Pediatric Behavioral Health-Related EMS Encounters: A Statewide Analysis

Abstract

Objectives

Pediatric behavioral health disorders and related emergency department visits are increasing, but effects on emergency medical services (EMS) are unknown. This study’s objective was to describe the statewide epidemiology of pediatric behavioral health-related EMS encounters in Florida, including mental health and substance use.

Methods

This analysis is a retrospective study of pediatric behavioral health-related EMS encounters from Florida’s statewide EMS Tracking and Reporting Systems Database from 2011–2016. Demographic, clinical, EMS, and geographic characteristics are described. We also compared characteristics between patients who did and did not receive an acute EMS behavioral / psychiatric intervention.

Results

There were 22,254 pediatric behavioral health-related EMS encounters during the study period, one-quarter of which were noted to have suspected or confirmed ingestion / substance use. The median age was 16 and the majority of patients were female and white. A total of 946 patients (4%) had an acute EMS behavioral / psychiatric intervention. EMS scene, ED turnaround, and total EMS time were significantly longer for intervention patients. Of the 14 counties in the top quartile of percentages of intervention patients, 7 were rural, 10 did not have any hospitals with child / adolescent psychiatric services, and 7 did not have any child psychiatrists.

Conclusions

Pediatric behavioral-health related EMS encounters had a significant proportion of suspected ingestions / substance use, and we found disproportionate effects on rural agencies. Increases in EMS resource utilization (including longer EMS times) occurred in certain settings with limited behavioral health infrastructure. Those findings suggest an opportunity for community paramedicine to alleviate EMS utilization and decrease the frequency of pediatric behavioral health emergencies.

Introduction

Recent years have seen significant increases in the number of emergency department (ED) visits for patients aged 15 years and older for behavioral health disorders (including mental health and substance use).¹ One study found that adult ED visits for mental or substance use disorders increased from 28 per 1,000 ER visits in 2005 to 35 in 2011.² For youth ages 18 years and younger, the prevalence of pediatric behavioral health disorders in the United States is also rising, with attendant increases in ED utilization.^3–6 For example, annual pediatric behavioral health-related emergency department visits rose 26% between 2001–2010 from 491,000 to 619,000.⁵

Accordingly, EMS are increasingly providing psychosocial care related to behavioral health.^7–9 Some have construed use of EMS for behavioral health emergencies as a misuse of emergency resources.^8,10 Others have argued that further work should be done to optimize EMS provider clinical decision-making related to behavioral health care.^9,11 Since EMS providers increasingly provide care for adult patients with behavioral health disorders,¹² EMS agencies have pursued adult direct transport protocols to psychiatric hospitals and community paramedicine initiatives to alleviate increases in EMS resource use such as repeat calls, and in turn reduce ED utilization that might be avoidable.^13–16 Additionally, community paramedicine models are being developed that include behavioral health.¹⁰ Such health care service delivery models expand paramedic and emergency medical technician (EMT) roles to work with adults with serious mental illnesses.¹⁰ However, there is little information on pediatric behavioral health-related EMS encounters in the United States. Therefore, this study’s objective was to 1) describe the statewide epidemiology of non-critically ill pediatric behavioral health-related encounters in Florida, including mental health and substance use; 2) to describe and compare characteristics of patients who received an acute prehospital intervention versus those who did not; and 3) to identify counties with the greatest proportion of patients receiving acute EMS intervention and describe the available pediatric mental health resources, in order to explore whether there might be a possible role for community paramedicine to address any unmet behavioral health needs.

