Accidental Injuries in Preschoolers: Are We Missing an Opportunity for Early Assessment and Intervention?

Abstract

Objective

Children with attention-deficit/hyperactivity disorder (ADHD) are at risk for accidental injuries, but little is known about age-related changes in early childhood. We predicted that ADHD would be associated with greater frequency and volume of accidental injuries. We explored associations between ADHD and injury types and examined age-related changes within the preschool period.

Methods

Retrospective chart review data of 21,520 preschool children with accidental injury visits within a large pediatric hospital network were examined. We compared children with ADHD (n = 524) and without ADHD (n = 20,996) on number of injury visits by age, total number of injury visits, injury volume, and injury type.

Results

Children with ADHD averaged fewer injury visits at age 3 and 90% more visits at age 6. Children with ADHD had injury visits in more years during the 3–6 age. There were no differences in injury volumes. Among patients with an injury visit at age 3, children with ADHD had 6 times the probability of a subsequent visit at age 6. At age 3, children with ADHD were estimated to have 50% fewer injury visits than children without ADHD, but by age 6, children with ADHD had an estimated 74% more injury visits than children without ADHD. Risk for several injury types for children with ADHD exceeded that for patients without ADHD by at least 50%.

Conclusions

Early identification and treatment of preschool ADHD following accidental injury may prevent subsequent injuries. Clinical implications and future directions are discussed with emphasis on the maintenance of parental monitoring into the older preschool years.

accidents and injuries, at risk youth, attention, hyperactivity, and ADHD, behavior problems, developmental perspectives, mental health, prevention science

Introduction

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental condition characterized by chronic inattention, hyperactivity, and/or impulsivity that leads to functional impairment, beginning early in life (American Psychiatric Association, 2013). An estimated 8.4% of youth in the U.S. (5.4 million children) meet criteria for ADHD in any given year, making it among the most common mental health conditions in childhood (Danielson et al., 2018).

Clinical practice guidelines exist for the psychosocial and medical management of ADHD symptoms (Barbaresi et al., 2020; Wolraich et al., 2019), driven in large part by recognition that ADHD is a clear risk factor for a wide range of serious proximal and distal outcomes. ADHD is associated with numerous experiences that directly impact physical health across development (Nigg, 2013) and emergent findings indicate that ADHD is associated with reductions in estimated life expectancy (Barkley & Fischer, 2019) and economic impacts to society ranging up to an estimated $52.4 billion (Pelham et al., 2007).

Embedded within the increased risk for other negative outcomes, ADHD diagnosis in childhood is associated with preventable pediatric injuries; however, this association is inconsistent across studies. Increased risk of physical injuries including traumatic brain injuries, burns, foreign body insertion, and bone fractures have been demonstrated for children with ADHD (Badger et al., 2008; Guo et al., 2016; Liou et al., 2018; Perera et al., 2009), whereas another recent study found that ADHD was specifically associated with increased risk of poisoning relative to physical injuries (Ruiz-Goikoetxea et al., 2018a). Age may explain some of these differences as studies included children across a wide range of development; however, even among studies limited to the preschool age range, some have failed to find an association between injury risk and ADHD (Byrne et al., 2003; Garzon et al., 2008), whereas others have found an association between ADHD symptoms and head trauma (Altun & Altun, 2018). Kang et al. (2013) demonstrated that children with ADHD were particularly vulnerable to injuries between the ages of 4 and 6. Taken together, these findings indicate that several factors may contribute to inconsistencies including age, sample size, method of assessment of ADHD symptoms (i.e., rating scales, diagnostic interviews, or chart review), use of single vs. combined injury categories, methodology (i.e., insurance utilization studies vs. retrospective parental reports of injury), and setting (e.g., Emergency Department).

