Incremental Utility of 24-Month Autism Spectrum Disorder Screening After Negative 18-Month Screening

Abstract

The American Academy of Pediatrics recommends Autism Spectrum Disorder (ASD) screening at 18 and 24 months. However, utility of rescreening at 24 months, after a negative 18-month screening, remains unknown. We identified cases of ASD detected at 24 months after a negative 18-month screening (i.e., Catch-24 group; n = 10) and compared them to toddlers detected by 18-month screening (i.e., Early Diagnosis group; n = 203). Repeated ASD-specific screening at 24 months detected children who were missed at their 18-month screening. Thus, our findings support repeated screening for ASD at both at 18 and 24 months in order to maximize identification of toddlers with ASD and other neurodevelopmental disorders who require intervention.

Autism Spectrum Disorder (ASD) is an increasingly prevalent neurodevelopmental disorder, with current estimates indicating that one in every 59 children meets diagnostic criteria (Centers for Disease Control and Prevention (CDC), 2018). Early diagnosis, followed by targeted, intensive intervention services, improves outcomes for children with ASD (Baird et al., 2001; Lipkin & Hyman, 2011). Indeed, young children with ASD show large developmental gains, including improved scores on measures of intellectual functioning and language use, following participation in intervention services that are initiated prior to the child’s third birthday (Anderson, Liang, & Lord, 2014; Harris & Handleman, 2000; MacDonald, Parry-Cruwys, Dupere, & Ahearn, 2014; Orinstein et al., 2014; Rogers & Vismara, 2008). Further, with intensive ASD-specific services, a minority of children may improve such that they no longer meet ASD diagnostic criteria (Fein et al., 2013; Orinstein et al., 2014). However, early access to these intensive intervention services, and the resultant benefits for children and families, are contingent upon early identification (Baird et al., 2001; Duby et al., 2006; Gupta et al., 2007; Johnson & Myers, 2007).

Based on the clear benefits of early diagnosis, as well as the historical finding that many children in the United States were diagnosed with ASD years after their symptoms emerged (Mandell, Novak, & Zubritsky, 2005), the American Academy of Pediatrics (AAP) recommended routine developmental surveillance and population-based screening to identify young children at risk of developmental delays (Duby et al., 2006). As part of developmental surveillance, pediatricians are encouraged to elicit and discuss caregiver concerns and to document both caregiver report and their own observations of child development during all well-child care visits (Duby et al., 2006). Population-based screening using a validated measure was recommended for the early detection of developmental delays at 9, 18, and 24 or 30 months. ASD-specific screening was recommended at 18 and 24 months given evidence of stable diagnoses as early as 18 months, as well as concern for possible regression between 18 and 24 months of age (Duby et al., 2006; Johnson & Myers, 2007).

Routine use of validated screening instruments in the context of a broader developmental surveillance program may promote the identification of at-risk children before concerns become apparent through other means (e.g., spontaneously expressed caregiver concerns or in-office behavioral observations), particularly with regards to identifying subtle social and fine motor delays (Baird et al., 2001; Crais & Watson, 2014; Gupta et al., 2007; Johnson & Myers, 2007; Keil, Breunig, Fleischfresser, & Oftedahl, 2014; Zuckerman, Lindly, & Sinche, 2015; Zwaigenbaum et al., 2015). Screening may also help provide an accurate representation of a child’s development, as some atypical behaviors are difficult for pediatric providers to accurately assess within the confines of brief office visits (Gabrielsen et al., 2015). Furthermore, implementation of AAP ASD-specific screening recommendations reduces the average age at first diagnosis (including for Black children, who otherwise tend to be diagnosed later than White children) and improves access to very early ASD-specific intervention services, all without significantly increasing the length of well-child care visits (Herlihy et al., 2014; Lipkin & Hyman, 2011; Mandell, Listerud, Levy, & Pinto-Martin, 2002).

Since the AAP published their most recent screening recommendation in 2007, subsequent literature has supported the practice of screening for ASD at both 18 and 24 months (Barton, Dumont-Mathieu, & Fein, 2012; Zwaigenbaum et al., 2015). Screening at 18 months is vital because symptoms of ASD typically emerge by this age (Duby et al., 2006), and waiting to diagnose a child at 24 months precludes at least six months of possible early intervention at a crucial age of maximum neural plasticity (MacDonald et al., 2014; Rogers et al., 2014). Rescreening at 24 months is supported by research suggesting that approximately 20–40% of children with ASD appear to develop typically into their second year of life and then experience either a regression or a later emergence of ASD symptoms (Barton et al., 2012; Gupta et al., 2007; Ozonoff, Heung, Byrd, Hansen, & Hertz-Picciotto, 2008). In addition, limited parental knowledge of symptoms of ASD or the age at which children typically achieve developmental milestones, or limited understanding of screening questions, may contribute to children screening negative at 18 months and then positive at 24 months. Indeed, some children demonstrate symptoms of ASD at 18 months, even if they screen negative on an ASD-specific screening tool at that time (Øien et al., 2018). Repeated screenings for ASD at 18 and 24 months can therefore facilitate earlier referral to intervention services, while increasing the likelihood of identifying children missed by an earlier screening (Barton et al., 2012; Crais et al., 2014; Crais & Watson, 2014; Zwaigenbaum et al., 2015).

