Predictors of emergency department use by adolescents and adults with autism spectrum disorder: a prospective cohort study.

OBJECTIVES: To determine predictors of emergency department (ED) visits in a cohort of adolescents and adults with autism spectrum disorder (ASD).
DESIGN: Prospective cohort study.
SETTING: Community-based study from Ontario, Canada. PARTICIPANTS: Parents reported on their adult sons and daughters with ASD living in the community (n=284). MAIN OUTCOME MEASURES: ED visits for any reason, ED visits for medical reasons and ED visits for psychiatric reasons over 1 year.
RESULTS: Among individuals with ASD, those with ED visits for any reason were reported to have greater family distress at baseline (p<0.01), a history of visiting the ED during the year prior (p<0.01) and experienced two or more negative life events at baseline (p<0.05) as compared with those who did not visit the ED. Unique predictors of medical versus psychiatric ED visits emerged. Low neighbourhood income (p<0.01) and living in a rural neighbourhood (p<0.05) were associated with medical but not psychiatric ED visits, whereas a history of aggression (p<0.05) as well as being from an immigrant family (p<0.05) predicted psychiatric but not medical emergencies.
CONCLUSIONS: A combination of individual and contextual variables may be important for targeting preventative community-based supports for individuals with ASD and their families. In particular, attention should be paid to how caregiver supports, integrative crisis planning and community-based services may assist in preventing or minimising ED use for this vulnerable population. © Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted.

This is one of the first published studies to date to compare the profiles of individuals with autism spectrum disorder who did and did not visit the emergency department.

A convenience sample in a setting with universal health insurance was used and may be limited in its generalisability.

Data were based on parent self-report rather than on chart abstraction and so may be susceptible to biases.

Introduction

Across childhood, adolescence and adulthood, individuals with autism spectrum disorder (ASD) are more likely to visit the emergency department (ED) than their typically developing peers for both medical and psychiatric reasons.1–6 The high-intensity environment of an ED visit can be extremely challenging for individuals with ASD, who may present with sensory, social communication and behavioural difficulties.7 Perhaps representative of these challenges, individuals with ASD are restrained or sedated in approximately one in four ED visits.8 ASD knowledge gaps and inadequate professional training often compound these negative experiences for affected individuals and their family caregivers.9 10 From a systemic perspective, frequent ED use is costly and ED overcrowding and long wait times are key system level challenges in most jurisdictions.11 12 The number of ED visits by individuals with ASD is increasing,5 and it is important to better understand predictors of ED visits among individuals with ASD to assist in appropriate provision of community-based care.

Individuals with ASD often have complex care needs and can visit the ED for a host of psychiatric and medical concerns. In addition to an increased risk of developing co-occurring mental health concerns such as anxiety and depression,13 14 youth and adults with ASD often display problem behaviours, including elopement, aggression and self-harming behaviours,15–17 which frequently necessitate emergency care.2 3 8 Gastrointestinal issues (eg, abdominal pain, nausea and vomiting) and neurological symptoms and disorders (eg, seizures and epilepsy) are common comorbid medical concerns among individuals with ASD and account for a large proportion of medical related ED visits in this population.1 2 5 18 19 Youth with ASD may also be at high risk for ED visits due to accidents and injury.20–22 Medical issues can be further complicated by atypical clinical presentations, such that problem behaviours can sometimes mask underlying medical conditions.23 24

In recent years, several studies have compared ED patterns in those with and without ASD1–6 ; however, we know very little about what differentiates individuals with ASD who use the ED compared with those with ASD who do not use the ED. Additionally, the majority of what is known about emergency service use patterns in individuals with ASD comes from the analysis of administrative data. This type of analysis is limited in its ability to describe situational and family centred variables that are not typically documented in insurance claims or medical records. Our previous retrospective, cross-sectional study on ED use in ASD8 demonstrated that factors such as previous ED visits, lack of meaningful day time activities and a history of hurting others were related to emergency service use, whereas demographic variables were not. This study, however, was limited in that it did not explore how these variables impact ED use over a longer period of time.

