Implementing the Get SET Early Model in a Community Setting to Lower the Age of ASD Diagnosis

Abstract

Objective:

To implement a validated, university-based early detection program, the Get SET Early model, in a community-based setting. Get SET, was developed to improve Screening, Evaluation, and Treatment referral practices. Specifically, its purpose was to lower the age of diagnosis and enable toddlers with autism spectrum disorder (ASD) to begin treatment by 36 months.

Methods:

109 primary healthcare providers (PHP) were recruited to administer the CSBS-IT Checklist at 12-, 18- and 24-month well-baby visits and referred toddlers whose scores indicated the need for a developmental evaluation. Licensed psychologists were trained to provide diagnostic evaluations to toddlers as young as 12 months. Mean age of diagnosis was compared to current population rates.

Results:

In 4-years, 45,504 screens were administered at well-baby visits and 648 children were evaluated at least one time. The median age for ASD diagnosis overall was 22 months, significantly lower than the median age reported by the CDC (57 months). For children screened at 12 months, the age of first diagnosis was significantly lower at 15 months. Of the 350 children who completed at least one follow up evaluation, 323 were diagnosed with ASD or another delay, and 239 (74%) were enrolled in a treatment program.

Conclusions:

Toddlers with ASD were diagnosed nearly 3 years earlier than the most recent CDC report, which allowed children to start a treatment program by 36 months. Overall, Get SET Early was an effective strategy for improving the current approach to screening, evaluation, and treatment. Efforts to demonstrate sustainability are underway.

Despite research indicating that symptoms of autism are present as early as 12-months^1–6, and that autism can be diagnosed soon thereafter ^4,7, the median age of diagnosis is 51 months ⁸ and, in many states, even older. Reasons for late diagnosis include limited understanding of warning signs among healthcare professionals and parents, and limited access to diagnostic evaluations ^9–10. Since diagnosis enables access to early intensive intervention, which improves long-term outcomes ^11–14, effective screening practices are critical.

While research on improving early identification of ASD tends to focus on the efficacy of screening tools such as the M-CHAT, screening merely creates the opportunity for early diagnosis and intervention ¹⁵. The American Academy of Pediatrics (AAP) recommends screening at 18- and 24-month visits ¹⁶. Approximately 50% of pediatricians routinely screen with validated tools, but they often do not refer a toddler with a failed screen for an evaluation ^17–18. Other pediatricians rely on clinical judgment or parent concerns, which is problematic given that standardized screening is more accurate ¹⁹. Possible reasons for not using a screening tool include lack of time (e.g. to administer screens and discuss with parents), uncertainty about autism-specific screening tools, reimbursement issues, and limited referral options ^{9, 20–22}. Without specific training and procedural guidance, the benefit of screening may not be realized.

Improving the screening to treatment pathway will allow more children to benefit from early intervention. University researchers developed the 1-Year Well-Baby Check-Up Approach ³ to reduce the mean age of ASD detection and treatment referral. The project demonstrated the direct benefits of minimal PHP training and providing a clear path for an evaluation based on the results of a screening questionnaire rather than clinical judgment. Several key elements lead to the success of this approach. For one, PHPs and autism specialists formed a unified goal to improve early detection. Relationships among various professionals can overcome barriers¹⁹, increase pediatrician awareness of detection methods, and improve the referral process ²¹. Use of a validated screening tool to identify developmental delays was also vital ²¹.

While effective in reducing the age of diagnosis, screening only at 12 months may have missed a substantial percentage of children who would later show delays. To improve the program’s efficacy, Pierce and colleagues revised this approach and developed the Get SET Early model ²³, where S = Screen, E=Evaluate, and T = Treat, and introduced repeat screening at the 12-, 18-, and 24-month well-baby visits. Using this revised model, the overall mean age of 1^st diagnosis and treatment referral was 19 months, and for infants screened at 12 months the mean was 15 months. While highly effective, such successes may be due to the broad resources and experience often associated with large university-based programs. To have widespread positive effects, a variety of centers that serve the community at large must be able to implement the model.

