Factors Associated With Police Bringing Children to a Psychiatric Emergency Room

In September 2019, GQ headlined with “Florida Police Officer Arrested and Handcuffed a 6-Year-Old Black Girl for a Tantrum in Class” (1). The story recounted that she “was throwing a tantrum in class and kicked a staff member”; the school resource officer put her in the back of his cruiser and drove her to a juvenile detention facility. In February 2021, NBC News reported that police officers responded to a report of “family trouble” and “were made aware that a 9-year-old [girl] indicated that she wanted to kill herself and she wanted to kill her mom” (2). Subsequent footage showed authorities handcuffing the girl while she repeatedly screamed for her father and refused to get into the vehicle. Officers can be heard saying they would “pepper-spray her if she continued to resist.” Their threat did not deescalate the situation, and the child was pepper-sprayed.

These stories highlight several questions that have received little attention: Why are police called, for example, to bring young children to emergency rooms (ERs) (i.e., “police arrivals”)? Are the events that precipitate child ER visits prompted by a 911 call different from other child psychiatric emergencies? Can we find better ways to help families manage situations that caused the 911 call?

Although some literature has examined police arrivals among adults (3–7), no study has focused specifically on young children. One study compared police arrivals with nonpolice arrivals among youths. In the sample of 1,779 assessments from two New York hospitals between 1993 and 1995 (mean age=14 years) (8), 3% of the ER visits were police arrivals. Compared with youths not arriving via police, youths brought in by police had shown more assaultive and destructive behavior as well as more drug use but not alcohol use. No significant differences were found in discharge diagnoses or hospitalization rates.

Two chart review studies of child psychiatric emergency referrals examined repeated ER visits and reported rates of police arrivals (9, 10). In one study of 855 children and adolescents presenting to a comprehensive psychiatric emergency program (CPEP) in New York City in 2011–2012, about 26% were ≤11 years old (9). The overall 6-month revisit rate was 21%, and 19% of those with a revisit were accompanied by police or social services. Rates of outward aggression (41%) were higher than rates of suicidal behavior (25%) and rates of both aggression and suicidality (12%), and predictors of revisits were female sex, more severe symptoms, disruptive behavior, previous suicide attempt, receipt of outpatient mental health services, and involvement of child welfare agencies.

A recent study from Michigan included 1,654 children ages 5–12 years seen between 2012 and 2018 (10). The authors reported rates of suicidal ideation (46%) and self-harm (32%), but aggression was not studied. The revisit rate was 24%, and 18% of revisits were police arrivals.

The present study adds to this literature by characterizing children ≤12.9 years old who were brought to the CPEP of a suburban university hospital either by police in response to a 911 call or by caregivers. Specifically, we addressed two questions: What clinical and demographic factors differentiate a child’s first police arrival from family or caregiver (i.e., nonpolice) arrivals? Among children who experience a first police arrival, what factors predict having another?

Because of the important roles of demographic factors such as race and social disadvantage (e.g., not living in a two-parent family or having Medicaid) in serious childhood mental health problems, we posed the following hypotheses (11–13): suicidal ideation or behavioral and emotional outbursts during which the child destroyed property or was verbally or physically aggressive would be associated with police arrival to the CPEP; Black race, evidence of social disadvantage, and lack of previous treatment would characterize police arrivals (9, 14–16); and, compared with children with a single police arrival, children with multiple police arrivals would have more types of aggression and suicidality and would have had more treatment because of previous CPEP encounters recommending such treatment (9, 10).

Methods

Study Design

Following a protocol approved by the Stony Brook University Institutional Review Board, we reviewed the electronic medical records (EMRs) of all children ages 5.0–12.9 years evaluated in the study CPEP from September 9, 2017, to April 30, 2018. These children composed the index sample and were evaluated during the academic school year, when child referrals are at their height. The study period preceded the start of a countywide initiative to divert less severe pediatric referrals to a different program. However, all previous (the earliest CPEP encounter by a child in the index sample was April 2013) and subsequent encounters (through March 2020, when COVID-19 shutdowns began) for the index sample were also reviewed to determine the total frequency of CPEP encounters and the reasons for these encounters.

For the duration of this study and the follow-up period, the CPEP was the only option for child psychiatric emergencies in the county, whereas other options were available for adolescents ages ≥13 years. The police brought children to the CPEP when someone called 911 and the responding officers felt that evaluation in the CPEP was warranted.