The current study investigates predictors of ED use in adolescents and adults with ASD based on parent reported data. As a secondary exploratory aim, we sought to separately examine predictors of medically related ED visits and predictors of psychiatric related ED visits to determine if they were distinct. For these analyses, a modified version of Andersen’s Behavioural Model of Health Services Use25 was used to group predictors into three categories: (1) predisposing factors (ie, static demographic characteristics that describe the likelihood an individual will seek healthcare in a given situation), (2) enabling factors (ie, environmental characteristics and available resources) and (3) need variables (ie, events or medical/psychological conditions that might facilitate a greater need for services).

Methods

Participants

Parents of adolescents and adults with ASD age 12 years and up were recruited from across Ontario, Canada, between December 2010 and October 2012. Parents were recruited from flyers and websites of collaborating community agencies, recreation programs and schools that support individuals with ASD, as well as through every local chapter of the provincial ASD organisation. To be eligible for the study, individuals with ASD needed (1) to be living in Ontario, (2) to be 12 years of age or older, (3) to have a clinical diagnosis of ASD and (4) to exceed the recommended cut-off score of 12 on the Social Communication Questionnaire–Lifetime Version (SCQ),26 27 a standardised ASD screening tool. A total of 514 families consented to participate in the study and 462 completed the baseline survey; 396 of these participants met the age and SCQ cut-offs. Families were contacted every 2 months to report on their ED use and were contacted approximately 1 year after baseline to fill out a final survey. To be included in the current study, parents were required to complete the final survey within 18 months of their baseline date. The final sample for the current study consisted of 284 individuals with ASD (mean age 18.41 years, SD 6.16; 223 males, 61 females). There were no significant differences between the final sample (n=284) and those who did not complete the final survey within the required time period (n=112) with respect to most individual and parent demographics (individual age, individual gender, individual place of residence, parent age, parent marital status, parent education and neighbourhood income) or clinical needs (individual ASD severity, individual ID status, individual medical and psychiatric diagnoses; all p>0.05). However, a significantly greater proportion of non-responders identified as non-Caucasian (26.6% vs 17.0%, p=0.03).

The majority of survey respondents were mothers (93.7%), and nearly half reported that their dependent with ASD was 18 years or older (44.7%). Most individuals with ASD lived with family (91.2%) at the time of baseline survey completion, and the remaining were reported to reside in group homes or other supported living situations (6.4%), in treatment facilities (0.7%) or independently (1.8%).

Predictor variables

All predictor variables were collected as part of the baseline survey.

Predisposing variables

Individual predisposing variables included age, gender, intellectual disability (ID) status and ASD severity. ID status was split into those who had an ID (ie, mild to profound ID) and those who did not (ie, gifted, normal and borderline intelligence). ASD severity was measured using the SCQ,26 with higher scores reflective of more severe symptomology. The SCQ is reported to have good internal consistency (alpha coefficients 0.81–0.93); and external validity.27 In the current study, internal consistency for the overall scale was strong (Cronbach’s alpha=0.81).

Enabling variables

Parent enabling variables included marital status (married/living with partner or single), education level and if they were born in Canada (yes/no). Parent education level was dichotomised into high education (college diploma or higher) or low education (some college or university education or lower). Household income of each family was estimated based on the median income of the respondent’s postal code region according to census data28 (range $33 030.00–$107 742.00, mean $65 597.12, SD $15 970.27). Since the neighbourhood income in our sample was higher than the national mean ($53 634), we compared families whose neighbourhood income fell into the bottom quartile with those that fell in the top three quartiles of our sample. Postal codes were also used to determine rural status (rural vs urban) based on Statistics Canada’s definition of rurality.29 Individuals were classified as either having or not having ‘structured’ daytime activities based on parents’ responses to the question “What does your child do during his/her typical day? (ie, school, working, volunteering or job training, day programme, no structured daytime activity or other)”. Family distress was measured at baseline using the 10-point Brief Family Distress Scale,30 which asks parents to indicate their level of distress on a scale from 1 ‘everything is fine we are not in crisis at all’ to 10 ‘we are currently in crisis, and it could not get any worse’.