In the present study, Get SET Early was implemented from 2014 to 2019 at a community-based center for autism treatment and research in a city that consistently has one of the highest median ages of diagnosis according to the Autism and Developmental Disorders Monitoring Network (ADDM) since 2002. In the 2016 surveillance year, the prevalence in Arizona was 1 in 63 and the median age of diagnosis was 57 months, the highest age among the 11 states in the network ⁸. The center is a non-profit that serves only the autism community and is sustainable on philanthropy, insurance revenue, and research grants and contracts.

Successful implementation involved developing a network of PHPs, creating an expert autism diagnostic program to evaluate referred toddlers, and providing families with rapid referrals for treatment when appropriate ²³. The model was evaluated by investigating: 1) the change in PHP screening perceptions and behaviors; 2) the number of children screened and evaluated; 3) mean age of first diagnosis, and 4) the number of children who began treatment before 36 months.

METHOD

Participants

Pediatric Healthcare Providers (PHP).

Study staff recruited PHPs by contacting practices that had at least one office within a 30-mile radius of the study site. Areas with racial and ethnic diversity were deliberately targeted. PHPs who responded with interest were recruited. The network included 109 PHPs (79 pediatricians, 17 nurse practitioners, 9 physician assistants, and 4 registered nurses) from 13 practices (30 locations). According to 2017 census data by zip code, the median household income ranged from $20,000 to $47,000 for 11 offices, $50,000 to $70,000 for 11 offices, and $71,000 to $121,000 for 8 offices.

A comparison network included 18 independent PHPs (6 practices). Consent was obtained, but demographics of PHPs were not collected as the intent is to compare screening procedures only. To avoid bias that could inflate the numbers of screenings performed by PHPs in this network, they were given no knowledge that the studies’ purpose was to improve screening and detection of ASD.

Evaluated Children.

Children were considered enrolled once parents provided informed consent to participate in the study. Psychologists evaluated 648 children and all were invited back for an annual follow up evaluation until the child turned 36 months; 350 children completed a second evaluation.

Procedures

Network PHP Practice Measurement.

To measure changes in PHP behaviors and attitudes about screening and resources for individuals with ASD, all PHPs in the network (n = 109) were asked to anonymously complete Physician Screening Practices Questionnaire (PSPQ) pre- and post-model implementation twice. Pre-implementation, 78% (n=85) completed the PSPQ, and two-year post implementation 53% (n=58) of network PHPs completed the questionnaire.

Comparison PHP Practice Measurement.

To quantify actual screens (and not attitudes), only comparison network PHPs completed this questionnaire at each 12-, 18-, and 24-month well visit for a period of 1 year.

Get SET Early Training and Establishment of the Model.

The site investigator provided training on implementing Get SET to pediatricians, front desk personnel, nurses, and MAs during a lunch hour, including the determination of a positive screening. Components of Get SET (Pierce et al., 2021) were reviewed (Supplemental Figure 1) as were the early warning signs of autism. Get SET includes guidelines for execution and monitoring of progress. Key components include: 1) a PHP network from diverse communities that screens at 12-, 18-, and 24-month well visits with a validated broadband screening tool rather than an ASD-specific tool or clinical judgment alone; 2) a rapid referral of failed screens to an expert evaluation center; and 3) treatment referral.

Screening and Referral.

Network PHPs agreed to screen all children at the 12-, 18-, and 24-month well-visits. Front desk personnel distributed it to parents at check in. Parents completed the questionnaire in the waiting room and signed the consent form to share their contact information with the study center. MAs scored it, marked the areas of concern, and then put it in the patient’s chart for PHP review with the parent. PHPs referred children who failed the screen to the study by providing an informational card about the evaluation to the parent and encouraged them to call for an appointment. For parents who did not call within 2 weeks, study personnel called to schedule the evaluation. As a result, all evaluations were scheduled within two months of screening. Further, network pediatricians also referred children between 25 and 36 months (beyond the validated age range of the CSBS) to the center for an evaluation if ASD was suspected.

Creation of an Infant/Toddler Autism Evaluation Center.

A central tenet of Get SET Early is an evaluation center with expertise in diagnosing ASD in toddlers. A licensed psychologist performed evaluations. The lead psychologist at the San Diego Get SET site (C.C.), a credentialed ADOS trainer, established reliability with the lead psychologist at the study center (E.S.).

Measures

Physician Screening Practices Questionnaire (PSPQ).