Despite the clear benefits of routine screening, a number of factors prevent many pediatric healthcare providers from screening children for ASD at 18 and 24 months (Arunyanart et al., 2012; Zuckerman et al., 2013). Limited time, resources (e.g., trained office staff), training specific to ASD, knowledge of and access to validated ASD-specific screening tools, and office-based systems for making referrals, billing, and monitoring outcome all serve as barriers (Crais et al., 2014; Zwaigenbaum et al., 2015). It is also possible that both pediatric healthcare providers and caregivers may react negatively to rescreening, given that they screened six months earlier; this hesitation to rescreen may be particularly apparent in the case of children who initially screen negative (i.e., not at-risk) at 18 months. Given these concerns, there is an ongoing need for research to inform screening practices and reduce research-to-practice gaps, with a particular need to determine whether there is a significant benefit to screening at both 18 and 24 months.

There is no literature, to our knowledge, examining the incremental utility of 24-month screening after negative 18-month screening, although a direct comparison of initial 18- versus 24-month screening for ASD showed better positive predictive value (PPV) at 24 months, especially for children drawn from a general population (i.e., low-risk) sample (Pandey et al., 2008; Sturner, Howard, Bergmann, Stewart, & Afarian, 2017). Thus, the present study aimed to extend the existing literature by investigating the incremental utility of rescreening children for ASD at 24 months who initially screened negative at 18 months. In addition, we compared children diagnosed early (i.e., following a positive 18-month screening; Early Diagnosis group) to children diagnosed with ASD who had a negative 18-month screening followed by a positive 24-month screening (Catch-24 group). We examined potential demographic differences, as well as performance on measures of developmental functioning, adaptive skills, and ASD symptomatology. Based on previous research suggesting that children identified younger show more severe symptoms than those identified later (Mandell et al., 2005; Twyman, Maxim, Leet, & Ultmann, 2009), we hypothesized that children in our Early Diagnosis group would be more severely affected than toddlers in the Catch-24 group. Our findings have implications for ASD screening in pediatric primary care settings and can provide data to refine or reinforce the current AAP ASD-specific screening guidelines.

Methods

Participants

Participants were toddlers recruited as part of a larger, federally-funded study on the early detection of ASD. The study’s catchment area included Connecticut and metropolitan Atlanta. All children were screened with either the Modified Checklist for Autism in Toddlers with Follow-Up (M-CHAT/F; Robins, Fein, & Barton, 1999; Robins, Fein, Barton, & Green, 2001) or its revision, the Modified Checklist for Autism in Toddlers, Revised with Follow-Up (M-CHAT-R/F; Robins, Fein, & Barton, 1999; Robins et al., 2014), at their 18-month pediatric well-child care visit (n = 19,685; Figure 1), or at both their 18- and 24-month well-child care visits (n = 7,993; Figure 2). Children were excluded from the current study if they were screened using a Spanish version of the M-CHAT(-R)/F, as there were insufficient numbers to allow for comparison with those screened using the English version of the M-CHAT(-R)/F.

Figure 1.

Origins of Early Diagnosis group. Participants screened with the M-CHAT(-R)/F at 18 months only are presented; positive screens on either the initial questionnaire or Follow-Up are shaded in light gray, and those children subsequently diagnosed with ASD (i.e., Early Diagnosis group; n = 203) are shaded in dark gray. Both typically developing children and those diagnosed with a developmental disorder other than ASD are included in the Non-ASD category.

Figure 2.

Origins of Catch-24 group. Participants who screened negative on the M-CHAT(-R)/F at 18 months and were then rescreened at 24 months are presented; those children who screened negative on either the initial questionnaire or Follow-Up at 18 months are shaded in light gray, and those who then screened positive at 24 months on both the initial questionnaire and Follow-Up are presented in medium gray. Of those participants, the children who subsequently received am ASD diagnosis at 24 months (i.e., Catch-24 group, n = 10) are shaded in dark gray. Typically developing children (n = 4) and those diagnosed with a developmental disorder other than ASD are both included in the Non-ASD category.

