Improving Early Identification and Access to Diagnosis of Autism Spectrum Disorder in Toddlers in a Culturally Diverse Community with the Rapid Interactive screening Test for Autism in Toddlers

The Early Identification of Autism Spectrum Disorder (ASD) is an evolving research area. Benefits of early intervention on improving the course and outcomes of the disorder are well known (Suma et al. 2016). However, ASD is still not detected early enough in the United States (Baio et al. 2018), and especially for children from culturally diverse and underserved communities (Elder et al. 2016; Zuckerman et al. 2014). While there are significant basic research advances occurring, such as eye-tracking studies (Wan et al. 2018), electroencephalogram (EEG) markers (Bosl et al. 2011), and genetic or biochemical biomarkers (Bridgemohan et al. 2019; Yuen et al. 2017), these approaches are still far from clinical application and generalization. Improving the early identification of ASD remains largely a clinical challenge.

The American Academy of Pediatrics continues to support screening for ASD in primary care (Johnson and Myers 2007). Other programs, such as the Centers for Disease Control (CDC) “ Learn the Signs. Act Early” Campaign, launched in 2004, have developed educational material targeting early childhood providers and parents. These materials are designed to improve developmental monitoring and the detection of early signs of ASD (Gadomski et al. 2018; “Learn the Signs. Act Early” CDC website 2019). Improving the early identification of ASD and access to diagnosticians, are urgent public health goals, as they will lead to early intensive interventions to improve outcomes. Currently, ASD diagnosticians are in very limited supply and overwhelmed with the volume of referrals. Patient wait times for evaluations vary by state, with reports of 12 months or more in many areas of the United States (Austin et al. 2016). Referrals and wait times are typically further increased among culturally diverse children with limited options through health insurance and other obstacles to their care (Rea et al. 2019). In addition, wait lists, for diagnostic evaluations in toddlers with ASD, include other young children who present with developmental delay without the presence of ASD. The merging of children, who do not have ASD with those who do, dramatically slows the process for those who will need more specialized intervention than existing early childhood programs can provide.

In our clinical center, the average wait time for a toddler aged 18–36 months referred for evaluation because of concerns of ASD, varies between 3 and 5 months with an average wait of 4 months or 16 weeks. To improve access and early identification of ASD in our culturally diverse community, we implemented and assessed a clinical approach that utilizes a two-level ASD screening model, in combination with a close partnership with the largest Early Intervention (EI) program in our region.

The research team reached out to the THOM Early Intervention (EI) Program, the largest EI program in Worcester, MA where this study was completed. The town of Worcester, MA has a culturally diverse population, with 21% of the population living below the federal poverty level and close to 35% with a primary language other than English (US Census Bureau QuickFacts 2019). We trained four THOM EI providers reliably on the RITA-T. The training was completed in person and inter-rater reliability was calculated. THOM was starting an internal training for their staff on the MCHAT-R/F at the same time. Toddlers, 18–36 months of age, enrolling in EI or already in EI, were administered the MCHAT-R/F as part of their evaluation or re-evaluation within 6 months, or if their provider had concerns for ASD. The MCHAT-R/F was administered in an interactive way, with parent and provider discussing each question. For this study, after consenting the families as per our Institutional Review Board approved protocol, when the MCHAT-R/F had a score above 2, or if the child’s providers had concerns for ASD, the RITA-T was administered by one of the trained providers. Results of the MCHAT-R/F and the RITA-T were sent to the research assistant (RA) in the research team in a sealed envelope. Those results were then entered in the study database without the PI or the research psychologist being aware of them. The family’s EI provider then began the conversation with the family about concerns for ASD. Children and their families were then referred to the research psychologist on the team, who was blinded to the results of the MCHAT-R/F and the RITA-T. The psychologist obtained a brief developmental history, observed the DSM-5 criteria for ASD and administered the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) (Lord and Rutter 2012) as well as the Mullen Scales of Early Learning (MSEL) (Mullen 1995). The ADOS-2 Toddler module or Module 1 was administered depending on the age of the child. The EI provider and the RA scheduled the child and his/her family for an appointment with the PI after the child was evaluated by the research psychologist.

The research psychologist met with the Principal Investigator (PI) on a weekly basis and reviewed autism and developmental testing with the PI without disclosure of previous MCHAT R/F and RITA-T results. A “best estimate diagnosis” was made based on diagnostic tests and history obtained by the research psychologist. Two groups were identified: toddlers with ASD diagnosis and those with a non-ASD/Developmental Delay (non-ASD/DD) diagnosis. Autism and developmental testing results and algorithms scoring documents were shared with the RA who then entered them in the database. At the time of the appointment with the PI, the PI met with the family, child, EI provider, and an interpreter as needed. The PI then completed a medical and developmental history, observed behavior and play, completed a physical exam and reviewed previous test results and current observations with the family. All previous testing was then available to the PI, including previous results of the MCHAT R/F and the RITA-T. The PI then reviewed and discussed diagnoses, recommendations and services while providing a letter to initiate services. The best estimate diagnosis was provided to the RA who entered it in the database. A follow up was arranged within 1–3 months with a family support specialist in our center.

We collected data on time-in-days between referral from EI after administration of RITA-T to meeting with the PI for review of testing and final diagnosis and compared this with wait time to be evaluated if a toddler was referred outside of this model to our center. To further enrich the sample of toddlers enrolled from EI, we also enrolled 12 toddlers, ages 18–36 months, from EI who had no concerns for ASD. They were administered the RITA-T, the Battelle Developmental Inventory, Second Edition (BDI-2) (Newborg 2005) as per EI protocol, and a brief developmental history was obtained. In addition, a DSM-5 checklist and a MCHAT-R/F were completed by their EI provider and family respectively.