This 30-item survey was designed for this study to assess PHP’s screening practices (e.g. Please identify your method of screening (i.e. clinical judgment, screening tool); beliefs and confidence about detecting ASD at the 12-, 18-, and 24-month well visits (e.g. In your opinion can [autism, developmental delays, or language delays] be detected at 12 months?), and referral practices (e.g “Where do you refer for developmental evaluation?”).

Comparison Screening Procedures Questionnaire (CSPQ).

Comparison network PHPs anonymously completed a 9-item survey developed for this study that assessed their standard well-visit procedures (e.g. Was height and weight measured? Were immunizations given? Did you perform developmental screening?). It also collected information about a parent report of concerns, pediatrician concerns, and referral for an evaluation.

Communication and Symbolic Behavior Scales Developmental Profile Infant-Toddler Checklist (CSBS-ITC).

Parents completed this scale on paper to screen children for developmental delays. This broadband screening tool detects various delays, including autism ²⁴. The 24-item parent-report screener examines social communication, expressive speech/language, and symbolic functioning, can be completed in approximately 5 minutes and scored in approximately 2 minutes. The CSBS-ITC has high sensitivity and specificity for detecting delays related to autism spectrum disorders ²⁴ with a positive predictive value of 75% ³.

Diagnostic Evaluations.

Evaluations included the Autism Diagnostic Observation Schedule (ADOS-2; Module T, 1, or 2 ^25–27; Mullen Scales of Early Learning (MSEL ²⁸); and Vineland Adaptive Behavior Scales, Second Edition (VABS-II; Sparrow, Cicchetti, & Balla 29) across a 3–4 hour visit as necessary. Evaluations resulted in one of six diagnostic judgment categories: 1) ASD; 2) ASD-features (prominent DSM-5 ASD criteria, but not fully satisfied); 3) language delay (receptive and/or expressive); 4) developmental delay (two domains delayed) 5) global developmental delay (three domains delayed) but did not meet full criteria for ASD; and 6) typical. Any child showing signs of ASD or non-ASD delays were referred for treatment at the end of the visit. Families were asked to return to the Center for follow-up evaluations every 9–12 months until the child turned 3 years.

Treatment Survey.

Treatment surveys used in previous work ³ were sent to enrolled families prior to all follow-up appointments. Caregivers were asked to report type of intervention or service received, treatment provider and setting, start date, treatment frequency, and opinion of progress for each of the following four domains: general therapies (e.g. speech, physical, occupational therapy), autism treatments (e.g. ABA, Pivotal Response Training, social skills classes), alternative therapies (e.g. special diet, animal therapy), and education (e.g. typical class, special day class, toddler school).

Data Analysis

Changes in PHP attitudes and behaviors were determined by observing pre- and post-implementation responses on the PSPQ. Actual screens at well-visits in the comparison network were measured with the CSPQ and the likelihood of screening at each age group was determined with chi-square. Percentages of screening, evaluation, and diagnostic outcomes by screening age group documented results of Get SET implementation, and results were compared to metrics generated by the ADDM network.

RESULTS

PHP Attitudes and Behaviors – Get SET and Comparison Network PHPs

Survey Results, Get SET.

At baseline, 85 of 109 network PHPs completed the PSPQ (see Supplementary Table 1 for PHP characteristics). Respondents included 70% pediatricians, < 50 years of age, and in practice 1–20 years. While 64% believed it was possible to detect autism at 12 months, only 15% used a screening tool (n = 13), and 24% (n = 20) referred children for evaluation. At the two-year follow-up, 58 PHPs completed the PSPQ. Behaviors associated with the 12-month visit showed notable improvement. While PHPs who believed autism could be identified at 12 months increased by only 5%, use of a screening tool increased to 74% (n = 63 of 85), and evaluation referral increased to 79% (n = 45 of 58) (Figure 1). Even after implementing Get SET, the majority of PHPs still believed there were inadequate resources for diagnosis and treatment (despite a marginal increase from 8 to 12% across all age groups; see Supplementary Figure 2). Results must be interpreted with caution due to the change in sample size from pre to post sampling.

Figure 1.

Effect of the Get SET Early model on screening habits of pediatric healthcare providers (PHPs) at the 12-, 18-, and 24-month well visit. Due to the change in sample size from pre to post, % change may be overstated. Results should be interpreted with caution. Figure A. shows the change in percent of PHPs who reported using a screening tool at each visit. Figure B. shows the change in percent of PHPs who referred for an evaluation at each visit.

