Brief Report: Stability of ADHD symptoms in early childhood

Abstract

Objective:

We evaluated the stability of attention-deficit/hyperactivity (ADHD) symptoms from the toddler (24 months of age) to the preschool period (36–64 months of age) in a sample enriched for varied neurodevelopmental risk to ensure a range of ADHD symptoms.

Method:

Participants (n=256) included infants with a family history of ADHD (n=66), autism (n=115), or neither (n=75) who were prospectively followed over the first several years of life. At 24 and 36–64 months of age, parents completed the Preschool ADHD Rating Scale. At the preschool visit, children were classified into one of three mutually exclusive outcome groups: ADHD Concerns (i.e., elevated symptoms, clinician concern), Autism, or Comparison. ADHD symptom stability from the toddler to the preschool period was assessed, along with variability in early symptoms and change over time by preschool clinical presentation.

Results:

Symptoms were moderately stable, with summary score correlations of 0.56–0.60. Total scores were significantly higher at 24 months among those with later concerns for ADHD or diagnoses of autism, and increased in these two groups over time, but decreased in the Comparison group.

Conclusions:

Symptom measurement at age 2 may be useful for identifying toddlers at higher risk for later behavioral challenges, with implications for early screening. Future work in this area can help delineate boundaries of developmental typicality versus atypicality, evaluate the specificity of early symptom elevations to longer-term outcomes, and identify patterns of symptom stability from the toddler period over time that may aid in identifying children at greatest risk for persistent/increasing challenges.

Attention-deficit/hyperactivity disorder (ADHD) affects up to 8–10% of children in the U.S. (Danielson et al., 2018). While typically diagnosed around age 7, the high prevalence and associated long-term impairment necessitate improvement in identifying risk much earlier. Moreover, parents often identify ADHD-related concerns several years prior to the typical age of diagnosis (Miller, Iosif, Young, Hill, & Ozonoff, 2018), and emerging research suggests early behavioral and brain-based differences in the first years of life (Shephard et al., 2022), highlighting the potential utility of identifying children at risk for ADHD as early as possible. The ability to identify very young children at risk for ADHD could provide opportunities for the application of lower-intensity supportive interventions that reduce risk before substantial impairment develops.

One major barrier concerns our ability to adequately measure ADHD symptomatology in early childhood, a challenge due to the normative nature of high levels of activity, impulsivity, and inattentive behavior in young children. As such, determining developmental atypicality in these areas becomes increasingly difficult the younger the child. Indeed, questions about the stability and meaning of ADHD symptoms in early childhood are abundant. However, there is evidence that early symptom measurements are meaningful. For example, although categorical ADHD presentation classifications are not particularly stable from preschool (ages 4–6 years) to middle childhood (Lahey, Pelham, Loney, Lee, & Willcutt, 2005), stability in overall classification of ADHD (i.e., any presentation) is high, ranging from 70–89% (Bufferd, Dougherty, Carlson, Rose, & Klein, 2012; Lahey et al., 2005; Law, Sideridis, Prock, & Sheridan, 2014; Overgaard et al., 2023; Riddle et al., 2013).

There is also evidence of stability at the symptom level in population-based cohorts, with moderate correlations between ages 3 and 8 years (Overgaard et al., 2023), and indications that symptoms are quite stable from a mean age of 4.9 years through middle school (Leopold, Christopher, Olson, Petrill, & Willcutt, 2019). While symptom severity decreased between baseline (mean age of 4.4 years) and initial follow-up (mean age of 7.4 years) in a large clinically-referred, treated sample of preschoolers with moderate-to-severe ADHD (Preschoolers with ADHD Treatment Study), most symptom severity scores remained above clinical cutoffs with continued impairment from preschool into middle childhood, suggesting minimal clinically significant changes over time despite ongoing treatment (Riddle et al., 2013). Prior research in prospective birth cohorts has also shown that preschool symptoms at ages 3 and 5 predict later ADHD diagnosis (Oerbeck et al., 2020).

Existing studies suggest that certain symptom domains or individual symptoms may be more or less developmentally meaningful in the preschool period. For example, in a preschool-aged sample with and without early behavior problems, children eventually diagnosed with ADHD at age 6 exhibited increases in inattention from 3–6 years of age, but stable hyperactive-impulsive symptoms (Harvey, Lugo-Candelas, & Breaux, 2015). This study also found that preschoolers who ultimately met diagnostic criteria for ADHD differed in patterns of change across individual symptoms, with some behaviors remaining stable (e.g., “doesn’t listen when spoken to directly”, “difficulty engaging in leisure activities quietly”), some increasing (e.g., “easily distracted”, “blurts out answers”), and some showing variable trajectories (e.g., “difficulty sustaining attention”, “seems restless”) between 3–6 years of age. Moreover, certain symptoms were associated with greater sensitivity, specificity, and predictive power than others across the preschool period. Whether different clinical phenotypes exhibit similar or different developmental trends in symptom stability remains underexplored but could be clinically meaningful.