More data on injury in younger children are needed to inform models for early intervention. Health records from hospital systems may yield datasets that capture diagnostic outcomes (including other neurodevelopmental conditions to serve as a comparison) and injury recurrence. Capturing these data from the source rather than relying on retrospective parent report is also advantageous for data quality. Finally, providers in medical settings might also be well-positioned to facilitate referrals to embedded or community behavioral health supports.

Therefore, this study aims to replicate and extend previous investigations of increased risk of accidental injury in youth with ADHD by focusing on changes in frequency and volume of injuries within a large sample of preschool children. In addition, the use of injury data gathered directly from acute care settings (Emergency Department and Urgent Care) provides an advantage over insurance utilization data or parental retrospective report of injury and permits a greater range of severity of injuries relative to studies that rely exclusively on Emergency Department visits. We seek to clarify whether ADHD is associated with greater frequency of injury within the preschool period and if so, whether there are specific ages/windows of development in which preschool children with ADHD are most vulnerable. We also seek to clarify whether there are differences in the volume or types of injuries for children with ADHD as well as explore rates of subsequent injuries by age. We predicted that children with ADHD would have more hospital and urgent care visits for accidental injuries; we also predicted greater volume of injuries (as measured by the number of diagnostic codes documented per visit) compared to the injuries of children without a diagnosis of ADHD. Finally, to obtain a more complete picture of injury risk, we explored associations between ADHD and specific injury types and examined age-related changes within the preschool period in the hope that emerging patterns could inform future early identification models.

Methods

Participants and Procedures

The present study is a retrospective chart review utilizing a de-identified dataset obtained from electronic medical records at Children’s Mercy Kansas City (CMKC). Study activities were designated as nonhuman subjects research by the local Institutional Review Board at Children’s Mercy Research Institute.

The dataset includes clinical encounters that took place at two CMKC emergency department and four urgent care locations spread across urban and suburban settings between 2016 and 2019 with patients between the ages of 3 and 6 at the time of visit. Encounters were included in the study if the medical record had note of an ICD-10 injury diagnosis code (S00-T98). Encounters were excluded from the dataset if the recorded codes for that visit were not accidental injuries including child abuse and neglect, allergic reactions, and motor vehicle accidents (please see Supplementary material for the full list of exclusionary diagnosis codes).  After identifying children with an injury visit in the 3–6 age range, we searched their medical record for encounters at any CMKC location when an ICD-10 psychiatric diagnosis code (F1-F99) was recorded, including visits occurring between the ages of 1–9, consistent with the methodology used in previous studies (Byrne et al., 2003; Kang et al., 2013) though most non-injury visits (80%) occurred in the 3–6 age range. We took this approach to ensure that we captured psychiatric diagnoses that occurred before or after the age range of interest in light of research indicating that the average age of ADHD diagnosis is 6.2 years-old (Visser et al., 2014). Children with more than one psychiatric condition were coded as positive in all applicable psychiatric diagnostic categories.

Measures

Encounter-specific information captured in the dataset include visit location, ICD-10 diagnosis codes recorded, and patient demographics. To better understand potential between-group differences, variables were created to categorize patients’ psychiatric history and injury type. Please see Supplementary materials for the full list of psychiatric codes and injuries.

We removed data for two patients with missing data on gender and for 1,107 patients in race categories with fewer than three ADHD patients, leaving 21,520 patients in the final dataset. There were no other missing data. In this sample, 2.4% of participants had a diagnosis of ADHD in the study window, which was captured by a medical record search for the presence of ICD-10 diagnoses code for ADHD (F90.0, F90.1, F90.2, F90.8, F90.9) within any visit included in the data. The ADHD group accounted for 1.2% of patients with a visit at age 3, 2.3% with a visit at age 4, 3.3% with a visit at age 5, and 4.3% with a visit at age 6. Other study participant demographic information is summarized in Table I.

Table I.