Screening Effects: Network and Comparison PHPs

Network PHPs (n = 109) performed a total of 45,504 screens: 40,760 screenings from well-visits at 12-, 18-, and 24-months, and an additional 4,744 children between 25 and 36 months with suspected developmental delays. Approximately 6,150 screens (13.5%) indicated concerns and should have prompted a referral. As screening occurred at multiple well-visits, “total screenings” does not represent unique individuals (best estimate = 27,832). Children over 24 months were referred because Get SET was designed to get children evaluated and in treatment by 36 months, therefore PHPs referred children up to age 36 months. The greatest number of screens (16,714) occurred at the 12-month visit (Figure 2), roughly 35% more than the 18- and 24-month visit. Of the total sample screened, 46% gave consent to be contacted by the study team and provided contact information, 2,681 failed screening, and 832 (31%) expressed interest in participating in the developmental evaluation.

Figure 2.

Flow chart of participants in the Get SET Early Model implementation. 1 = 45,504 screens does not represent unique children.

Comparison PHPs completed 1,319 CSPQs: 36% were completed for the 12-month well visit, 29% for the 18-month visit, 32% for the 24-month visit, and 3% over 24 months (see Supplementary Table 2). Chi square analyses indicated a significant association between age group and referral for an evaluation (Χ² (3) = 24.86, p < .0001; Cramer’s V = .137) indicating that the non-network PHPs were less likely to refer children for a developmental evaluation from the 12-month well visit than for the 18- or 24-month visits.

Diagnostic Effects

Psychologists diagnosed 648 toddlers within 7 weeks of screening (mean = 40.3 days). Upon first evaluation, 91.7% of the toddlers met criteria for one of the five delay categories (see Table 1). Only 8.3% had a false positive screening result. For toddlers with any developmental delay, the mean age of diagnosis was 22.1 months (SD = 6.56; range 12 to 38 months). Collapsed across screening age, toddlers with ASD were diagnosed at an average age of 22.9 months. However, the average age of diagnosis for toddlers who were screened at 12 months was 15.52 months, which is nearly identical to the results from the Get SET implementation in San Diego²³.

Table 1.

Get SET Participant characteristics and diagnostic outcomes.