Although there is evidence to suggest that ADHD symptoms are relatively stable from preschool (with samples typically recruited beginning around age 4) to school-age, little is known about stability across early childhood, specifically from the toddler period (i.e., under age 3) into the preschool period (i.e., 3–5 years of age). Moreover, prior studies have primarily focused on samples of preschoolers clinically referred for ADHD/externalizing behavior problems, preschoolers enrolled in treatment studies which selected for especially high levels of impairment, or population-based birth cohorts with no known risk for ADHD. These approaches may suffer from selection biases and/or result in restricted ranges within samples (i.e., most cases having either elevated or low scores) that fail to represent the entire continuum of symptom variability. Finally, drawing from clinically-referred samples may impact generalizability to the broader population. Thus, there is a gap in terms of what is known about the stability of early childhood ADHD symptoms in non-referred samples recruited before symptoms emerge. Samples enriched for a range of phenotypic variation and which are recruited well before the typical onset of ADHD symptoms and concomitant parental concerns may be better suited to assessing ADHD symptom stability in early childhood. Prospective longitudinal studies of infants at familial risk for neurodevelopmental conditions like ADHD and autism provide an opportunity to evaluate this due to the high degree of co-occurrence and cross-aggregation between these conditions (Lai et al., 2019; Miller et al., 2019; Musser et al., 2014), elevated rates of eventual diagnoses (Miller et al., 2019; Ozonoff et al., 2011), and the presence of both subthreshold phenotypes and typical development in these samples.

In the present investigation, we examine ADHD symptom stability from the toddler period into the preschool period in a sample of young children with varied neurodevelopmental risk. We expected (1) parent-rated ADHD symptoms would be moderately stable from 24 to 36–64 months of age at both the subscale and item levels, (2) ADHD symptoms at 24 months of age would distinguish children with preschool ADHD Concerns or Autism outcomes from Comparison children, and (3) the degree of symptom stability/change from the toddler period to preschool would vary based on preschool clinical presentation (concerns for ADHD, autism).

Method

Overview of Procedure

This study utilizes data from three prospective longitudinal investigations of infants at average and elevated likelihood of developing ADHD symptoms based on family history. All studies were approved by the University’s Institutional Review Board. Parental informed consent was obtained prior to conducting assessments. All participants were enrolled by 18 months of age, with 72.2% of the sample having completed their first assessment by 9 months of age and 92.5% by 12 months of age. As part of the larger studies, children were assessed by Master- or Ph.D.-level examiners unaware of familial risk group membership or prior results. Parents completed questionnaires at up to 5 ages (depending on age at enrollment): 6 or 9, 12, 18, 24, and 36 months of age. At 36 months, participants were classified into one of three outcome groups: ADHD Concerns, Autism, or Comparison (described below). The primary measures of interest for the present analyses were obtained at the 24- (i.e., toddler) and 36-month (i.e., preschool) assessments. Due to the COVID-19 pandemic, 36-month data collection was delayed for one of the cohorts. As a result, data from this final timepoint spans 36–64 months of age (mean=38.3, SD=5.6 months).¹

Participants

A total of 439 infants with an older sibling diagnosed with autism (autism-risk group; n=193), an older sibling or parent diagnosed with ADHD (ADHD-risk group; n=135), or no family history of either (low-risk group; n=111) participated across the three longitudinal studies. For the ADHD-risk group, older sibling or parent (i.e., proband) diagnoses were confirmed with an intake screener (Diagnostic and Statistical Manual, 5^th Edition ADHD checklist) and clinician diagnostic reports/documentation of treatment for ADHD. If medical records were unavailable for sibling probands, we conducted a diagnostic evaluation. When records were unavailable for parent probands, eligibility was based on self-reported prior ADHD diagnosis and a T-score ≥65 on the ADHD Index from the Conners Adult ADHD Rating Scale (CAARS; Conners et al., 1999), rated by partner/spouse. ADHD-risk group exclusion criteria consisted of birth before 32-weeks’ gestation; autism in first-, second-, or third-degree relatives; or a known genetic condition in the proband. For the autism-risk group, sibling proband diagnoses were confirmed using the Autism Diagnostic Observation Schedule-Second Edition (ADOS-2; Lord et al., 2012) and Social Communication Questionnaire (SCQ; Rutter et al., 2003). Autism-risk group exclusion criteria included birth prior to 32-weeks’ gestation and a known genetic condition in the proband. Infants in the low-risk group were required to have a typically developing older sibling and no family history of autism or ADHD in any first-, second-, or third-degree relative. Additional low-risk group exclusion criteria included birth before 36-weeks’ gestation and any other developmental, learning, or medical conditions in any older sibling.