The following data were available from the EMRs for all encounters: child’s age, month and year of referral, sex, race-ethnicity (declared by the child’s admitting caregiver at the time of registration), living situation (two-parent family vs. other arrangements, including single parent, living with other family members, or foster or residential care), who brought the child to CPEP (police or family), who referred the child to CPEP (family, school, or mental health or other medical professional), any past or present psychiatric treatment at the time of the child’s arrival (yes or no), disposition (hospitalization or other disposition), and CPEP discharge diagnoses.

The evaluation of the CPEP attending psychiatrist was based on information from the caregiver, child, and, where relevant, police report. From these sources, reasons for CPEP encounters were coded as aggression related, suicide related, both, or other. Discharge diagnoses were attention-deficit hyperactivity disorder (ADHD), oppositional defiant/conduct disorder, anxiety, depression, autism spectrum or developmental disorder, and psychosis. Adjustment disorder was recorded if that diagnosis had been given and no other primary diagnosis preempted it.

Suicidality (suicide-related referrals) was coded for verbal behavior (ideation or threat) and for any kind of suicidal behavior or action (including nonsuicidal self-injurious behavior and aborted or interrupted suicide attempts). Aggressiveness, usually in the context of severe emotional outbursts, was coded for presence of verbal or physical aggression or destruction of property. Summary scores were created by using the number of classes of behaviors the child exhibited (i.e., exhibiting both suicidal ideation and behavior was scored as 2; a child who showed verbal and physical aggression and destroyed property received an aggressiveness score of 3). Raters coding the CPEP notes were blind to child race-ethnicity (not included in CPEP notes). Pairs of six raters were trained to a kappa of 0.8 for each of the behavior ratings (suicidal ideation or attempt or property destruction or verbal or physical aggression). Once rater agreement reached a kappa of 0.8, raters independently reviewed notes. Video calls were held biweekly to reconcile disagreements or, after training was complete, to resolve scoring issues.

Approximately one-third of children had multiple CPEP encounters. Several characteristics of the child were time invariant (e.g., sex, race, and whether the child was ever on Medicaid [an indicator of socioeconomic status]), but others (e.g., the child’s age when the encounter occurred, suicidality, aggression, living situation, treatment history, and whether the CPEP encounter was a police arrival) could vary from encounter to encounter.

Statistical Analysis

Children who had at least one police arrival to the CPEP were compared with those who had none. We used t tests, Fisher’s exact tests, and chi-square tests for the time-invariant characteristics to compare the two groups on numeric, binary, and multicategory variables, respectively. For time-varying characteristics, we used multilevel linear, logistic, and generalized logit models to compare the groups; heuristically, these models were used to estimate each child’s mean level of the characteristic across all encounters through March 2020 (N=757) and then to average these means across all children within a group. For binary and categorical variables, the analysis was used to estimate the average of the child-specific percentages across all children within a group.

To examine factors associated with a first police arrival, we estimated two multilevel logistic regression models for all CPEP encounters up to and including the first police arrival. The first model included the child’s time-invariant characteristics as well as age at the time of encounter and whether the child was living with two parents; collectively, these characteristics could be considered sociodemographic factors. The second model added the two behavioral factors (i.e., aggression and suicidality), treatment status, and the number of previous nonpolice arrivals to the CPEP (all time-varying factors).

For the 103 children with a first police arrival, we used ordinary logistic regression analysis to assess which factors predicted having a subsequent police arrival to the CPEP. This analysis included the same variables used in the previous analysis. Values for the time-varying factors were used from the first police arrival; therefore, this was a prospective analysis (using retrospectively coded data).

All analyses were performed with SAS, version 9.4. A two-tailed p<0.05 was considered statistically significant. The Veall and Zimmermann pseudo-R² was used for each logistic regression model.

Results

Over the 8-month index period (September 2017 through April 2018), 435 CPEP encounters occurred for 339 children ≤12.9 years; of the total number of CPEP referrals (N=11,447), child encounters constituted 3.8%. Additionally, 22% (N=96) of the 435 encounters during this period were revisits and 28% (N=121) were police arrivals.

Of the 339 children, 103 (30%) had at least one police arrival through March 2020, including 33 (10%) with a single police arrival; 25 (7%) with multiple CPEP encounters, which were all police arrivals; and 45 (13%) with both police and nonpolice arrivals. Of those with only nonpolice arrivals (N=236), 165 (70%) had a single encounter, and 71 (30%) had multiple nonpolice arrivals.

In addition to the 435 CPEP encounters during the index period, the children involved in these encounters had an additional 108 encounters that had occurred before the index period and 214 that occurred after (Table 1). The analysis of factors associated with a first police arrival was based on 548 CPEP encounters, up to and including the first police arrival, spanning the preindex, index, and postindex periods (Table 1).