Pediatricians in two large practices in Worcester, one hospital based and one large private practice, were made aware about this project. Both practices referred children with developmental concerns to THOM EI. Both practices administered the MCHAT-R/F at 18 and 24 months and usually referred as well to our center for evaluations. When toddlers were administered the RITA-T by EI trained providers, their pediatrician was made aware of the results and the referral for an evaluation by the research team.

Over a period of 12 months, 81 toddlers were enrolled in this project. Of those, 57 received a diagnosis of ASD and 24 had a diagnosis of non-ASD/DD. The groups had similar gender distribution and age in months; race and income were representative of the area where this study was completed and were similar in the two groups. The RITA-T mean scores were significantly higher in the ASD group than in the non-ASD/DD group with a mean score of 20.1 (SD = 3.9) in those diagnosed with ASD and a mean score of 9.7 (SD = 2.9) in those diagnosed as non-ASD/DD (p < 0.0001). The MSEL composite scores were also significantly different between groups, which can be interpreted as meaning that those ultimately diagnosed with ASD demonstrated greater delays than those without a diagnosis of ASD. The ADOS-2 total scores were significantly different between both groups [ASD: 17.4 (SD = 3.9) vs. non-ASD/DD/: 3.9 (SD = 2.8), p < 0.0001]. The MCHAT R/F mean scores were significantly different between those in the ASD group vs. those in the non-ASD/DD group. Interestingly, among those with MCHAT R/F > 2, 50 out from the 57 children in this study were diagnosed with ASD yielding a PPV of 87.7% which is much higher than the previously reported PPV in a high-risk group of 61–79% (Pandey et al.2008). Furthermore, the MCHAT R/F scores seemed to be associated with a diagnosis of ASD in this current study independently from the RITA-T scores (Table 1).

In looking at the relationship between the ADOS-2 and DSM-5 with the RITA-T, we assessed the correlation between their scores. The Pearson correlation coefficient for ADOS (either Module 1 or Toddler) by RITA-T was 0.70 (p < 0.0001). This illustrates that as the RITA-T total score was increased, the ADOS-2 total score also increased, and the relationship between the two scores was moderately strong. In order to investigate the association between ASD diagnosis and RITA-T score, we modeled diagnosis as a function of RITA-T score. Using unadjusted logistic regression to predict ASD diagnosis, we found that the odds of being diagnosed with ASD is almost 3 times as likely when compared to non-ASD/DD [OR 2.96, 95% CI (1.42, 6.15)].

To determine the best cut-point for RITA-T in determining ASD diagnosis, we investigated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). These statistics were calculated for each possible RITA-T score (Table 2).

In our sample, a cut-off score of 16 had best PPV (1) and specificity (1), meaning that all children with a score higher than 16 were ultimately diagnosed with ASD.

Wait time between referral to final diagnosis varied between 14 to 105 days with a mean of 6 weeks. Wait time in this center for those 18–36 months referred with a question of ASD and not part of this current model varied between 3 and 5 months with an average of 4 months or 16 weeks. We also looked at inter-rater reliability and asked the 4 EI providers to each independently score 3 videos of administration of the RITA-T. Scores were compared with the PI’s scores and results showed an inter-rater reliability of 80% for total scores among each provider and among each provider and the PI.

A short survey was completed with EI providers about the usefulness of this model and for feedback. The RITA-T allowed providers to: (1) evaluate for the constructs delayed in ASD; and (2) empowered them to start a conversation with the family regarding concerns of ASD.

This is the third in a series of studies to determine the RITA-T sensitivity, specificity and predictive value. The initial study (Choueiri and Wagner 2015) was exploratory and hospital based entirely. That study included 61 toddlers, 19 of which had no developmental concerns at the time of the study. The second RITA-T study looked at the psychometrics of the RITA-T in 239 toddlers within a hospital based triaging model including the RITA-T (Lemay et al. 2018, 2020). In the current study, we included the RITA-T in a two-level screening model and partnered with an Early Intervention program in this culturally diverse community. Specifically, our goals were to study the feasibility and improved access in those children 18–36 months old with concerns for ASD. The current study showed improvement with access to diagnosis with an average wait time of 6 weeks compared to an average of 16 weeks in those children referred for an ASD evaluation who were not part of this model. The RITA-T identified those at-risk for ASD and diverted them in a timely manner for a formal ASD diagnostic evaluation and subsequent specialized services. Reliable training and inter-rater reliability were achieved after 3 h of in-person training. In addition, the RITA-T correlated strongly with autism diagnostic measures, including the ADOS-2, DSM-5 criteria for ASD, and clinician judgment.

It is interesting to note the difference in developmental levels between both groups. The MSEL was used as a measure of developmental levels. This difference could be because the MSEL relies on language and the child to be focused during its administration, which can be challenging in those with an ultimate diagnosis of ASD. A cognitive factor, however, has been reported in previous studies looking at autism screening and diagnostic measures which highlights the complexity of autism clinical presentations and developmental delays. This may be especially important in those children from different cultures (Esler et al. 2017; Stone et al. 2004). Previous studies have also shown that children of immigrants from countries with a low human resource index have a higher incidence of intellectual disability and ASD compared to other groups (Esler et al. 2017). Thus, it is essential that autism evaluations include medical and developmental history, clinical presentation, screening results, and necessary additional diagnostic testing to provide the best-estimate clinical diagnosis.