Need variables

Parents were asked to provide current medical (eg, gastrointestinal and skin conditions) and psychiatric diagnoses (eg, anxiety and depression) for their child with ASD. Comorbid medical and psychiatric diagnoses were each dichotomised as has been done in many large-scale parent report surveys.31 32 Parents indicated (yes/no) whether their child had a history of aggression and whether they visited the ED in the year prior to the baseline survey.

Parents also completed a modified version of the negative life events checklist from the Psychiatric Assessment for Adults with Developmental Disabilities Checklist (PAS-ADD33), in which they were to check off any of the significant life events that their child experienced in the 2 months prior to baseline (eg, death of a family member, change in residence and loss of a job). For this analysis, the ‘injury to self’ item was excluded as injuries may directly cause an ED visit. The checklist was dichotomised into individuals who had two or more negative life events prior to beginning the survey and those who had fewer than two, similar to past studies using this measure.34 35

Dependent variables

Primary outcome variable (ED visit)

In each of the follow-up surveys, parents were asked to indicate (yes/no) whether their child used the ED in the last 2 months. Parents were also asked in the final survey about whether their child had any ED visits in the previous year. Individuals with ASD were dichotomised into individuals with and without any ED visit during the past year. To ensure that the sampling period for visits was standardised across participants, only ED visits during the year that preceded participants’ final survey were included. All participants were therefore sampled for 12 months.

Secondary outcome variables (medical or psychiatric ED visits)

Parent descriptions of visits were coded independently by two authors (MP-S and AT) as being either medical or psychiatric in nature with 98% inter-rater agreement. Any disagreements were resolved through discussion with the senior author (YL). If parents provided detailed descriptions of more than one ED visit, each visit was coded separately. Thus, the same individual could be classified as having both a medical emergency and a psychiatric emergency (n=3). Some parents failed to provide enough descriptive information about emergencies in their bimonthly form for these emergencies to be coded (n=17), so these cases were excluded from the secondary analysis of emergency type.

Procedure

The York University and Centre for Addiction and Mental Health research ethics boards approved this research. Consenting parents were sent their baseline survey electronically or on paper. The majority of participants (n=223; 78.5%) completed online surveys. Parents who completed the surveys online (mean 48.67 years, SD 6.88) were significantly younger than parents who completed the surveys via mail or telephone interview (n=61; mean 52.36 years, SD 8.44, p=0.001); no other parent or individual demographic variables differed according to survey response method. Participants had a unique identifier that was linked to their information, and every 2 months following the completion of their last survey, they received another follow-up survey. At their final time point, they received a longer measure to complete. Participants received payment ($20) on completing baseline information and subsequent payment ($20) when they completed their final measure.

Data analysis

A bivariate analysis compared the groups with and without ED visits using single predictor logistic regression models. From these models, we derived ORs to demonstrate the relationship between each predictor variable and the likelihood of having any ED use. To identify the relative influence of these variables, we then ran two adjusted logistic regression explanatory models predicting any ED use. The first included all the variables that had a p value of 0.20 or lower in the bivariate analysis, as well as the age variable that did not meet this threshold but was included since it is a potential confounder. The second adjusted logistic regression model included all predictor variables.

Subsequently, we examined if different variables were associated with medical ED visits compared with psychiatric ED visits. To do this, we again ran single predictor logistic regression models on two different subsets of the sample. First, these models were used to compare individuals with any medical ED visits with individuals with no medical ED visits. Second, we used these models to compare individuals with any psychiatric ED use relative to individuals with no psychiatric ED use. Adjusted logistic regressions could not be conducted as part of this analysis due to the small number who presented with each type of visit.

For all of the models, individuals with missing data for ID status (n=31) were included, but individuals with missing data for parental education (n=2) were excluded since it was not sufficiently large to be included as its own category. Analyses were conducted using SPSS version 24.

Results

ED use

Of the cohort of 284 individuals, 63 (22.2%) reported ED use during the year following their baseline data collection and 221 (77.8%) reported no ED use during that year. Participant demographic and clinical characteristics according to ED outcome are shown in table 1.