Methods

Study Design and Setting

This analysis was a retrospective observational study of pediatric patients ages 2 – 18 years with a behavioral health-related EMS encounter from 2011–2016. We selected this age range to capture the fullest picture of pediatric EMS utilization and because are infrequently diagnosed in patients less than two years of age.¹⁷ We identified patients from Florida’s EMS Tracking and Reporting System (EMSTARS) database, which contains EMS encounters from over 100 EMS agencies and includes 74% of all statewide 911 EMS patient calls during the study period.¹⁸ Most counties are served by one EMS agency.¹⁸ EMSTARS data for the study period covered 64 of Florida’s 67 counties, of which 28 counties were classified as rural by the Florida Department of Health (DOH).¹⁹ The 3 counties not included were all classified as rural.¹⁹ In Florida, each EMS agency operates under its own protocols, in contrast to other statewide systems. Accordingly, any official relationship between EMS and mental health resources, and/or EMS protocols for behavioral health encounters, vary widely. Individual agencies submit patient care reports (collected via National EMS Information Systems (NEMSIS)-compliant and state-validated ePCR software) to the Florida DOH Bureau of EMS which integrates the data into EMSTARS. EMSTARS during the study period contained Florida-specific elements (both data variables and their values) and national elements which conform to the NEMSIS-National Highway Traffic Safety Administration (NEMSIS-NHTSA) Version 2.2.1 Data Dictionary elements.²⁰ We obtained county-level data on child psychiatric resources (number of hospitals with child / adolescent psychiatric services and child psychiatric providers in 2016) from the Health Resources and Services Administration’s Area Health Resource Files (AHRF).²¹ The University of Florida and Florida DOH Institutional Review Boards approved the study (IRB201702645 and Protocol 180000U11, respectively).

Inclusion & Exclusion Criteria

We included patients with one or more behavioral health-related EMS provider primary and/or secondary impression(s), and excluded those with other provider impressions indicative of critical medical or traumatic illnesses (Table 1). The inclusion/exclusion criteria based on provider primary and/or secondary impression were made a priori while examining all potential impressions available in EMSTARS. EMS encounters without transport to a facility were excluded. Interfacility transports were excluded as the intent was to characterize primary scene calls.

Table 1:

Included and Excluded EMS Provider Primary and Secondary Impressions^*

Inclusion	Exclusion
Any primary or secondary impression of:	Any primary or secondary impression of:
Behavioral / psychiatric disorder	Cardiac arrest
Intentional drug use	Airway obstruction
Alcohol related problems / Delirium tremens (DTs)	Vaginal hemorrhage
	Pregnancy / OB delivery
	Hypovolemia / shock
	Obvious Death
	Respiratory arrest
	Fever related problems / symptoms
	Heat related illness
	Sepsis
	Sickle cell crisis
	Stroke
	Electrocution

^*All of the listed phrases in both inclusion and exclusion criteria are values for EMS provider primary and secondary impression in EMSTARS

EMS = emergency medical services, EMSTARS = EMS tracking and reporting system

Data Variables

We abstracted demographic, clinical, EMS, and geographic characteristics. EMSTARS contains a unique patient identifier for each encounter, which allows for unique patient encounter identification even if multiple EMS units respond to the same scene. Each vital sign’s first recorded measurement was used (as many EMS decisions are made based on the initial assessment).²² EMS times were calculated as minute intervals. We considered negative time intervals as miscoded and those were excluded. Other variables were abstracted directly from EMSTARS. After examining all the EMS medications and procedures administered to the study sample patients, we characterized patients by those who had one or more of certain a posteriori-defined acute behavioral / psychiatric intervention(s) by EMS: administration of activated charcoal, diazepam, haloperidol, lorazepam, midazolam, or naloxone, or the procedures: decontamination (referring to the patient), psychological first aid, restraints – pharmacological, and/or restraints – physical.

Data Analyses

We report descriptive statistics and compared characteristics between patients who did and did not receive an acute behavioral / psychiatric intervention (“intervention”). We performed an analysis specifically examining the distribution of intervention and non-intervention patients by age. Continuous variables were compared using unpaired t-test and Wilcoxon Rank Sum tests, as appropriate. Categorical variables were compared using the Chi-Square test, or Fisher’s Exact Test when there were less than 10 observations (a conservative approach).²³ The Kolmogorov-Smirnov test evaluated whether variables were normally distributed (sample size greater than 2,000).²⁴

We conducted geographic analyses classifying patients by the treating EMS agency’s home county. We first classified counties as urban or rural (rural being 100 persons or less per square mile).¹⁹ For each county, we calculated the percent of patients who received one or more acute behavioral / psychiatric EMS interventions. Since percent by county was not normally distributed, we divided counties into quartiles by the percent of patients who received an acute intervention. Next, we described the behavioral health resources available for each county and also by each quartile using the AHRF’s 2016 county-level data.