Sample Characteristics and Participant Demographics

Characteristic	Level	Total sample (n = 21,520)	ADHD (n = 524)	No ADHD (n = 20,996)
Gender	Female	9,377 (44%)	122 (23%)	9,255 (44%)
Race	Black	3,709 (17%)	131 (25%)	3,578 (17%)
	Hispanic	2,336 (11%)	47 (9%)	2,289 (11%)
	Multiracial	1,441 (7%)	38 (7%)	1,403 (7%)
	White	14,034 (65%)	308 (59%)	13,726 (65%)
Medicaid (any visit)		9,337 (43%)	361 (69%)	8,976 (43%)
Visit at age 3		6,795 (32%)	84 (16%)	6,711 (32%)
Visit at age 4		5,836 (27%)	132 (25%)	5,704 (27%)
Visit at age 5		5,520 (26%)	180 (34%)	5,340 (25%)
Visit at age 6		5,358 (25%)	229 (44%)	5,129 (24%)

Data Analysis

We compared ADHD (n = 524) and non-ADHD (n = 20,966) patients on the count of injury visits at each age (3, 4, 5, and 6), total number of injury visits, and number of years (1, 2, 3, or 4) in the data set (i.e., years with an injury visit). Testing for between-group differences was carried out using approximate Fisher-Pitman Permutation Tests in R.

In a multivariate analysis, we examined patient characteristics as potential correlates of patients’ yearly counts of injury visits. We created a stacked, longitudinal data set for this analysis, with each patient having one row of data for each of the four ages (3, 4, 5, and 6). The value of the dependent variable in each row was the total count of the patient’s injury visits for the year of age corresponding to that row. Using Poisson regression, a form of generalized linear modeling used for count outcomes (McCullagh & Nelder, 1989), we modeled the dependent variable as a function of ADHD diagnosis, age, ADHD diagnosis × age interaction, race, Medicaid history (any vs. none), any anxiety diagnosis, any autism spectrum disorder (ASD) diagnosis, any conduct disorder diagnosis, any depression diagnosis, any intellectual disability (ID), any oppositional defiant disorder (ODD) diagnosis, and any other Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) psychiatric diagnosis (list available as Supplementary material). Age was treated as a categorical variable with levels 3, 4, 5, and 6 to allow for non-linear effects.

Because of the clustered nature of the data we initially fit a Poisson mixed model with a random patient intercept to account for correlation between observations within patient (Stroup, 2012). The intra-class correlation estimate was zero, however, indicating no effect of clustering; this made the mixed model approach unnecessary. We also considered negative binomial regression to allow for over-dispersion, which occurs in count data when the variance exceeds the mean, in violation of the Poisson distribution assumption, but found no evidence of over-dispersion (McCullagh & Nelder, 1989). All regression models were fit using the GLIMMIX Procedure in SAS.

In a second multivariate analysis, we computed the count of ICD-10 injury diagnoses coded for each injury visit and modeled this dependent variable as a function of the same set of patient characteristic variables used in the previous model. The stacked, longitudinal data set for this analysis included one row per injury visit. The value of the dependent variable in each row was the number of injury diagnoses recorded for the given visit. We modeled the dependent variable using Poisson regression, including the patient characteristic variables above, as well as visit setting (Emergency Department vs. urgent care), as explanatory variables. We also fit a Poisson mixed model to account for clustering of visits within patient and a negative binomial model to allow for over-dispersion in the injury diagnosis counts, but as in the previous multivariate analysis, there was no evidence of intra-patient correlation or over-dispersion.

Finally, we compared the ADHD and non-ADHD groups on the percentage of patients treated for each of 20 specific injuries. For each injury, we computed the risk ratio for ADHD vs. non-ADHD patients and the p-value for either a Chi-square test or, for cases with expected cell count(s) not meeting the minimum threshold, a Fisher’s exact test. We carried out these tests in R.