Table 2 presents descriptive statistics for the full sample, including children with nonpolice arrivals and children with police arrivals. The latter group had significantly more CPEP encounters than children with only nonpolice arrivals. Children were referred to the CPEP by families (60%), school personnel (who either called the police or told parents to take the child to the CPEP) (32%), or a mental health or other medical professional advising that the family take the child to the CPEP (8%).

No significant differences in sex or age were found between children with or without police arrivals. Rates of ever having Medicaid insurance and coming from families without two parents were substantially higher for children with a police arrival than for children with only nonpolice arrivals.

Children who were brought by police showed significantly more aggressive and suicidal behaviors than children who never arrived with police, as evidenced by large statistically significant differences in most of the encounter-level binary measures. Outbursts of verbal and physical aggression or destruction of property were significantly more common in encounters among children with at least one police arrival (71%) than among those with no police arrivals (30%) (OR=5.63). Although most children with outbursts had calmed by the time police responded to the 911 call, according to the police reports, two boys and three girls continued to show aggression in the presence of the police officers. Finally, suicidal behavior, but not suicidal ideation, was more than twice as high among children with any police arrivals (23%) than those with only nonpolice arrivals (10%) (OR=2.56).

Compared with children with only nonpolice arrivals, children with a police arrival were more likely to have had either current or past treatment and to be hospitalized after their CPEP visits (Table 2). Police-arrived children were significantly more likely to have discharge diagnoses of ADHD and oppositional defiant/conduct disorder and were less likely to have a diagnosis of anxiety-related or adjustment disorder. Rates of depression and developmental disorder did not differ significantly between the two groups. Psychotic-like symptoms were rare, with only 13 instances of auditory or visual hallucinations reported; unspecified psychosis was diagnosed once.

Although Black children and Latinx children had significantly more police arrivals than White children and Asian or other children, the initial multilevel logistic regression model (Table 3, model 1) revealed that only Black race, older age, male gender, not living in a two-parent family, and having Medicaid insurance were independently associated with a first police arrival. With the addition of clinical factors (aggression and suicidal behavior severity, model 2) into the analysis, Black race and insurance status were no longer statistically significant factors. Table 4 shows that Black children were most likely to have multiple police arrivals and were least likely to have a two-parent family. Asian children and those from other race-ethnicities (i.e., Indigenous American) were most likely to have a single police arrival and were least likely to use Medicaid, compared with children in the other racial-ethnic groups. Black children were brought to the ER for outbursts or aggression more often than their peers, whereas Asian or other children were more often brought for suicidality.

Table 5 shows that young age at the time of a first police arrival predicted an increased likelihood of a future police arrival (model 1). Although the OR for age in model 2 was similar to that of model 1, age was not statistically significant in model 2 because of multicollinearity with the added variables. Severity of outbursts and aggression, severity of suicidality, and a history of treatment were the strongest predictors of a second police arrival.

Discussion

In this study, we used police arrival data from the only psychiatric ER setting for children in the United States for which data were available at the time, finding that 28% of CPEP arrivals were brought in by police in response to a 911 call. This rate is higher than those reported in studies conducted in New York City and Michigan (9, 10), even for revisits (18%–19%). It is unclear whether this difference represents variation in police practices, samples, 911 call use, or treatment options. Moreover, comparison with previous literature is difficult because of disparate study goals; only one study of children and adolescents examined police arrivals (8). In other studies, aggressive behavior was not addressed in the clinical information examined (10, 17), settings included both inner cities and other environments (8–10), study periods ranged from the 1990s (8) to recent years (9, 10), and sample ages ranged from children (10) to adolescents (8, 10). Finally, police arrival data from studies of adults entail very different developmental considerations (e.g., parents and schools are not usually involved in arrivals) and diagnostic characteristics (e.g., intoxication, psychosis, disorientation, and involuntary commitment) (3–7).

The children in this study who arrived at the CPEP with police one or more times showed highly dysregulated behaviors and represented a socially marginalized group. Consistent with some of our hypotheses, important demographic correlates were Black race, Medicaid insurance, not living in a two-parent home, male sex, and older age. These associations were statistically significant before the clinical variables were added to the model. However, the clinical variables appeared to account for differences in police versus nonpolice arrivals and were associated with both first and subsequent police arrivals. This was especially true for aggressive outbursts, which prompted more police calls among Black families, who were more likely than families of other race-ethnicities to have the aforementioned demographic characteristics. In this sample, suicidality was a less common reason for police encounters than were aggressive outbursts, with suicidal behavior but not ideation differentiating police arrivals from nonpolice arrivals.