Participants who received an ASD diagnosis were split into two groups based on timing of their positive screening. The Early Diagnosis group (n = 203) were children who screened positive at 18 months and received an ASD diagnosis (see Figure 1). The Catch-24 group (n = 10) was composed of children who initially screened negative at 18 months, but then screened positive (i.e., at risk) at 24 months and received an ASD diagnosis (see Figure 2). Demographic characteristics of both groups are presented in Table 1.

Table 1.

Participant Demographic Characteristics

Variable	Catch-24 Group	Early Diagnosis Group	p
Gender
Male	8	151	.98
Female	2	52
Race/Ethnicity
White	5	125	.35
Black	1	26
Asian	1	8
Biracial	1	12
Hispanic	2	19
Other	0	1
Not reported	0	12
Maternal Education
No high school degree	2	8	.21
High school degree	0	16
Vocational/technical	0	1
Some college	4	34
College degree	2	24
Advanced degree	0	23
Not reported	2	97
Annual Household Income
Median income bracket	$27,500	$45,000	.51
Not reported	6	71

Note. Child gender, child race/ethnicity, maternal education, and the number of participants for whom data was not reported are presented as frequency counts(n). Income is presented as median income in dollars.

Procedures

Participants were screened with the M-CHAT(-R) at their pediatrician’s office between the ages of 16 and 30 months; specifically, the 18-month screening window was 16 months to 21 months, 30 days, and the 24-month screening range was 22 months to 30 months. In line with AAP recommendations, pediatric providers were encouraged to screen at both 18 and 24 months; however, in order to maximize the number of children screened, practices that chose to screen only at 18 months were also included. For children who screened positive on the initial questionnaire, a member of the research team contacted caregivers by telephone to administer the Follow-Up questions in an interview format (i.e., M-CHAT(-R)/F). Children who screened positive on the two-stage M-CHAT(-R)/F were offered a free developmental and diagnostic evaluation. A licensed clinical psychologist or a developmental-behavioral pediatrician and a doctoral student completed all evaluations. Evaluations generally took place at the research team’s university clinics. However, families without transportation were provided with either free taxi service to the university, evaluations in their pediatrician’s office, or evaluations in their homes. Study diagnoses were assigned based on clinical best estimate judgment, incorporating data from observation, developmental history, and direct testing. Given the years in which participants were evaluated, all diagnoses were assigned according to DSM-IV-TR criteria (American Psychiatric Association, 2000). Since a diagnosis of ASD, based on an at risk-screening as early as 16 months, is considered reliable (Kleinman et al., 2008), children who were diagnosed based on a positive 18-month screening were not rescreened at 24 months.

Measures

Modified Checklist for Autism in Toddlers with Follow-Up (M-CHAT/F; Robins et al., 1999; Robins et al., 2001) and Modified Checklist for Autism in Toddlers, Revised with Follow-Up (M-CHAT-R/F; Robins et al., 2009; Robins et al., 2014).

The M-CHAT/F (23 items) and its revision, the M-CHAT-R/F (20 items), are yes/no caregiver-report measures used to screen for ASD in toddlers between the ages of 16 and 30 months. When caregivers’ initial responses indicate concern for ASD, a structured Follow-Up interview is administered, during which caregivers are asked more detailed questions about at-risk items to solicit examples of atypical child behaviors. A positive screen on the M-CHAT(-R)/F suggests elevated risk for ASD. This two-stage screening process yields a PPV of 0.48 for an ASD diagnosis and a PPV of 0.95 for any developmental delay or concern (Robins et al., 2014). The M-CHAT(-R)/F is the most widely used screener for ASD, and it has been recommended for ASD screening by organizations like the AAP and Autism Speaks (Duby et al., 2006; Ibañez, Stone, & Coonrod, 2014; Sturner et al., 2016).

Demographics and History Forms.

Forms developed by the research team were used to collect child race/ethnicity, maternal education, and annual household income. To ascertain income, caregivers were asked to select the category that best represented their annual household income; brackets ranged from “less than $10,000” to “greater than $100,000”. Alternatively, caregivers had the option to report their monthly household income, which was converted to yearly income. The final income metric consisted of 15 categories, with one representing the lowest annual household income (below $10,000) and 15 representing the highest annual income (greater than $95,000).