Table 1

Demographic and clinical characteristics of 284 youth and adults with ASD living in Ontario with and without ED visits

	Total sample (N=284)	Visited ED (n=63)	Did not visit ED (n=221)
	n (%)	n (%)	n (%)
Predisposing
Age
Mean (SD)	18.4 (6.2)	18.8 (5.9)	18.3 (6.3)
Gender
Female	61 (21.5)	19 (30.2)	42 (19.0)
Male	223 (78.5)	44 (69.8)	179 (81.0)
SCQ score
Mean (SD)	22.6 (6.3)	22.5 (6.7)	22.6 (6.2)
ID
ID	132 (46.5)	29 (46.0)	103 (46.6)
No ID	121 (42.6)	30 (47.6)	91 (41.2)
Missing	31 (10.9)	4 (6.3)	27 (12.2)
Enabling
Parent marital status
Unmarried	67 (23.6)	18 (28.6)	49 (22.2)
Married/living with partner	217 (76.4)	45 (71.4)	172 (77.8)
Parent immigrated
Yes	67 (23.6)	15 (23.8)	52 (23.5)
No	217 (76.4)	48 (76.2)	169 (76.5)
Parental education
High education	211 (74.3)	48 (76.2)	163 (73.8)
Low education	71 (25.0)	56 (23.8)	56 (25.3)
Missing	2 (0.7)	0	2 (0.9)
Neighbourhood income
Low	59 (20.8)	18 (28.6)	41 (18.6)
High	225 (79.2)	45 (71.4)	180 (81.4)
Daytime activity
Unstructured	39 (13.7)	13 (20.6)	26 (11.8)
Structured	245 (86.3)	50 (79.4)	195 (88.2)
Family distress
Mean (SD)	4.2 (1.8)	5.0 (1.9)	3.9 (1.7)
Urban-rural residence
Urban	254 (89.4)	52 (82.5)	202 (91.4)
Rural	30 (10.5)	11 (17.5)	19 (8.6)
Need
Medical comorbidity
Yes	98 (34.5)	23 (36.5)	75 (33.9)
No	186 (65.5)	40 (63.5)	146 (66.1)
Psychiatric comorbidity
Yes	132 (46.5)	36 (57.1)	96 (43.4)
No	152 (53.5)	27 (42.9)	125 (56.6)
On medication
Yes	162 (57.0)	46 (73.0)	116 (52.5)
No	122 (43.0)	17 (27.0)	105 (47.5)
History of aggression
Yes	180 (63.4)	42 (66.7)	138 (62.4)
No	104 (36.6)	21 (33.3)	83 (37.6)
ED visit in past year
Yes	67 (23.6)	30 (47.6)	37 (16.7)
No	217 (76.2)	33 (52.4)	184 (83.3)
2+ Negative life events
Yes	68 (23.9)	26 (41.3)	42 (19.0)
No	216 (76.9)	37 (58.7)	179 (81.0)

ASD, autism spectrum disorder; ED, emergency department; ID, intellectual disability; SCQ, Social Communication Questionnaire.

In the bivariate analysis, higher family distress, living in a rural area, being on medication, having an ED visit in the past year and having two or more negative life events were significant at the 0.05 level. In addition, gender, neighbourhood income, daytime activity and psychiatric comorbidity had p values larger than 0.05 but less than 0.20 (table 2). Despite the two adjusted models, including different predictors (the first included predictors with a p value of less than 0.20 in the bivariate analysis, and the second model included all predictor variables), both models yielded similar results. In both models, clinical need variables (having an ED visit in the year prior to the study and reporting two or more significant life events at baseline) and one enabling variable (family distress) predicted having any ED use, when controlling for other variables (see table 2).