	Diagnostic Category (N=648)
	ASD	ASD Features	LD	DD	GDD	TD
	(n=308)	(n=85)	(n=58)	(n= 128)	(n=15)	(n=54)
Sex, M/F	247/61	65/20	40/18	87/41	9/6	29/25
* Screen Age 12 months (n = 199)	n=73	n=30	n=23	n=49	n=2	n=22
Age at Screen M(SD)	12.59 (1.19)	12.80 (1.19)	12.74 (1.21)	12.47 (1.19)	13.50 (2.12)	12.50 (1.01)
Age at 1st Diagnosis M(SD)	15.52 (3.40)	15.93 (4.92)	14.57 (2.35)	16.45 (6.54)	17.50 (4.95)	18.55 (8.87)
* Screen Age 18 months (n = 190)	n=87	n=19	n=20	n=40	n=5	n=19
Age at Screen M(SD)	18.01 (.69)	18.17 (.86)	18.15 (.37)	18.03 (.70)	18.00 (.71)	18.11 (.66)
Age at 1st Diagnosis M(SD)	20.31 (2.67)	20.26 (1.76)	20.35 (1.63)	20.10 (1.78)	20.40 (1.67)	20.11 (1.94)
* Screen Age 24 months (n = 142)	n=73	n=23	n=8	n=22	n=4	n=12
Age at Screen M(SD)	23.15 (1.43)	22.74 (1.57)	23.63 (1.06)	22.59 (1.79)	24 (0)	23.50 (.91)
Age at 1st Diagnosis M(SD)	25.36 (3.05)	25.00 (3.05)	26.13 (1.36)	25.27 (2.62)	26.50 (1)	25.25 (2.34)
a >24 months (n = 117)	n=75	n=13	n=7	n=17	n=4	n=1
Age at 1st Diagnosis M(SD)	30.85 (3.41)	32.77 (3.44)	30.57 (4.43)	29.71 (2.85)	35.25 (3.10)	33 (−)
Ethnicity
Hispanic/Latino %	34.1	22.4	41.4	46.9	66.7	27.8
Not-Hispanic/Latino %	62.9	74.1	56.9	53.1	26.7	70.4
Not Reported %	3	3.5	1.7	0	6.6	1.8
Race
Caucasian %	68.2	72.9	58.6	60.2	46.7	72.2
Black/African Amer. %	5.5	8.2	0	6.2	0	5.6
Asian %	4.2	4.7	1.7	2.4	6.6	3.7
Pacific Islander %	0.3	0	1.7	0	0	0
Native Amer./Alaska %	1.3	0	0	2.4	0	1.8
Mixed Race %	7.8	3.5	5.2	6.2	0	5.6
Not Reported %	12.7	10.7	32.8	22.6	46.7	11.1
MSEL T Score
Visual Reception M (SD)	35.02 (13.63)	47.69 (13.34)	49.34 (8.45)	44.34 (13.02)	23.13 (4.31)	55.65 (11.11)
Fine Motor M(SD)	30.10 (10.88)	41.18 (10.37)	46.88 (9.47)	40.26 (12.28)	22.60 (6.66)	48.89 (7.74)
Receptive Language M(SD)	29.11 (11.87)	39.76 (12.06)	39.78 (12.99)	37.17 (11.84)	23.67 (5.19)	49.70 (9.13)
Expressive Language M(SD)	29.39 (11.88)	38.40 (11.59)	33.53 (8.95)	33.88 (10.43)	20.67 (1.49)	50.31 (6.90)
ELC M(SD)	66.25 (17.95)	84.78 (17.27)	85.41 (11.84)	79.55 (17.57)	52.33 (4.56)	102.39 (12.34)
Vineland II Stan Score
Communication M(SD)	71.73 (14.37)	85.56 (12.92)	83.88 (8.42)	81.73 (13.90)	60.60 (10.56)	95.76 (8.47)
Daily Living M(SD)	79.29 (12.86)	93.53 (11.01)	100.29 (10.65)	89.62 (12.62)	68.80 (8.99)	101.39 (9.21)
Socialization M(SD)	75.64 (11.88)	89.24 (11.85)	94.03 (9.60)	87.07 (11.25)	68.20 (8.66)	98.15 (9.66)
Motor Skills M(SD)	81.95 (11.54)	89.08 (10.43)	94.02 (8.57)	85.14 (10.51)	66.27 (8.52)	95.65 (9.43)
Adap Beh Com M(SD)	74.16 (11.44)	87.19 (10.94)	91.40 (8.52)	83.41 (11.24)	63.00 (7.47)	97.02 (8.60)
ADOS Score
SA
Module T M(SD)	13.28 (3.92)	7.95 (2.78)	3.61 (1.98)	4.65 (2.84)	-	2.78 (1.67)
Module 1 M(SD)	10.86 (3.54)	5.05 (2.28)	2.56 (1.81)	2.70 (2.11)	5.13 (4.63)	2.00 (1.67)
Module 2 M(SD)	8.86 (2.81)	4.73 (2.26)	2.67 (0.58)	3.00 (1.23)	-	2.19 (1.57)
RRB
Module T M(SD)	5.50 (1.62)	4.51 (1.73)	2.26 (1.20)	2.74 (1.44)	-	2.67 (1.41)
Module 1 M(SD)	6.30 (1.16)	4.23 (1.31)	2.44 (1.42)	3.30 (1.41)	4.47 (1.46)	3.17 (1.17)
Module 2 M(SD)	5.86 (1.19)	4.15 (1.05)	2.67 (0.58)	2.92 (0.76)	-	2.57 (1.29)
Total
Module T M(SD)	18.78 (4.85)	12.46 (3.31)	5.87 (2.21)	7.38 (2.97)	-	5.44 (2.14)
Module 1 M(SD)	17.17 (4.04)	9.27 (1.78)	5.00 (1.41)	6.00 (2.41)	9.60 (4.79)	5.17 (1.72)
Module 2 M(SD)	14.71 (3.14)	8.88 (2.36)	5.33 (0.58)	5.92 (1.19)	-	4.76 (1.87)

Note.