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

The MSEL is a developmental assessment of cognitive, motor, and language abilities in children aged one to 68 months. The current study used scores in the domains of Visual Reception, Fine Motor, Receptive Language, and Expressive Language. MSEL T scores were not normally distributed because a large number of children received the lowest possible standardized score (i.e., T = 20). In order to utilize parametric statistical tests without violating their assumptions, age-equivalent (AE) scores in each domain were converted to developmental quotient scores according to the formula mental age (i.e., AE scores) divided by chronological age, multiplied by 100 (Reitzel et al., 2013).

Vineland Adaptive Behavior Scales: Interview Edition, Survey Form (VABS; Sparrow, Balla, & Cicchetti, 1984) and Vineland Adaptive Behavior Scales, Second Edition: Survey Interview Form (VABS-II; Sparrow, Cicchetti, & Balla, 2005).

The VABS and its revision, the VABS-II, are semi-structured caregiver interviews that assess adaptive behaviors in the domains of Communication, Daily Living Skills, Socialization, and Motor Skills. Standard scores in all domains were normally distributed and were used to quantify adaptive functioning.

Autism Diagnostic Observation Schedule-Generic (ADOS; Lord, Risi, & Lambrecht, 2000) and Autism Diagnostic Observation Schedule, Second Edition: Toddler Module (ADOS-2; Lord, Luyster, Gotham, & Guthrie, 2012).

The ADOS and its revision, the ADOS-2, are semi-structured play-based assessments designed to diagnose ASD and measure autism severity. The current study used ADOS-2 Toddler Module, designed for pre-verbal children aged 12 to 30 months, ADOS Module 1, designed for children with no words or single words, or ADOS Module 2, designed for children who are not verbally fluent but have some phrase speech. Raw scores were converted into calibrated severity scores for Social Affect and Restricted/Repetitive Behaviors, as well as an overall severity score, based on published algorithms (Esler et al., 2015; Gotham, Pickles, & Lord, 2009; Hus, Gotham, & Lord, 2014). Severity scores range from 1 to 10, and higher scores indicate greater ASD severity.

Childhood Autism Rating Scale (CARS; Schopler, Reichler, & Renner, 1988) and Childhood Autism Rating Scale, Second Edition (CARS2; Schopler, Van Bourgondien, Wellman, & Love, 2010).

The CARS and its revision, the CARS2, are 15-item clinician rating scales measuring ASD symptom severity based on observation and caregiver report. A total score, based on the sum of individual items classifies a child as non-autistic, mildly-moderately autistic, or severely autistic.

Data Analytic Plan

To address our first aim, we calculated descriptive statistics (i.e., counts, percentages) to determine how many children fell into the Catch-24 group (i.e., screened negative at 18 months, but then screened positive at 24 months and were diagnosed with ASD).

To address our second aim, we compared participants in the Catch-24 group to children in the Early Diagnosis group to explore potential differences in demographics, cognitive functioning, adaptive skills, and ASD severity. Independent samples t-tests were performed to analyze continuous variables (i.e., age, MSEL, VABS(-II), ADOS(-2), CARS(2)); Chi square tests of independence or Fisher’s exact tests (in cases of small cell counts) were used to examine categorical data (i.e., gender, race/ethnicity, annual income).

All analyses were run using RStudio Version 3.

Results

Aim 1: Incremental Utility

Our first aim sought to examine the utility of rescreening children at 24 months who had previously screened negative on the M-CHAT(-R)/F at 18 months. Thus, our sample of interest for this question was narrowed to participants who screened negative at 18 months, either on the initial questionnaire or after Follow-Up, and were rescreened at 24 months (n = 7,781; light gray in Figure 2). Of these children, 0.4% (n = 32) then screened positive on the two-stage M-CHAT(-R)/F at 24 months, indicating risk for ASD (medium gray in Figure 2). As shown in dark gray in Figure 2, 10 of these children, or 0.1% of the entire rescreening sample, ultimately received an ASD diagnosis based on their positive 24-month screening; accordingly, these 10 participants make up our Catch-24 group.

Of the 32 children who screened negative at 18 months and positive at 24 months, 31% (n = 10) received an ASD diagnosis (i.e., Catch-24 group), 18% (n = 6) received a non-ASD developmental diagnosis, and 13% (n = 4) were judged to be typically developing. The remaining 38% (n = 12) of these participants were not evaluated following their positive 24-month screening for various reasons (e.g., lost to follow-up, refused to participate, significant motor impairment preventing accurate testing), and thus their diagnostic status could not be confirmed. However, even considering only children who attended an evaluation (n = 20), 80% received a developmental diagnosis (including ASD) warranting intervention, 50% were diagnosed with ASD, and only 20% were typically developing.