Table 2

ORs and 95% CIs from bivariate and multivariate analyses predicting ED visits among 284 youth and adults with ASD living in Ontario

	Bivariate analysis	Significant predictor only +age model	Fully adjusted model
	OR (95% CI)*	OR (95% CI)†	OR (95% CI‡
Predisposing
Age (years)	1.0 (1.0 to 1.1)	1.0 (1.0 to 1.1)	1.0 (0.9 to 1.1)
Gender
Female	1.8 (1.0 to 3.5)§	1.8 (0.9 to 3.7)§	1.8 (0.9 to 3.8)§
Male	1.0	1.0	1.0
SCQ score	1.0 (0.9 to 1.0)		1.0 (0.9 to 1.0)
ID status
ID (mild to profound ID)	1.0		1.0
No ID	1.2 (0.7 to 2.1)		1.1 (0.5 to 2.3)
Missing	0.5 (0.2 to 1.6)		0.5 (0.1 to 1.7)
Enabling
Parent marital status
Unmarried	1.4 (0.8 to 2.6)		1.2 (0.5 to 2.5)
Married/living with partner	1.0		1.0
Parent immigrated
Yes	1.0 (0.5 to 2.0)		1.2 (0.6 to 2.6)
No	1.0		1.0
Parental education
High education	1.1 (0.6 to 2.1)		1.4 (0.6 to 2.9)
Low education	1.0		1.0
Neighbourhood income
Low	1.8 (0.9 to 3.3)§	1.8 (0.9 to 3.8)§	1.7 (0.8 to 3.6)§
High	1.0	1.0	1.0
Daytime activity
Unstructured	2.0 (0.9 to 4.1)§	1.1 (0.4 to 2.9)	1.1 (0.4 to 3.2)
Structured	1.0	1.0	1.0
Family distress
Urban-rural residence	1.4 (1.2 to 1.6)***	1.3 (1.1 to 1.5)**	1.3 (1.1 to 1.5)**
Urban	1.0	1.0	1.0
Rural	2.3 (1.0 to 5.0)*	2.0 (0.8 to 4.9)§	2.1 (0.8 to 5.4)§
Need
Medical comorbidity
Yes	1.1 (0.6 to 2.0)		0.8 (0.4 to 1.7)
No	1.0		1.0
Psychiatric comorbidity
Yes	1.7 (1.0 to 3.1)§	1.0 (0.5 to 2.0)	0.9 (0.4 to 1.9)
No	1.0	1.0	1.0
On medication
Yes	2.5 (1.3 to 4.5)**	1.4 (0.7 to 2.9)	1.6 (0.8 to 3.5)§
No	1.0	1.0	1.0
History of aggression
Yes	1.2 (0.7 to 2.2)		0.8 (0.4 to 1.7)
No	1.0		1.0
ED visit in past year
Yes	4.5 (2.5 to 8.3)***	2.7 (1.4 to 5.5)**	2.9 (1.4 to 5.9)**
No	1.0	1.0	1.0
2+ Negative life events
Yes	3.0 (1.6 to 5.5) ***	2.5 (1.2 to 5.2)*	2.5 (1.2 to 5.3)*
No	1.0	1.0	1.0

*Derived from single predictor models.

†Derived from a model that included predictors with a p value of 0.20 or lower in the bivariate analysis: age; gender; neighbourhood income, typical day; family distress, urban-rural residence; psychiatric comorbidity; ED visit in past year; 2+ negative life events.

‡Derived from a model that included all predictors, n=282 due to missing data.

§p<0.20, *p<0.05,** p<0.01, ***p<0.001.

ASD, autism spectrum disorder; ED, emergency department; ID , intellectual disability; SCQ, Social Communication Questionnaire.

Medical and psychiatric ED visits

Forty-six of the 63 parents who reported that their child visited the ED provided a detailed description of at least one such visit. Three parents described separate psychiatric and medical visits. For the purposes of the following analyses, we included only the first medical and the first psychiatric visit described by each parent for a total of 49 descriptions of ED visits. Of the 49 described ED visits, 31 visits related to medical issues and 18 visits related to psychiatric issues. The most common type of medical visit was injury (n=10), followed by gastrointestinal issues (n=4) and infections (n=4). The most common reason for psychiatric ED visits was related to self-harm and/or suicidality (n=5) followed by physical assault towards family members (n=3) and three incidents where parents described individuals as both a danger to self and others.