^*If a toddler was screened at an age that was not 12, 18 or 24 months, they were placed in a category closest to their screen age. For example, a toddler screened at 19 months would be placed in the 18-month category.

^aNot screened, but referred for evaluation.

In further analysis, the age of first evaluation from the Arizona site of the ADDM network 2014 surveillance year (N = 349) ³¹ were compared to the data from the Get Set Early implementation (n = 648). The ADDM data set included 8-year-old children with autism (n = 311) and children with ASD-NOS (n = 38; see Table 2). The mean age of first evaluation (M = 49.09, SD = 20.98) was significantly older than the mean age from Get SET (M = 22.94, SD = 6.36); t (386.22) = 23.70, p < .001.

Table 2.

ADDM comparison data collected during the study period (2014 −2019: Baio 2018)

CDC/ADDM Sample
	8 year olds*		4 year olds
	Autism	Autism-NOS	Autism	Autism-NOS
	(n=311)	(n=38)	(n = 271)	(n = 33)
Sex, M/F	236/75	32/6	225/46	27/6
Age group 12 months	n=12	n=1	n=21	n=2
Age at Evaluation M(SD)	10.08 (4.30)	10.00 (−)	9.81 (4.59)	2.50 (3.54)
Age group 18 months	n=5	n=0	n=12	n=1
Age at Evaluation M(SD)	18.60 (.89)		18.17 (.84)	16.00 (−)
Age group 24 months	n=31	n=0	n=34	n=3
Age at Evaluation M(SD)	22.10 (1.33)		21.85 (1.42)	21.67 (1.53)
Age group > 24 months	n=263	n=37	n=204	n=27
Age at Evaluation M(SD)	51.80 (17.38)	70.30 (15.06)	38.99 (12.84)	46.41 (18.55)
Ethnicity/Race (%)
White, Non-Hisp.	55.30	71.10	61.60	60.60
Hispanic	30.20	18.40	25.50	21.20
Black, Non-Hisp.	7.70	5.30	6.30	6.10
Asian/Pacific Isl., Non-Hisp.	3.20	0.00	2.20	3.00
Native Amer./Alaska, Non-Hisp.	1.30	0.00	2.20	3.00
Mixed Race	1.90	2.60	1.80	6.10
Not Reported	0.40	2.60	0.40	0.00

^*If a toddler was screened at an age that was not 12, 18 or 24 months, they were placed in a category closest to their screen age. For example, a toddler screened at 19 months would be placed in the 18-month category.

As the ADDM sample includes children who are 8 years old in the surveillance year and the Get SET sample includes children up to age 3 years, an ADDM sample collected from 4-year-old children (n = 304; age range 0 to 4 years, 9 months; see Table 2) offered a better comparison. Only the ASD or ASD-features cases from the Get SET Early program (n = 393) were included in this analysis. The age of first evaluation from the ADDM 4-year-old children (M = 31.51, SD = 11.36) was significantly older than the age from the Get SET Early program (M = 22.72, SD = 6.44); t (427.41) = 11.69, p < .001, d = .99). Results should be interpreted with caution as Get SET yields data from a self-selected sample.

Treatment

Of the 350 children (54% of the total sample) who completed at least one follow up evaluation, 323 of them (92%) were diagnosed with one of the 5 developmental delay categories. Of those who were diagnosed, 239 (74%) were enrolled in variety of treatment programs: 56% started speech therapy by an average of 3.47 months after evaluation (range = 0.03 to 24.22, SD = 3.99). Children diagnosed with ASD = 201 (57%), 91% were enrolled in a treatment program within an average of 4.52 months after diagnosis (range = .26 to 12.26, SD = 3.11) : 74% enrolled in speech, but only 22% enrolled in ABA (e.g., discreet trial, pivotal response training, parent training, or incidental teaching) (see Supplementary Table 3).

DISCUSSION

This study demonstrated that a community-based center successfully implemented the Get SET Early model. PHP’s demonstrated improved screening and referral behaviors and resulted in a significantly lower age of diagnosis (22 months) for 648 toddlers. Thus, Get SET is effective at encouraging early screening and diagnosis.