Missing data were excluded by individual variable (e.g., for a patient encounter missing ethnicity data, that encounter was excluded from ethnicity analysis only, but the patient encounter was still included for other analyses). We excluded missing data since we were not able to determine whether EMSTARS’ data were missing at random due to its aggregation of data from multiple heterogeneous agencies.¹⁸ Statistical analysis was performed using SAS^® version 9.4 (Cary, NC). Geospatial analysis was performed using ArcGIS Desktop 10.4.1 (Redlands, CA).

Results

From 2011 to 2016, there were 3,491,241 patients of all ages in EMSTARS and 388,187 patients ages 2–18 transported by EMS, of which 22,254 were pediatric behavioral health-related EMS encounters (Figure 1). Most patients (90%, n=20,023) had at least one provider impression of behavioral / psychiatric disorder. One-quarter of patients (25%, n=5,670) had at least one impression of poisoning / drug ingestion, intentional drug use, or alcohol-related problems. The top three provider primary impressions were “behavioral / psychiatric disorder” (n=13,739, 62%), “intentional drug use; related problems” (n=1,974, 9%), and “alcohol related problems / delirium tremens” (n=1,020, 5%). There were 981 acute behavioral / psychiatric interventions by EMS in 946 patients (4% of total patients, Table 2). Of those 946 patients, 911 had only one intervention and 35 patients had 2 interventions.

Figure 1:

Study Patient Selection

Table 2:

Number of Acute EMS Behavioral / Psychiatric Interventions, Total N=981 for N=946 patients

Intervention	N (%)
EMS Administered Medication
Activated Charcoal	23 (2%)
Diazepam	53 (6%)
Haloperidol	71 (8%)
Lorazepam	55 (6%)
Midazolam	110 (12%)
Naloxone	187 (20%)
EMS Procedure
Decontamination	1 (<1%)
Psychological First Aid	227 (24%)
Restraints – Pharmacologic	8 (1%)
Restraints – Physical	246 (26%)

Patient and EMS characteristics, by those who received one or more acute intervention(s) versus those who received no acute behavioral / psychiatric interventions are displayed in Tables 3a and 3b, respectively. For both groups, the median age was 16 and the majority were female and white. Intervention patients had significantly higher heart and respiratory rates, and decreased total Glasgow Coma Scores (GCS) and level of alertness (all p<0.0001). Missing EMSTARS data is also documented in Tables 3a and 3b, with race, ethnicity, respiratory effort, and level of alertness having the highest percentage of missing values. Prehospital blood glucose levels were missing in 60% of patients, which precluded further meaningful analysis between intervention and non-intervention patients. However, of the 3,847 patients with a GCS less than 15, 3,370 (88%) did not have a blood glucose level recorded.

Table 3a:

Patient characteristics of pediatric behavioral health-related EMS encounters in Florida between 2011–2016, by those who did and did not receive an acute behavioral / psychiatric intervention

	Intervention (N = 946)	No Intervention (N = 21,308)	P-Value	Missing
Demographics
#Age (Years), [Median (IQR)]	16 (14–18)	16 (14–17)	<0.0001	0 (0%)
Female Gender	499 (53%)	13,340 (63%)	<0.0001	14 (<1%)
Race			0.0021	1928 (9%)
White	472 (53%)	11078 (57%)
Black or African American	236 (27%)	5356 (28%)
Other	169 (19%)	2817 (15%)
Asian	10 (1%)	130 (<1%)
American Indian or Alaskan Native	3 (<1%)	34 (<1%)
Native Hawaiian or other Pacific Islander	0 (0%)	21 (<1%)
Ethnicity			0.1137	4334 (20%)
Non-Hispanic or Latino	650 (79%)	13881 (81%)
Hispanic or Latino	173 (21%)	3216 (19%)
Clinical Characteristics
DBP, [mean (SD)]	78 (17)	78 (15)	0.927	949 (4%)
SBP, [mean (SD)]	125.7 (22)	125.9 (19)	0.826	761 (3%)
Heart Rate (beats per minute), [mean (SD)]	108.2 (26)	100.9 (23)	<0.0001	443 (2%)
Respiratory Rate (breaths per minute), [median (IQR)]	20 (16–24)	18 (16–22)	<0.0001	696 (3%)
Respiratory Effort			0.0109	5376 (24%)
Normal	746 (94%)	15489 (96%)
Labored	28 (4%)	434 (3%)
Fatigued	16 (2%)	148 (1%)
Absent	1 (<1%)	16 (<1%)
#Pulse Oximetry, [median (IQR)]	99 (97–100)	99 (98–100)	<0.0001	1884 (9%)
#Glasgow Coma Score, [median (IQR)]	15 (13–15)	15 (15–15)	<0.0001	1182 (5%)
Level of Alertness			<0.0001	3197 (14%)
Alert	655 (82%)	17071 (94%)
Verbal	54 (7%)	647 (4%)
Painful	57 (7%)	378 (2%)
Unresponsive	30 (4%)	165 (1%)
Condition of Patient at Destination			<0.0001	872 (4%)
Improved	467 (50%)	6176 (30%)
Unchanged	463 (50%)	14226 (70%)
Worse	5 (<1%)	45 (<1%)