As shown in table 3, bivariate analyses indicate that medical ED visits were associated with living in a lower-income neighbourhood, living in a rural area and family distress at baseline (enabling factors), along with two or more negative life events, being on medication at baseline and visiting the ED in the year prior to data collection (need factors). Two enabling factors (parent immigrant status and family distress) predicted having a psychiatric ED visit in the follow-up period (see table 3). Clinical need variables, such as being on medication, having a history of aggression and having two or more negative life events were also significantly associated with psychiatric emergencies.

Table 3

ORs and 95% CIs from bivariate analyses predicting medical and psychiatric ED visits among 284 youth and adults with ASD living in Ontario

	Medical ED visit			Psychiatric ED visit
	Medical ED visit (N=31), n (%)	No Medical ED visit (N=236), n (%)	OR (95% CI)	Psychiatric ED visit (N=18), n (%)	No psychiatric ED visit (N=249), n (%)	OR (95% CI)
Predisposing
Age
Mean (SD)	17.7 (4.5)	18.3 (6.3)	1.0 (0.9 to 1.1)	18.6 (6.6)	18.2 (6.6)	1.0 (0.9 to 1.1)
Gender
Female	7 (22.6)	47 (19.9)	1.2 (0.5 to 2.9)	6 (33.3)	48 (19.3)	2.1 (0.7 to 5.9)
Male	24 (77.4)	189 (80.1)	1.0	12 (66.7)	201 (80.7)	1.0
SCQ score
Mean (SD)	22.5 (7.2)	22.6 (6.1)	1.0 (0.9 to 1.1)	20.8 (4.7)	22.7 (6.3)	1.0 (0.9 to 1.0)
ID status
ID (mild to profound ID)	14 (45.2)	110 (46.6)	1.00	8 (44.4)	116 (46.6)	1.0
No ID	15 (48.4)	98 (41.5)	1.2 (0.6 to 2.6)	9 (50.0)	104 (41.8)	1.3 (0.5 to 3.4)
Missing	2 (6.5)	28 (11.9)	0.6 (0.1 to 2.6)	1 (5.6)	29 (11.6)	0.5 (0.1 to 4.2)
Enabling
Parent marital status
Unmarried	10 (32.3)	53 (22.5)	1.6 (0.7 to 3.7)	7 (38.9)	56 (22.5)	2.2 (0.8 to 5.9)
Married/living with partner	21 (67.7)	183 (77.5)	1.0	11 (61.1)	193 (77.5)	1.0
Parent immigrated
Yes	4 (12.9)	60 (25.4)	0.4 (0.2 to 1.3)	8 (44.4)	56 (22.5)	2.8 (1.0 to 7.3)*
No	27 (87.1)	176 (74.6)	1.0	10 (55.6)	193 (77.5)	1.0
Parental education
High education	25 (80.6)	173 (73.3)	1.47 (0.6 to 3.8)	12 (66.7)	186 (74.7)	0.7 (0.2 to 1.8)
Low education	6 (19.4)	61 (25.8)	1.0	5 (33.3)	61 (24.5)	1.0
Missing	0	2 (0.8)		0	2 (0.8)
Neighbourhood income
Low	13 (41.9)	41 (17.4)	3.4 (1.6 to 7.6)**	2 (11.1)	52 (20.9)	0.5 (0.1 to 2.1)
High	18 (58.1)	195 (82.6)	1.0	16 (88.9)	197 (79.1)	1.0
Daytime activity
Unstructured	6 (19.4)	207 (87.7)	1.7 (0.7 to 4.5)	5 (27.8)	30 (12.0)	2.8 (0.9 to 8.4)
Structured	25 (80.6)	29 (12.3)	1.0	13 (72.2)	219 (88.0)	1.0
Family distress
Mean (SD)	5.1 (2.1)	4.0 (1.7)	1.4 (1.1 to 1.6)**	5.3 (1.5)	4.0 (1.8)	1.4 (1.1 to 1.8)**
Urban-rural residence
Urban	24 (77.4)	216 (91.5)	1.0	16 (88.9)	224 (90.0)	1.0
Rural	7 (22.6)	20 (8.5)	3.2 (1.2 to 8.2)*	2 (11.1)	25 (10.0)	1.1 (0.2 to 5.2)
Need
Medical comorbidity
Yes	12 (38.7)	79 (33.5)	1.3 (0.6 to 2.7)	4 (22.2)	87 (34.9)	0.5 (0.2 to 1.7)
No	19 (61.3)	157 (66.5)	1.0	14 (77.8)	162 (65.1)	1.0
Psychiatric comorbidity
Yes	18 (58.1)	105 (44.5)	1.7 (0.8 to 3.7)	11 (61.1)	112 (45.0)	1.9 (0.7 to 5.1)
No	13 (41.9)	131 (44.5)	1.0	7 (38.9)	137 (55.0)	1.0
On medication
Yes	23 (74.2)	129 (54.7)	2.4 (1.0 to 5.6)*	16 (88.9)	136 (54.6)	6.7 (1.5 to 29.5)*
No	8 (25.8)	107 (45.3)	1.0	2 (11.1)	113 (45.4)	1.0
History of aggression
Yes	21 (67.7)	151 (64.0)	1.2 (0.5 to 2.6)	16 (88.9)	156 (62.7)	4.8 (1.1 to 21.2)*
No	10 (32.3)	85 (36.0)	1.0	2 (11.1)	93 (37.3)	1.0
ED visit in past year
Yes	15 (48.4)	41 (17.4)	4.5 (2.0 to 9.7)***	7 (38.9)	49 (19.7)	2.6 (1.0 to 7.1)
No	16 (51.6)	195 (82.6)	1.0	11 (61.1)	200 (80.3)	1.0
2+ Negative life events
Yes	14 (45.2)	49 (20.8)	3.1 (1.5 to 6.8)**	8 (44.4)	55 (22.1)	2.8 (1.1 to 7.5)*
No	17 (54.8)	187 (79.2)	1.0	10 (55.6)	194 (77.9)	1.0