To be included in the present analyses, children were required to have ADHD Rating Scale (ADHD-RS) data at both the toddler and preschool visits and outcome classification at the preschool visit. Children who had not yet aged into the 36-month visit window (n=47), did not have a 36-month outcome classification (n=55), or did not have ADHD-RS data available at the toddler or preschool visit (n=1 no visit, n=35 24-month data only, n=45 36-month data only) were excluded. This resulted in a final analyzed sample of 256 (n=75 low-risk, n=66 ADHD-risk, n=115 autism-risk). Table 1 presents sample characteristics by outcome group; Table S1 provides characteristics by familial risk group.

Table 1.

Participant characteristics by outcome group.

	Autism (n=41)	ADHD Concerns (n=50)	Comparison (n=165)	p-value
Risk Group, n (%)				<0.001
Autism-Risk	38 (93%)	14 (28%)	63 (38%)
ADHD-Risk	2 (5%)	24 (48%)	40 (24%)
Low-Risk	1 (2%)	12 (24%)	62 (38%)
Age (months), mean (SD)
Toddler visit (24 months)	24.2 (0.9)	24.2 (0.9)	24.2 (0.9)	0.78
Preschool visit (36–64 months)	38.7 (6.1)	39.0 (6.3)	38.0 (5.3)	0.37
Male Gender, n (%)	28 (68%)	31 (62%)	93 (56%)	0.35
Non-White Race,a,b n (%)	17 (43%)	12 (24%)	53 (34%)	0.20
Hispanic Ethnicity,c n (%)	8 (20%)	10 (21%)	23 (14%)	0.45
Maternal Education,d n (%)				<0.001
Less than College	20 (50%)	18 (37%)	31 (19%)
College or Higher	20 (50%)	31 (63%)	134 (81%)
Income,e n (%)				0.23
≤$100,000	18 (49%)	23 (53%)	56 (38%)
$100,001 to $150,000	10 (27%)	12 (28%)	42 (28%)
≥$150,001	9 (24%)	8 (19%)	51 (34%)
Scores at Preschool visit, mean (SD)
ADHD-RS Inattentive	9.4 (5.1)*	9.5 (4.5)*	4.2 (3.3)#	<0.001
ADHD-RS Hyperactive-Impulsive	10.7 (6.1)*	12.0 (5.1)*	4.9 (3.1)#	<0.001
ADHD-RS Total	20.1 (10.5)*	21.5 (9.0)*	9.1 (5.9)#	<0.001
Overall DQf	74.2 (22.2)*	93.5 (11.9)#	104.4 (13.7)†	<0.001
Verbal DQf	70.7 (26.8)*	92.1 (12.9)#	103.1 (14.3)†	<0.001
Nonverbal DQg	77.7 (20.1)*	94.6 (13.9)#	105.8 (16.0)†	<0.001
BOSAh	9.7 (1.0)*	2.6 (1.7)#	2.4 (1.9)#	<0.001
ADOS-2 Comparisoni	6.8 (1.9)*	2.0 (1.1)#	1.6 (0.9)#	<0.001

Note. Due to rounding, percentages may not sum to exactly 100. ADHD, Attention-Deficit/Hyperactivity Disorder; SD, Standard Deviation; DQ, Developmental Quotient; ADOS-2, Autism Diagnostic Observation Schedule, 2^nd Edition; ADHD-RS, Attention-Deficit/Hyperactivity Disorder Rating Scale; BOSA, Brief Observation of Symptoms of Autism.

^aNon-White includes n=3 American Indian or Alaskan Native, n=13 Asian, n=4 Black or African American, n=1 Native Hawaiian or Other Pacific Islander, and n=61 more than one race.

Data missing:

^bn=1 Autism, n=1 ADHD Concerns, n=9 Comparison

^cn=1 Autism, n=2 ADHD Concerns, n=4 Comparison

^dn=1 Autism, n=1 ADHD Concerns

^en=4 Autism, n=7 ADHD Concerns, n=16 Comparison

^fn=1 Comparison, n=2 ADHD Concerns

^gn=1 Comparison.

^h,iChildren received either BOSA or ADOS.

n=9 Autism, n=10 ADHD Concerns, n=20 Comparison received BOSA. n=32 Autism, n=40 ADHD Concerns, n=145 Comparison received ADOS.

^*,#,†Overall group differences were assessed using nonparametric Kruskal-Wallis tests for age, one-way ANOVA for other continuous variables, and Chi-square tests for categorical variables. For continuous scores with overall group differences (p-values <0.05), we conducted follow-up pairwise group comparisons; groups with different superscripts differ significantly after Tukey–Kramer adjustment for multiple comparisons.

