontaneous hypothermia may also serve as a predictor of poor neurologic prognosis in children with unintentional injuries. When shock and spontaneous hypothermia occur in physically abused and unintentionally injured children, the harmful changes become additive, and a higher mortality rate is ensued.
Herein, we also found that convulsion was not only a common post-trauma complication but also a risk factor for poor neurologic outcomes in both the physically abused and unintentionally injured groups. The correlation between convulsion and poor neurologic outcomes may be attributed to the consequent increases in metabolic demands, intracranial pressure, and release of neurotransmitters arising from secondary brain injury (14,25). Therefore, early detection using continuous electroencephalographic monitoring and antiepileptic drug administration could be recommended for pediatric TBI to avoid progression of poor neurologic outcomes (26,27). Although traumatic causes may result in a pattern of retinal hemorrhage, this type of hemorrhage is still an important manifestation of AHT and frequently observed particularly during dilated eye examination in approximately 85% of children with AHT. In our study, retinal hemorrhage was found to be a risk factor for poor neurologic outcomes among the physically abused children but not among the unintentionally injured children. The presence of retinal hemorrhage may have a nonspeci c pattern in children with AHT; however, once retinal hemorrhages are noted in children with AHT, poor neurologic outcomes may be expected. Therefore, primary clinicians should pay more attention to performing dilated eye examinations in children suspected to have experienced child maltreatment because the presence of retinal hemorrhage may not only identify the existence of child abuse but also indicate the possibility of poor neurologic outcomes.

Conclusions
Initial presentation with shock and spontaneous hypothermia on ICU admission could serve as strong predictors of mortality and shock and convulsion as strong predictors of poor neurologic outcomes in physically abused and unintentionally injured children. In addition, retinal hemorrhage and spontaneous hypothermia could serve as a risk factor for poor neurologic outcomes in physically abused and unintentionally injured children, respectively.