*p<0.05, ** p<0.01, ***p<0.001.

ASD, autism spectrum disorder; ED, emergency department; ID , intellectual disability; SCQ, Social Communication Questionnaire.

Discussion

Principal findings

Approximately 22% of individuals with ASD went to the ED at least once in a 12-month period. The results suggest that a combination of enabling (family distress) and need variables (visiting the ED in the year prior and having two or more negative life events at baseline) predicted ED use. In the secondary analysis, family distress, life events and being on medication were associated with having both medical and psychiatric emergencies at the bivariate level. Interestingly, low income and living in a rural area were associated with medical but not psychiatric emergencies. Conversely, having a history of aggression, as well as being from an immigrant family, predicted psychiatric but not medical emergencies.

Strengths and limitations

This is the first study to examine ED visits in adolescents and adults with ASD prospectively based on families’ reports. Other studies have focused primarily on emergencies in children and youth with ASD without including both family and individual variables. We examined how a wide range of predisposing, enabling and need factors contributed to ED use in this population. We also identified unique contributors to medical and to psychiatric emergencies, although our sample size did not allow for these results to be based on an adjusted analysis.

Although the findings were robust as demonstrated by two different adjusted models yielding the same results, caution is needed when interpreting the study findings due to important limitations. Our sample largely consisted of parent respondents who identified as Caucasian with high levels of education. Additionally, we were reliant on parent report of ASD severity, ID status and medical and psychiatric comorbid diagnoses. Similarly, categorisation of ED visit type was based on parent report, not chart audits. Although we tried to improve accuracy of parent reporting by engaging in regular follow-up, their responses may still be vulnerable to recall bias. Further research combining parent perspectives with administrative health data would be valuable. Finally, our results are based on a small cohort, and replication with a larger sample is warranted. Future work with larger sample sizes would also enable more detailed analyses, such as ED predictors separated by age group as well as by outcome (eg, hospital admission) or ID status.