Aim 2: Comparison of Early Diagnosis and Catch-24 Children

Children in the Early Diagnosis group were compared to those in the Catch-24 group in order to understand how toddlers who received an ASD diagnosis based on their 18-month screening (n = 203) differed from those identified and diagnosed with ASD based on a positive 24-month screening, after a negative 18-month screening (n = 10).

Demographics.

At the time of evaluation, there was an approximately six month difference in child’s age between Catch-24 (M = 27.87 months, SD = 2.46) and Early Diagnosis (M = 22.32 months, SD = 3.21) children, consistent with the later timing of the Catch-24 group’s positive M-CHAT(-R)/F result (t(209) = −5.39, p = < .001). Groups did not differ on child gender, (χ²(1, N = 213) = .0007, p = .98), race/ethnicity (p = .35, Fisher’s exact test), maternal education (p = .22, Fisher’s exact test), or annual household income (Mann-Whitney U = 213.00, p = .51). Additional demographic information is presented in Table 1.

Developmental functioning.

On the MSEL, children in both groups scored in the borderline to impaired ranges across all areas of cognitive development. Groups did not differ on Fine Motor (t(183) = 1.26, p = .21, η² = .01) or Expressive Language (t(182) = .62, p = .54, η² = .002) skills. However, the difference between groups on Visual Reception (t(183) = 1.84, p = .07, η² = .02) and Receptive Language (t(182) = 1.93, p = .05, η² = .02) approached significance, with small effect sizes; children in the Catch-24 group performed lower, on average, than those in the Early Diagnosis group (see Table 2).

Table 2.

Developmental Functioning

Measure	Catch-24 Group	Early Diagnosis Group	t	df	p	η2
Mullen Scales of Early Learning
Visual Reception (n = 184)	59.12 (22.63)	71.50 (19.52)	1.84	183	.07	.02
Fine Motor (n = 184)	71.82 (12.11)	78.89 (16.55)	1.26	183	.21	.01
Receptive Language (n = 183)	33.51 (28.91)	46.83 (19.66)	1.93	182	.05	.02
Expressive Language (n = 183)	45.82 (28.17)	50.18 (20.21)	0.62	182	.54	.002
Vineland Adaptive Behavior Scales
Communication (n = 206)	71.00 (15.53)	69.95 (10.13)	0.31	205	.76	.0005
Daily Living Skills (n = 206)	81.50 (11.31)	78.24 (13.53)	−0.75	205	.46	.003
Socialization (n = 206)	77.30 (10.61)	76.39 (9.71)	−0.29	205	.77	.0004
Motor Skills (n = 206)	87.70 (11.25)	86.51 (11.13)	−0.33	205	.74	.0005

Note. All scores are presented as M (SD). Developmental quotients and standard scores are presented for MSEL and VABS(-II), respectively. Of the children missing data, one child in the Catch-24 group is missing MSEL results; all children in the Catch-24 group have complete VABS(-II) data.

VABS(-II) domains are also shown in Table 2. There were no significant differences between groups on Communication (t(205) = 0.31, p = .76, η² = .0005), Daily Living Skills (t(205) = −0.75, p = .46, η² = .003), Socialization (t(205) = −0.29, p = .77, η² = .0004), or Motor Skills (t(205) = −0.33, p = .74, η² = .0005) domains.

Autism symptom severity.

Calibrated severity scores (CSS) on the ADOS(-2) are presented in Table 3. ADOS(-2) Combined CSS significantly differed between the two groups (t(186) = −2.73, p = .01, η² = .04), with a small-to-medium effect, with greater overall ASD severity in the Catch-24 group. This difference appeared to be driven by a trend-level difference on ADOS(-2) Social Affect CSS (t(186) = −2.00, p = .05, η² = .02); again, children in the Catch-24 group had more severe scores, but with a small effect. There was no group difference on ADOS(-2) Restricted/Repetitive Behaviors (RRB) CSS (t(186) = 0.18, p = .89, η² = .0002), indicating that Catch-24 toddlers were no more likely to show RRBs than children in the Early Diagnosis group.

Table 3.

Autism Symptom Severity

Measure	Catch-24 Group	Early Diagnosis Group	t	df	p	η2
Autism Diagnostic Observation Schedule (n = 187)
Combined Severity Score	8.00 (1.31)	6.04 (2.01)	−2.73	186	.01*	.04
Social Affect	7.63 (1.41)	6.31 (1.83)	−2.00	186	.05	.02
Restricted/Repetitive Behaviors	5.88 (1.89)	6.06 (2.77)	0.18	186	.89	.0002
Childhood Autism Rating Scale (n = 204)	31.75 (4.40)	32.67 (5.29)	0.54	203	.59	.001

Note. All scores are presented as M (SD). ADOS(-2) calibrated severity scores (CSS; non-spectrum = 1–3, autism spectrum = 4–5, autism = 6–10); CARS(2) total scores (non-autistic = 15–29.5, mildly-moderately autistic = 30–36.5, severely autistic = 37–60). Of the children missing data, two children in the Catch-24 group are missing ADOS(-2) data, and all children from the Catch-24 group have CARS(2) data.