In keeping with calls for reform from the statistics community, including the American Statistical Association, we did not dichotomize p-values based on an arbitrary criterion (e.g., p < 0.05) or attempt to categorize a subset of findings as “statistically significant” (Amrhein et al., 2019; Hurlbert et al., 2019; McShane et al., 2019; Wasserstein et al., 2019). Instead, we focus on effect sizes, report p-values as continuous (not binary) measures of incompatibility with the hypothesized null model, and weigh this evidence along with confidence intervals (CIs) in selecting some of the more pronounced effects to highlight. In reporting regression model results, we highlight differences between groups of 10% or more along with meaningful time trends, noting lack of precision in estimation as appropriate. Risk ratios for specific injuries are highlighted if they (or their reciprocal) exceed 1.15 except in a few cases where the small number of injuries resulted in an imprecise CI allowing for a substantial effect in either direction.

Results

There were 26,655 injury visits for 21,520 children ages 3–6 in the data set. Sample characteristics are provided in Table I. Not surprisingly, males accounted for a disproportionate share of the ADHD group (77% vs. 56% for the non-ADHD group). Black patients were also disproportionately represented among ADHD patients, accounting for 25% of the ADHD group and 17% of the non-ADHD group. Most (69%) patients in the ADHD group had some history of Medicaid as their insurance payer, whereas most (57%) in the non-ADHD group did not.

As shown in Table II, children in the ADHD group averaged fewer injury visits than children in the non-ADHD group at age 3, about the same number at age 4, 41% more visits at age 5, and 90% more visits at age 6. Among patients with an injury visit at age 3, children in the ADHD group had nearly four times the probability as children in the non-ADHD group of a subsequent visit at age 4 (32.1% vs. 8.2%), 2.8 times the probability of a subsequent visit at age 5 (11.9% vs. 4.2%), and six times the probability of a subsequent visit at age 6 (6.0% vs. 1.0%). The ADHD group also had injury visits in more years during the 3–6 age range than the non-ADHD group, on average.

Table II

. Injury Visit Statistics for Total Sample, ADHD Patients, and Non-ADHD Patients

	Unique patients	Total sample (n = 21,520)	ADHD (n = 524)	No ADHD (n = 20,996)	Standardized mean difference	Za (p-value)
Injury visits, age 3	6795	0.36 (0.50)	0.20 (0.50)	0.36 (0.57)	−0.29	6.5 (<0.001)
Injury visits, age 4	5836	0.31 (0.60)	0.30 (0.60)	0.31 (0.54)	−0.01	0.2 (0.869)
Injury visits, age 5	5520	0.29 (0.64)	0.41 (0.64)	0.29 (0.53)	0.23	5.1 (<0.001)
Injury visits, age 6	5358	0.28 (0.70)	0.53 (0.70)	0.28 (0.52)	0.48	10.8 (<0.001)
Total injury visits		1.24 (0.84)	1.44 (0.84)	1.23 (0.57)	0.36	8.2 (<0.001)
Years with injury visit(s)		1.12 (0.57)	1.24 (0.57)	1.11 (0.40)	0.30	6.8 (<0.001)

a

Z-statistic and p-value for approximate Fisher-Pitman Permutation Test for comparing ADHD and non-ADHD patients.

Results for the Poisson regression model for yearly count of injury visits are shown in Table III. Controlling for the other explanatory variables, the average number of injury visits decreased each year for the non-ADHD group but increased each year for the ADHD group, consistent with the results of the cross-sectional analysis (see Table II). At age 3, children with ADHD were estimated to have 50% [95% CI (41%, 61%)] fewer injury visits on average than children without ADHD. At age 6, by contrast, children with ADHD had an estimated 74% [95% CI (53%, 97%)] more injury visits on average than children without ADHD. Conduct Disorder diagnosis and other psychiatric diagnosis were respectively associated with 10% and 13% more yearly injury visits, on average. There was some evidence of differences related to ID and ODD, but the parameter estimates were too imprecise to rule out an effect in either direction.

Table III.