Measures

Attention-Deficit/Hyperactivity Disorder Rating Scale, Preschool Version (ADHD-RS; McGoey et al., 2007).

Caregivers completed the preschool version of the ADHD-RS at both the toddler and preschool visits. Parents were asked to rate 18 inattentive and hyperactive-impulsive behaviors on a scale of 0 (Never or Rarely), 1 (Sometimes), 2 (Often), or 3 (Very Often). McGoey et al. (2007) provide normative and reliability/validity data in preschoolers, indicating acceptable internal consistency (0.88), test-retest reliability (0.80), and validity (0.75). Similar adaptations have been used in 24-month-olds (Gimpel & Kuhn, 2000). The total parent-rated continuous subscale (inattention, hyperactivity-impulsivity) and total scores were obtained by summing item-level scores. Subscale scores could range from 0–27, while total scores could range from 0–54. Teacher/daycare provider ratings were obtained whenever possible and incorporated into the clinical best estimate (CBE) outcome classification process (see below).

ADOS-2 (Lord et al., 2012).

This semi-structured standardized observation of social communication and repetitive behaviors was used to determine autism outcome classification at the preschool timepoint for participants evaluated before the COVID-19 pandemic. The ADOS-2 has adequate reliability (mean agreement scores 78–98%) and validity (sensitivity=83–91%, specificity=86–94%).

Brief Observation of Symptoms of Autism (BOSA; Dow et al., 2022).

If the ADOS-2 could not be administered due to pandemic-related restrictions, the BOSA was used to inform CBE classification (Lord et al., 2020). Using ADOS-2 coding and materials, the BOSA allows for observation of social communicative behaviors with minimal face-to-face contact between children/parents and the examiner. Consistent with the BOSA manual, the examiner observed from behind a one-way mirror and provided real-time guidance to parents via a bug-in-the-ear. Parents and children were unmasked during their interactions. BOSA scores have been shown to align with ADOS-2 scores, making the BOSA a viable assessment option under pandemic restrictions (Dow et al., 2022). The BOSA has demonstrated adequate initial reliability (interrater 0.90–0.95, test-retest 0.95) and validity (0.74). A total of n=39 participants were administered the BOSA instead of the ADOS-2 at the preschool visit.

Mullen Scales of Early Learning (MSEL; Mullen, 1995).

This standardized developmental test for children from birth to 68 months was administered at each visit to evaluate cognitive functioning. MSEL subscales have excellent internal consistency (median 0.91) and test-retest reliability (median 0.84). Four subscales were administered: Visual Reception, Fine Motor, Receptive Language, and Expressive Language, each resulting in separate T-scores. Two developmental quotient (DQ) composite scores were calculated by dividing each Mullen subscale age-equivalent score by the child’s chronological age and multiplying by 100: Nonverbal DQ (i.e., average of Fine Motor and Visual Reception DQs) and Verbal DQ (i.e., average of Receptive and Expressive Language DQs).

CBE outcome.

At the end of the preschool visit, participants were classified into one of three outcome groups by a licensed psychologist: ADHD Concerns (n=50), Autism (n=41), or a Comparison group (n=165). Participants were classified into the “ADHD Concerns” group if they (a) received an examiner-rated clinical best estimate outcome of “ADHD Concerns,” and (b) exhibited ≥4 DSM-5 ADHD symptoms within any one symptom domain (i.e., inattentive or hyperactive-impulsive) or ≥5 DSM-5 symptoms across both symptom domains (i.e., inattentive and hyperactive-impulsive combined) across reporters (i.e., parent and teacher report on the ADHD Rating Scale Preschool version, examiner observation); at least one of the endorsed symptoms must have been reported on the ADHD-RS by the parent or teacher as evidence of some degree of cross-situational presence of symptoms. Participants were classified into the Autism group if they met DSM-5 criteria for autism spectrum disorder and received an ADOS-2 calibrated severity score ≥4 (Lord et al., 2012) or a BOSA algorithm score ≥9 (Dow et al., 2022). Participants who met criteria for both autism and our algorithmic criteria for “ADHD Concerns” (n=17) were included in the Autism group for primary analyses; secondary sensitivity analyses excluded these participants. All remaining participants were classified into the Comparison group.

Statistical Analysis

Stability was evaluated in several ways. First, we used Spearman’s rank-order- correlations to assess the consistency of the ranking order of continuous scores across the two timepoints at the subscale, total score, and item levels. Second, we conducted separate generalized linear mixed-effects models at the subscale, total score, and item levels to evaluate developmental change in these scores over time (McCulloch, Searle, & Neuhaus, 2008). We used the identity link and a normal variance function to model square-root transformed subscale and total scores. Items were rated on a 0–3 Likert scale, with most showing floor effects (16–88% scored 0 at 24 months, 24–77% at 36 months). Thus, for item-level data, we employed proportional-odds mixed-effects models to estimate the log-odds (or cumulative logit) of receiving a higher rating, reflecting more severe symptoms. For items with <30 participants in the highest category, we merged the highest two (or three) categories.