IQR = interquartile range, SD = standard deviation

^#P-value significant due to large sample size and differences in distributions (not median). For age, the third quartile is different between intervention and non-intervention patients. For pulse oximetry, the first quartile is different between intervention and non-intervention patients. For Glasgow Coma Score, the first and second quartiles are different between intervention and non-intervention patients.

Table 3b:

EMS characteristics of pediatric behavioral health-related EMS encounters in Florida between 2011–2016, by those who did and did not receive an acute behavioral / psychiatric intervention

	Intervention (N = 946)	No Intervention (N = 21,308)	P-Value	Missing
EMS Characteristics
Response Mode			<0.0001	0 (0%)
Lights & Sirens	751 (79%)	15464 (73%)
No Lights or Sirens	137 (15%)	3658 (17%)
Lights & Sirens then downgrade	48 (5%)	1952 (9%)
No Lights or Sirens then upgrade	10 (1%)	234 (1%)
Transport Mode			<0.0001	18 (<1%)
No Lights or Sirens	521 (55%)	13744 (65%)
Lights & Sirens	398 (42%)	6050 (28%)
Lights & Sirens then downgrade	20 (2%)	1313 (6%)
No Lights or Sirens then upgrade	7 (<1%)	183 (1%)
Incident Location			<0.0001	912 (4%)
Home/Residence	541 (59%)	10116 (50%)
Health Care Facility	39 (4%)	2000 (10%)
Public Building	58 (6%)	2094 (10%)
Street or Highway	111 (12%)	2697 (13%)
Business	86 (9%)	1614 (8%)
Place of Recreation	18 (2%)	410 (2%)
Residential Institution^	13 (1%)	431 (2%)
Other	55 (6%)	1059 (5%)
Reason for Choosing Destination			<0.0001	520 (2%)
Closest Facility	550 (59%)	10322 (50%)
Patient/Family Choice	207 (22%)	6251 (30%)
Law Enforcement Choice	43 (5%)	739 (4%)
Protocol	61 (7%)	1495 (7%)
Specialty Resource Center	56 (6%)	1403 (7%)
Patients Physicians Choice	7 (<1%)	279 (1%)
Other	9 (1%)	312 (2%)
#Response time (minutes), [median (IQR)]	7 (5–10)	7 (5–10)	0.2506	39 (<1%)
#Scene time (minutes), [median (IQR)]	16 (12–21)	13 (10–18)	<0.0001	47 (<1%)
#Transport time (minutes), [median (IQR)]	13 (9–21)	13.3 (9–20)	0.5149	17 (<1%)
#ED turnaround time (minutes), [median (IQR)]	25 (18–34)	19.8 (14–27)	<0.0001	39 (<1%)
#Total EMS time (minutes), [median (IQR)]	66 (53–80)	58 (46–72)	<0.0001	26 (<1%)

EMS = emergency medical services, IQR = interquartile range, SD = standard deviation

^{^}Nursing home, jail, prison, juvenile detention, etc.