Results from Poisson Regression Model for Yearly Count of Injury Visits

Explanatory variable	exp(β)a	95% CIb for exp(β)	P
ADHD	1.74	1.53, 1.97	<0.001
Age
Age 3	1.30	1.26, 1.35	<0.001
Age 4	1.11	1.07, 1.15	<0.001
Age 5	1.05	1.01, 1.08	0.015
Age 6	Referent
ADHD × age interaction
ADHD × Age 3	0.29	0.23, 0.36	<0.001
ADHD × Age 4	0.51	0.42, 0.63	<0.001
ADHD × Age 5	0.74	0.62, 0.89	0.001
ADHD × Age 6	Referent
Female	0.98	0.95, 1.00	0.076
Race
Black	0.97	0.93, 1.00	0.055
Hispanic	0.96	0.92, 1.00	0.034
Multiracial	0.99	0.95, 1.04	0.780
White	Referent
Medicaid (any visit)	1.05	1.03, 1.08	<0.001
Anxiety diagnosis	0.97	0.87, 1.09	0.604
Autism diagnosis	1.03	0.93, 1.14	0.581
Conduct Disorder diagnosis	1.13	1.03, 1.24	0.010
Depression diagnosis	1.08	0.85, 1.37	0.550
Intellectual disability diagnosis	1.10	0.88, 1.37	0.412
Oppositional Defiant Disorder diagnosis	0.90	0.72, 1.13	0.371
Other psychiatric diagnosis	1.10	1.06, 1.15	<0.001

a

Exponentiated regression coefficient: estimated multiplicative effect on mean count of yearly injury visits, controlling for other model variables. For example, exp(β) = 1.1 would indicate the mean is estimated to be 10% higher than the mean for an otherwise equivalent patient in the referent group for the given variable.

b

95% confidence interval.

A different pattern emerged from the model for count of injury diagnosis codes recorded at visit (Table IV). There was little evidence of meaningful differences between the ADHD and non-ADHD groups on this outcome, and the count of injury diagnoses increased with age for both groups, with patients averaging 11% fewer diagnoses at age 3 than at age 6, controlling for the other explanatory variables. Hispanic patients received an estimated 15% fewer diagnoses on average than non-Hispanic children, and patients with any history of Medicaid received an estimated 39% more diagnoses on average than comparable patients without a history of Medicaid. Whereas patients with an ID averaged 20% fewer diagnoses per injury visit than patients without such a disability, children with ODD averaged 14% more diagnoses per visit than children without ODD, though the latter effect was estimated with limited precision [95% CI (−3%, 33%), p = 0.105]. Not surprisingly, Emergency Department visits were associated with an estimated 13% more injury diagnoses than urgent care visits on average, all else being equal.

Table IV.

Results from Poisson Regression Model for Count of Injury Diagnosis Codes Recorded at Visit

Explanatory variable	exp(β)a	95% CIb for exp(β)	P
ADHD	0.97	0.88, 1.07	0.535
Age
Age 3	0.89	0.86, 0.91	<0.001
Age 4	0.91	0.89, 0.94	<0.001
Age 5	0.97	0.94, 0.99	0.011
Age 6	Referent
ADHD × age interaction
ADHD × Age 3	1.06	0.89, 1.25	0.532
ADHD × Age 4	0.96	0.82, 1.11	0.571
ADHD × Age 5	1.01	0.88, 1.16	0.879
ADHD × Age 6	Referent
Female	1.00	0.98, 1.02	0.756
Race
Black	0.97	0.95, 1.00	0.027
Hispanic	0.85	0.83, 0.88	<0.001
Multiracial	0.98	0.95, 1.02	0.356
White	Referent
ED (vs. urgent care)	1.13	1.11, 1.16	<0.001
Medicaid (any visit)	1.39	1.36, 1.42	<0.001
Anxiety diagnosis	1.00	0.92, 1.09	0.955
Autism diagnosis	0.97	0.90, 1.05	0.485
Conduct Disorder diagnosis	1.02	0.95, 1.10	0.527
Depression diagnosis	1.03	0.86, 1.22	0.776
Intellectual disability diagnosis	0.80	0.67, 0.96	0.018
Oppositional Defiant Disorder diagnosis	1.14	0.97, 1.33	0.105
Other psychiatric diagnosis	0.99	0.96, 1.02	0.473