^*p <.05

On the CARS(2), there was no significant difference in autism severity between the Catch-24 and Early Diagnosis groups (t(203) = 0.54, p = .59, η² = .001). Mean total scores by group are presented in Table 3.

All analyses were rerun on a sub-sample matched for gender, maternal education, and race/ethnicity (n = 20). Results remained consistent, except that MSEL Visual Reception scores changed from a trend-level difference between groups to not being statistically significant (t(16) = 1.57, p = .14).

Discussion

Our primary study aim was to investigate the incremental utility of rescreening children for ASD at 24 months after they screened negative at 18 months in the course of routine pediatric care. There is extensive literature documenting the benefits of routine ASD-specific screening, both in leading to early intervention that improves outcomes and in decreasing inequities in age at diagnosis between minority and non-minority children (Crais et al., 2014; Gupta et al., 2007; Herlihy et al., 2014; MacDonald et al., 2014; Nadel & Poss, 2007; Pandey et al., 2008; Rogers et al., 2014). However, no research to date has specifically evaluated the utility of repeated screening, per AAP guidelines. In light of the increasing brevity of pediatric well-child care visits and the number of tasks to be completed during that limited time, it is important to evaluate whether screening children at both 18 and 24 months yields results worthy of the additional effort. The benefits of early identification and intervention dictate screening at 18 months; therefore, understanding what value is added from rescreening children at 24 months who screened negative at 18 months is an important question.

In the current sample, 0.4% of children who screened at both 18 and 24 months screened positive at 24 months following a prior negative screening (i.e., 32 out of 7,993). Of these 32 children, 16 were diagnosed with a neurodevelopmental disorder (i.e., 10 ASD, six non-ASD), four were judged to be typically developing, and 12 refused an evaluation or were lost to follow-up. Therefore, although only a small percentage of children who rescreened at 24 months were diagnosed with ASD (i.e., 0.1%; 10 out of 7,993), 80% of the children who were evaluated based on a second screening met criteria for a neurodevelopmental delay that warranted intervention. Given this high percentage, it is likely that some of the 12 toddlers who were not evaluated would also have demonstrated symptoms that warranted a diagnosis and intervention. Although the yield of ASD cases detected based on repeated screening was small in absolute terms, one could posit that even a small number of children identified earlier than might otherwise occur is worth the necessary effort. It should be noted that rescreening at 24 months is important because the frequency of regular well-child care visits begins to decrease after the child is 24 months of age (Duby et al., 2006).

The second aim of this study compared children in the Early Diagnosis group, diagnosed with ASD after 18-month screening, to Catch-24 participants, who were identified after rescreening at 24 months. The Catch-24 group was approximately six months older at the time of evaluation than the Early Diagnosis group, as would be expected. The two groups did not differ by demographics such as gender, race/ethnicity, maternal education, or annual income, nor did they significantly differ on overall developmental functioning. Both groups were delayed on both the MSEL and VABS(-II). Although we observed trend-level differences on MSEL Visual Reception and Receptive Language skills, with the Catch-24 children exhibiting slightly more substantial delays than the Early Diagnosis group, the effects were small and, overall, groups performed similarly. Furthermore, the two groups did not differ on adaptive skills as reported by their caregivers using the VABS(-II). Regarding ASD severity, children in the Catch-24 group differed from the Early Diagnosis group on the ADOS(-2), primarily due to increased social deficits (but not RRBs) in the Catch-24 group. The size of this effect was small, however, and not supported by groups’ performance on our other measure of ASD symptom severity, the CARS(2). Accordingly, this result may be a measurement artifact and should be interpreted with caution, pending replication.