PHP surveys indicated improved attitudes and behavior about early screening after implementing Get SET. Increased use of a standardized screening tool at all screening ages is a clear indicator of the model’s positive impact. The 60% increase in screening at the 12-month visit is especially important because it led to a 15 month mean age of ASD diagnosis for these children. Evidence shows that a diagnosis at 14 months is stable throughout childhood ⁴. Without Get SET, screening occurs primarily at older ages, and less regularly as data from the comparison network indicated substantially fewer screenings without the model.

Determining reasons for the consistently high median age of ASD diagnosis in Arizona as reported by the CDC ^{30, 8} is beyond the scope of this study, but the lack of regular structured screening at early ages is a clear contributing factor. The perception of limited resources is an additional factor. Without a reliable diagnostic center, pediatricians are less likely to refer a very young toddler for an evaluation until they are certain significant delays are present. PHP’s may be hesitant in general to refer children of any age if their symptoms are subtle considering that 117 evaluated children were older than 24 months (18% of the sample), the ceiling age of the CSBS-ITC. Thus, the creation of an early diagnostic center is a key factor in successful implementation and should be the first step in every city to replicate Get SET.

A primary goal of Get SET is to facilitate treatment engagement for all toddlers with ASD by 36 months. Of the 350 children who completed a second evaluation, 74% began a treatment program. Still, approximately 26% of children in this study were without treatment 1 year after diagnosis. Possible reasons include parent denial, lack of concern because of mild impairment, services being denied, unsuccessful navigation of state systems, or long waitlists. Further, while 74% were enrolled in a treatment program only 22% of children who were diagnosed with ASD began an ABA program, which may indicate limited resources.

Future Directions

Perhaps the strongest evidence that Get SET is an effective model is the communities’ ability to maintain the process after the study. In Arizona, a majority of the pediatric sites in the study adopted the structured procedures for screening and referral. That is, 62 pediatricians are screening at the 12-, 18-, and 24-month well visits with the CSBS-ITC and have referral relationships with the study center. At the time of this report, an additional 16 practices have joined.

To accommodate the expanding PHP network, the study center developed a diagnostic evaluation network, which includes licensed psychologists with experience in the early diagnosis of ASD who accept a variety of payment options. A network coordinator receives referrals from the pediatrician and connects the family with a provider who accepts their form of payment. Expedited access is provided in most cases because providers hold open spots in their schedule that will be filled by the network coordinator. In the future, the center will grow the diagnostic network by providing training, insurance credentialing, and insurance billing for new providers, reserving a percentage of the total amount billed per case to sustain the network coordinator position. This translational approach will allow the center to maintain Get SET procedures, grow the pediatrician and provider networks, and expand the positive results of this project to the entire state.

Limitations

Get SET generated 45,504 screens that were completed and scored by hand which adds challenges to data analysis. Electronic screening methods would prove more efficient ³¹. Get SET was implemented digitally in San Diego and their results illustrate the many benefits beyond accurate tracking²³.

The obvious disparity between the number of failed screens and evaluations indicates that a substantial number of parents do not follow through with an evaluation based on a failed screening. The CSBS-ITC can have a high 25% false positive rate ³, which may explain some of the disparity. Parents may not see delays in their child despite screening results. Other parents may see differences, but their child’s behavior does not cause problems for the individual or family, so there is limited motivation to seek an evaluation. Longitudinal follow up to determine the number of children eventually diagnosed would be valuable data to support the efficacy of this process and perhaps encourage parents and professionals to act earlier.

As there were only comparisons to estimates of community practice (comparison PHP network and ADDM) regarding screening and diagnosis and no true control group, it is possible that PHP practices improved during the study period due to general growing awareness of the importance of early ASD detection. Given that there have not been significant improvements in population data since the end of this project in 2019, this possibility seems unlikely. Still, the inclusion of a true control group would have empirically eliminated that possibility and enhanced the interpretation of results.

Conclusions

Overall, results from the current study indicate a successful implementation of Get SET by a community-based center in an independent city with limited resources for early evaluations and treatment. There is evidence for efficacy on changing PHP attitudes and behaviors on screening, lowering the age of ASD diagnosis, and allowing families to begin treatment before 36 months. Replication would help lower the national age of ASD diagnosis and give families several additional years to participate in treatment. Given the enormous potential of neural plasticity in the first years of life ³², Get SET Early is critical for creating opportunity for people with ASD to begin treatment earlier and reach their individualized optimal functioning.