Although lack of previous treatment is often associated with excessive ER use (14, 17–19), treatment history was not significantly associated with a first police arrival. Of note, treatment, including hospitalization, at the time of the first police arrival significantly predicted an increased likelihood of a second police arrival. Treatment and medication use have also been associated with repeat ER use (9, 10). We surmise that police might be called, possibly on the recommendation of treatment providers, because families, schools, and residential programs lack other viable options and do not know what else to do.

Unlike in adult samples (3–7), psychosis, intoxication, and severe mood disorder were rare to absent in our sample of children. The diagnosis of ADHD, however, was made for 43% (N=146) of the children, and more often for Black children (N=42, 66%) than for White (N=80, 43%), Latinx (N=24, 29%), or Asian or other youths (N=0) (Table 4). ADHD is a common diagnosis among children with aggressive outbursts, and several recent studies have highlighted the importance of emotion dysregulation and impulsivity in this disorder (20–24). Moreover, higher temper outburst scores and levels of aggression among children with ADHD appear to be linked to their ADHD symptoms rather than to their mood symptoms (22). Detection of ADHD alone or in addition to other behavior disorders has important therapeutic implications, because adequate ADHD treatment significantly reduces aggression (23). Perhaps the child and adolescent psychiatrists in this CPEP, who were preferentially dispatched to see young referrals, were attentive to making a treatable diagnosis regardless of the patient’s race-ethnicity. By contrast, other studies report that ADHD is underdiagnosed in Black patients (24–26).

This study included a review of records of police arrivals in one suburban, university-based CPEP/psychiatric ER, and thus, generalizability of its findings is limited. However, no other comparable studies with samples of children are available. Although many studies have investigated pediatric ER referrals (17), they do not describe police involvement or do not ascertain presence of aggressive behavior (10, 17).

Socioeconomic information was limited in this study. Aggression and suicide severity were assessed after the fact on the basis of the extent of aggressive behavior (verbal and physical as well as destruction of property) or suicidality (verbal or behavioral) reported in the psychiatric intake history and police reports. On the other hand, as noted above, many studies have not acknowledged outbursts and aggressive behavior, focusing only on suicidality (10, 17). The lack of a consistent method of identifying outbursts and aggressive behavior limits gathering of data on this factor (12). Since this study was completed, the American Psychiatric Association’s DSM-5 committee has recently approved the use of an R code, R45.89, to be used for impairing emotional outbursts. We hope that this designation will improve data gathering in the future if clinicians use this code for children with outbursts.

Although study raters were blind to participants’ race-ethnicities, the clinicians in CPEP writing the notes were not. It is possible that implicit bias in the information was recorded in the context of police involvement. Similarly, recording of greater aggressiveness in the EMR may be used to justify police involvement. However, convergence of multiple sources of information, the greater frequency of CPEP revisits, and higher rates of hospitalization suggest that levels of behavioral severity were valid.We did not know the frequency of instances in which police were called but no ER referral was made, because local precincts did not have these data. Racial-ethnic and other biases may have influenced how police managed these situations.

Conclusions

This study describes the characteristics of children with outbursts of aggression and suicidal behavior that prompt families, school systems, and providers (12) to call the police, setting in motion a cascade of inpatient and outpatient treatments. In half of police arrivals to the CPEP, the treatment did not prevent a subsequent police arrival. Involving police adds an additional layer of risk for Black children, because previous research has shown that police interaction yields particularly poor health outcomes for this population (27).

Our data are consistent with reports of serious behavioral health crises among children and adolescents that are often not adequately addressed by ERs, inpatient units, and outpatient treatment. These crises thus lead to law enforcement referrals that are used disproportionately for individuals who are marginalized and publicly insured (9, 14, 15, 18, 19).

The National Suicide Hotline Designation Act of 2020, which established a dedicated crisis telephone number (988), could provide an immediate response to psychiatric emergencies, obviating the initial need for police involvement. Hoover and Bostic (28) further recommend school-based and community crisis response services, specific child and adolescent training for the 988 initiative, and development of programs for emergency psychiatric response technicians that include specialized training to help families and schools better manage mental health emergencies in the community. In addition, outbursts of aggression and suicidal behavior are often chronic problems, judging by treatment recidivism (14, 19–21, 29). These problems need better identification and sustained, effective, and long-term interventions with particular attention to underserved youths and families (30–33). We submit that these interventions are best designed by child mental health experts who have the skills to address the clinical and psychosocial factors leading to psychiatric emergency visits by children (34).