Similarities and differences to prior research

The likelihood of ED use in the current study was not related to age, ID status or ASD severity. Previous research on predictors of ED use among individuals with ASD and other populations has shown mixed results with respect to the role of predisposing demographic factors.1–6 For instance, some studies report a greater likelihood of ED visits in older individuals with ASD,2 4 but others do not.6 Our cohort may be too small to detect these types of patterns, or indeed, they may not be the drivers of such visits when multivariate analyses include a wider combination of variables.

Only one enabling variable predicted ED use while controlling for predisposing and need variables. Parents who reported that their families were experiencing significant distress at baseline reported that their son or daughter was more likely to visit the ED during the study period. Higher rates of family stress are associated with increased ED use among parents of youth with psychiatric conditions36 and higher rates of general health service use among parents of youth with ASD.37 This is the first study to report an association between family distress and future ED use in the ASD literature.

With respect to clinical need variables, ED use in the previous year was the strongest predictor of ED visits in the year studied. This is similar to other studies that examine ED use in individuals with ID38 and our previous study that examined ED use in individuals with ASD over a 2-month period.8 Reporting two or more negative life events at baseline more than doubled the likelihood that an individual would visit the ED in the follow-up period. This aligns with previous research showing that life events predicted ED visits in individuals with ID.34 Individuals with ASD experience a high frequency of negative life events, trauma and greater stress when compared with typically developing peers.39–41 These stressors, in turn, are associated with decreases in social functioning39 and depression,42 which may lead to greater ED use.

When looking at unique predictors of each of medical and psychiatric ED visits, distinct enabling variables, like low income and living in a rural area, were associated with medical but not psychiatric emergencies. It is well recognised that some healthcare services are less available in rural areas of Ontario.43 In the general population, having lower social economic status is strongly correlated with adverse health outcomes.44 Neighbourhood income also predicts preventable hospitalisations in adults with ID.45 In our study, it may be that individuals with greater financial means and those who live in urban areas are more able to access specialised medical care, and as such may require less ED support. Not surprisingly, a history of aggression uniquely predicted psychiatric but not medical emergencies. This aligns closely with previous work looking at the predictors of psychiatric hospitalisations in children with ASD, which also pinpoint aggression as a strong predictor.46 47 Interestingly, being from an immigrant family was associated with psychiatric ED use but was not associated with medical ED visits. This begs the question of how well these families are engaged with and proactively accessing mental health services for their sons or daughters. Individuals who have recently immigrated might have less knowledge of the mental healthcare system and resources to effectively find accessible and appropriate community support services for their child with ASD.48

Implications

Our findings replicate what others have reported regarding frequent ED use among individuals with ASD1–6 and identified a combination of enabling and need variables that may be important for targeting preventative community-based supports. For instance, parents who reported higher family distress were more likely to bring their son or daughter to the ED. Caring for a child with ASD can bring many challenges, which persist across the life course.49 When parents indicate that they are approaching crisis, this can also signal future emergencies, and community-based clinicians need to proactively help such families. Brief family distress measures can assist clinicians in determining the severity of the crisis state and in the selection of appropriate interventions.30 Likewise, family members themselves can be taught to self-monitor their own distress levels and effective methods of seeking help apart from visiting the ED. From a systemic perspective, policy practices need to expand beyond the identified individual with ASD and include family focused care. Community-based services that provide relief to families, such as respite care, can help reduce stress on families and, in turn, reduce emergency service use for individuals with ASD.50

It is equally important for community-based service providers to acknowledge the impact of negative life events on individuals with ASD and their families. Yerkey and Wildman51 found that primary care physicians improved their identification rates for youth mental health problems using knowledge of negative life events. Increased clinical knowledge of the difficulties many individuals with ASD experience during transitions and provision of appropriate community mental healthcare targeted at building resiliency when exposed to negative life events may alleviate some need for ED services among this population.

Considering the link between previous and future ED visits, effective care plans need to be in place following the first ED visit to help deal with future crises. Clinical practice guidelines for this population52 indicate for ED staff to recommend families work with their community care providers to prepare proactive care plans as part of the discharge process. Further, to ensure individuals with ASD receive appropriate care in the ED and potentially negate the need for repeat visits, families can make use of hospital passports and other patient information tools to assist ED staff in better understanding an individual’s specific care needs.53