Unadjusted models included a fixed effect for visit (Toddler, Preschool) and a random intercept to account for the within-child dependence. Adjusted models included fixed effects for outcome group (Autism, ADHD Concerns, or Comparison), visit (Toddler, Preschool), and their interaction, along with a random intercept to account for within-child dependence. Significant overall tests for the outcome group by visit interaction were followed by linear contrasts to examine pairwise differences in trajectories (intercepts and slopes) between outcome groups and to assess outcome group differences at each visit. In these models, we estimated group-specific intercepts, representing average symptom levels at the Toddler visit, and slopes, capturing developmental change over time (i.e., the average difference in symptoms between the Preschool and Toddler visits). If the outcome group by visit interaction was not significant, it was removed from the reported models. In those cases, slopes reflect a common change in symptom scores across all groups (i.e., the average difference between preschool and toddler visits) and group differences are interpreted as overall average differences between the groups across both visits. A sensitivity analysis was conducted by rerunning the unadjusted and adjusted models after removing the participants with both Autism and ADHD Concerns outcomes. All hypothesis tests were two-sided, and p-values <0.05 were deemed statistically significant. All analyses were implemented in SAS Version 9.4 (SAS Institute Inc., Cary, NC).

Results

Symptom stability

Continuous symptoms.

Spearman’s rank-order correlations of subscale scores between the toddler and preschool periods were moderate and statistically significant for both inattention (r_s=0.56, p<0.001) and hyperactivity-impulsivity (r_s=0.57, p<0.001) (Table 2). The pattern was similar for overall total scores (r_s=0.60, p<0.001). In all cases, higher scores in toddlerhood were associated with higher scores in the preschool period. There was sufficient variability in scores, therefore these correlations do not appear to be artificially attenuated by a narrow range: At the toddler visit, scores ranged from 0 to 20 for inattention (mean=5.81, SD=4.52), from 0 to 24 for hyperactive-impulsive (mean=7.18, SD=5.02), and from 0 to 43 for total (mean=12.99, SD=9.02). Similarly, at the preschool visit, scores ranged from 0 to 24 for inattention (mean=6.06, SD=4.64), from 0 to 25 for hyperactive-impulsive (mean=7.24, SD=5.18), and from 0 to 46 for total (mean=13.30, SD=9.34).

Table 2.

Spearman’s rank-order correlations between toddler and preschool visits ratings.

	Toddler Visit (n=256)	Preschool Visit (n=256)	Spearman Rank-Order Correlation	p-value
Subscale/Item	Median [Range]	Median [Range]
Inattentive	5 [0 – 20]	5 [0 – 24]	0.56	<0.001
Hyperactive-Impulsive	6 [0 – 24]	6 [0 – 25]	0.57	<0.001
Total score	12 [0 – 43]	11.5 [0 – 46]	0.60	<0.001
Inattentive items
Poor attention to detail	1 [0 – 3]	1 [0 – 3]	0.29	<0.001
Difficulty sustaining attention	1 [0 – 3]	1 [0 – 3]	0.45	<0.001
Doesn’t listen	1 [0 – 3]	1 [0 – 3]	0.49	<0.001
Lacks follow through	1 [0 – 3]	1 [0 – 3]	0.38	<0.001
Difficulty organizing	1 [0 – 3]	0 [0 – 3]	0.24	<0.001
Avoids sustained mental effort	0 [0 – 3]	0 [0 – 3]	0.21	0.001
Loses things	0 [0 – 3]	0 [0 – 3]	0.14	0.02
Easily distracted	1 [0 – 3]	1 [0 – 3]	0.46	<0.001
Forgetful	0 [0 – 3]	0 [0 – 3]	0.15	0.02
Hyperactive-impulsive items
Fidgety	1 [0 – 3]	1 [0 – 3]	0.50	<0.001
Leaves seat	1 [0 – 3]	1 [0 – 3]	0.35	<0.001
Runs/climbs	1 [0 – 3]	1 [0 – 3]	0.55	<0.001
Loud	0 [0 – 3]	0 [0 – 3]	0.34	<0.001
On the go	1 [0 – 3]	1 [0 – 3]	0.61	<0.001
Talks excessively	0 [0 – 3]	0.5 [0 – 3]	0.33	<0.001
Blurts out	0 [0 – 3]	0 [0 – 2]	0.07	0.26
Difficulty waiting	1 [0 – 3]	1 [0 – 3]	0.35	<0.001
Interrupts	1 [0 – 3]	1 [0 – 3]	0.30	<0.001

Note. Toddler visit was completed at 24 months and the preschool visit was completed between 36–64 months.