a

Exponentiated regression coefficient: estimated multiplicative effect on mean count of injury diagnoses, controlling for other model variables. For example, exp(β) = 1.1 would indicate the mean is estimated to be 10% higher than the mean for an otherwise equivalent patient in the referent group for the given variable.

b

95% confidence interval.

In the analysis of specific injuries, most types of injury were more frequent among children in the ADHD group (Table V). Estimated risk for this group exceeded that for children in the non-ADHD group by at least 65% for abrasion, bike injury, concussion, poisoning, and strain/sprain; p-values for these differences were in the 0–0.09 range. In addition, risk for the ADHD group was 36% higher for contusion, 28% higher for foreign body injury, and 18% higher for unspecified injury, but 21% lower for fracture.

Table V.

Frequencies (%) for Specific Injuries

Injury type	Total sample (n = 21,520)	ADHD (n = 524)	No ADHD (n = 20,996)	Risk ratioa (95% CIb)	Pc
Abrasion	1,566 (7.3)	62 (11.8)	1,504 (7.2)	1.65 (1.27, 2.07)	<0.001
Bike injury	340 (1.6)	14 (2.7)	326 (1.6)	1.72 (0.91, 2.70)	0.064
Burn	481 (2.2)	13 (2.5)	468 (2.2)	1.11 (0.56, 1.75)	0.814
Concussion	259 (1.2)	11 (2.1)	248 (1.2)	1.78 (0.80, 2.93)	0.089
Contusion	2,594 (12.1)	85 (16.2)	2,509 (11.9)	1.36 (1.11, 1.63)	0.004
Crushing injury	101 (0.5)	3 (0.6)	98 (0.5)	1.23 (0.00, 2.86)	0.739d
Dislocation	124 (0.6)	3 (0.6)	121 (0.6)	0.99 (0.00, 2.34)	1.000d
Electrocution	12 (0.1)	1 (0.2)	11 (0.1)	3.64 (0.00, 15.03)	0.256d
Constriction	19 (0.1)	0 (0.0)	19 (0.1)	0.00 (NA, NA)	1.000d
Foreign body	1,964 (9.1)	61 (11.6)	1,903 (9.1)	1.28 (0.99, 1.60)	0.052
Fall	244 (1.1)	4 (0.8)	240 (1.1)	0.67 (0.15, 1.42)	0.547
Fracture	3,994 (18.6)	77 (14.7)	3,917 (18.7)	0.79 (0.62, 0.96)	0.025
Laceration	6,849 (31.8)	162 (30.9)	6,687 (31.8)	0.97 (0.85, 1.10)	0.685
Open bite	247 (1.1)	9 (1.7)	238 (1.1)	1.52 (0.64, 2.61)	0.302
Open wound	172 (0.8)	4 (0.8)	168 (0.8)	0.95 (0.22, 1.98)	1.000d
Pedestrian injury	66 (0.3)	2 (0.4)	64 (0.3)	1.25 (0.00, 3.45)	0.676d
Poison	170 (0.8)	14 (2.7)	156 (0.7)	3.60 (1.81, 5.76)	<0.001d
Puncture	258 (1.2)	8 (1.5)	250 (1.2)	1.28 (0.49, 2.29)	0.621
Sprain/strain	613 (2.8)	25 (4.8)	588 (2.8)	1.70 (1.08, 2.41)	0.011
Superficial injury	200 (0.9)	3 (0.6)	197 (0.9)	0.61 (0.00, 1.42)	0.640d
Toxic effect	229 (1.1)	10 (1.9)	219 (1.0)	1.83 (0.76, 3.11)	0.091
Unspecified injury	5,176 (24.1)	148 (28.2)	5,028 (23.9)	1.18 (1.02, 1.35)	0.026

a

Risk ratio: proportion of ADHD patients with injury/proportion of non-ADHD patients with injury. For example, risk ratio = 2 would indicate a rate of injury for ADHD patients twice that for non-ADHD patients.