There are several reasons that children might be identified at 24 months following a negative 18-month screening. One possibility is that children in the Catch-24 group had a regressive type of ASD, and as a result, the positive 24-month screening reflected loss of previously acquired skills or emergence of ASD symptoms between the two screenings. Indeed, this pattern is reflected in their increasing total scores on the M-CHAT-(R)/F from 18 (M = 1.3) to 24 (M = 6.4) months. This difference in scores suggests an increase in atypical social behavior for these children between the two time points, as reported by caregivers. Extant literature supports that between 20% and 40% of children with ASD present with a regressive subtype of ASD (Barton et al., 2012; Kalb, Law, Landa, & Law, 2010; Ozonoff et al., 2008). In the case of regression, screening results at both time points would be considered valid. However, this study did not include a reliable measure of regression as part of the evaluation, as children were not evaluated at both time points, and caregiver report of regression does not always correlate with objective evidence of loss of skills (Ozonoff et al., 2010). Therefore, future research is necessary to examine this hypothesis.

A second possibility is that some children begin developing typically, but then reach a plateau of skill development at approximately two years of age (Kalb et al., 2010; Ozonoff et al., 2008). These children maintain the skills that were previously acquired, but they then fail to develop further skills, causing the gap between expected and actual development to widen over time. As such, the Catch-24 participants may have screened negative at 18 months because they were developing appropriately at that time, yet then screened positive at 24 months due to a greater discrepancy between their developmental level and that expected of a typical two-year-old child.

A third possibility is that child or family factors that we were unable to assess may contribute to some children being detected later based on screening. For example, caregiver readiness to acknowledge developmental difficulties, misinterpretation of M-CHAT(-R) questions, and limited knowledge of typical and atypical development may all contribute to caregivers not indicating concerns at 18 months (Gupta et al., 2007; Johnson & Myers, 2007). Caregivers may also become increasingly aware of their child’s delays as the child gets older, due to increased opportunities to observe their child interacting with typically-developing peers and due to the widening gap in social skills between typical and atypical development by two years of age. Even when caregivers are concerned, they sometimes delay discussing their observations of their child’s atypical behaviors with their pediatrician for several months, perhaps because they fear that the pediatrician will dismiss their concerns (Johnson & Myers, 2007) or provide a passive or reassuring response (Zuckerman, Lindly, & Sinche, 2015). The AAP acknowledges this issue by stating that a lack of overt caregiver concerns, as communicated directly to the child’s healthcare provider, “does not preclude the possibility of serious developmental delays” (Duby et al., 2006, p. 408).

All of these reasons suggest the importance of rescreening at 24 months, following a negative 18-month screening. The clinical benefits of identifying a greater number of children with ASD, particularly those who may have a regressive or otherwise severe subtype, may offset some of the barriers to rescreening. In addition, there are practical benefits to repeated screening. Specifically, the PPV of 24-month screening with the M-CHAT(-R)/F is higher than that at 18 months (Pandey et al., 2008; Sturner et al., 2017), thereby reducing the number of false positive screens.

Limitations

There are some limitations to this study that should be noted. It was not feasible to systematically evaluate children who screened negative on the M-CHAT-(R)/F, thus we do not know how participants in the Catch-24 group would have presented in terms of their developmental functioning and ASD symptoms at 18 months. Such information would have helped us better understand the reason children screened negative at 18 months but then positive at 24 months. Specifically, evaluating children at both time points would have enabled us to compare performance and determine if a child lost previously acquired skills (i.e., regression), failed to gain new skills between 18 and 24 months (i.e., plateauing), or if symptoms were present even though they were not reported by the caregiver at the 18-month screening.

Additionally, a number of children at both time points screened at-risk but were not available for evaluation. This may bias our results, as there are likely meaningful differences that we were unable to ascertain between children who presented for an evaluation and those who did not. For example, caregivers who decline a free evaluation for their child may not be ready to address their child’s developmental difficulties, have less awareness about the milestones expected in typically developing children, or feel less urgency because their child has fewer symptoms. Relatedly, since 12 of the 32 children who screened positive at 24 months did not attend an evaluation, we do not know whether the 80% of children diagnosed with a neurodevelopmental disorder would be an over- or under-estimate of the children who screened positive, if this were extrapolated to the general population. Specifically, it is possible that some children who did not attend the evaluation would have met criteria for a diagnosis, in which case our calculation may be an under-estimate. On the other hand, caregivers may have declined an evaluation for their child if they thought that their child did not exhibit concerning symptoms, in which case extrapolating the number of ASD cases based on repeated screening at 24 months could be an over-estimation.

Although we started with a large sample (i.e., over 27,000 children), narrowing our analyses to children who screened negative at 18 months, then screened positive at 24 months, and ultimately were evaluated, produced small sample sizes for our second aim (i.e., n = 10 for Catch-24 group; n = 203 for Early Diagnosis group). Although the ratio of children diagnosed with ASD and other developmental disorders is consistent with other M-CHAT(-R)/F screening literature and with the prevalence of ASD, comparable data from other studies is necessary in order to obtain a stable estimate of how many children with ASD and other developmental disorders can be detected with repeated screening.