^#Response time = Time from Dispatch notification of EMS unit to time of EMS unit arrival on scene; Scene time = time from EMS arrival on scene to when EMS left the scene; Transport time = time from EMS leaving the scene to ED arrival, ED turnaround time = time from ED arrival to when EMS unit marked themselves as back in service; Total EMS time = time from dispatch notification to time when EMS unit marked themselves as back in service

With respect to EMS characteristics, a higher percentage of behavioral / psychiatric intervention patients were transported with lights and sirens (Table 3b, p<0.0001). Intervention patients had significantly longer scene times (median 16 vs 13 minutes), ED turnaround times (median 25 vs 20 minutes), and total EMS encounter times (median 66 vs 58 minutes) (all p<0.0001). The a priori-excluded negative time intervals comprised less than 1% of all times. More patients with an intervention were noted to be improved clinically by the EMS provider at the time of ED arrival (p<0.0001). Aside from those medications which defined an intervention, both groups received few other medications aside from oxygen (8% intervention, 13% non-intervention) and fluids (<1% intervention, 6% non-intervention). Intravenous (IV) access was low and did not differ between groups (5% in both groups). When patients were divided into age categories reflective of different developmental and school stages, the proportion of patients receiving an acute behavioral / psychiatric intervention varied significantly amongst categories (p<0.0001) (Figure 2).

Figure 2:

Percent Study and Intervention Patients by Age Category

The EMS agencies’ county data were available for all but 71 patients (<1% of total). There were no encounters in 3 of Florida’s 67 counties (all rural). As between urban and rural counties, there were no significant difference in response time, scene time, and transport time (all p>0.05). Notably, rural counties had shorter ED turnaround times and total EMS times (both p<0.0001). The median county percentage of acute intervention patients was 2%, with an interquartile range of 0% – 7%, and not normally distributed (Shapiro-Wilk w value 0.8, p-value <0.0001). Figure 3 maps county-level intervention quartiles and number of hospitals with child / adolescent psychiatric services. Of the 14 counties in the highest intervention percentage quartile, 7 were rural, and 7 were urban. Of the 7 rural counties, 6 did not have any hospitals with child / adolescent psychiatric services, and 5 did not have any child psychiatrists in the AHRF database. Of the 7 urban counties, 4 did not have any hospitals with child / adolescent psychiatric services, and 2 did not have any child psychiatrists.

Figure 3.

Pediatric Mental Health Resources and Proportion of Pediatric EMS Behavioral Health-Related Encounters that Received an Acute Behavioral / Psychiatric Intervention by County from 2011–2016.

Discussion

This study describes the epidemiology of non-critically ill pediatric behavioral health-related EMS encounters in the state with the fourth-largest pediatric population.²⁵ The vast majority of study and intervention patients were adolescents (median age 16 years), although the thousands of younger patients who met inclusion criteria merit future study. Females had a higher proportion of encounters than males. That finding is consistent with pediatric studies showing a higher female prevalence of mental health disorders such as anxiety and depression, and a higher proportion of psychiatric hospitalizations for suicidal ideation or self-harm.^3,4

One-quarter of patients had EMS provider impressions indicating confirmed or suspected ingestion / substance use. Given the 187 administrations of naloxone, some are likely pediatric opioid overdoses. The large proportion of suspected or confirmed ingestions in this study likely decreases the yield and safety of a pediatric psychiatric direct transport protocol.^13–16 However, certain other non-ingestion encounters may be appropriate for direct transport to psychiatric facilities.

The large proportion of substance use when paired with 62% of patients missing prehospital blood glucose measurements (including 88% of patients with a GCS less than 15) prompts concern for unrecognized hypoglycemia. Low rates of blood glucose testing may be due to EMS provider safety concerns with potentially aggressive and unpredictable patients. However, EMS education should encourage blood glucose testing whenever feasible, as hypoglycemia may cause or exacerbate patient agitation.