b

95% bootstrap confidence interval.

c

p-value for Chi-square test except where noted.

d

p-value for Fisher’s exact test.

Discussion

The current study compared rates of accidental injuries in preschool children with and without ADHD within a large children’s hospital network in the Midwestern United States, including two emergency department locations and four urgent care locations spread across urban and suburban settings. The current study adds to the literature through the inclusion of actual injury data across urgent and emergent care rather than relying on retrospective parent report of injury or insurance utilization data, which may be less accurate/representative. Additionally, the sample was large (>20,000 children) and diverse, including 4 years of injury data. Notably, all children with an injury visit within the preschool age range (3–6 years-old) were included regardless of medical or psychiatric comorbidities; additionally, to make sure we did not miss any psychiatric diagnoses that may have fallen outside the 3–6 age range, we searched the medical record for all psychiatric encounters between 1 and 9 years-old, permitting a more complete view of mental health vulnerability outside of the preschool period. Our findings replicate and extend previous research demonstrating increased risk of accidental injury in youth with ADHD by focusing on age-related changes in frequency and volume of injuries within the preschool period, a critical period for early intervention.

In line with our hypotheses, preschool children with ADHD averaged more injury visits overall. This was due to their having substantially more visits at ages 5 and 6; children in the ADHD group had fewer visits on average than those in the non-ADHD group at age 3. Regarding trends across the 4-year age range, the mean number of injury visits per year increased steadily with age for children with ADHD but decreased with age for children without ADHD. Contrary to our hypotheses, ADHD was not associated with more injury diagnoses being recorded at visits. However, children with ADHD were at substantially greater risk for many types of accidental injuries relative to children without ADHD and children with ADHD had injury visits in more years than children without ADHD.

These findings make sense within the broader developmental context. As children age, parenting behaviors change (e.g., decreased monitoring, supervision, and supports) and opportunities for independence increase across settings. As independence increases, so does the opportunity to engage in activities that also carry greater risk of physical harm (e.g., contact sports, bike riding).

Our results are also consistent with the broader literature in which neuroimaging research suggests that ADHD is likely related to delayed brain development (Hoogman et al., 2019; Shaw et al., 2015). The increased rate of injury observed in older preschool children may therefore be related to the delayed neuronal (i.e., prefrontal cortex) development and associated delays in key executive functions (e.g., working memory, response inhibition) relative to peers without ADHD. When children are less able to inhibit responses or store and manipulate multiple pieces of information, they may place themselves at greater risk for injury. Our findings suggest that parents of older preschool children with ADHD, particularly those with a history of injury at a young age, may need to maintain levels of supervision, monitoring, and support consistent with that for young preschool children.

We also found that children with Medicaid history have an estimated 39% more injury codes on average than those without a Medicaid history. Previous studies have found similar associations between lower socioeconomic conditions and injury in children as well (Dal Santo et al., 2004; Laflamme et al., 2010) and poverty reduction efforts are a promising approach for reducing injury risk (D’Souza et al., 2008).

Hispanic children averaged 15% fewer diagnostic injury codes per injury visit compared to non-Hispanic children. Several factors may limit providers’ efforts to elicit complete evaluations for Hispanic families including a provider’s implicit bias (i.e., stereotypes or attitudes toward others that may unconsciously influence decision making and behavior), a family’s limited English proficiency, and limited access to in-person professional interpreters within acute care settings (Boylen et al., 2020; Riera & Walker, 2010; Alvarado & Modesto-Lowe, 2016). Hispanic families may be less forthcoming during clinical interviews given that minorities are at higher risk of being reported for abuse/neglect compared to White parents (Hymel et al., 2018; Laskey et al., 2012).