Further, socioeconomic status was not available for a large number of participants. For example, annual household income was not reported for 60% of the Catch-24 group and 35% of the Early Diagnosis group. Although we cannot state with certainty the reason for this missing information, we can speculate that this was a sensitive question or a low-priority item for caregivers towards the end of a lengthy Demographics and History form. However, these missing data may signify a meaningful pattern. For instance, it is possible that individuals with lower incomes may have skipped this question. Socioeconomic status has important implications for screening, since lower socioeconomic status predicts later age at evaluation and greater ASD severity (Herlihy et al., 2014). Thus, in future studies, it would be useful to make efforts to obtain this information from a larger number of participants.

The Catch-24 group included children who screened negative on the M-CHAT-(R)/F screening at 18 months (n = 7,424), as well as those who initially screened positive but then screened negative on the Follow-Up interview (n = 357). These children form important study sub-groups, with potentially unique and informative developmental trajectories. For example, it is possible that caregivers of children who initially screened positive at 18 months, but then had a negative screen following the Follow-Up interview (n = 3), could either have been reluctant to disclose the extent of their child’s delays to an unknown interviewer by telephone, or could have misinterpreted M-CHAT(-R) questions initially and subsequently changed their responses when provided with additional examples and probes on Follow-Up. However, these sub-groups within the Catch-24 sample were too small to allow for analyses of any potential differences between children who screen negative “outright” on the M-CHAT(-R) versus those who screen negative following a structured Follow-Up telephone interview; this would be an interesting and informative area of further study.

Finally, a few factors limit generalizability of our results to other populations. First, the overwhelming majority of our sample was White. This may limit generalizability of screening results to non-White toddlers and their caregivers, as minority families are less likely to complete the M-CHAT(-R) Follow-Up telephone interview (Khowaja, Hazzard, & Robins, 2014). Still, research suggests that screening with the M-CHAT(-R) leads to more timely diagnosis and helps reduce age at diagnosis across racial groups (Herlihy et al., 2014).

Second, participants in this study who screened positive on the M-CHAT(-R) received the Follow-Up telephone interview, which was administered with fidelity by trained research staff. However, in community settings, the M-CHAT(-R) Follow-Up interview may be administered differently, and sometimes is not administered at all. On the other hand, pediatric practices are increasingly moving towards electronic screening; eliminating the lag between M-CHAT(-R) administration and Follow-Up may streamline the screening process and reduce the number of families who are lost to follow-up after a positive screening result (Brooks, Haynes, Smith, McFadden, & Robins, 2016).

Third, participants were provided with a timely, high quality diagnostic evaluation if they screened positive on the M-CHAT(-R)/F; unfortunately, this service is unavailable to many children who screen at-risk for ASD. Delay to diagnosis, and the difficulty of connecting families to appropriate evaluation services, compounds the stress that pediatricians associate with screening. Indeed, pediatric providers report that a lack of office-based systems for making referrals, helping their patients receive evaluations, and monitoring outcome are among the primary barriers to screening (Zwaigenbaum et al., 2015). This may all lead to difficulties adhering to the dual screening model recommended by the AAP. More resources are needed to assist pediatricians in the screening and referral process, as well as increased access to timely, high quality diagnostic and intervention services for children who screen at-risk for ASD.

Conclusion

Despite the challenges posed to busy pediatricians in the administration of two ASD-specific screenings in addition to repeated general developmental screening and ongoing developmental surveillance, the opportunity to identify additional children who would benefit from early intervention services, in our view, far outweighs the risks and possible inconveniences of executing this repeated ASD screening procedure. This study supports the current AAP recommendations that all children be routinely screened at both their 18- and 24-month well-child care visits, in addition to the developmental surveillance that should be occurring at every visit. Screening at 18 months is important, as symptoms of ASD often emerge by this age and can be reliably diagnosed, thereby facilitating early intervention. Screening at 24 months is also critical for children who regress, plateau, or are missed at earlier ages for various other reasons that remain undetermined. In addition, the PPV of 24-month screening with the M-CHAT(-R)/F is higher than that at 18 months (Pandey et al., 2008; Sturner et al., 2017), thereby reducing the number of false positive screens. Screening at both time points may be particularly feasible as pediatric practices increasingly move to electronic screening (Brooks et al., 2016). Our findings suggest that when this repeated ASD-specific screening procedure is carried out, the number of children diagnosed with ASD and other neurodevelopmental conditions requiring early intervention services is maximized, thus promoting improved outcomes.