The linear mixed-effects models at the subscale and total score levels were generally consistent with the correlation findings across the full sample (r_s>0.55), showing relative stability (Table S2). However, as shown in Tables 3 and S2, symptoms in the toddler period differed based on preschool outcome classification. Specifically, at 24 months of age, all symptom scores were significantly higher among those with preschool ADHD Concerns and Autism outcomes relative to the Comparison group (Table 3). Moreover, the change in symptom scores between the toddler and preschool periods differed by outcome, with the ADHD Concerns and Autism groups showing increases in inattentive, hyperactive-impulsive, and total symptoms and the Comparison group showing either no change or declines (Table S2). Symptom scores were similar between children with ADHD Concerns and those with Autism outcomes during both the toddler and preschool periods, as was the change in scores between the two periods (Tables 1, 3).

Table 3.

Summary of the outcome group differences^a in toddler levels (intercepts) and change from toddler to preschool age (slopes), for ADHD-RS symptom scores and items.

	Autism vs. Comparison		ADHD Concerns vs. Comparison		ADHD Concerns vs. Autism
	Intercept (SE)	Slope (SE)	Intercept (SE)	Slope (SE)	Intercept (SE)	Slope (SE)
Inattentive b	0.60 (0.17)***	0.52 (0.16)**	0.73 (0.16)***	0.46 (0.15)**	0.13 (0.21)	−0.07 (0.20)
Hyperactive-Impulsive b	0.57 (0.16)***	0.45 (0.16)**	0.84 (0.15)***	0.48 (0.15)**	0.27 (0.19)	0.03 (0.19)
Total Score b	0.80 (0.21)***	0.66 (0.20)**	1.07 (0.20)***	0.65 (0.19)***	0.27 (0.26)	−0.01 (0.25)
Inattention items
Poor attention to detailc	1.41 (0.33)***	–	1.29 (0.30)***	–	−0.12 (0.38)	–
Difficulty sustaining attention	2.11 (0.48)***	1.60 (0.56)**	2.15 (0.45)***	0.94 (0.50)	0.04 (0.55)	−0.66 (0.64)
Doesn’t listen	2.52 (0.52)***	0.61 (0.55)	1.13 (0.45)*	1.25 (0.50)*	−1.38 (0.59)*	0.64 (0.64)
Lacks follow-throughc	1.91 (0.35)***	–	1.74 (0.32)***	–	−0.17 (0.39)	–
Difficulty organizing	0.40 (0.38)	1.22 (0.50)*	0.77 (0.35)*	0.61 (0.46)	0.37 (0.46)	−0.61 (0.59)
Avoids sustained mental effortc	0.53 (0.34)	–	1.55 (0.31)***	–	1.02 (0.39)**	–
Loses things	−1.01 (0.50)*	1.61 (0.61)**	0.62 (0.37)	0.56 (0.47)	1.63 (0.57)**	−1.05 (0.69)
Easily distracted	1.19 (0.47)*	1.38 (0.54)*	1.90 (0.45)***	1.27 (0.51)*	0.71 (0.57)	−0.11 (0.65)
Forgetful	−0.88 (0.51)	1.42 (0.63)*	0.12 (0.39)	1.02 (0.52)	0.99 (0.59)	−0.39 (0.72)
Hyperactive-Impulsive items
Fidgety	1.00 (0.48)*	1.16 (0.54)*	1.90 (0.47)***	1.27 (0.52)*	0.90 (0.59)	0.11 (0.66)
Leaves seat	0.34 (0.41)	1.38 (0.50)**	1.26 (0.38)**	1.33 (0.45)**	0.92 (0.49)	−0.05 (0.58)
Runs/climbs	1.61 (0.54)**	1.67 (0.58)**	2.50 (0.52)***	1.12 (0.51)*	0.89 (0.63)	−0.55 (0.66)
Loud	1.35 (0.46)**	1.05 (0.55)	1.24 (0.42)**	1.41 (0.51)**	−0.11 (0.53)	0.36 (0.63)
On the go	1.94 (0.55)***	0.92 (0.55)	2.72 (0.51)***	1.38 (0.49)**	0.78 (0.62)	0.46 (0.63)
Talks excessivelyc	−0.52 (0.36)	–	1.05 (0.31)**	–	1.56 (0.44)***	–
Blurts outc	0.11 (0.37)	–	1.05 (0.31)***	–	0.94 (0.41)*	–
Difficulty waiting	1.32 (0.41)**	1.47 (0.50)**	1.21 (0.37)**	1.04 (0.45)*	−0.11 (0.48)	−0.42 (0.59)
Interruptsc	1.08 (0.32)***	–	1.24 (0.30)***	–	0.16 (0.38)	–

Note. ADHD, Attention-Deficit/Hyperactivity Disorder; SE, Standard Error; ADHD-RS, Attention-Deficit/Hyperactivity Disorder Rating Scale.