Regarding our geographic objective of identifying counties with disproportionate numbers of intervention patients, half of the counties with the highest percentage of acute behavioral / psychiatric intervention patients (fourth quartile) were rural. 5 of those 7 rural counties had neither a hospital that offered child psychiatric services nor a child psychiatrist. Behavioral health workforce shortages have disproportionately impacted rural counties.²⁶ Limitations with existing behavioral health infrastructure combined with increased interventions may be problematic for rural counties where adequate EMS workforce staffing is already strained.²⁷ EMS scene time, ED turnaround time, and total EMS time were all significantly longer for pediatric patients with acute behavioral / psychiatric interventions. Longer EMS times decrease available units for the next dispatch call. The 8-minute increase in median total EMS times for intervention patients may or may not be operationally significant depending on the EMS agency. Another explanation for interventions by rural EMS agencies may be safety concerns for longer transports, however there was not a significant difference in EMS transport time between intervention and non-intervention patients, as well as between urban and rural counties. This still may reflect concerns EMS providers have regarding patient and crew safety for transports in counties where there are few or no behavioral health services that have potentially engaged or counseled the EMS agency on behavioral health training. Alternatively, the percentage of interventions in rural counties may reflect poorly controlled behavioral health disorders due to a lack of community behavioral health resources.

The high proportion of substance use/ingestions, the impact on EMS encounter times - particularly for patients requiring interventions, and our analysis showing that some counties with high acute intervention rates did not have access to many hospital-based resources suggests that community paramedicine may provide a potential solution. Possible roles for EMS providers under a community paramedicine initiative may include patients and families counseling on the benefits and logistics of outpatient psychiatric care, assisting with care coordination, and identifying available community behavioral health resources.^10,28 Those interventions may reduce EMS utilization through the decrease in avoidable and/or repeat EMS pediatric behavioral health-related encounters. However, research involving community paramedicine in this role is needed.

Whether urban or rural, this study’s findings stress the need to train EMS providers to provide pediatric-specific behavioral health care during acute encounters or as part of nascent community paramedicine programs. Local public health and EMS leadership in areas with high or disproportionate levels of behavioral health-related EMS encounters should examine the utility of direct transport protocols, community paramedicine, and care coordination with the (decreasing) number of pediatric mental health providers.²⁹ Future research can assess how such programs reduce EMS resource utilization and avoidable pediatric EMS behavioral health-related encounters.

Limitations

This study is limited by its retrospective design and its setting is limited to one state, although Florida has the fourth-largest pediatric population in the United States.²⁵ Missing data varied greatly by variable and should be taken into consideration when interpreting results. EMSTARS collects data from individual agencies that use different data capture systems, and therefore does not have the same degree of data capture, quality, and “cleaning” as does a registry.¹⁸ EMSTARS is a population-based dataset with voluntary participation (albeit covering 64 of 67 counties), and is therefore subject to both selection and information bias.¹⁸ EMSTARS contains both state-specific and NEMSIS-compliant data, therefore not every variable and/or value (including EMS interventions such as “psychological first aid”) may apply in other settings and EMS systems.^20,18 Additionally, our definitions of a behavioral health-related encounter and acute behavioral / psychiatric intervention were made using clinical judgement. There may be relevant EMS encounters or interventions not captured by those definitions. Further to this, we chose to exclude acutely ill patients, which may not capture patients suffering critical medical illnesses or trauma secondary to a behavioral/psychiatric problem. For example, our selection criteria would exclude a patient with a cardiopulmonary arrest due to an opioid overdose. However, the intention of this study was to characterize non-critically ill behavioral health EMS patients. We were unable to ascertain the timing of administered medications or procedures with respect to with each patient’s first documented vital signs, and therefore we are unable to present whether or not any abnormal vital signs were before or after an EMS intervention. ED outcomes were not available, limiting assessment of the results of EMS interventions and each patient encounter. While we used the most recent AHRF data available to characterize psychiatric resources, it is possible that new facilities were opened during or after the study period that would not have been operative for the entire study period.²¹

Conclusion

Pediatric behavioral-health related EMS encounters have a significant proportion of suspected ingestions / substance use, and disproportionate effects on rural agencies. Increases in EMS resource utilization are occurring in settings with limited behavioral health infrastructure. Those findings may represent an opportunity for direct transport protocols, community paramedicine, and/or coordination with local or regional behavioral health resources to decrease the frequency of EMS utilization and pediatric behavioral health emergencies for patients and families.