We found evidence that ODD was associated with increased injury severity (14% increase in injury diagnostic codes per injury visit). Children with ODD may be less likely to comply with parental limit-setting, increasing the likelihood of injury. Interestingly, neither ASD nor ID were associated with increased injury frequency or severity. These findings may reflect the diagnostic categorization approach used in the current study (i.e., inclusion in the ADHD group if an ADHD code was present, regardless of co-occurring conditions), further suggesting that ADHD and other co-occurring symptoms are important risk factors of increased injury risk for youth with developmental disabilities (DiGuiseppi et al., 2018; Jain et al., 2014).

There are important limitations in this study worthy of consideration. We did not have access to medical records outside the hospital setting, which likely resulted in underreporting of psychiatric and developmental conditions diagnosed and treated within the community, such as ADHD treated by primary care physicians. We did not evaluate the impact of psychiatric comorbidity (e.g., ADHD + ODD), focusing instead on the presence of ADHD itself as a driver of injury risk. Future studies should evaluate co-existing psychiatric conditions, as there is some evidence that preschool children with ODD are at greater risk of subsequent injuries regardless of whether they also had ADHD and that ADHD alone was not associated with increased injury relative to age-matched controls (Schwebel et al., 2002). Additional limitations include the inability to evaluate the timing of the ADHD diagnoses relative to injury visits, as well as our inability to ascertain the quality and methodology used for coding psychiatric conditions and injury visits. Regarding toxic effect injuries, we are unable to ascertain how many were the result of parental errors in administering medication relative to child-driven accidental ingestions. Lastly, it is important to note that every child within our sample had at least one injury visit and therefore our results may not generalize to all children with ADHD.

Regarding clinical implications, our finding that the risk of subsequent injuries increases across development for children with ADHD relative to children without ADHD highlights the occurrence of preschool injury as a timepoint for action, potentially including routine referral for ADHD evaluation for preschoolers presenting with accidental injuries in acute care settings, particularly those with repeated injury requiring medical intervention. At minimum, this approach could prompt anticipatory guidance offered in primary care (Woods, 2006); ideally, early identification would facilitate referral for behavioral parent training to increase parental monitoring, supervision, and limit setting to reduce subsequent injury risk. A recent meta-analysis suggests that medication management may play a role in reducing injury risk for youth with ADHD (Ruiz-Goikoetxea et al., 2018b), but clinical implications lack clarity given the limited FDA-approved options for youth in the preschool range when injury risk might be identified. In any case, psychosocial treatments should continue to be prioritized for preschool children and broader dissemination of existing injury prevention models (Brixey et al., 2014; Gittelman et al., 2015) must occur. Mental health providers should routinely (and repeatedly) address accidental injury prevention when treating youth at higher risk due to ADHD symptoms, particularly older preschool children.

In terms of future directions, longitudinal studies are needed to inform risk for accidental injuries, particularly those that can tease apart the unique contributions of co-existing psychiatric conditions. The use of continuous symptom rating scales may provide greater sensitivity to emerging ADHD diagnoses in a young sample and prospective studies could identify optimal timing for ADHD screening relative to injury visits. Quality improvement initiatives targeting improved communication in the handoff from acute care settings may aid in the linkage of mental health treatments that could subsequently reduce risk. Lastly, additional studies are needed to better understand the impact of systematic factors (e.g., access to in-person interpreter services) and implicit bias that impact access and quality of healthcare for minority families.

Supplementary Data

Supplementary data can be found at: https://academic.oup.com/jpepsy.

Conflicts of interest: None declared.