^aEstimated differences between groups in intercepts (levels at the toddler visit) or slopes (change between the toddler and preschool visits) were derived from linear (for scores) and proportional-odds (for items) mixed-effects models that included terms for visit (Toddler, Preschool), outcome group (Autism, ADHD Concerns, Comparison) and their interaction, and a random intercept for child to account for the within-child dependence.

^bSquare-root transformed for analyses.

^cIn these models, the outcome group by visit interaction was not significant and, therefore, not retained in the model. Thus, the slopes represent a common change in symptoms across all groups (i.e., the average difference between Preschool and Toddler visits), and group differences were interpreted as overall average differences between the groups across the two visits.

^*p<0.05

^**p<0.01

^***p<0.001

Item-level.

We next examined stability at the item level to assess whether specific individual symptoms in the toddler period are particularly unstable or stable into the preschool period. Spearman’s rank-order correlations of most items between toddler and preschool periods were moderate and statistically significant, except for “blurts out” (Table 2). One-third of the items (33.3%) showed stronger correlations (rs ≥ 0.40), and these higher-scoring items were distributed evenly across the two symptom domains. Overall, higher scores in the toddler period were associated with higher scores in the preschool period.

Across the full sample, children exhibited significant increases in the odds of being in a higher category for four of the 18 items (“loses things”, “forgetful”, “talks excessively”, “blurts out”) and significant decreases for two items (“on the go”, “difficulty waiting”; Table S2, unadjusted models). The adjusted proportional-odds mixed-effect models indicated that the pattern of item-level change varied by outcome group, as shown in Tables S2 and 3. Specifically, the Comparison group exhibited significant decreases in the odds of being in a higher category for six items (“difficulty sustaining attention”, “difficulty organizing”, “leaves seat”, “runs/climbs”, “on the go”, “difficulty waiting”) and an increase in two (“talks excessively”, “blurts out”; Table S2). The ADHD Concerns group, in contrast, exhibited significant increases in seven items (“doesn’t listen”, “loses things”, “forgetful”, “leaves seat”, “loud”, “talks excessively”, “blurts out”) and decreases in none. Likewise, the Autism group exhibited significant increases in five items (“loses things”, “forgetful”, “leaves seat”, “talks excessively”, “blurts out”) and decreases in none. Of note, the odds of being in a higher category for two items (“talks excessively”, “blurts out”) increased across all outcome groups, while three items (“loses things”, “forgetful”, “leaves seat”) increased in both the ADHD Concerns and Autism groups, and two items (“doesn’t listen”, “loud”) increased in only the ADHD Concerns group.

When directly testing for outcome group differences during the toddler period, the ADHD Concerns group had greater odds of being rated higher than the Comparison group for every item except “loses things” and “forgetful”. The ADHD Concerns group also had greater odds of being rated higher than the Autism group for four items (“avoids sustained mental effort”, “loses things”, “talks excessively”, “blurts out”; Table 3). The Autism group had greater odds of being rated higher than the Comparison group for 11 items (“poor attention to detail”, “difficulty sustaining attention”, “doesn’t listen”, “lacks follow-through”, “easily distracted”, “fidgety”, “runs/climbs”, “loud”, “on the go”, “difficulty waiting”, and “interrupts”) and lower odds of being rated higher than the Comparison group for one item: “loses things”.

Additionally, the ADHD Concerns group had a greater change in the odds of being rated higher from the toddler to preschool period relative to the Comparison group for eight items (“doesn’t listen”, “easily distracted”, “fidgety”, “leaves seat”, “runs/climbs”, “loud”, “on the go”, “difficulty waiting”; Table 3). The Autism group had a greater change in odds relative to the Comparison group for nine items (“difficulty sustaining attention”, “difficulty organizing”, “loses things”, “easily distracted”, “forgetful”, “fidgety”, “leaves seat”, “runs/climbs”, “difficulty waiting”). There were no slope differences between the Autism and ADHD Concerns groups for any item.

Excluding the 17 autistic children who also met algorithmic criteria for ADHD Concerns outcomes did not substantially alter the intercept and slope estimates within outcome groups (Table S4). In these sensitivity analyses, the Autism group’s estimates occasionally decreased and were no longer significant, although one decrease resulted in significance (Table S4). Most significant outcome group differences (or lack thereof) were retained in the sensitivity analyses, although some discrepancies were observed, including attenuated differences between the Autism and Comparison groups (Table S3).

Discussion

In a transdiagnostic-risk sample originally recruited before concerns about ADHD tend to emerge and enriched for a range of ADHD symptom variability, symptoms of ADHD appear to be moderately stable from the toddler period to preschool age, with correlations for summary scores ranging from 0.56 to 0.60. These correlations are higher than those reported in population-based birth cohorts (e.g., 0.36–0.46; Overgaard et al., 2023), potentially due to the shorter duration between measurements in our study (i.e., average of 1.2 years vs. 5 years). Sufficient variability in scores at each timepoint suggested that such stability was not simply due to artificial attenuations as the result of a narrow range, suggesting that meaningful variability in ADHD symptoms is measurable at these early ages. Our findings also suggest relatively similar stability across ADHD symptom dimensions, consistent with prior research in older samples finding no differences in patterns of change in inattentive versus hyperactive-impulsive scores from preschool to later follow-up timepoints (Riddle et al., 2013).

Although further empirical testing is required, when considered in the context of the larger literature indicating that preschool ADHD symptoms are relatively stable into middle childhood, our findings could imply that symptom measurement at age 2 provides similar predictive utility to symptoms measured at ages 3–5, or at least may be useful for identifying toddlers at higher risk for later behavioral challenges. Our finding that those with preschool outcomes indicating some degree of neurodevelopmental concern (i.e., ADHD Concerns, Autism) exhibit higher symptoms than those in the Comparison group at 24 months of age and experience increases in symptoms between the toddler and preschool period (vs. decreases in the Comparison group) further suggests that ADHD symptom measurements at 24 months of age have some degree of clinically meaningful predictive utility, at least in the short-term.

Consistent with work in an older sample (3–6 years; Harvey et al., 2015), despite relative stability at the symptom domain level, findings from item-level analyses indicated some variability in the stability of individual ADHD symptoms, especially when considering preschool outcome classification. Across the full sample, we found some evidence of item-level stability between the toddler and the preschool periods, with significant developmental change for one third of the total items, most of which were either more relevant to later developmental periods in terms of school-related demands or expectations (i.e., increases in “loses things,” decreases in “difficulty waiting”), or potentially indicative of typical developmental trends in language acquisition (i.e., increases in “talks excessively,” “blurts out”).

However, when examining item-level change by outcome group, differences in patterns were evident, which could inform our understanding of the boundaries of developmental appropriateness/inappropriateness at these early ages (Nili et al., 2023). Specifically, increases in talking excessively between the toddler and preschool periods—which occurred in all groups—may well be developmentally appropriate in this period and, therefore, should not be interpreted as concerning. In contrast, patterns of change across a subset of other items between the toddler and preschool periods may be particularly useful with respect to identifying early risk for ADHD, as these items specifically increased in the ADHD Concerns outcome group, but did not change in the Autism and Comparison outcome groups. This includes inattentive behaviors like not listening and hyperactive-impulsive behaviors like loudness. It is also worth noting that while, for example, certain symptoms were higher in toddlerhood in the ADHD Concerns group than the Autism group, there was overlap in patterns of item-level change between the groups, as well as some similarities in presentation of these behaviors at young ages, which should be replicated in other samples. Ultimately, determining the most predictive combinations of items in the toddler period, better discerning the boundary of developmental typicality versus atypicality, and examining the specificity of early symptom elevations to longer-term outcomes will require further follow-up in these samples, but could help inform the development of early screening measures.

Overall, our findings suggest that there is sufficient variability in ADHD symptom measurements at age 2 to indicate that early individual differences in inattentive and hyperactive-impulsive behavior may be meaningful. These results also indicate that overall symptoms are relatively stable from 2 years of age into the preschool period in a non-referred sample comprising varying degrees of neurodevelopmental risk. Limitations of this study include partial confounding of ADHD-RS scores at the preschool timepoint with ADHD Concerns outcome group classification, which we approached by focusing group-based analyses on 24-month data and change over time rather than group differences in symptoms at the preschool visit. Additionally, whether the caregiver was the child’s mother or father, or whether the same caregiver provided ratings at the two timepoints, was not considered. The sample sizes for two of the outcome groups are modest, potentially reducing our ability to detect more subtle developmental changes, and approximately 41% of the Autism outcome group also met our algorithmic criteria for ADHD Concerns, limiting our ability to fully assess specificity. We attempted to address this by running sensitivity analyses excluding this group, which suggested that co-occurring ADHD concerns may have contributed to the differences between the Autism and Comparison groups. However, 34% of the Autism group was excluded from the sensitivity analyses, reducing power. Finally, the overall sample was predominantly male (59.4%), White (66.5%), non-Hispanic (83.5%), and advantaged. Future work will benefit from planned longitudinal follow-up of the reported samples to examine patterns of symptom stability from age 2 to middle childhood, along with examination of predictive utility to later diagnosis. Of particular interest is the use of data-driven approaches to identify subgroups showing different ADHD symptom stability profiles from early to middle childhood, which may help to identify specific young children at greatest risk for persistent or increasing symptoms over time, as well as subsequent diagnosis.